Monthly Archives: July 2020
- Google has the ability to measure content quality signals like never before, and with each new Core Update, they understand content quality more as humans do.
- Google uses different search models depending on the search query, meaning ranking signals can vary depending on search intent.
- In a recent correlation analysis we did in the sports ticketing industry, the top factors that correlated with good rankings were long-form landing page content and high domain authority.
- Given Google’s guidance about their forthcoming 2021 web vitals update, page experience and performance are likely going to become more essential for ranking on page one.
- LinkGraph’s CTO explains how machine learning is helping us better understand how Google finds and evaluates content quality signals when ranking web pages.
What makes a web page rank on the first page of Google? Historically, the best correlation studies regressed thousands of search factors against page one rankings in an attempt to understand the primary drivers of SERP performance.
But these ranking factors are weighted differently depending on the type of search. Google uses varying search models depending on the search intent behind the question. Local searches with the map pack, higher economic value searches with high CPC, informational queries with high search volume, searches where the most relevant results may be rich media like videos/images, and even search within highly regulated industries like health and money, can all weight ranking factors differently
To make correlation studies even more challenging, Google updates its core algorithm several times a year, meaning those signals continue to evolve. As Google continues to refine its ability to measure and analyze those signals, their bots are getting better at understanding website quality the way that humans do.
Opening Google’s black box and the North star
In the last decade, the academic fields of machine learning and natural language processing have made great strides. Starting around 2012, Google’s search algorithms evolved beyond regression-based models towards deep learning. Google’s ranking algorithm has now become a black box, and even their engineers find it challenging at times to understand why their models produce the results they do.
Ultimately, we know their goal has always been to bring the highest quality search results to users, and they’ve always done the best they could with the technologies and datasets available to them at the time.
Every core update is consistent with what they have been telling us for years: they reward high-quality pages and penalize low-quality sites and spam. Google’s updates have always been marching in the same direction, and it’s well understood among SEOs that there is a North Star towards:
- Higher-quality content
- Fast, snappy user experiences
- Increasing site authority and reputation
What can be frustrating for webmasters is Google’s lack of specific guidance about which specific ranking factors matter the most in their specific industries? When studying how certain properties like page speed, text content, and backlinks relate to rankings, advanced SEOs are turning to correlation analysis and machine learning tools to better understand how Google finds these signals through the noise.
What we learned about the primary drivers of SERP rankings
We look into an industry-specific correlation analysis, and what we learned about the primary drivers of SERP rankings.
To better understand these signals, we studied the correlation of 18 important ranking factors across 200 searches in the sports ticketing vertical. Each keyword had a CPC of over $ 8.
What we discovered was that the most important factors that were correlated with top rankings were high Domain Authority and long-form landing page content. Domain organic traffic value, URL organic traffic, and page load speed were among the weaker correlations.
Unfiltered backlink count and Referring Domain counts themselves were not as strongly correlated as metrics like Moz’s Domain Authority, which differentiates between low-quality links and high-value links. Additionally, Majestic’s Trust Flow and Citation Flow metrics proved less correlated to ranking on the first page then Moz’s Domain Authority.
For the startup client, we were working with, this analysis taught us that maximizing the amount of page rank on their most competitive landing pages was critical in order to contend with the heavyweights in their space, more than anything else they could do for their SEO.
In our analysis, we were surprised to see that page load speeds weren’t as strong of a factor as we might have expected. Nevertheless, given Google’s guidance about their forthcoming 2021 web vitals update, performance signals are likely to become more important. Even though these signals may not be as important as site authority in the algorithm today, I would strongly advise brands to begin preparing for the update and begin making on-page content improvements and page speed improvements right now and understand how your site stacks up in the ‘Chrome User Experience’ report.
What this means for ranking on page one in your industry
If you’re an upstart and looking to break into page one of Google, most of these Google algorithm updates should come as very good news because they create a more democratic SEO landscape. In every correlation analysis, we’ve performed, site authority has always been the most influential factor. It’s also the hardest ranking factor to improve because link building takes such a concerted amount of effort and can take years in big industries.
Seeing other ranking factors demonstrating strong influence over rankings – such as content length and quality, page speed, and web vitals, and high-quality UX. It means that new entrants gain more opportunities to succeed in SEO on the merits of their web pages, not just because they are incumbents with massive backlink profiles.
Even if you don’t have the tools to do your own comprehensive correlation analysis, you can manually reverse engineer the SERPs in your industry. Studying how your competitors to benchmark on these important search factors can reveal valuable tactical insights that help you direct your team’s SEO efforts towards the most impactful work possible and deliver faster-ranking improvements.
Manick Bhan is the founder and CTO of LinkGraph, an award-winning digital marketing and SEO agency that provides SEO, paid media, and content marketing services. He is also the founder and CEO of SearchAtlas, a software suite of free SEO tools. He is the former CEO of the ticket reselling app Rukkus.
The post 1000 Ranking factors: How Google finds signals through the noise appeared first on Search Engine Watch.
Twitter announced Tuesday that many accounts spreading the pervasive right-wing conspiracy theory known as QAnon would no longer be welcome on its platform.
Citing concerns about “offline harm,” the company explained that it would begin treating QAnon content on the platform differently, removing related topics from its trending pages and algorithmic recommendations and blocking any associated URLs. Twitter also said that it would permanently suspend any accounts tweeting about QAnon that have previously been suspended, coordinate harassment against individuals or amplify identical content across multiple accounts.
We will permanently suspend accounts Tweeting about these topics that we know are engaged in violations of our multi-account policy, coordinating abuse around individual victims, or are attempting to evade a previous suspension — something we’ve seen more of in recent weeks.
— Twitter Safety (@TwitterSafety) July 22, 2020
Twitter says the enforcement will go into effect this week and that the company would continue to provide transparency and additional context as it makes related platform policy choices going forward. According to a Twitter spokesperson, the company believes its action will affect 150,000 accounts and more than 7,000 QAnon-related accounts have already been removed for breaking the rules around platform manipulation, evading a ban and spam.
QAnon emerged in the Trump era and the conspiracy’s adherents generally fervently support the president, making frequent appearances at his rallies and other pro-Trump events. QAnon’s supporters believe that President Trump is waging a hidden battle against a secretive elite known as the Deep State. In their eyes, that secret battle produces many, many clues that they claim are encoded in messages sprinkled across anonymous online accounts and hinted at by the president himself.
QAnon is best known for its connection to Pizzagate, a baseless conspiracy that accused Hillary Clinton of running a sex trafficking ring out of a Washington D.C. pizza place. The conspiracy inspired an armed believer to show up to the pizza shop, where he fired a rifle inside the restaurant, though no one was injured.
While the conspiracy theory is elaborate, odd, and mostly incoherent, it’s been popping up in other mainstream places. Last week, Ed Mullins, the head of one of New York City’s most prominent police unions, spoke live on Fox News with a mug featuring the QAnon logo within clear view of the camera. In Oregon, a QAnon supporter won her primary to become the state’s Republican nominee for the Senate.
From defense contractors to videogame companies, the indictment details an astonishing array of victims.
Feed: All Latest
At Microsoft Inspire today, the company made several Dynamics 365 announcements, including Dynamics 365 Customer Voice, a real-time customer feedback tool that could compete with Qualtrics, the company SAP bought in 2018 for a cool $ 8 billion.
Microsoft General Manager Brenda Bown says that as more customers move online during the pandemic, it’s more important than ever to capture real-time customer feedback that you can combine with other data to build a more complete picture of the customer that could lead to more successful interactions in the future.
“Customer Voice is a feedback management solution, and it’s designed to empower businesses and organizations to build better products, deliver better experiences to customers, and really build the relationships for the customers with that feedback management tool,” Bown told TechCrunch.
The data gets shared with Microsoft’s customer data platform (CDP), and is built on top of Dynamics 365 and the Power Platform. The latter provides a way to customize the Customer Voice tool to meet the needs of an individual company.
Brent Leary, partner and co-founder at CRM Essentials, says this solves the problem of getting feedback as the interaction is happening. He adds that being able to share that data directly with the CDP makes it even more valuable.
“Customer feedback has to be done as close to the interaction/transaction as possible and as frictionless as possible for it to really work, or else customers won’t give it to you. And then the data has to be integrated into the CDP with all the other data automatically to really be of use. And having a platform to handle both the feedback capture and the data integration optimizes the likelihood of this happening,” Leary said.
The company also announced Dynamics 365 Connected Store, a set of tools designed to help stores manage in-store and curbside traffic, among other things. As the pandemic limits the number of people in a store at one time, using sensors and cameras, Connected Store can help managers understand and manage the number of people inside the store at any given time to help aid in social distancing.
It can also help add a level of automation to curbside pickup, letting an employee know when the customer has pulled up. “It alerts the employee and they can bring out the order for a more seamless and quick pickup. And obviously this scenario is super important today because of [more people wanting] contactless pick up,” Bown said.
Finally, the company announced a fraud protection component. She says that Dynamics 365 Fraud Protection helps protect businesses online or in physical stores from fraudulent activities, which she says is even more important as more transactions are conducted digitally. New capabilities include account protection and loss prevention tooling.
Inspire is the company’s annual partner conference, which is being held virtually this year. Bown says by running it virtually, the company can involve even more partners than a typical in-person conference because companies that couldn’t previously attend because of cost and distance, are able to participate this year.
- If you are thinking about having a site designed or redesigned, it is important to understand the complex relationship between SEO and web design.
- While designers focus on the aesthetic look of a website, and SEOs focus on optimizing the site for search engine rankings, the desired result is the same – a site that gives visitors what they want and helps your business grow.
- SEO Expert at UENI, Javier Bello discusses how SEO and web design can work together to create a successful business and attract more visitors to your site.
Do you consider SEO and website design as two separate elements of your website? You shouldn’t. There is no point developing a slick website if nobody will be able to find the site online, and with over 1.2 billion websites on the internet, it can be hard for your website to stand out. To avoid having a website that is not ‘search engine friendly’, we spoke to SEO expert, Javier Bello at UENI, who outlines five ways SEO and Web Design can go together.
Research shows that mobile devices generate over half of global website traffic, so having a website that is mobile-friendly is essential to reach over half your audience. Google made mobile-friendliness a ranking factor in 2015, and mobile indexing was then introduced in 2017, which meant that Google predominantly uses the mobile version of the content for indexing and ranking.
Therefore, to reach a wider customer base and rank higher on Google, you should spend some extra time working on the mobile version of your site. Is your mobile site just as usable as the desktop version? Are your pictures placed in the same format? Is everything tagged the same way as it is on the desktop version? Understanding that mobile matters will make your SEO better and build your Google rapport.
2. Website speed
There is nothing that will make a visitor click off more quickly than a slow website. All your web pages should load instantly, otherwise, this can impact user experience and SEO performance. If you find that your website speed is lagging, chances are it could have something to do with your website design. Devote your time to optimizing images, removing unnecessary plugins, allowing browser caching, and so on. One study found that 88% of consumers are unlikely to return if they had a bad experience with your site, so it is crucial for your success that you have a high website speed.
3. User friendly
Everyone knows that content strategy is an essential part of SEO, but did you know that the presentation of it can affect your rankings too? Not only is user-friendly, intuitive design an underrated component of SEO, but it will ensure good user experience so that your customers do not click off your web page. Too many hyperlinks, text that is difficult to read against backgrounds, images that take too long to load, and pages with blocks of content in strange places are all examples of bad website design. This might erase the audience you have worked so hard to bring to your site. Treat Google like a regular customer and make it easy for consumers to take in your content, so they stay and click through your site.
4. Site maps
A sitemap is essentially a blueprint of your website that helps search engines find, crawl and index your website’s content. A good sitemap will allow search engine crawlers to more intelligently crawl your site, therefore improving Google rankings and bringing in more traffic. That is if your content is already well prepared and appealing to web users. To create a site map that makes sense, ensure that your design intuitively leads people to the right place and that your internal linking structure makes sense to you. This may sound like a lot of work, but effective SEO is a process and can be considered a technical art.
5. Fresh and engaging content
Without good content, it becomes difficult to create informative web pages or rank highly in search engines. Google loves sites that have a clear content structure, with easy to follow pages and keywords positioned strategically. Also, when existing content is updated it indicates to Google that your site is ‘alive’ and greater crawling frequency is achieved. Popular content included how-to posts, FAQs, and case studies. Use paragraphs, headings, and signal words to display your content nicely on your webpage, allowing for greater user experience. Lastly, if you display or link to related content on your site, consumers can click on another landing page from your site for more information, rather than looking elsewhere.
Javier Bello is an SEO expert at UENI.
As Marketers our job is to not only interpret analytics data, but to also provide a summary of the performance and apply recommendations for future strategies, forecasting and on-going testing. However, this standard metric of decoding is not enough and we need to find a better way to communicate successes and failures that the client can understand. That is why storytelling is just as important now than it was when we are in Kindergarten when the teacher read us a story in a circle.
In this post, I will highlight the importance of storytelling with the client which not only helps the client understand, but also reinforces the client-agency relationship.
Storytelling is also a Science
As marketers, early on we are classically trained to become proficient in Excel, Powerpoint and (my personal favorite) writing on whiteboards so that we can be perceived as smartest one in the room. These elements of communication comprise of bullet points, summarizations, goals and objectives, sales vs. cost projections, etc… On the contrary, we are most likely doing it all wrong. There have been many studies and published articles that debunk this MBA/classroom method and reinforce the one of oldest and most fundamental communication methods.
In an very “eye-opening” article by Lifehacker.com published back in 2012 entitled “The Science of Storytelling: Why Telling a Story is the Most Powerful Way to Activate Our Brains“, author Leo Widrich states “It’s in fact quite simple. If we listen to a powerpoint presentation with boring bullet points, a certain part in the brain gets activated. Scientists call this Broca’s area and Wernicke’s area. Overall, it hits our language processing parts in the brain, where we decode words into meaning. And that’s it, nothing else happens. When we are being told a story, things change dramatically. Not only are the language processing parts in our brain activated, but any other area in our brain that we would use when experiencing the events of the story are too.“ So in essence, telling stories not only puts our entire brain to work it also allows the storyteller to put ideas and thoughts into the listeners brain as well.
Complexities of Storytelling
For most clients, they do not care too much about CTR%, AVG positions, bounce rates, etc… they want to know what is causing their cash register to ring below are some of the common questions they are mostly concerned about:
- What’s working and why?
- Whats not working and why?
- Why are sales down this month as compared to last month?
- How can we generate more sales without increasing the budget, etc…
Because of this difference in understanding success metrics, marketers need to take all of the Analytics data (which are considered very complex by clients) and transform them into a story/language that they can understand. For example, lets suppose that the client saw a 50% increase in sales coming from their “Brand Terms” in Adwords as compared to the previous month. Instead of just providing them with increased performance metrics such as CTR%, Conversion rates, etc.., marketers need to do a little digging around and form a story that they can understand.
A story would be something like:
“Well, since we added more generalized “non-branded” terms as well as your interview on the local TV station, a larger audience of people who were not familiar with your brand before, typed your brand into Google and clicked on the PPC Text Ads. ” It is this type of success story that can create that “light bulb” in the heads of the client to ensure them that they are prospering their investment in you or your agency.”
Leveraging Web Analytics Data to Feed the Story
Just looking at common performance data is simply not enough to tell a story. Marketers need to look at various layers of data to comprise a story that can makes sense to the client. Identifying these interesting and important metrics such as hour of day, day of the week. GEO by state, metro area, city, direct/bookmark, conversion funnels, etc… These are examples of the metrics, combined with overall performance data is what makes up the holistic story that the client needs to hear. Moreover, these stories often lead to future optimization strategies and testing which is great for the client-agency relationship.
Trying to explain all of the intricate metrics and what they mean to a client is hard enough. But simplifying the data and creating a story around it, even as an “ice-breaker” at the beginning of the conversation, helps the client feel like they made the right choice in hiring you. The one thing we need to remember is that a story, if broken down into the simplest form, is a connection of cause and effect and that is what clients need to understand.
Changes to How Google Might Rank Image Search Results
We are seeing more references to machine learning in how Google is ranking pages and other documents in search results.
That seems to be a direction that will leave what we know as traditional, or old school signals that are referred to as ranking signals behind.
It’s still worth considering some of those older ranking signals because they may play a role in how things are ranked.
As I was going through a new patent application from Google on ranking image search results, I decided that it was worth including what I used to look at when trying to rank images.
Images can rank highly in image search, and they can also help pages that they appear upon rank higher in organic web results, because they can help make a page more relevant for the query terms that page may be optimized for.
Here are signals that I would include when I rank image search results:
- Use meaningful images that reflect what the page those images appear on is about – make them relevant to that query
- Use a file name for your image that is relevant to what the image is about (I like to separate words in file names for images with hyphens, too)
- Use alt text for your alt attribute that describes the image well, and uses text that is relevant to the query terms that the page is optimized for) and avoid keyword stuffing
- Use a caption that is helpful to viewers and relevant to what the page it is about, and the query term that the page is optimized for
- Use a title and associated text on the page the image appears upon that is relevant for what the page is about, and what the image shows
- Use a decent sized image at a decent resolution that isn’t mistaken for a thumbnail
Those are signals that I would consider when I rank image search results and include images on a page to help that page rank as well.
A patent application that was published this week tells us about how machine learning might be used in ranking image search results. It doesn’t itemize features that might help an image in those rankings, such as alt text, captions, or file names, but it does refer to “features” that likely include those as well as other signals. It makes sense to start looking at these patents that cover machine learning approaches to ranking because they may end up becoming more common.
Machine Learning Models to Rank Image Search Results
Giving Google a chance to try out different approaches, we are told that the machine learning model can use many different types of machine learning models.
The machine learning model can be a:
- Deep machine learning model (e.g., a neural network that includes multiple layers of non-linear operations.)
- Different type of machine learning model (e.g., a generalized linear model, a random forest, a decision tree model, and so on.)
We are told more about this machine learning model. It is “used to accurately generate relevance scores for image-landing page pairs in the index database.”
We are told about an image search system, which includes a training engine.
The training engine trains the machine learning model on training data generated using image-landing page pairs that are already associated with ground truth or known values of the relevance score.
The patent shows an example of the machine learning model generating a relevance score for a particular image search result from an image, landing page, and query features. In this image, a searcher submits an image search query. The system generates image query features based on the user-submitted image search query.
That system also learns about landing page features for the landing page that has been identified by the particular image search result as well as image features for the image identified by that image search result.
The image search system would then provide the query features, the landing page features, and the image features as input to the machine learning model.
Google may rank image search results based on various factors
Those may be separate signals from:
- Features of the image
- Features of the landing page
- A combining the separate signals following a fixed weighting scheme that is the same for each received search query
This patent describes how it would rank image search results in this manner:
- Obtaining many candidate image search results for the image search query
- Each candidate image search result identifies a respective image and a respective landing page for the respective image
- For each of the candidate image search results processing
- Features of the image search query
- Features of the respective image identified by the candidate image search result
– Generating an image search results presentation that displays the candidate image search results ordered according to the ranking
– Providing the image search results for presentation by a user device
Advantages to Using a Machine Learning Model to Rank Image Search Results
If Google can rank image search query pairs based on relevance scores using a machine learning model, it can improve the relevance of the image search results in response to the image search query.
This differs from conventional methods to rank resources because the machine learning model receives a single input that includes features of the image search query, landing page, and the image identified by a given image search result to predicts the relevance of the image search result to the received query.
This process allows the machine learning model to be more dynamic and give more weight to landing page features or image features in a query-specific manner, improving the quality of the image search results that are returned to the user.
By using a machine learning model, the image search engine does not apply the same fixed weighting scheme for landing page features and image features for each received query. Instead, it combines the landing page and image features in a query-dependent manner.
The patent also tells us that a trained machine learning model can easily and optimally adjust weights assigned to various features based on changes to the initial signal distribution or additional features.
In a conventional image search, we are told that significant engineering effort is required to adjust the weights of a traditional manually tuned model based on changes to the initial signal distribution.
But under this patented process, adjusting the weights of a trained machine learning model based on changes to the signal distribution is significantly easier, thus improving the ease of maintenance of the image search engine.
Also, if a new feature is added, the manually tuned functions adjust the function on the new feature independently on an objective (i.e., loss function, while holding existing feature functions constant.)
But, a trained machine learning model can automatically adjust feature weights if a new feature is added.
Instead, the machine learning model can include the new feature and rebalance all its existing weights appropriately to optimize for the final objective.
Thus, the accuracy, efficiency, and maintenance of the image search engine can be improved.
The Rank Image Search results patent application can be found at
Ranking Image Search Results Using Machine Learning Models
US Patent Application Number 16263398
File Date: 31.01.2019
Publication Number US20200201915
Publication Date June 25, 2020
Applicants Google LLC
Inventors Manas Ashok Pathak, Sundeep Tirumalareddy, Wenyuan Yin, Suddha Kalyan Basu, Shubhang Verma, Sushrut Karanjkar, and Thomas Richard Strohmann
Methods, systems, and apparatus including computer programs encoded on a computer storage medium, for ranking image search results using machine learning models. In one aspect, a method includes receiving an image search query from a user device; obtaining a plurality of candidate image search results; for each of the candidate image search results: processing (i) features of the image search query and (ii) features of the respective image identified by the candidate image search result using an image search result ranking machine learning model to generate a relevance score that measures a relevance of the candidate image search result to the image search query; ranking the candidate image search results based on the relevance scores; generating an image search results presentation; and providing the image search results for presentation by a user device.
The Indexing Engine
The search engine may include an indexing engine and a ranking engine.
The indexing engine indexes image-landing page pairs, and adds the indexed image-landing page pairs to an index database.
That is, the index database includes data identifying images and, for each image, a corresponding landing page.
The index database also associates the image-landing page pairs with:
- Features of the image search query
- Features of the images, i.e., features that characterize the images
- Features of the landing pages, i.e., features that characterize the landing page
Optionally, the index database also associates the indexed image-landing page pairs in the collections of image-landing pairs with values of image search engine ranking signals for the indexed image-landing page pairs.
Each image search engine ranking signal is used by the ranking engine in ranking the image-landing page pair in response to a received search query.
The ranking engine generates respective ranking scores for image-landing page pairs indexed in the index database based on the values of image search engine ranking signals for the image-landing page pair, e.g., signals accessed from the index database or computed at query time, and ranks the image-landing page pair based on the respective ranking scores. The ranking score for a given image-landing page pair reflects the relevance of the image-landing page pair to the received search query, the quality of the given image-landing page pair, or both.
The image search engine can use a machine learning model to rank image-landing page pairs in response to received search queries.
The machine learning model is a machine learning model that is configured to receive an input that includes
(i) features of the image search query
(ii) features of an image and
(iii) features of the landing page of the image and generate a relevance score that measures the relevance of the candidate image search result to the image search query.
Once the machine learning model generates the relevance score for the image-landing page pair, the ranking engine can then use the relevance score to generate ranking scores for the image-landing page pair in response to the received search query.
The Ranking Engine behind the Process to Rank Image Search Results
In some implementations, the ranking engine generates an initial ranking score for each of multiple image—landing page pairs using the signals in the index database.
The ranking engine can then select a certain number of the highest-scoring image—landing pair pairs for processing by the machine learning model.
The ranking engine can then rank candidate image—landing page pairs based on relevance scores from the machine learning model or use those relevance scores as additional signals to adjust the initial ranking scores for the candidate image—landing page pairs.
The machine learning model would receive a single input that includes features of the image search query, the landing page, and the image to predict the relevance (i.e., relevance score, of the particular image search result to the user image query.)
We are told that this allows the machine learning model to give more weight to landing page features, image features, or image search query features in a query-specific manner, which can improve the quality of the image search results returned to the user.
Features That May Be Used from Images and Landing Pages to Rank Image Search Results
The first step is to receive the image search query.
Once that happens, the image search system may identify initial image-landing page pairs that satisfy the image search query.
It would do that from pairs that are indexed in a search engine index database from signals measuring the quality of the pairs, and the relevance of the pairs to the search query, or both.
For those pairs, the search system identifies:
- Features of the image search query
- Features of the image
- Features of the landing page
Features Extracted From the Image
These features can include vectors that represent the content of the image.
Vectors to represent the image may be derived by processing the image through an embedding neural network.
Or those vectors may be generated through other image processing techniques for feature extraction. Examples of feature extraction techniques can include edge, corner, ridge, and blob detection. Feature vectors can include vectors generated using shape extraction techniques (e.g., thresholding, template matching, and so on.) Instead of or in addition to the feature vectors, when the machine learning model is a neural network the features can include the pixel data of the image.
Features Extracted From the Landing Page
These aren’t the kinds of features that I usually think about when optimizing images historically. These features can include:
- The date the page was first crawled or updated
- Data characterizing the author of the landing page
- The language of the landing page
- Features of the domain that the landing page belong to
- Keywords representing the content of the landing page
- Features of the links to the image and landing page such as the anchor text or source page for the links
- Features that describe the context of the image in the landing page
- So on
Features Extracted From The Landing Page That Describes The Context of the Image in the Landing Page
The patent interestingly separated these features out:
- Data characterizing the location of the image within the landing page
- Prominence of the image on the landing page
- Textual descriptions of the image on the landing page
More Details on the Context of the Image on the Landing Page
The patent points out some alternative ways that the location of the image within the Landing Page might be found:
- Using pixel-based geometric location in horizontal and vertical dimensions
- User-device based length (e.g., in inches) in horizontal and vertical dimensions
- An HTML/XML DOM-based XPATH-like identifier
- A CSS-based selector
The prominence of the image on the landing page can be measured using the relative size of the image as displayed on a generic device and a specific user device.
The textual descriptions of the image on the landing page can include alt-text labels for the image, text surrounding the image, and so on.
Features Extracted from the Image Search Query
The features from the image search query can include::
- Language of the search query
- Some or all of the terms in the search query
- Time that the search query was submitted
- Location from which the search query was submitted
- Data characterizing the user device from which the query was received
- So on
How the Features from the Query, the Image, and the Landing Page Work Together
- The features may be represented categorically or discretely
- Additional relevant features can be created through pre-existing features (Relationships may be created between one or more features through a combination of addition, multiplication, or other mathematical operations.)
- For each image-landing page pair, the system processes the features using an image search result ranking machine learning model to generate a relevance score output
- The relevance score measures a relevance of the candidate image search result to the image search query (i.e., the relevance score of the candidate image search result measures a likelihood of a user submitting the search query would click on or otherwise interact with the search result. A higher relevance score indicates the user submitting the search query would find the candidate image search more relevant and click on it)
- The relevance score of the candidate image search result can be a prediction of a score generated by a human rater to measure the quality of the result for the image search query
Adjusting Initial Ranking Scores
The system may adjust initial ranking scores for the image search results based on the relevance scores to:
- Promote search results having higher relevance scores
- Demote search results having lower relevance scores
- Or both
Training a Ranking Machine Learning Model to Rank Image Search Results
The system receives a set of training image search queries
For each training image search query, training image search results for the query that are each associated with a ground truth relevance score.
A ground truth relevance score is the relevance score that should be generated for the image search result by the machine learning model (i.e., when the relevance scores measure a likelihood that a user would select a search result in response to a given search query, each ground truth relevance score can identify whether a user submitting the given search query selected the image search result or a proportion of times that users submitting the given search query select the image search result.)
The patent provides another example of how ground-truth relevance scores might be generated:
When the relevance scores generated by the model are a prediction of a score assigned to an image search result by a human, the ground truth relevance scores are actual scores assigned to the search results by human raters.
For each of the training image search queries, the system may generate features for each associated image-landing page pair.
For each of those pairs, the system may identify:
(i) features of the image search query
(ii) features of the image and
(iii) features of the landing page.
We are told that extracting, generating, and selecting features may take place before training or using the machine learning model. Examples of features are the ones I listed above related to the images, landing pages, and queries.
The ranking engine trains the machine learning model by processing for each image search query
- Features of the image search query
- Features of the respective image identified by the candidate image search result
- Features of the respective landing page identified by the candidate image search result and the respective ground truth relevance that measures a relevance of the candidate image search result to the image search query
The patent provides some specific implementation processes that might differ based upon the machine learning system used.
Take Aways to Rank Image Search Results
I’ve provided some information about what kinds of features Google May have used in the past in ranking Image search results.
Under a machine learning approach, Google may be paying more attention to features from an image query, features from Images, and features from the landing page those images are found upon. The patent lists many of those features, and if you spend time comparing the older features with the ones under the machine learning model approach, you can see there is overlap, but the machine learning approach covers considerably more options.
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
A set of mini apps has gone live on Snapchat platform, marking the beginning of a new chapter for Los Angeles-headquartered firm as it aims to emulate aspects of the popular Chinese “super-app” model.
Unveiled last month, Snap Minis are lightweight, simplified versions of apps that live within Snap’s Chat section. These apps — built with HTML — are designed to improve engagement among users by enabling them to perform a range of additional tasks without leaving Snap app.
Four of the seven “Minis” that Snap unveiled last month are now available across the platform. These mini apps that are going live today are: Meditation service Headspace, studying collaboration tool Flashcards, an “interactive messaging experience” service called Prediction Master, and Let’s Do It, a mini app developed by Snap itself that allows users to make decision with their friends.
Mini apps unveiled by Coachella that would allow users to plan festival trip, Atom’s movie ticketing, and Saturn, which is aimed at helping students share and compare their class schedules are yet to go live.
The rollout on Monday is nonetheless an important shift in Snap’s strategy to boost engagement on its ephemeral messaging app, which has amassed over 229 million daily users.
Though a relatively new concept in the U.S. and UK, mini apps model is quite popular in Asian markets. Tencent’s WeChat has attracted over a million miniature apps that allow users to perform a range of tasks.
In India, mobile payments services PhonePe and Paytm have rolled out several such in-apps, too, that allow users to book flight and movie tickets and order food and cabs.
Snapchat has previously said that its relationship with Tencent, an investor in the Los Angeles firm, has been influential in its decision to replicate the super-app offering.
The strategy looks promising — at least on paper. It’s a win-win scenario for both Snap and the developers who make these mini-apps. By gaining access to these mini-apps, Snap can potentially see a boost in user engagement, and developers are able to cater to a whole set of new audience.
But whether this model finds home with users in the U.S. and the UK and other markets where Snap has made inroads — and regions that unlike China are open — remains a mystery. As my colleague Lucas pointed out last month, Facebook has attempted to replicate the WeChat model through chatbots on Messenger over the years to little success.
- Why You Should Layer Affinity Audiences on your Google Ads Search Campaigns
- Twitter now lets everyone limit replies to their tweets
- Maybe Netflix and Amazon Should Just Buy Theater Chains
- Parsable scores $60M Series D as pandemic forces faster digitization of industrial sector
- Defining value stream management for SEO agencies business owners