American Online Phonebook

Tag: From

Daily Crunch: Apple removes Fortnite from the App Store

August 14, 2020 No Comments

Epic Games takes on Apple, Instagram fixes a security issue and Impossible Foods raises $ 200 million. This is your Daily Crunch for August 13, 2020.

The big story: Apple removes Fortnite from the App Store

The controversy over Apple’s App Store policies has expanded to include Epic Games and its hit title Fortnite. The company introduced a direct payment option for its in-game currency on mobile, leading Apple to remove the app for violating App Store rules.

“Epic enabled a feature in its app which was not reviewed or approved by Apple, and they did so with the express intent of violating the App Store guidelines regarding in-app payments that apply to every developer who sells digital goods or services,” Apple said.

Epic, meanwhile, said it’s taking legal action against Apple, and that the game’s removal is “yet another example of Apple flexing its enormous power in order to impose unreasonable restraints and unlawfully maintain its 100% monopoly over the iOS In-App Payment Processing Market.”

The tech giants

Bracing for election day, Facebook rolls out voting resources to US users — The hub will centralize election resources for U.S. users and ideally inoculate at least some of them against the platform’s ongoing misinformation epidemic.

Instagram wasn’t removing photos and direct messages from its servers — A security researcher was awarded a $ 6,000 bug bounty payout after he found Instagram retained photos and private direct messages on its servers long after he deleted them.

Slack and Atlassian strengthen their partnership with deeper integrations — At the core of these integrations is the ability to get rich unfurls of deep links to Atlassian products in Slack.

Startups, funding and venture capital

Impossible Foods gobbles up another $ 200 million — Since its launch the plant-based meat company has raised $ 1.5 billion from investors.

Omaze raises $ 30 million after expanding beyond celebrity campaigns — The Omaze model has shifted away from celebrity-centric campaigns to include fundraisers offering prizes like an Airstream Caravel or a trip to the Four Seasons resort in Bora Bora.

We’re exploring the future of SaaS at Disrupt this year — We’re bringing Canaan Partners’ Maha Ibrahim, Andreessen Horowitz’s David Ulevitch and Bessemer Venture Partners’ Mary D’Onofrio together to help explain how the landscape has changed.

Advice and analysis from Extra Crunch

How to get what you want in a term sheet — Lior Zorea discusses the reality of term sheets.

Five success factors for behavioral health startups — Courtney Chow and Justin Da Rosa of Battery Ventures argue that behavioral health is particularly suited to benefit from the digitization trends COVID-19 has accelerated.

Minted.com CEO Mariam Naficy shares ‘the biggest surprise about entrepreneurship’ — Naficy got into the weeds with us on topics that founders don’t often discuss.

Everything else

Digital imaging pioneer Russell Kirsch dies at 91 — It’s hard to overstate the impact of his work, which led to the first digitally scanned photo and the creation of what we now think of as pixels.

AMC will offer 15-cent tickets when it reopens 100+ US theaters on August 20 — The theater juggernaut announced plans to reopen more than 100 theaters in the U.S. on August 20.

The Daily Crunch is TechCrunch’s roundup of our biggest and most important stories. If you’d like to get this delivered to your inbox every day at around 3pm Pacific, you can subscribe here.

Mobile – TechCrunch

Facebook extends coronavirus work from home policy until July 2021

August 9, 2020 No Comments

Facebook has joined Google in saying it will allow employees to work from home until the middle of next year as a result of the coronavirus pandemic.

“Based on guidance from health and government experts, as well as decisions drawn from our internal discussions about these matters, we are allowing employees to continue voluntarily working from home until July 2021,” a spokeswoman told the Reuters news agency.

Facebook also said it will provide employees with an additional $ 1,000 to spend on “home office needs.”

Late last month Google also extended its coronavirus remote work provision, saying staff would be able to continue working from home until the end of June 2021.

Both tech giants have major office presences in a number of cities around the world. And despite the pandemic forcing them into offering more flexible working arrangements than they usually do, the pair have continued to build out their physical office footprints, signaling a commitment to operating their own workplaces. (Perhaps unsurprisingly, given how much money they’ve ploughed in over the years to turn offices into perk-filled playgrounds designed to keep staff on site for longer — with benefits such as free snacks and meals, nap pods, video games arcade rooms and even health centers.)

Earlier this month, Facebook secured the main office lease on an iconic building in New York, for example — adding 730,000 square feet to its existing 2.2 million square feet of office space. Google has continued to push ahead with a flagship development in the U.K. capital’s King’s Cross area, with work resuming last month on the site for its planned London “landscraper” HQ.

In late July, Apple said staff won’t return to offices until at least early 2021 — cautioning that any return to physical offices would depend on whether an effective vaccine and/or successful therapeutics are available. So the iPhone maker looks prepared for a home-working coronavirus long haul.

As questions swirl over the future of the physical office now that human contact is itself a public health risk, the deepest pocketed tech giants are paradoxically showing they’re not willing to abandon the traditional workplace altogether and go all in on modern technologies that allow office work to be done remotely.

Twitter is an exception. During the first wave of the pandemic the social network firmly and fully embraced remote work, telling staff back in May that they can work from home forever if they wish.

Whether remote work played any role in the company’s recent account breach is one open question. It has said phone spear phishing was used to trick staff to gain network access credentials.

Certainly, security concerns have been generally raised about the risk of more staff working remotely during the pandemic — where they may be outside a corporate firewall and more vulnerable to attackers.

A Facebook spokeswoman did not respond when we asked whether the company will offer its own staff the option to work remotely permanently. But the company does not appear prepared to go so far — not least judging by signing new leases on massive office spaces.

Facebook has been retooling its approach to physical offices in the wake of the COVID-19 pandemic, announcing in May it would be setting up new company hubs in Denver, Dallas and Atlanta.

It also said it would focus on finding new hires in areas near its existing offices — including in cities such as San Diego, Portland, Philadelphia and Pittsburgh.

Facebook CEO Mark Zuckerberg said then that over the course of the next decade half of the company could be working fully remotely. Though he said certain kinds of roles would not be eligible for all-remote work — such as those doing work in divisions like hardware development, data centers, recruiting, policy and partnerships.


Social – TechCrunch

How to use trending keywords from current events in content marketing

July 26, 2020 No Comments

30-second summary:

  • Companies can rank for trending keywords by combining content marketing with current events.
  • There are some specific Dos and Don’ts for creating this content in a tasteful manner.
  • Dos include questioning your motives, thinking of clients first, keeping content value-driven, and giving company updates.
  • Don’ts include pretending nothing is going on, posting irrelevant content, abandoning your existing SEO strategy, or giving up.
  • VP of sales and marketing at Strategic Sales & Marketing, helps you achieve SEO value while maintaining sensitivity using trending keywords.

A good content marketing strategy should already be backed up by a list of target keywords or keyphrases your business is trying to rank for. But alongside these evergreen keywords, incorporating trending search terms into content marketing can boost SEO significantly. There’s already plenty of advice for finding trending keywords, but how do we harness their power without being distasteful?

Approaching serious current events in content marketing can be a tricky wire to walk. In 2020 alone, businesses and marketing teams everywhere have struggled to decide how to use keywords related to the COVID-19 crisis, the Black Lives Matter movement, Brexit, increasing climate change, and even a sighting of murder hornets. To avoid embarrassment and truly rank for important trending searches in an authentic way, follow these dos and don’ts.

Do: Question your motives

A national or global news story breaks, new keywords start trending immediately, and the first thing you think is…

If the answer was “Ooo I can use this”, then you may not have the best motives for adding that current event to your content marketing. Stopping to question your motives is the first step to harnessing the power of trending keywords without making a marketing blunder. It should and will take deep thought as to how to approach a trending topic with sensitivity.

Ranking for trending keywords shouldn’t be just about good SEO. Your keyword usage serves as a connection between what is important to both a brand and its consumers. When the great toilet paper shortage of 2020 occurred, bath tissue companies could have easily changed their marketing to fuel the need and push profits even more. But brands like Cottonelle took the exact opposite approach with their #ShareASquare social media campaign and partnership with United Way. Content marketing should always be motivated first and foremost by your business values, not your bottom line.

Don’t: Pretend nothing is going on

Some companies read the news and move right along with business as usual. That’s okay if you don’t have the right motives or anything significant to add value for your audience. However, the conversation surrounding authentic branding is growing, and more and more consumers are wanting to see companies respond to important events with transparency and empathy. 

In a 2019 Deloitte study, 55% of survey respondents reported believing that “businesses today have a greater responsibility to act on issues related to their purpose”. It’s okay to ignore which Kardashian is getting divorced when it comes to your marketing. But issues like climate change, world health, and racism cannot be ignored or you risk alienating consumers who equate silence with complicity. 

Do: Think of your clients first

Any good company already has their customers at the center of their content marketing strategy. Now more than ever, businesses need to refocus their marketing and make sure they are putting their customers’ needs first. This can be difficult in a climate where world events are creating new and different needs seemingly daily.

Consumers are shifting their needs and preferences to new types of content and new ways of engaging with companies. In particular, educational content is gaining popularity and driving “how to” searches. Think of what your clients need to know or hear in order to interact with your business, products, or services. The client always comes first.

Don’t: Provide irrelevant content

Anything you share concerning current events needs to be connected to your business in some way or another. Posting something completely irrelevant or out of the blue may throw off and leave a sour taste with customers and followers. If you can’t tie the content directly to your customer’s needs, then it’s likely not relevant enough to share.

One way to connect your business with seeming-distant trending topics is to think of secondary trending searches still related to your industry or product. COVID-19 has changed the way people live which has led to a wide variety of spiking trends such as gardening or home haircuts. There’s always an authentic way to connect a current event to your company. 

Do: Keep your content marketing value-driven

Even if you’re making a commentary or stating a position or opinion, you still need to add value to your content. Every message should have a takeaway that readers can apply to better themselves or their lives. Sometimes the value is in buying your service or product, but other times the value lies in the emotional connection imparted to the reader.

Think, what does value mean to your clients right now? Nike is a great example of a company providing new value to consumers. With gyms everywhere shutting down, Nike released their workout content for free and ramped up posting blogs to its apps and website. Their latest trend-focused content features celebrities challenging at-home exercisers to various workouts. Creating this value-driven content allows them to rank for many trending 2020 searches such as “at-home workout” as well as various trending athletes and celebrities.

Source: Google Trends

Don’t: Abandon your content marketing strategy

In turbulent times, your audience needs some nuggets of normalcy. A good content marketing strategy will provide the flexibility to adapt to sudden world changes or important events that need commenting on without abandoning the original plan. That being said, you’ll probably need to pivot on a few things, or at least give your audience a heads up about why they may still be seeing the content you already had planned.

Following travel restrictions due to Covid-19, Travel Zoo issued an email and blog statement explaining why they were going to continue on with their email series promoting travel, even when they knew their audience wouldn’t need their services right then. They add that they will be offering the same great content and deals they had planned, so email subscribers will “continue to find experiences that inspire and enlighten, whether in fantastic locations around the world or through something new you can try right in your home”.

Abandoning evergreen keywords for trending ones can lead to a big drop in your overall SEO. Remember, these searches are “trending” for a reason. That means, just as quickly as consumers are finding your recent content marketing, they are moving on to new trends and keywords.

Do: Give company updates

Once you decide to approach a trending topic in a blog, email, or any piece of content, it’s good to give clients company updates related to your first statement. This creates more opportunities for using related trending searches without keyword-stuffing your original content. This also shows consumers that you follow through on your promises which builds customer–brand rapport.

A great example of a constant commitment to this is Ben & Jerry’s Issues We Care About blog. These articles keep consumers interested in the company while ranking their content high in trending search results. 

Ben & Jerry's blog example

Source: benjerry.com

Don’t: Give up

Using trending keywords related to current events is key to helping consumers find your content. Just because it may take more consideration to create does not mean it’s worth skipping. “Going dark” can hurt exposure, engagement, and sales. And even businesses that weather the storm will face further recovery time if their company was out of mind due to lack of marketing.

Times are tough, there’s no doubt about it. But when it comes to keeping your business running in the face of global catastrophe, you cannot give up, especially on your marketing. Rather, content marketing and reaching clients and consumers at home requires your doubled commitment. You can make your content a place people turn to for wisdom and perspective—all while scoring those trending keyword SEO points.

Gregg Schwartz is the VP of sales and marketing at Strategic Sales & Marketing, a lead-generation firm based in Connecticut.

The post How to use trending keywords from current events in content marketing appeared first on Search Engine Watch.

Search Engine Watch

Lessons from lockdown: Four content types that users really engage with

July 21, 2020 No Comments

30-second summary:

  • Life in lockdown led to a huge uptake in media consumption, including a 44% worldwide increase in social media use according to Statista.
  • While the COVID-19 pandemic has shell shocked the world, many industries were affected
  • Many businesses not only stayed afloat but actually managed to attract new audiences through their intelligent use of content marketing. 
  • CEO of Go Up Ltd. shares four engaging content types that have seen businesses succeed during the lockdown.

The COVID-19 pandemic and subsequent lockdown restrictions have meant businesses have needed to react quickly and adapt strategies in order to continue trading. Retailers were forced online and restaurants had to offer a takeaway-only service, with limited menus, for months. Some industries, such as pubs and the hospitality sector, ground to a complete halt. 

However, many businesses not only stayed afloat but actually managed to attract new audiences through their intelligent use of content marketing. Here are four engaging types of content that have seen businesses succeed during the lockdown. 

1. Community content

Once upon a time, brands — particularly small, independent businesses — relied on their existing customer base purchasing items or placing pre-orders at times when business was not able to run as usual. However, as the “new normal” set in and attention focused predominantly on COVID-19, business owners were forced to adapt as swiftly as possible to their customers’ changing priorities.  

As such, existing marketing plans were shelved, making way for new coronavirus-aware content strategies that placed emphasis on empathy over commercialism. With so much financial uncertainty, customers were no longer interested in the latest products to hit the market. Instead, businesses needed to step up and show that they care about both their consumers and staff, demonstrating how they can provide value and help make a difference during the outbreak. 

Building community with your audience, adjusting to the new normal, and showing how you’re sticking together behind the scenes is impactful. Aldi, for example, shared updates on how they were investing in supporting local communities during the outbreak. While this type of content doesn’t yield a huge amount of income in the short term, it does help to build brand affinity, and ultimately sales, in the long term.

A 2019 study found that content that focuses on brand purpose triggers a more positive physical and emotional reaction in consumers than those that focus on a product, with 83% of consumers more likely to be loyal to a brand that does so.  

2. Educational content

Many people have used their newfound free time during lockdown as an opportunity for continued education. Not only has this been an opportunity to develop professional skills, but it’s also been a method of remaining engaged and keeping positive mental health during the lockdown. As such, more people than ever before have been engaging with educational content online.  

During the pandemic, LinkedIn has seen its highest levels of engagement, with users watching over 4 million hours of content on LinkedIn Learning soon after lockdown went into effect. LinkedIn professionals are hosting live chats where they share data-driven and real insight with their audiences. 

Other brands have provided free-to-use resources to help users upskill during this time. Moz, for example, allowed users to access their usually paid-for SEO academy courses for free. Meanwhile, recognizing that customers would be unable to go on holiday as normal, Pasta Evangelists’ Italy at Home campaign provided users with authentic recipes to try that would help to bring some Italian escapism into their homes. While these content strategies have no immediate commercial value, those that did access the training sessions or recipes during the lockdown will have a stronger affinity towards each brand and will be more likely to become a paying customer in the future.  

3. Uplifting and entertaining content

Negative news stories have been hard to avoid, whether it be related to coronavirus deaths, political upheaval, or social injustices. It’s, therefore, no surprise that many consumers have sought some relief from reality, with lighthearted and entertaining becoming more popular than ever before.  

According to a survey by Channel Factor, 80% of consumers head to their favorite vloggers on YouTube to improve their mood, with 69% of respondents finding that content more uplifting than those on other channels. Of the most popular videos streamed on YouTube, almost half of them were entertainment videos (48%), while 33% were comedy focused.  

Innocent Smoothies, well known for their relaxed approach to content marketing, set daily work from home challenges to keep their community entertained, while The Woodland Trust provided tips on how to keep children engaged with nature despite being forced to stay at home during the lockdown.  

Creating fun and lighthearted content with no commercial interest can work well in times of uncertainty, growing your social media audience, and building brand affinity in the process. Alternatively, brands can invest in influencer marketing and collaborate with content creators who have their own established audiences to help grow their own audience. 

4. Omnichannel content

Life in lockdown led to a huge uptake in media consumption, including a 44% worldwide increase in social media use according to Statista. With more people browsing through various social media channels, it’s more important than ever for brands to be where their audience is, especially as online commerce continues to rise. 

It’s been predicted that digital experiences will be even more important following the pandemic. It’s become clear that many aspects of customer interaction will need to be digitized to abide by social distancing measures. So now more than ever, it’s crucial that brands create a seamless online experience for their audiences in order to avoid any confusion between platforms and encourage shoppers with targeting ads and posts. 

Edward Coram James is an SEO professional and the Chief Executive of Go Up Ltd, an international agency dedicated to helping its clients navigate the complexities of global SEO and the technical aspects of delivering location-specific pages to targeted audiences.

The post Lessons from lockdown: Four content types that users really engage with appeared first on Search Engine Watch.

Search Engine Watch

Puppet announces $40 million debt round from BlackRock

July 19, 2020 No Comments

Puppet, the Portland, Oregon-based infrastructure automation company, announced a $ 40 million debt round today from BlackRock Investments.

CEO Yvonne Wassenaar says the company sees this debt round as part of a longer-term relationship with BlackRock . “What’s interesting, and I think part of the reason why we decided to go with BlackRock, is that typically when you look at how they invest this is the first step of a much longer-term relationship that we will have with them over time that has different elements that we can tap into as the company scales,” Wassenaar told TechCrunch.

In terms of the arrangement, rather than BlackRock taking a stake in the company, Puppet will pay back the money. “We’ve borrowed a sum of money that we will pay back over time. BlackRock does have a board observer seat, and that’s really because they’re very interested in working with us on how we grow and accelerate the business,” Wassenaar said.

Puppet has been in the process of rebuilding its executive team, with Wassenaar coming on board about 18 months ago. Last year she brought in industry veterans Erik Frieberg and Paul Heywood as CMO and CRO, respectively. This year she brought in former Cloud Foundry Foundation director Abby Kearns to be CTO.

All of these moves are with an eye to a future IPO, says Wassenaar. “We’re looking at how do we progress ultimately, ideally on a path to an IPO, and what is it going to take for Puppet to go through that journey,” she said.

She points out that in some ways, the pandemic has forced companies to look more closely at automation solutions like the ones that Puppet provides. “What’s really interesting is […] that the pandemic in many ways has put wind in our sails in terms of the need for corporations to automate and think about how they leverage and extend from a technology perspective going forward,” she said.

As Puppet continues to grow, she says that diversity is a core organizational value, and that while the company has made progress from a gender perspective (as illustrated by the presence of her and Kearns in the C Suite), they still are working at being more racially diverse.

“Where I believe we have a lot more work and there’s a lot more focus right now is further complementing that [gender diversity] from a racial perspective. And it’s an area that I have personally taken on, and I’m committed to making changes in the company as we go forward to support more racial diversity as well,” she said.

Previously the company had raised almost $ 150 million, with the most recent round being a $ 42 million Series F in 2018, according to Crunchbase data. The company previously took $ 22 million in debt financing in 2016, prior to Wassenaar coming on board.

Puppet names former Cloud Foundry Foundation executive director Abby Kearns as CTO


Enterprise – TechCrunch

CIO Cynthia Stoddard explains Adobe’s journey from boxes to the cloud

June 29, 2020 No Comments

Up until 2013, Adobe sold its software in cardboard boxes that were distributed mostly by third party vendors.

In time, the company realized there were a number of problems with that approach. For starters, it took months or years to update, and Adobe software was so costly, much of its user base didn’t upgrade. But perhaps even more important than the revenue/development gap was the fact that Adobe had no direct connection to the people who purchased its products.

By abdicating sales to others, Adobe’s customers were third-party resellers, but changing the distribution system also meant transforming the way the company developed and sold their most lucrative products.

The shift was a bold move that has paid off handsomely as the company surpassed an $ 11 billion annual run rate in December — but it still was an enormous risk at the time. We spoke to Adobe CIO Cynthia Stoddard to learn more about what it took to completely transform the way they did business.

Understanding the customer

Before Adobe could make the switch to selling software as a cloud service subscription, it needed a mechanism for doing that, and that involved completely repurposing their web site, Adobe.com, which at the time was a purely informational site.

“So when you think about transformation the first transformation was how do we connect and sell and how do we transition from this large network of third parties into selling direct to consumer with a commerce site that needed to be up 24×7,” Stoddard explained.

She didn’t stop there though because they weren’t just abandoning the entire distribution network that was in place. In the new cloud model, they still have a healthy network of partners and they had to set up the new system to accommodate them alongside individual and business customers.

She says one of the keys to managing a set of changes this immense was that they didn’t try to do everything at once. “One of the things we didn’t do was say, ‘We’re going to move to the cloud, let’s throw everything away.’ What we actually did is say we’re going to move to the cloud, so let’s iterate and figure out what’s working and not working. Then we could change how we interact with customers, and then we could change the reporting, back office systems and everything else in a very agile manner,” she said.


Enterprise – TechCrunch

Robocallers face $225M fine from FCC and lawsuits from multiple states

June 10, 2020 No Comments

Two men embodying the zenith of human villainy have admitted to making approximately a billion robocalls in the first few months of 2019 alone, and now face an FCC fine of $ 225 million and a lawsuit from multiple attorneys general that could amount to as much or more — not that they’ll actually end up paying that.

John Spiller and Jakob Mears, Texans of ill repute, are accused of (and have confessed to) forming a pair of companies to make millions of robocalls a day with the aim of selling health insurance from their shady clients.

The operation not only ignored the national Do Not Call registry, but targeted it specifically, as it was “more profitable to target these consumers.” Numbers were spoofed, making further mischief as angry people called back to find bewildered strangers on the other end of the line.

These calls amounted to billions over two years, and were eventually exposed by the FCC, the offices of several attorneys general and industry anti-fraud associations.

Now the pair have been slapped with a $ 225 million proposed fine, the largest in the FCC’s history. The lawsuit involves multiple states and varying statutory damages per offense, and even a conservative estimate of the amounts could exceed that number.

Unfortunately, as we’ve seen before, the fines seem to have little correlation with the amounts actually paid. The FCC and FTC do not have the authority to enforce the collection of these fines, leaving that to the Department of Justice. And even should the DoJ attempt to collect the money, they can’t get more than the defendants have.

FTC smacks down robocallers, but the penalties don’t match their heinous crimes

For instance, last year the FTC fined one robocaller $ 5 million, but he ended up paying $ 18,332 and the market price of his Mercedes. Unsurprisingly, these individuals performing white-collar crimes are no strangers to methods to avoid punishment for them. Disposing of cash assets before the feds come knocking on your door is just part of the game.

In this case the situation is potentially even more dire: the DoJ isn’t even involved. As FCC Commissioner Jessica Rosenworcel put it in a statement accompanying the agency’s announcement:

There’s something missing in this all-hands effort. That’s the Department of Justice. They aren’t a part of taking on this fraud. Why not? What signals does their refusal to be involved send?

Here’s the signal I see. Over the last several years the FCC has levied hundreds of millions in fines against robocallers just like the folks we have here today. But so far collections on these eye-popping fines have netted next to nothing. In fact, it was last year that The Wall Street Journal did the math and found that we had collected no more than $ 6,790 on hundreds of millions in fines. Why? Well, one reason is that the FCC looks to the Department of Justice to collect on the agency’s fines against robocallers. We need them to help. So when they don’t get involved—as here—that’s not a good sign.

While the FCC’s fine and the lawsuit will certainly put these robocallers out of business and place further barriers to their conducting more scam operations, they’re not really going to be liable for nine figures, because they’re not billionaires.

It’s good that the fines are large enough to bankrupt operations like these, but as Rosenworcel put it back in 2018 when another enormous fine was levied against a robocaller, “it’s like emptying the ocean with a teaspoon.” While the FCC and states were going after a pair of ne’er-do-wells, a dozen more have likely popped up to fill the space.

Industry-wide measures to curb robocalls have been underway for years now, but only recently have been mandated by the FCC after repeated warnings and delays. Expect the new anti-fraud frameworks to take effect over the next year.

Mobile – TechCrunch

Search Suggestions from Previously Submitted Searcher Queries

May 20, 2020 No Comments

I came across an interesting Search Engine Land post last week. It inspired me to search and see if I could find a patent that might be related to it from Google:

Google is suggesting searches based on users’ recent activity

I tried to reproduce the search suggestions that were being shown to the author of the Search Engine Land article, but Google would not return those to me. Google may be experimenting with a limited number of searchers rather than showing those results to everyone. I did find a patent discussing search suggestions that were similar.

When Google shows a search suggestion about something you may have searched for in the past, that predicted suggestion is likely related to a patent I’ve written about before, Autocompletion using previously submitted query data.

I wrote about that patent being updated in a continuation patent, but hadn’t provided much in the way of details about how it works at: How Google Predicts Autocomplete Query Suggestions is Updated.

There are some interesting parts about how search suggestions are identified and ranked, which inspired me to write this post.

Search Suggestions Based on Previously Submitted Query Data

The description of this patent starts off by telling us that it is about: “using previously submitted query data to anticipate a user’s search request.”

That pinpoints that Google has a long memory, and it remembers a lot about what someone might search for.

This patent description also includes a lot of the assumptions that search engineers make about searchers (often an interesting reason to read through patents). Here are some from this patent that are worth thinking about:

Internet search engines aim to identify documents or other items that are relevant to a user’s needs and to present the documents or items in a manner that is most useful to the user. Such activity often involves a fair amount of mind-reading–inferring from various clues what the user wants. Certain clues may be user-specific. For example, the knowledge that a user is making a request from a mobile device, and knowledge of the location of the device, can result in much better search results for such a user.

Clues about a user’s needs may also be more general. For example, search results can have elevated importance, or inferred relevance, if a number of other search results link to them. If the linking results are themselves highly relevant, then the linked-to results may have particularly high relevance. Such an approach to determining relevance may be premised on the assumption that, if authors of web pages felt that another web site was relevant enough to be linked to, then web searchers would also find the site to be particularly relevant. In short, the web authors “vote up” the relevance of the sites.

Other various inputs may be used instead of, or in addition to, such techniques for determining and ranking search results. For example, user reactions to particular search results or search result lists may be gauged, so that results on which users often click will receive a higher ranking. The general assumption under such an approach is that searching users are often the best judges of relevance, so that if they select a particular search result, it is likely to be relevant, or at least more relevant than the presented alternatives.

A Summary of the Search Suggestions Process Based on Previous Submitted Queries

Like most patents, the Description for this one starts out with a summary section that provides an overview of how the process defined in the patent works. It is followed by a “Detailed Description” section that goes into more depth and provides details about how search at Google works, and how specific aspects of search at Google power this search suggestion process. So read about how search suggestions might be provided based upon user queries that have been searched for previously, and then read for the more detailed explanation, which goes way beyond autocomplete.

In the summary section of the description for the patent, we are told about how the patent may address those assumptions:

When anticipating user search requests, responding to the algorithm in this patent can involve certain methods for processing query information. Those include:

  • Receiving query information at a server system, with a portion of a query from a searcher
  • Obtaining a set of predicted queries relevant to the portion of the searcher’s query based on query and data indicative of the searcher relative to previously submitted queries
  • Providing the set of predicted queries to the searcher

The patent also points out additional features involved in the process such as obtaining the predicted queries including ordering the set of predicted queries based upon ranking criteria.

Those ranking criteria may be based upon the data indicative of searcher’s behavior relative to previously submitted queries.

Data about the searcher’s behavior regarding those previously submitted queries may include:

  • Click data
  • Location-specific data
  • Language-specific data
  • Other similar types of data

The patent points out the following as advantages of following the process described in the patent:

A search assistant receives query information from a search requestor, prior to a searcher completely inputting the query.

Information associated with previous user (or users) searches (such as click data associated with search results) is collected. From the query information and the previous search information, a set of predicted queries is produced and provided to the search requestor for presentation.

The patent can be found at:

Autocompletion using previously submitted query data
Inventors: Michael Herscovici, Dan Guez, and Hyung-Jin Kim
Assignee: Google Inc.
US Patent: 9,740,780
Granted: August 22, 2017
Filed: December 1, 2014

Abstract

A computer-implemented method for processing query information includes receiving query information at a server system. The query information includes a portion of a query from a search requestor. The method also includes obtaining a set of predicted queries relevant to the portion of the search requestor query based upon the portion of the query from the search requestor and data indicative of search requestor behavior relative to previously submitted queries. The method also includes providing the set of predicted queries to the search requestor.

Analysis of Ranking and Selection of Search Suggestions Based Upon Previous Query Data

The “Detailed Description” section of this search suggestions patent provides some insightful analysis about search at Google.

Relevance and Backlinks and a Rank Modifying Engine Lead to Ranking For Many Results at Google

This patent points out some of how search works at Google. It tells us that:

  1. The purpose of the process in the patent is to “improve the relevance of results obtained from submitting search queries.”
  2. It describes the ranking of documents for a query as something that can be “performed using traditional techniques for determining an information retrieval (IR) score for indexed documents in view of a given query.” And relevance of a particular document with respect to a query term may be determined by a technique, such as looking at the general level of back-links to a document that contain matches for a search term that may be used to infer a document’s relevance. As the patent tells us:

    In particular, if a document is linked to (e.g., is the target of a hyperlink) by many other relevant documents (e.g., documents that also contain matches for the search terms), it can be inferred that the target document is particularly relevant. This inference can be made because the authors of the pointing documents presumably point, for the most part, to other documents that are relevant to their audience.

  3. We are given more details about some results being even more relevant than ones with backlinks. We are told that:

    If the pointing documents are in turn the targets of links from other relevant documents, they can be considered more relevant, and the first document can be considered particularly relevant because it is the target of relevant (or even highly relevant) documents. Such a technique may be the determinant of a document’s relevance or one of multiple determinants. The technique is exemplified in some systems that treat a link from one web page to another as an indication of quality for the latter page, so that the page with the most such quality indicators is rated higher than others. Appropriate techniques can also be used to identify and eliminate attempts to cast false votes so as to artificially drive up the relevance of a page.

  4. There is another step that could potentially make some results even more relevant that involve what is referred to as a rank modifier engine:

    To further improve such traditional document ranking techniques, the ranking engine can receive an additional signal from a rank modifier engine to assist in determining an appropriate ranking for the documents. The rank modifier engine provides one or more prior models, or one or more measures of relevance for the documents based on one or more prior models, which can be used by the ranking engine to improve the search results’ ranking provided to the user. In general, a prior model represents a background probability of document result selection given the values of multiple selected features, as described further below. The rank modifier engine can perform one or more of the operations described below to generate the one or more prior models, or the one or more measures of relevance based on one or more prior models.

    5. This is a more detailed description of ranking than we normally see at Google. The section above references a Rank Modifier Engine that will be described in more detail further down this post

    Indexing, Scoring, Ranking, and Rank Modifier Engine

    ranking search suggestions

    The information retrieval system from this patent includes a number of different components:

    • Indexing engine
    • Scoring engine
    • Ranking engine
    • Rank modifier engine

    The indexing engine can function as described in the section above for the indexing engine.

    Scoring Engine

    In addition, a scoring engine may provide scores for document results based on many different features including:

    • Content-based features that link a query to document results
    • query-independent features that generally indicate the quality of document results

    Content-based features include aspects of document format, such as Query matches to title or anchor text in an HTML (HyperText Markup Language) page.

    The query-independent features can include aspects of document cross-referencing, such as a rank of the document or the domain.

    Moreover, the particular functions used by the scoring engine can be tuned, to adjust the various feature contributions to the final IR score, using automatic or semi-automatic processes.

    Ranking Engine

    A ranking engine can produce a ranking of document search results for display to a searcher based on IR scores received from the scoring engine and possibly one or more signals from the rank modifier engine.

    A tracking component may be used to record information about individual searcher selections of the search results presented in the ranking. The patent describes how selections may be tracked using javascript or a proxy system or a toolbar plugin:

    For example, the tracking component can be embedded JavaScript code included in a web page ranking that identifies user selections (clicks) of individual document results and also identifies when the user returns to the results page, thus indicating the amount of time the user spent viewing the selected document result. In other implementations, the tracking component can be a proxy system through which user selections of the document results are routed, or the tracking component can include pre-installed software at the client (e.g., a toolbar plug-in to the client’s operating system). Other implementations are also possible, such as by using a feature of a web browser that allows a tag/directive to be included in a page, which requests the browser to connect back to the server with message(s) regarding link(s) clicked by the user.

    That selection information may also be logged, which could capture for each selection:

    • the query (Q)
    • the document (D)
    • the time (T) on the document
    • the language (L) employed by the user
    • the country (C) where the user is likely located (e.g., based on the server used to access the IR system).

    Also other information may be recorded about a searcher’s interactions with presented rankings:

    • Negative information, such as the fact that a document result was presented to a user, but was not clicked
    • Position(s) of click(s) in the user interface
    • IR scores of clicked results
    • IR scores of all results shown before the clicked result
    • Titles and snippets shown to the user before the clicked result
    • The user’s cookie
    • Cookie age
    • IP (Internet Protocol) address
    • User agent of the browser
    • Etc

    More information can be recorded (as described in this post below) about building a prior model.

    Rank Modifier Engine

    Similar information (e.g., IR scores, position, etc.) may be recorded for an entire session, or multiple sessions of a searcher, including possibly recording it for every click that occurs both before and after a current click.

    Information that is stored in the result selection logs may be used by the rank modifier engine to generate one or more signals to the ranking engine.

    Information stored in the search results selection logs along with the information collected by the tracking component may also be accessible by a search assistant, which is also a component of the information retrieval system.

    Along with receiving information from these components, the search assistant could also monitor a user’s entry of a search query.

    On receiving a partial search query, the query along with the information (e.g., click data) from the tracking component and the results selection log(s) may be used to predict a searcher’s contemplated complete query.

    Based on this information, predictions may be ordered according to one or more ranking criteria before being presented to assist the user in completing the query.

    Presentation of a Search Suggestion

    As a searcher enters a search query, the searcher’s input is monitored.

    Before the searcher signals that they have completed entering the search query, a portion of the query is sent to the search engine.

    Also, data such as click data (or other types of previously collected information) may also be sent with the query portion.

    The portion of the query sent may be:

    • A few characters
    • A search term
    • More than one search term
    • Any other combination of characters and terms

    The search engine receives the partial query and the data (e.g., click data) for processing and makes predictions) as to the searcher’s contemplated complete query.

    Relevant information may be retrieved for processing with the received partial query to produce search suggestions predictions.

    Predictions may be ordered according to one or more ranking criteria.

    So, queries that have been submitted at a higher frequency may be ordered before queries submitted at lower frequencies.

    The search engine may also use various types of information for ranking and ordering predicted queries as search suggestions.

    Information about previously entered search queries may be used to make ordered predictions.

    Previous queries may include search queries associated with the same user, another user, or from a community of users.

    If one of the predicted queries is what the searcher intended as the desired query, the searcher may select that predicted query and proceed without having to finish entering the desired query.

    Alternatively, if the predicted queries do not reflect what the searcher had in mind, then the searcher can continue entering the desired search query, which could trigger one or more other sets of search suggestions.

    Ranking User Submitted Previous Queries as Search Suggestions

    The patent tells us that a few different processes may be used in ranking and ordering predicted search queries:

    • Predicted search queries may be ordered in accordance with a frequency of submission by a community of users
    • Time constraints may also be used with search queries ordered in accordance with the last time/date value that the query was submitted
    • Personalization information or community information may be used such as information about subjects, concepts or categories of information that are of interest to the user (from prior search or browsing information)
    • Personalization may also be from a group that the searcher is associated with or belongs to (a member or an employee.)
    • According to a first ranking criteria, such as predefined popularity criteria, and then possibly reordered if any of the predicted search queries match the user personalization information of the user, to place the matching predicted search queries at or closer to the top of the ordered set of predicted search queries
    • Information provided by the tracking component and the result selection log(s) might be used for ranking and ordering the predicted search queries. (click data, language-specific, and country-specific data.)
    • Processed click data (e.g., aggregated click data for a given query) could be used for ranking and ordering predicted search queries – or each query a score may be calculated by summing click data (e.g., weighted clicks, etc.) on documents associated with the query, and predicted queries may be ordered based upon the score (e.g., higher values representing better)

    An Information Model Based On Previously Submitted Query Data to Obtain Search Suggestions Predictions

    This model can be used to predict what query data might satisfy a searcher the most by looking at long click information. A timer can be used to track how long a user views or “dwells” on a document.

    The amount of time is referred to as “click data”.

    A longer time spent dwelling on a document, would be termed a “long click”, and can indicate that a user found the document to be relevant for their query.

    A brief period viewing a document would be termed a “short click”, and can be interpreted as a lack of document relevance.

    Click data is a count of each click type (e.g., long, medium, short) for a particular query and document combination.

    This aggregated click data from model queries for a given document can be used to create a quality of result statistic for that document to enhance a ranking of that document.

    Quality of result statistic can be a weighted average of the count of long clicks for a given document and query.

    This description from the patent tells us about how click data might be stored in tuples:

    A search engine (e.g., the search engine) or other processes may create a record in the model for documents that are selected by users in response to a query or a partial query. Each record within the model (herein referred to as a tuple: ) is at least a combination of a query submitted by users, a document reference selected by users in response to that query, and an aggregation of click data for all users that select the document reference in response to the query. The aggregate click data can be viewed as an indication of document relevance. In various implementations, model data can be location-specific (e.g. country, state, etc) or language-specific. For example, a country-specific tuple would include the country from where the user query originated from in whereas a language-specific tuple would include the language of the user query. Other extensions of model data are possible.

    The model may also include Post-click behavior that has been tracked by the tracking component.

    This patent does include a lot of information about how Google might use click tracking data when ranking search suggestion predictions. It tells us about the data that could be collected about clicks:

    The information gathered for each click can include:

    (1) the query (Q) the user entered,
    (2) the document result (D) the user clicked on,
    (3) the time (T) on the document,
    (4) the interface language (L) (which can be given by the user),
    (5) the country (C) of the user (which can be identified by the host that they use, such as www-store-co-uk to indicate the United Kingdom), and
    (6) additional aspects of the user and session.

    The time (T) can be measured as the time between the initial click through to the document result until the time the user comes back to the main page and clicks on another document result. Moreover, an assessment can be made about the time (T) regarding whether this time indicates a longer view of the document result or a shorter view of the document result, since longer views are generally indicative of quality for the clicked through result. This assessment about the time (T) can further be made in conjunction with various weighting techniques.

    Beyond Long Clicks

    We are also told that document views from the selections can be weighted based on viewing length information to produce weighted views of the document result.

    So, rather than simply distinguishing long clicks from short clicks, a wider range of click through viewing times can be included in the assessment of result quality, where longer viewing times in the range are given more weight than shorter viewing times.

    Predicted Search Suggestions

    Google will sometimes display search suggestions using autocomplete and also based upon user data from previous queries from a searcher’s previous search history, or the history of someone whom the searcher may be associated with, such as a fellow member of an organization or a co-worker.

    While results related to those previous queries were ranked based upon such things as relevance and backlinks, the search suggestions may include results that searchers spent long clicks upon, including long times viewing.

    So pursuant to this patent, predictions about search suggestions chosen using autocomplete may best meet a searcher’s informational needs by being searches that include results remembered as resulting in long clicks and long viewing times.

Copyright © 2020 SEO by the Sea ⚓. This Feed is for personal non-commercial use only. If you are not reading this material in your news aggregator, the site you are looking at may be guilty of copyright infringement. Please contact SEO by the Sea, so we can take appropriate action immediately.
Plugin by Taragana

The post Search Suggestions from Previously Submitted Searcher Queries appeared first on SEO by the Sea ⚓.


SEO by the Sea ⚓

Search Suggestions from Previously Submitted Searcher Queries

May 17, 2020 No Comments

I came across an interesting Search Engine Land post last week. It inspired me to search and see if I could find a patent that might be related to it from Google:

Google is suggesting searches based on users’ recent activity

I tried to reproduce the search suggestions that were being shown to the author of the Search Engine Land article, but Google would not return those to me. Google may be experimenting with a limited number of searchers rather than showing those results to everyone. I did find a patent discussing search suggestions that were similar.

When Google shows a search suggestion about something you may have searched for in the past, that predicted suggestion is likely related to a patent I’ve written about before, Autocompletion using previously submitted query data.

I wrote about that patent being updated in a continuation patent, but hadn’t provided much in the way of details about how it works at: How Google Predicts Autocomplete Query Suggestions is Updated.

There are some interesting parts about how search suggestions are identified and ranked, which inspired me to write this post.

Search Suggestions Based on Previously Submitted Query Data

The description of this patent starts off by telling us that it is about: “using previously submitted query data to anticipate a user’s search request.”

That pinpoints that Google has a long memory, and it remembers a lot about what someone might search for.

This patent description also includes a lot of the assumptions that search engineers make about searchers (often an interesting reason to read through patents). Here are some from this patent that are worth thinking about:

Internet search engines aim to identify documents or other items that are relevant to a user’s needs and to present the documents or items in a manner that is most useful to the user. Such activity often involves a fair amount of mind-reading–inferring from various clues what the user wants. Certain clues may be user-specific. For example, the knowledge that a user is making a request from a mobile device, and knowledge of the location of the device, can result in much better search results for such a user.

Clues about a user’s needs may also be more general. For example, search results can have elevated importance, or inferred relevance, if a number of other search results link to them. If the linking results are themselves highly relevant, then the linked-to results may have particularly high relevance. Such an approach to determining relevance may be premised on the assumption that, if authors of web pages felt that another web site was relevant enough to be linked to, then web searchers would also find the site to be particularly relevant. In short, the web authors “vote up” the relevance of the sites.

Other various inputs may be used instead of, or in addition to, such techniques for determining and ranking search results. For example, user reactions to particular search results or search result lists may be gauged, so that results on which users often click will receive a higher ranking. The general assumption under such an approach is that searching users are often the best judges of relevance, so that if they select a particular search result, it is likely to be relevant, or at least more relevant than the presented alternatives.

A Summary of the Search Suggestions Process Based on Previous Submitted Queries

Like most patents, the Description for this one starts out with a summary section that provides an overview of how the process defined in the patent works. It is followed by a “Detailed Description” section that goes into more depth and provides details about how search at Google works, and how specific aspects of search at Google power this search suggestion process. So read about how search suggestions might be provided based upon user queries that have been searched for previously, and then read for the more detailed explanation, which goes way beyond autocomplete.

In the summary section of the description for the patent, we are told about how the patent may address those assumptions:

When anticipating user search requests, responding to the algorithm in this patent can involve certain methods for processing query information. Those include:

  • Receiving query information at a server system, with a portion of a query from a searcher
  • Obtaining a set of predicted queries relevant to the portion of the searcher’s query based on query and data indicative of the searcher relative to previously submitted queries
  • Providing the set of predicted queries to the searcher

The patent also points out additional features involved in the process such as obtaining the predicted queries including ordering the set of predicted queries based upon ranking criteria.

Those ranking criteria may be based upon the data indicative of searcher’s behavior relative to previously submitted queries.

Data about the searcher’s behavior regarding those previously submitted queries may include:

  • Click data
  • Location-specific data
  • Language-specific data
  • Other similar types of data

The patent points out the following as advantages of following the process described in the patent:

A search assistant receives query information from a search requestor, prior to a searcher completely inputting the query.

Information associated with previous user (or users) searches (such as click data associated with search results) is collected. From the query information and the previous search information, a set of predicted queries is produced and provided to the search requestor for presentation.

The patent can be found at:

Autocompletion using previously submitted query data
Inventors: Michael Herscovici, Dan Guez, and Hyung-Jin Kim
Assignee: Google Inc.
US Patent: 9,740,780
Granted: August 22, 2017
Filed: December 1, 2014

Abstract

A computer-implemented method for processing query information includes receiving query information at a server system. The query information includes a portion of a query from a search requestor. The method also includes obtaining a set of predicted queries relevant to the portion of the search requestor query based upon the portion of the query from the search requestor and data indicative of search requestor behavior relative to previously submitted queries. The method also includes providing the set of predicted queries to the search requestor.

Analysis of Ranking and Selection of Search Suggestions Based Upon Previous Query Data

The “Detailed Description” section of this search suggestions patent provides some insightful analysis about search at Google.

Relevance and Backlinks and a Rank Modifying Engine Lead to Ranking For Many Results at Google

This patent points out some of how search works at Google. It tells us that:

  1. The purpose of the process in the patent is to “improve the relevance of results obtained from submitting search queries.”
  2. It describes the ranking of documents for a query as something that can be “performed using traditional techniques for determining an information retrieval (IR) score for indexed documents in view of a given query.” And relevance of a particular document with respect to a query term may be determined by a technique, such as looking at the general level of back-links to a document that contain matches for a search term that may be used to infer a document’s relevance. As the patent tells us:

    In particular, if a document is linked to (e.g., is the target of a hyperlink) by many other relevant documents (e.g., documents that also contain matches for the search terms), it can be inferred that the target document is particularly relevant. This inference can be made because the authors of the pointing documents presumably point, for the most part, to other documents that are relevant to their audience.

  3. We are given more details about some results being even more relevant than ones with backlinks. We are told that:

    If the pointing documents are in turn the targets of links from other relevant documents, they can be considered more relevant, and the first document can be considered particularly relevant because it is the target of relevant (or even highly relevant) documents. Such a technique may be the determinant of a document’s relevance or one of multiple determinants. The technique is exemplified in some systems that treat a link from one web page to another as an indication of quality for the latter page, so that the page with the most such quality indicators is rated higher than others. Appropriate techniques can also be used to identify and eliminate attempts to cast false votes so as to artificially drive up the relevance of a page.

  4. There is another step that could potentially make some results even more relevant that involve what is referred to as a rank modifier engine:

    To further improve such traditional document ranking techniques, the ranking engine can receive an additional signal from a rank modifier engine to assist in determining an appropriate ranking for the documents. The rank modifier engine provides one or more prior models, or one or more measures of relevance for the documents based on one or more prior models, which can be used by the ranking engine to improve the search results’ ranking provided to the user. In general, a prior model represents a background probability of document result selection given the values of multiple selected features, as described further below. The rank modifier engine can perform one or more of the operations described below to generate the one or more prior models, or the one or more measures of relevance based on one or more prior models.

    5. This is a more detailed description of ranking than we normally see at Google. The section above references a Rank Modifier Engine that will be described in more detail further down this post

    Indexing, Scoring, Ranking, and Rank Modifier Engine

    ranking search suggestions

    The information retrieval system from this patent includes a number of different components:

    • Indexing engine
    • Scoring engine
    • Ranking engine
    • Rank modifier engine

    The indexing engine can function as described in the section above for the indexing engine.

    Scoring Engine

    In addition, a scoring engine may provide scores for document results based on many different features including:

    • Content-based features that link a query to document results
    • query-independent features that generally indicate the quality of document results

    Content-based features include aspects of document format, such as Query matches to title or anchor text in an HTML (HyperText Markup Language) page.

    The query-independent features can include aspects of document cross-referencing, such as a rank of the document or the domain.

    Moreover, the particular functions used by the scoring engine can be tuned, to adjust the various feature contributions to the final IR score, using automatic or semi-automatic processes.

    Ranking Engine

    A ranking engine can produce a ranking of document search results for display to a searcher based on IR scores received from the scoring engine and possibly one or more signals from the rank modifier engine.

    A tracking component may be used to record information about individual searcher selections of the search results presented in the ranking. The patent describes how selections may be tracked using javascript or a proxy system or a toolbar plugin:

    For example, the tracking component can be embedded JavaScript code included in a web page ranking that identifies user selections (clicks) of individual document results and also identifies when the user returns to the results page, thus indicating the amount of time the user spent viewing the selected document result. In other implementations, the tracking component can be a proxy system through which user selections of the document results are routed, or the tracking component can include pre-installed software at the client (e.g., a toolbar plug-in to the client’s operating system). Other implementations are also possible, such as by using a feature of a web browser that allows a tag/directive to be included in a page, which requests the browser to connect back to the server with message(s) regarding link(s) clicked by the user.

    That selection information may also be logged, which could capture for each selection:

    • the query (Q)
    • the document (D)
    • the time (T) on the document
    • the language (L) employed by the user
    • the country (C) where the user is likely located (e.g., based on the server used to access the IR system).

    Also other information may be recorded about a searcher’s interactions with presented rankings:

    • Negative information, such as the fact that a document result was presented to a user, but was not clicked
    • Position(s) of click(s) in the user interface
    • IR scores of clicked results
    • IR scores of all results shown before the clicked result
    • Titles and snippets shown to the user before the clicked result
    • The user’s cookie
    • Cookie age
    • IP (Internet Protocol) address
    • User agent of the browser
    • Etc

    More information can be recorded (as described in this post below) about building a prior model.

    Rank Modifier Engine

    Similar information (e.g., IR scores, position, etc.) may be recorded for an entire session, or multiple sessions of a searcher, including possibly recording it for every click that occurs both before and after a current click.

    Information that is stored in the result selection logs may be used by the rank modifier engine to generate one or more signals to the ranking engine.

    Information stored in the search results selection logs along with the information collected by the tracking component may also be accessible by a search assistant, which is also a component of the information retrieval system.

    Along with receiving information from these components, the search assistant could also monitor a user’s entry of a search query.

    On receiving a partial search query, the query along with the information (e.g., click data) from the tracking component and the results selection log(s) may be used to predict a searcher’s contemplated complete query.

    Based on this information, predictions may be ordered according to one or more ranking criteria before being presented to assist the user in completing the query.

    Presentation of a Search Suggestion

    As a searcher enters a search query, the searcher’s input is monitored.

    Before the searcher signals that they have completed entering the search query, a portion of the query is sent to the search engine.

    Also, data such as click data (or other types of previously collected information) may also be sent with the query portion.

    The portion of the query sent may be:

    • A few characters
    • A search term
    • More than one search term
    • Any other combination of characters and terms

    The search engine receives the partial query and the data (e.g., click data) for processing and makes predictions) as to the searcher’s contemplated complete query.

    Relevant information may be retrieved for processing with the received partial query to produce search suggestions predictions.

    Predictions may be ordered according to one or more ranking criteria.

    So, queries that have been submitted at a higher frequency may be ordered before queries submitted at lower frequencies.

    The search engine may also use various types of information for ranking and ordering predicted queries as search suggestions.

    Information about previously entered search queries may be used to make ordered predictions.

    Previous queries may include search queries associated with the same user, another user, or from a community of users.

    If one of the predicted queries is what the searcher intended as the desired query, the searcher may select that predicted query and proceed without having to finish entering the desired query.

    Alternatively, if the predicted queries do not reflect what the searcher had in mind, then the searcher can continue entering the desired search query, which could trigger one or more other sets of search suggestions.

    Ranking User Submitted Previous Queries as Search Suggestions

    The patent tells us that a few different processes may be used in ranking and ordering predicted search queries:

    • Predicted search queries may be ordered in accordance with a frequency of submission by a community of users
    • Time constraints may also be used with search queries ordered in accordance with the last time/date value that the query was submitted
    • Personalization information or community information may be used such as information about subjects, concepts or categories of information that are of interest to the user (from prior search or browsing information)
    • Personalization may also be from a group that the searcher is associated with or belongs to (a member or an employee.)
    • According to a first ranking criteria, such as predefined popularity criteria, and then possibly reordered if any of the predicted search queries match the user personalization information of the user, to place the matching predicted search queries at or closer to the top of the ordered set of predicted search queries
    • Information provided by the tracking component and the result selection log(s) might be used for ranking and ordering the predicted search queries. (click data, language-specific, and country-specific data.)
    • Processed click data (e.g., aggregated click data for a given query) could be used for ranking and ordering predicted search queries – or each query a score may be calculated by summing click data (e.g., weighted clicks, etc.) on documents associated with the query, and predicted queries may be ordered based upon the score (e.g., higher values representing better)

    An Information Model Based On Previously Submitted Query Data to Obtain Search Suggestions Predictions

    This model can be used to predict what query data might satisfy a searcher the most by looking at long click information. A timer can be used to track how long a user views or “dwells” on a document.

    The amount of time is referred to as “click data”.

    A longer time spent dwelling on a document, would be termed a “long click”, and can indicate that a user found the document to be relevant for their query.

    A brief period viewing a document would be termed a “short click”, and can be interpreted as a lack of document relevance.

    Click data is a count of each click type (e.g., long, medium, short) for a particular query and document combination.

    This aggregated click data from model queries for a given document can be used to create a quality of result statistic for that document to enhance a ranking of that document.

    Quality of result statistic can be a weighted average of the count of long clicks for a given document and query.

    This description from the patent tells us about how click data might be stored in tuples:

    A search engine (e.g., the search engine) or other processes may create a record in the model for documents that are selected by users in response to a query or a partial query. Each record within the model (herein referred to as a tuple: ) is at least a combination of a query submitted by users, a document reference selected by users in response to that query, and an aggregation of click data for all users that select the document reference in response to the query. The aggregate click data can be viewed as an indication of document relevance. In various implementations, model data can be location-specific (e.g. country, state, etc) or language-specific. For example, a country-specific tuple would include the country from where the user query originated from in whereas a language-specific tuple would include the language of the user query. Other extensions of model data are possible.

    The model may also include Post-click behavior that has been tracked by the tracking component.

    This patent does include a lot of information about how Google might use click tracking data when ranking search suggestion predictions. It tells us about the data that could be collected about clicks:

    The information gathered for each click can include:

    (1) the query (Q) the user entered,
    (2) the document result (D) the user clicked on,
    (3) the time (T) on the document,
    (4) the interface language (L) (which can be given by the user),
    (5) the country (C) of the user (which can be identified by the host that they use, such as www-store-co-uk to indicate the United Kingdom), and
    (6) additional aspects of the user and session.

    The time (T) can be measured as the time between the initial click through to the document result until the time the user comes back to the main page and clicks on another document result. Moreover, an assessment can be made about the time (T) regarding whether this time indicates a longer view of the document result or a shorter view of the document result, since longer views are generally indicative of quality for the clicked through result. This assessment about the time (T) can further be made in conjunction with various weighting techniques.

    Beyond Long Clicks

    We are also told that document views from the selections can be weighted based on viewing length information to produce weighted views of the document result.

    So, rather than simply distinguishing long clicks from short clicks, a wider range of click through viewing times can be included in the assessment of result quality, where longer viewing times in the range are given more weight than shorter viewing times.

    Predicted Search Suggestions

    Google will sometimes display search suggestions using autocomplete and also based upon user data from previous queries from a searcher’s previous search history, or the history of someone whom the searcher may be associated with, such as a fellow member of an organization or a co-worker.

    While results related to those previous queries were ranked based upon such things as relevance and backlinks, the search suggestions may include results that searchers spent long clicks upon, including long times viewing.

    So pursuant to this patent, predictions about search suggestions chosen using autocomplete may best meet a searcher’s informational needs by being searches that include results remembered as resulting in long clicks and long viewing times.

Copyright © 2020 SEO by the Sea ⚓. This Feed is for personal non-commercial use only. If you are not reading this material in your news aggregator, the site you are looking at may be guilty of copyright infringement. Please contact SEO by the Sea, so we can take appropriate action immediately.
Plugin by Taragana

The post Search Suggestions from Previously Submitted Searcher Queries appeared first on SEO by the Sea ⚓.


SEO by the Sea ⚓

Search Suggestions from Previously Submitted Searcher Queries

May 15, 2020 No Comments

I came across an interesting Search Engine Land post last week. It inspired me to search and see if I could find a patent that might be related to it from Google:

Google is suggesting searches based on users’ recent activity

I tried to reproduce the search suggestions that were being shown to the author of the Search Engine Land article, but Google would not return those to me. Google may be experimenting with a limited number of searchers rather than showing those results to everyone. I did find a patent discussing search suggestions that were similar.

When Google shows a search suggestion about something you may have searched for in the past, that predicted suggestion is likely related to a patent I’ve written about before, Autocompletion using previously submitted query data.

I wrote about that patent being updated in a continuation patent, but hadn’t provided much in the way of details about how it works at: How Google Predicts Autocomplete Query Suggestions is Updated.

There are some interesting parts about how search suggestions are identified and ranked, which inspired me to write this post.

Search Suggestions Based on Previously Submitted Query Data

The description of this patent starts off by telling us that it is about: “using previously submitted query data to anticipate a user’s search request.”

That pinpoints that Google has a long memory, and it remembers a lot about what someone might search for.

This patent description also includes a lot of the assumptions that search engineers make about searchers (often an interesting reason to read through patents). Here are some from this patent that are worth thinking about:

Internet search engines aim to identify documents or other items that are relevant to a user’s needs and to present the documents or items in a manner that is most useful to the user. Such activity often involves a fair amount of mind-reading–inferring from various clues what the user wants. Certain clues may be user-specific. For example, the knowledge that a user is making a request from a mobile device, and knowledge of the location of the device, can result in much better search results for such a user.

Clues about a user’s needs may also be more general. For example, search results can have elevated importance, or inferred relevance, if a number of other search results link to them. If the linking results are themselves highly relevant, then the linked-to results may have particularly high relevance. Such an approach to determining relevance may be premised on the assumption that, if authors of web pages felt that another web site was relevant enough to be linked to, then web searchers would also find the site to be particularly relevant. In short, the web authors “vote up” the relevance of the sites.

Other various inputs may be used instead of, or in addition to, such techniques for determining and ranking search results. For example, user reactions to particular search results or search result lists may be gauged, so that results on which users often click will receive a higher ranking. The general assumption under such an approach is that searching users are often the best judges of relevance, so that if they select a particular search result, it is likely to be relevant, or at least more relevant than the presented alternatives.

A Summary of the Search Suggestions Process Based on Previous Submitted Queries

Like most patents, the Description for this one starts out with a summary section that provides an overview of how the process defined in the patent works. It is followed by a “Detailed Description” section that goes into more depth and provides details about how search at Google works, and how specific aspects of search at Google power this search suggestion process. So read about how search suggestions might be provided based upon user queries that have been searched for previously, and then read for the more detailed explanation, which goes way beyond autocomplete.

In the summary section of the description for the patent, we are told about how the patent may address those assumptions:

When anticipating user search requests, responding to the algorithm in this patent can involve certain methods for processing query information. Those include:

  • Receiving query information at a server system, with a portion of a query from a searcher
  • Obtaining a set of predicted queries relevant to the portion of the searcher’s query based on query and data indicative of the searcher relative to previously submitted queries
  • Providing the set of predicted queries to the searcher

The patent also points out additional features involved in the process such as obtaining the predicted queries including ordering the set of predicted queries based upon ranking criteria.

Those ranking criteria may be based upon the data indicative of searcher’s behavior relative to previously submitted queries.

Data about the searcher’s behavior regarding those previously submitted queries may include:

  • Click data
  • Location-specific data
  • Language-specific data
  • Other similar types of data

The patent points out the following as advantages of following the process described in the patent:

A search assistant receives query information from a search requestor, prior to a searcher completely inputting the query.

Information associated with previous user (or users) searches (such as click data associated with search results) is collected. From the query information and the previous search information, a set of predicted queries is produced and provided to the search requestor for presentation.

The patent can be found at:

Autocompletion using previously submitted query data
Inventors: Michael Herscovici, Dan Guez, and Hyung-Jin Kim
Assignee: Google Inc.
US Patent: 9,740,780
Granted: August 22, 2017
Filed: December 1, 2014

Abstract

A computer-implemented method for processing query information includes receiving query information at a server system. The query information includes a portion of a query from a search requestor. The method also includes obtaining a set of predicted queries relevant to the portion of the search requestor query based upon the portion of the query from the search requestor and data indicative of search requestor behavior relative to previously submitted queries. The method also includes providing the set of predicted queries to the search requestor.

Analysis of Ranking and Selection of Search Suggestions Based Upon Previous Query Data

The “Detailed Description” section of this search suggestions patent provides some insightful analysis about search at Google.

Relevance and Backlinks and a Rank Modifying Engine Lead to Ranking For Many Results at Google

This patent points out some of how search works at Google. It tells us that:

  1. The purpose of the process in the patent is to “improve the relevance of results obtained from submitting search queries.”
  2. It describes the ranking of documents for a query as something that can be “performed using traditional techniques for determining an information retrieval (IR) score for indexed documents in view of a given query.” And relevance of a particular document with respect to a query term may be determined by a technique, such as looking at the general level of back-links to a document that contain matches for a search term that may be used to infer a document’s relevance. As the patent tells us:

    In particular, if a document is linked to (e.g., is the target of a hyperlink) by many other relevant documents (e.g., documents that also contain matches for the search terms), it can be inferred that the target document is particularly relevant. This inference can be made because the authors of the pointing documents presumably point, for the most part, to other documents that are relevant to their audience.

  3. We are given more details about some results being even more relevant than ones with backlinks. We are told that:

    If the pointing documents are in turn the targets of links from other relevant documents, they can be considered more relevant, and the first document can be considered particularly relevant because it is the target of relevant (or even highly relevant) documents. Such a technique may be the determinant of a document’s relevance or one of multiple determinants. The technique is exemplified in some systems that treat a link from one web page to another as an indication of quality for the latter page, so that the page with the most such quality indicators is rated higher than others. Appropriate techniques can also be used to identify and eliminate attempts to cast false votes so as to artificially drive up the relevance of a page.

  4. There is another step that could potentially make some results even more relevant that involve what is referred to as a rank modifier engine:

    To further improve such traditional document ranking techniques, the ranking engine can receive an additional signal from a rank modifier engine to assist in determining an appropriate ranking for the documents. The rank modifier engine provides one or more prior models, or one or more measures of relevance for the documents based on one or more prior models, which can be used by the ranking engine to improve the search results’ ranking provided to the user. In general, a prior model represents a background probability of document result selection given the values of multiple selected features, as described further below. The rank modifier engine can perform one or more of the operations described below to generate the one or more prior models, or the one or more measures of relevance based on one or more prior models.

    5. This is a more detailed description of ranking than we normally see at Google. The section above references a Rank Modifier Engine that will be described in more detail further down this post

    Indexing, Scoring, Ranking, and Rank Modifier Engine

    ranking search suggestions

    The information retrieval system from this patent includes a number of different components:

    • Indexing engine
    • Scoring engine
    • Ranking engine
    • Rank modifier engine

    The indexing engine can function as described in the section above for the indexing engine.

    Scoring Engine

    In addition, a scoring engine may provide scores for document results based on many different features including:

    • Content-based features that link a query to document results
    • query-independent features that generally indicate the quality of document results

    Content-based features include aspects of document format, such as Query matches to title or anchor text in an HTML (HyperText Markup Language) page.

    The query-independent features can include aspects of document cross-referencing, such as a rank of the document or the domain.

    Moreover, the particular functions used by the scoring engine can be tuned, to adjust the various feature contributions to the final IR score, using automatic or semi-automatic processes.

    Ranking Engine

    A ranking engine can produce a ranking of document search results for display to a searcher based on IR scores received from the scoring engine and possibly one or more signals from the rank modifier engine.

    A tracking component may be used to record information about individual searcher selections of the search results presented in the ranking. The patent describes how selections may be tracked using javascript or a proxy system or a toolbar plugin:

    For example, the tracking component can be embedded JavaScript code included in a web page ranking that identifies user selections (clicks) of individual document results and also identifies when the user returns to the results page, thus indicating the amount of time the user spent viewing the selected document result. In other implementations, the tracking component can be a proxy system through which user selections of the document results are routed, or the tracking component can include pre-installed software at the client (e.g., a toolbar plug-in to the client’s operating system). Other implementations are also possible, such as by using a feature of a web browser that allows a tag/directive to be included in a page, which requests the browser to connect back to the server with message(s) regarding link(s) clicked by the user.

    That selection information may also be logged, which could capture for each selection:

    • the query (Q)
    • the document (D)
    • the time (T) on the document
    • the language (L) employed by the user
    • the country (C) where the user is likely located (e.g., based on the server used to access the IR system).

    Also other information may be recorded about a searcher’s interactions with presented rankings:

    • Negative information, such as the fact that a document result was presented to a user, but was not clicked
    • Position(s) of click(s) in the user interface
    • IR scores of clicked results
    • IR scores of all results shown before the clicked result
    • Titles and snippets shown to the user before the clicked result
    • The user’s cookie
    • Cookie age
    • IP (Internet Protocol) address
    • User agent of the browser
    • Etc

    More information can be recorded (as described in this post below) about building a prior model.

    Rank Modifier Engine

    Similar information (e.g., IR scores, position, etc.) may be recorded for an entire session, or multiple sessions of a searcher, including possibly recording it for every click that occurs both before and after a current click.

    Information that is stored in the result selection logs may be used by the rank modifier engine to generate one or more signals to the ranking engine.

    Information stored in the search results selection logs along with the information collected by the tracking component may also be accessible by a search assistant, which is also a component of the information retrieval system.

    Along with receiving information from these components, the search assistant could also monitor a user’s entry of a search query.

    On receiving a partial search query, the query along with the information (e.g., click data) from the tracking component and the results selection log(s) may be used to predict a searcher’s contemplated complete query.

    Based on this information, predictions may be ordered according to one or more ranking criteria before being presented to assist the user in completing the query.

    Presentation of a Search Suggestion

    As a searcher enters a search query, the searcher’s input is monitored.

    Before the searcher signals that they have completed entering the search query, a portion of the query is sent to the search engine.

    Also, data such as click data (or other types of previously collected information) may also be sent with the query portion.

    The portion of the query sent may be:

    • A few characters
    • A search term
    • More than one search term
    • Any other combination of characters and terms

    The search engine receives the partial query and the data (e.g., click data) for processing and makes predictions) as to the searcher’s contemplated complete query.

    Relevant information may be retrieved for processing with the received partial query to produce search suggestions predictions.

    Predictions may be ordered according to one or more ranking criteria.

    So, queries that have been submitted at a higher frequency may be ordered before queries submitted at lower frequencies.

    The search engine may also use various types of information for ranking and ordering predicted queries as search suggestions.

    Information about previously entered search queries may be used to make ordered predictions.

    Previous queries may include search queries associated with the same user, another user, or from a community of users.

    If one of the predicted queries is what the searcher intended as the desired query, the searcher may select that predicted query and proceed without having to finish entering the desired query.

    Alternatively, if the predicted queries do not reflect what the searcher had in mind, then the searcher can continue entering the desired search query, which could trigger one or more other sets of search suggestions.

    Ranking User Submitted Previous Queries as Search Suggestions

    The patent tells us that a few different processes may be used in ranking and ordering predicted search queries:

    • Predicted search queries may be ordered in accordance with a frequency of submission by a community of users
    • Time constraints may also be used with search queries ordered in accordance with the last time/date value that the query was submitted
    • Personalization information or community information may be used such as information about subjects, concepts or categories of information that are of interest to the user (from prior search or browsing information)
    • Personalization may also be from a group that the searcher is associated with or belongs to (a member or an employee.)
    • According to a first ranking criteria, such as predefined popularity criteria, and then possibly reordered if any of the predicted search queries match the user personalization information of the user, to place the matching predicted search queries at or closer to the top of the ordered set of predicted search queries
    • Information provided by the tracking component and the result selection log(s) might be used for ranking and ordering the predicted search queries. (click data, language-specific, and country-specific data.)
    • Processed click data (e.g., aggregated click data for a given query) could be used for ranking and ordering predicted search queries – or each query a score may be calculated by summing click data (e.g., weighted clicks, etc.) on documents associated with the query, and predicted queries may be ordered based upon the score (e.g., higher values representing better)

    An Information Model Based On Previously Submitted Query Data to Obtain Search Suggestions Predictions

    This model can be used to predict what query data might satisfy a searcher the most by looking at long click information. A timer can be used to track how long a user views or “dwells” on a document.

    The amount of time is referred to as “click data”.

    A longer time spent dwelling on a document, would be termed a “long click”, and can indicate that a user found the document to be relevant for their query.

    A brief period viewing a document would be termed a “short click”, and can be interpreted as a lack of document relevance.

    Click data is a count of each click type (e.g., long, medium, short) for a particular query and document combination.

    This aggregated click data from model queries for a given document can be used to create a quality of result statistic for that document to enhance a ranking of that document.

    Quality of result statistic can be a weighted average of the count of long clicks for a given document and query.

    This description from the patent tells us about how click data might be stored in tuples:

    A search engine (e.g., the search engine) or other processes may create a record in the model for documents that are selected by users in response to a query or a partial query. Each record within the model (herein referred to as a tuple: ) is at least a combination of a query submitted by users, a document reference selected by users in response to that query, and an aggregation of click data for all users that select the document reference in response to the query. The aggregate click data can be viewed as an indication of document relevance. In various implementations, model data can be location-specific (e.g. country, state, etc) or language-specific. For example, a country-specific tuple would include the country from where the user query originated from in whereas a language-specific tuple would include the language of the user query. Other extensions of model data are possible.

    The model may also include Post-click behavior that has been tracked by the tracking component.

    This patent does include a lot of information about how Google might use click tracking data when ranking search suggestion predictions. It tells us about the data that could be collected about clicks:

    The information gathered for each click can include:

    (1) the query (Q) the user entered,
    (2) the document result (D) the user clicked on,
    (3) the time (T) on the document,
    (4) the interface language (L) (which can be given by the user),
    (5) the country (C) of the user (which can be identified by the host that they use, such as www-store-co-uk to indicate the United Kingdom), and
    (6) additional aspects of the user and session.

    The time (T) can be measured as the time between the initial click through to the document result until the time the user comes back to the main page and clicks on another document result. Moreover, an assessment can be made about the time (T) regarding whether this time indicates a longer view of the document result or a shorter view of the document result, since longer views are generally indicative of quality for the clicked through result. This assessment about the time (T) can further be made in conjunction with various weighting techniques.

    Beyond Long Clicks

    We are also told that document views from the selections can be weighted based on viewing length information to produce weighted views of the document result.

    So, rather than simply distinguishing long clicks from short clicks, a wider range of click through viewing times can be included in the assessment of result quality, where longer viewing times in the range are given more weight than shorter viewing times.

    Predicted Search Suggestions

    Google will sometimes display search suggestions using autocomplete and also based upon user data from previous queries from a searcher’s previous search history, or the history of someone whom the searcher may be associated with, such as a fellow member of an organization or a co-worker.

    While results related to those previous queries were ranked based upon such things as relevance and backlinks, the search suggestions may include results that searchers spent long clicks upon, including long times viewing.

    So pursuant to this patent, predictions about search suggestions chosen using autocomplete may best meet a searcher’s informational needs by being searches that include results remembered as resulting in long clicks and long viewing times.

Copyright © 2020 SEO by the Sea ⚓. This Feed is for personal non-commercial use only. If you are not reading this material in your news aggregator, the site you are looking at may be guilty of copyright infringement. Please contact SEO by the Sea, so we can take appropriate action immediately.
Plugin by Taragana

The post Search Suggestions from Previously Submitted Searcher Queries appeared first on SEO by the Sea ⚓.


SEO by the Sea ⚓

©2018-2020 - American Online Phonebook - All Rights Reserved

Privacy Policy | Terms & Conditions