A new patent from Google dated December 28, 2021, focuses on how it interprets and resolves queries based on entity information.
I often link to Google patents in articles because I spend a lot of time learning from them.
Patents are filed to describe new inventions and stimulate innovation from potential competitors. They provide sufficient information to exclude others in the same business from copying the patent applicants’ intellectual property.
Oftentimes, we find interesting information about the assumptions patent creators make about search, researchers, and the web which can make Patents an interesting read, too.
As always, when I share highlights in an article like this, I encourage you to take a look at the patent itself.
I’m trying to explain what the patent might cover, but I don’t want to cover it in so much detail that my post might look like a copy of the patent for indexing software.
You may remember when Google search engineer Paul Haahr gave a presentation at SMX 2016 on “How Google Works”.
One interesting point was that Google is trying to limit when entities appear in queries sent by searchers.
This statement leads to the question of how Google can know which entity might be referred to in the query.
Google filed a patent where they explored this topic, and that’s what this post is about.
How to interpret queries better
Search has evolved to receive these search queries and return results in response to the query.
However, some search engines provide search results without understanding the search query.
For example, in response to [action movie with Tom Cruise]irrelevant search results such as [Last Action Hero] And [Tom and Jerry] It may be returned because part of the search query is included in the content fragments header.
Understanding your search query can help the search engine produce more meaningful results.
How can a search engine interpret queries?
The patent refers to these methods:
- Receive a query in the search field.
- Define search terms based on the query.
- Whether the search term matches the name of the entity.
- Find whether the name of the entity is from the metadata associated with the search field.
- Seeing that many of the entity names correspond to at least part of the number of search terms.
- Choose the entity type and entity degree associated with each of the corresponding entity name numbers.
- Finds a number of entity names by removing some matching entity names based on entity score and contextual information in the search query received.
- Perform a search in the search field with the remainder of the number of entity names.
- Where each named entity is searched for the remainder of the number of entity names matching the associated entity type.
This way of interpreting queries can also include:
- Receive a voice query in the search field.
- Choosing several voice recognition conditions based on the received voice query.
- Decide on each number of voice recognition terms.
- Whether at least part of the voice recognition term corresponds to the name of the entity.
- Where the entity name is derived from metadata associated with the research field and where the entity score is associated with the entity name.
- Determine a feasibility score for each voice recognition term based on the entity score.
- Ranking of the number of voice recognition terms based on the selected usefulness score.
- Select one of the categorized voice recognition term numbers to perform the voice query in the search field.
This query interpretation patent is located at:
Inventors: Youngsung Kim
Agent: Google LLC
US Patent: 11,210,289
Awarded: December 28, 2021
Submission date: May 5, 2017
Mechanisms are provided to interpret queries.
In some implementations, a way to parse queries is provided, which includes:
Receive a search query in the search field
Define search terms based on the search query
Determine, for each search term, whether the search term corresponds to the name of the entity,
Where the entity name is derived from the metadata associated with the search field.
In response specify the names of these entities that correspond to a portion of the search terms
Define the entity type and entity class associated with each of the corresponding entity names
Limit the remainder of the entity names by removing at least one of the matching entity names based on the entity score and contextual information in the search query
perform a search in the search field with the remainder of the entity names,
Each named entity is searched in the remainder of the entity names corresponding to the associated entity type.
Interpretation inquires about the conclusion of the patent
When a search engine identifies that an entity is in an article, it will attempt to determine specifically who the entity is.
A Google patent I wrote in the past indicated that an entity name like “Michael Jackson” might seem to identify only one person most people know. After all, he was a very famous musician and artist.
But there was another famous Michael Jackson who wasn’t quite like it at first; He was known as the Director of Homeland Security.
Google calculates confidence scores to determine which entity might be referred to when it is seen in a query.
This patent tells us how Google can determine which entity is being searched before returning results about that entity.
Keep in mind that when someone searches for “Lincoln” (an example from another Google patent), they could mean a Lincoln city car, former President Abraham Lincoln, or the city of Lincoln, Nebraska (also many other states).
If the search engine can correctly interpret the query, and display relevant answers to the searcher, it can satisfy the searcher’s informational or situational needs.
There is a lot of analysis of how this patent works in the patent description, but I wanted to show why it is needed and necessary.
There are a lot of risks of potential confusion if the search engine does not try to interpret a query correctly.
- Google: Patents are not always used in search
- Google patents and future proof your SEO [Podcast]
- Basic Web Basics: A Complete Guide
Featured image: Prabowo96/Shutterstock