How Google May Answer Questions in Queries with Rich Content Results

From the query, 'How to fix a clogged sink.'
From the query, ‘How to fix a clogged sink.’

This patent application from last week describes a search system query processing for queries that are questions.

Queries may be responded to with a wide variety of resources, such as image files, audio files, video files, and web pages. A search engine may identify resources in response to queries submitted by searchers and attempt to provide information in response “in a manner that is useful to the users.”

Searchers may look for an answer to a specific question, rather than a listing of links to other pages and resources.

They may want to know what the weather is like at a particular location, or a current quote for a stock, or the capital of a state, etc.

When a query is shaped in the form of a question, a search engine may work to response to that question format, by providing a response in the form of an “answer,” such as a “one box” answer to the question.

Last month, I wrote the post How Google May Trigger Answer Box Results For Queries, which provided a different description of how queries might be responded to with answer box results. It’s not surprising to have a number of patents that might look like responses to the same problem. This differs by attempting to provide a “rich” answer in response to a query, such as an image or audio or video.

There are a number of important steps related to the process in this patent, which include:

1) This patent works to receive a question query and return an answer responsive to the question query;
2) Using a contextual query that includes an element from the question query and an element from the answer; and submit that contextual query to a rich content search process and receiving rich content responsive to the contextual query, the rich content items ranked according to an order that is indicative of the relevance of each rich content item to a query for which the set was generated;
3) determining first rich content item in the first set of rich content items that meet a context condition that is indicative of a rich content item providing contextual information of both elements of the question query and the answer query; and
4) preferentially selecting from the first content items relative to the second rich content items to be provided as one or more answer images.

Advantages

This patent focuses upon providing rich content, such as an image, audio, or a video, to answer a question with. The rich content is chosen to convey to a searcher information regarding both the content of the question and the answer – this aims to “ensure that rich information provided to the user for an answer is responsive to the question and consistent with the answer provided.”

The patent application is:

Rich Content for Query Answers
Pub. No.: WO/2015/102869
International Application No.: PCT/US2014/070359
Publication Date: 09.07.2015
International Filing Date: 15.12.2014
Assigned to: Google
Invented by: Gal Chechik, Eyal Segalis, Yaniv Leviathan, and Yoav Izur

Abstract:

Methods and systems for providing rich content with an answer to a question query.

A method includes receiving a query determined to be a question query and a corresponding answer generated in response to the question query, generating a contextual query that includes an element relating to the question query and an element relating to the answer; submitting the contextual query to a rich content search process and receiving data specifying a first set of rich content items responsive to the contextual query, determining first rich content item in the first set of rich content items that meet a context condition that is indicative of a rich content item providing contextual information of both elements of the question query and the answer query; and preferentially selecting from the first content items relative to the second rich content items to be provided as one or more answer rich content items.

Question Queries versus Informational Queries

Searchers often look for “answers” rather than a range of links to web pages. An answer can be defined as a fact or relation responsive to a question. “Question queries” differ from an informational queries in that the question query is looking for the answer, while an informational query requests a variety of different types of information about a subject. As an example, a question query along the lines of “How high is Pikes Peak” is structured to clearly requests a fact – the height of Pikes Peak. A search engine can determine that the query is in a structured format and may interpret it as a question query, and return in response to the query, a result that is an answer to the question, as well as other results.

Rich Content Question Responses

A rich content question processing system provides an answer for a question query using rich information specific to the question query and the answer. The search engine returns a question query and an answer generated in response to the question query. Both the question query and the answer will have one or more terms. This system then creates a contextual query that includes elements relating to the question query and to the answer. For example, the contextual query can include at least one or more terms from the question query and one or more terms from the answer. Alternatively, the contextual query can specify entities that are identified by the question query and answer, or related terms derived from the question query and answer.

That contextual query may be submitted to a rich content search process, such as an image search process or a search for videos and audio. image rich content is used as an example.

An image search process would specify a first set of images in response to the contextual query, and then ranks those images in an order indicative of the relevance of each image to the question query. A first set would be provided to the rich content question processing system, which would determines first images and second images in the first set. The first images are images meeting a context condition indicative of an image providing contextual information of both terms of the question and terms of the answer, and the second images do not meet the context condition.

The first and second images may be chosen based in part on respective image sets for the question query and answer query, as well as a comparison of those image sets to the image set for the contextual query. They might also be chosen based upon image labels of each respective image.

contextual-query

Rich content in response to a question query might be chosen from many thousands of publisher websites.

The search engine crawls those publisher websites and indexes resources provided by the publisher websites, which are stored in a resource index.

Searcher requests are used to index and identify resources relevant to the queries, in the form of search results. A search result for a resource may include “a web page title, a snippet of text extracted from the web page, and a resource locator for the resource, e.g., the URL of a web page.”

These search results are ranked based on a mix of Information Retrieval (“IR”) scores and an “Authority” Score (such as PageRank ) related to the resources. The results are returned in that order based upon those scores.

Query Logs and Selection Logs

Queries submitted by searchers may be stored in query log files. The selection data for those queries and web pages referenced by those search results and selected by users could be stored in selection logs. This query log and selection information define search history data that includes data from and related to previous search requests associated with unique user/device identifiers. Those selection logs represent searcher actions taken in response to the search results that were provided by the search engine. This kind of information can include clicks on the search results. Both the query logs and selection logs can be used to map queries submitted by user devices to resources that were identified in search results and track actions taken by searchers when they are presented with the search results in response to the queries. Also, data may be associated with the identifiers from the search requests so that a search history for each identifier can be accessed. These selection logs and query logs can then be used by the search engine to track the respective sequences of queries submitted by the user devices, and the actions taken in response to the queries, and to track how often the queries have been submitted.

When personally identifiable information is collected about searchers, those searchers may be given a chance to control whether programs or features are used to collect user information from them, such as information about:

  • a user’s social network
  • social actions or
  • activities
  • profession
  • a user’s preferences, or
  • a user’s current location.

This is may be done to give them the option to protect the sharing and use of this information.

Question Query Processing

The patent describes when a query is in the form of a question, or an implicit question. An example of an implicit question might be the query “capital of California.”, short for “What is the capital of California?” A question may be more pecific, as in the query “What is the capital of California?” The search engine may have a “query question processor” to decide whether a query is a query question, and, if it is to decide whether there is an answer that is responsive to the question.

This query question processor may use a variety of different algorithms to determine whether a query is a question and also whether there is a particular answer responsive to the question.
These may involve language models, machine learned processes, knowledge graphs, grammars, search results and resources referenced by the search results, or combinations of all those, to determine question queries and answers.

If a query is determined to be a question query and there is a responsive answer, the query question processor may then get the rich content processor involved, which acts to identify rich content to be show with the answer. While that could be an image or set of images, it could also include video data, or audio data.

Responding with Rich Content to Queries

capital-california

A question such as “What is the capital of California” is answered by “Sacramento.” An ideal rich content piece responds to both the question and the answer, but showing a result that might include both, such as a text answer that identifies “sacremento’ and maybe an image of the capital building of California. A map image pinpointing Sacremento’s location in California might also be considered relevant as well.

Note that the images are in response to a combination of the question and its answer.
Note that the images are in response to a combination of the question and its answer.

The patent tells us that providing both the question and the answer to the rich content processor enables the question query respons to be “represented” by both “question query terms {QQT} and answer terms {AT}, respectively.”

This is a contextual query which includes elements from the question and from the answer. The contextual query can also “specify entities that are identified by the question query and answer, or related terms (synonyms, term expansions, etc.) derived from the terms of the question query and terms of the answer.”

It really can be as simple as having a question query of “Capital of California” and answer of “Sacramento,” and creating a the context query of “Capital of California Sacramento.” The patent tells us that other ways to generating a context query can also be used, and could involve removing stop words, substituting terms, and so on.

Rich content labels, such as tags, descriptors, and other text that may describe an image. may also be used in ranking these images. For instance, a logo of a sports team located in Sacremento may not rank as highly as other images, and may be demoted in rankings. The patent describes other instances of some rich content being demoted in rankings.

Take-aways

It’s good seeing some of the approaches and mechanisms that might be used by the search engine to provide responses to question answering queries. Seeing one that may be involved in results that show off videos or images in response to a question answering query is interesting. The idea of using a “contextual query” based upon both the question and the answer is also interesting.

Summary
Article Name
How Google May Answer Questions in Queries with Rich Content Results
Description
Google may respond to some queries with Answers. treating them as question-queries, and provide rich content in response such as images or audio or video.
Author

32 thoughts on “How Google May Answer Questions in Queries with Rich Content Results”

  1. Once again – brilliant Article! I wonder, when these types of results may appear in Germany. The implementation and experimental fashion, with which Google wokrs in US is fascinating andsomeimtes frustrating for SEOs that aren’t there.

    Thanks!
    Heinz

  2. Loved the way you describe everything in detail. It’d be interesting to have detailed answers to our simple questions. Looking forward to the implementation all over. Meanwhile can anyone suggest or predict the time by which these features will be available?

  3. 23. Great post Admin,
    An excellent example of how useful SEO can be for shaping your marketing strategy. Next time I need content ideas for a client I know where to look.

  4. Hi Heinz,

    No telling when these results might appear in Germany. Keep your eyes open and you may be the first to spot them.. 🙂

  5. Well explained content. Thanks for sharing such an informative blog post with us. Rich content have always given priority, as you described. It will
    help many of us.

  6. Hi Bill,

    I wonder from where you get this ‘insider’ sort of information regarding latest patents from the Google.

    Who wouldn’t love best and most relevant images along with answer box, to make the information complete and self explanatory.

  7. This is my first visit to your blog. Read your article about blogging. You man!! have completely nailed the Google Thing. A deep scanning about all google things need to know for bloggers.
    ya, Google is clearly sticking what it had said in the past that it wants its users to give most appropriate search results. Rich content results are giving the most exact answer in almost every case. Thanks for sharing the info. I liked the way you explain the things here. Keep up the good work.

  8. It’s interesting that G is considering returning one, and perhaps even two images to be included in the response, considering how high mobile use is getting. Thanks for all the info Bill.
    Andy 🙂

  9. Hi Bill,
    I must admit that you explained the content well. You’re so amazing for sharing such informative and valuable information with us. It’ll surely be helpful for many people.
    Regard,
    Edgar

  10. Hi Edgar.

    I feel like I have a much better grasp on the process described in the patent after having tried to share the details spelled out in the patent. I know a lot of people follow along, so I’m happy to help people who are, too.

  11. Yes, I believe the world it is also the fastest way to move forward. Thanks for posting such information. Keep sharing. This is quite helpful in business. It can help the business to grow better. I pray for you.

  12. I’ve noticed this on google too! I have one website in which the KW is a question. It’s fairly new, so it’ll be interesting to see how it does. It is a fairly low competition KW as well 🙂 Thanks for the info. Great article!

  13. Hi Kristin,

    This is a little different than optimizing a page for a keyword in the shape of a question. Hopefully that page will end up ranking well for the term you chose to optimize it for, but it might not end up showing up as a direct answer result for that query. Creating a page that contains the question, and the answer to the question upon it seems like it should be helpful as a step towards creating a page that will show up as an answer result like that. We will see. Hopefully it does rank well for that term.

  14. I like what what Google are doing in trying to provide relevant and useful information. How long before these Rich Content results become Sponsored?

    Thank for the article Bill, Always a joy to read although sometimes I need to read a few times to get my head around things:-)

  15. Creating a page that contains the question, and the answer to the question upon it seems like it should be helpful as a step towards creating a page that will show up as an answer result like that. I know a lot of people follow along, so I’m happy to help people who are, too.It’ll surely be helpful for many people.
    Regard,
    Akshay Garg

  16. Hi Syed,

    I am getting information from Google’s patents. This process is described by their inventors within the patent. It may be the way Google actually does this, but I can’t tell you for certain.

  17. It was really brilliant. I thought about this topics and ask many of my friends but they can’t able to give me answer but now I found my answer and many many thanks to Bill Slawski.

  18. I’ve seen these type Queries and Answer. I was amazed to see, as Query Answers shown by Google was from blog post which was ranking at number 8th position for that particular query.

    Why Google do such thing ?

  19. Hi Sumit,

    The question answering results that Google chooses to display in response to a question/query often comes from a high ranking page for the query term; but it doesn’t have to be the highest ranking page for the term. Some other decisions an play a role in the selection of an answer, such as how well it might answer the question posed in the query, and how well it might be formatted as an answer. It’s also possible that more than one answer may be a good match and Google might show one or the other, and see if people tend to click on a “read more” link that might be placed within that answer; if the answer gets a lot of clicks, that may be a sign indicating that viewers are interested in the answer.

  20. Hi Bill, some questions also revolving in mind on this topic like others but all the things are clarified here.

    Now a days i am seeing twitter tweets on google first for main keywords, can you explain it??

  21. Hi Techi,

    An interesting question. It’s possible that Google is showing twitter tweets as top results for certain keywords because they may feel that those queries are looking for very timely news or information, and tweets tend to carry very timely information for some topics really well.

  22. Hi Bill,
    It’s a nice post undoubtedly. Well, any guide to rank on answer box? I have listed my content in well manner way. Does Off Page SEO also matters in this case?

  23. Hi Basheer,

    One of Google’s patents about direct answers tells us that they will try to provide direct answers from sites they’ve decided are authoritative, so in that manner, offsite SEO likely matters to a degree.

  24. Hi Bill,
    Rich Content would be the best part for SERP rankings.
    Came to know one more time with your article.
    What about Less Content Sites or Blogs ranking for major keywords?.
    Is that because of site authority?. Else Good Content is the Only thing that we need?. Please clarify me and give me a correlation for Content and Authority 😀

Comments are closed.