Identifying interesting commonalities between entities

摘要

Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for generating descriptions of relationships between entities. In one aspect, a method includes identifying one or more related entities for a particular entity based at least in part on commonalities between the particular entity and the one or more related entities, sorting the commonalities according to a measure of uniqueness of each of the commonalities, and identifying a subset of the commonalities having a measure of uniqueness above a lower measure of uniqueness threshold. The identified subset of commonalities can include one or more commonalities. One or more commonalities can be selected from the subset of commonalities as indicative of a relationship to the particular entity, and a description of the relationship can be identified based on the selected one or more commonalities.

主权项

1. A method performed by data processing apparatus, the method comprising:
identifying one or more related entities for a particular entity based at least in part on (A) a plurality of commonalities between the particular entity and the one or more related entities and (B) a plurality of categories of potential commonalities, with each category of potential commonalities including a respective plurality of commonalities between the particular entity and one or more related entities, comprising:
sorting the plurality of commonalities according to a measure of uniqueness of each of the plurality of commonalities;selecting a particular category of potential commonalities, wherein the plurality of commonalities includes the respective plurality of commonalities for the particular category of potential commonalities, including:
ranking the plurality of categories of potential commonalities according to a contribution of the respective plurality of commonalities for each category of potential commonalities to a relatedness of the one or more related entities to the particular entity; andselecting the particular category of potential commonalities based on the ranking; identifying a subset of the sorted plurality of commonalities having a measure of uniqueness above a lower measure of uniqueness threshold, wherein the identified subset of commonalities includes one or more commonalities; selecting one or more commonalities from the subset of commonalities as indicative of a relationship to the particular entity; and identifying a description of the relationship based on the selected one or more commonalities.