Finally, I would like to discuss some Google patents that particularly caught my eye during my research and that support the approaches described above. All of the patents listed are still active, have a term until at least 2025 and are signed by Google.
Ranking search results based on entity metrics
This Google patent was transferred to Google in December 2017 and was updated in 2019. I think it is one of the most exciting patents because it describes how search results are arranged in order based on data from the Knowledge Graph. It describes the bridge between the world of the Knowledge Graph and the classic search index.
In the first step, the two most relevant results from the tunisia phone number data Knowledge Graph are determined based on a search query. These results can only be entities. A first series of metrics are determined for these entities. These metrics can be determined completely independently of the search query. How many metrics are selected from the first entity and how many from the second entity depends on the degree of relevance of the two entities in relation to the search query. The relevance score from the Knowledge Graph could play a role here.
In the next step, the respective entity types are determined for the two entities. The metrics are weighted based on the entity types. For example, the attribute “occupation” may be more important for the entity type “person” than the attribute “nationality”.
A combination of different metrics can be used such as
The relatedness metric can be determined based on the co-occurrence of an entity contained in a search query with the entity type of the entity on web pages. An entity type can be a defining characteristic and/or a categorization of an entity. For example, if the search query contains the entity "Empire State Building" for which the entity type "Skyscraper" has been defined, the relatedness metric can be determined by the co-occurrence of "Empire State Building" and "Skyscraper" in the content of a website. Co-occurrences, i.e. the parallel use of certain terms or entities or entity types, play a special role in this metric.
The entity type metric can be expressed as the ratio of a global popularity value to the rank of an entity type. This then expresses the importance of the entity in relation to, for example, a professional group, industry, etc.
The contribution metric is a type of influence metric based on the mention in top lists such as “The best 20 SEOs in Germany” or ratings and reviews.
The price metric is based on awards won. For example, a film may have won a variety of awards such as Oscars and Golden Globes, each of which has a certain value.
What is also interesting about the patent is the reference to the term domains, which describes a higher-level classification of different entity types.