20 Jun Google’s Related Questions Patent or ‘People Also Ask’ Questions
When you search at Google, the solutions you obtain typically now embrace extra questions, that usually have the label above them, “People Also Ask.” I used to be curious if I would have the ability to discover a patent about these questions, and I noticed that they had been typically known as “related questions.”
An article at Moz at this time on the subject was fascinating: Infinite ‘People Also Ask’ Boxes: Research and website positioning Opportunities. The solutions about how these associated questions are determined upon appears to have a less complicated origin as described in Google’s patent, however it’s fascinating evaluating the concepts from that publish with the patent.
I searched by means of Google patent seek for “related questions” and I got here up with a patent named, “Generating related questions for search queries”. When I seemed on the screenshots that accompanied the patent, they gave the impression to be similar to the “People also ask” kind questions Google exhibits us at this time in search outcomes.
The patent supplies some details about how Google will get these associated questions.
It seems that Google appears to be like at a question it receives and after receiving a lot of search outcomes, will determine upon one or extra subject units for every of the search outcome sources from “previously submitted search queries that have resulted in users selecting search results identifying the search result resource” and “selecting related questions from a question database using the topic sets.” The questions which can be chosen from these subject units could also be returned together with search outcomes for the question looked for.
The query database contains beforehand submitted search queries which were decided to be in query kind.
Deciding upon subject units for every of the search outcome useful resource includes:
(1) Identifying certified search queries for the search outcome useful resource – beforehand submitted search question that resulted in a person deciding on a search outcome that identifies the search outcome useful resource;
(2) Ranking the certified search queries based mostly both on a lot of instances every question has been submitted or based mostly on a lot of instances customers have chosen a search outcome figuring out the search outcome useful resource after submitting every question; and
(three) Selecting one or extra highest-ranked certified search queries as the subject units for the search outcome useful resource.
The questions chosen to be included with a set of search outcomes is perhaps in contrast with one another, and if they seem like equivalents of each other, the perfect model of that query would possibly substitute equal questions.
The patent provides us a purpose for displaying searchers associated questions:
Providing associated inquiries to customers may also help customers who’re utilizing un-common key phrases or terminology of their search question to determine key phrases or phrases which can be extra generally used to explain their intent. The person expertise might be improved by submitting the displayed content material of a associated query as a brand new search question and receiving a pre-determined, pre-formatted reply to the associated query as a part of a response from the search engine.
The patent is:
Generating associated questions for search queries
Inventors Yossi Matias, Dvir Keysar, Gal Chechik, Ziv Bar-Yossef, Tomer Shmiel
Publication quantity US9213748 B1
Granted date: Dec 15, 2015
Filing date Mar 14, 2013
Methods, techniques, and equipment, together with laptop applications encoded on laptop storage media, for figuring out associated questions for a search question is described. One of the strategies contains receiving a search question from a person gadget; acquiring a plurality of search outcomes for the search question supplied by a search engine, whereby every of the search outcomes identifies a respective search outcome useful resource; figuring out one or extra respective subject units for every search outcome useful resource, whereby the subject units for the search outcome useful resource are chosen from beforehand submitted search queries which have resulted in customers deciding on search outcomes figuring out the search outcome useful resource; deciding on associated questions from a query database utilizing the subject units; and transmitting knowledge figuring out the associated inquiries to the person gadget as a part of a response to the search question.
The Question Database
The query database could also be chosen from queries which can be in query kind, and a question could also be decided to have been in query kind based mostly upon a lot of options which may be related to it. It might need:
The predetermined set of query phrases can embrace one or extra of interrogative phrases, e.g., interrogative determiners, interrogative pronouns, and interrogative pro-adverbs, different perform phrases which can be steadily used to ask a query,
It may have punctuation marks, corresponding to query marks.
It might match sure query question templates, corresponding to, “why is [X] used,” the place [X] is a placeholder for one or extra question phrases.
The search queries which can be chosen might have to satisfy a threshold of getting been requested a sure variety of instances earlier than it’s chosen as a possible associated query.
Questions may be taken from different sources than simply queries, and will embrace questions from query and reply web sites.
The patent tells us concerning the number of search queries and sources of solutions to these queries. When questions are chosen, they could be ranked extra extremely based mostly upon issues corresponding to whether or not they had been supplied as questions that got here with solutions. Questions with solutions could also be promoted in rankings, and questions with out solutions could also be demoted as decisions of associated questions.
A high quality rating for an solutions could also be based mostly upon things like:
(1) A high quality rating generated by the search engine for the useful resource from which the reply is derived (the web page that’s the supply of the reply).
(2) The high quality of the reply could also be based mostly partially on a rating of a search outcome figuring out the useful resource from which the reply is derived in a rating of search outcomes generated by the search engine in response to the query being submitted as a search question.
(three) The high quality of the reply could also be based mostly partially on the size of the reply, i.e., the variety of tokens, phrases, or characters within the reply.
(four) If a number of solutions can be found for a given query, the standard of every reply might be based mostly partially on the quantity or proportion of phrases within the reply which can be repeated in different solutions for the query.