| CPC G06F 16/24575 (2019.01) [G06F 16/243 (2019.01)] | 20 Claims |
|
1. A system for extending retrieval of relevant information from a knowledge base, the system comprising:
a communication system;
a database comprising the knowledge base; and
a computing device operably connected to the database and the communication system, the computing device configured to:
access prior engagements including prior queries and prior intents resultant from the prior queries;
map the prior queries to the prior intents respectively;
create a first bipartite graph comprising links of the prior queries and the prior intents;
apply a graph embedding method over the first bipartite graph to measure respective distances between the prior intents;
group the prior intents together into a plurality of similar intent groups based at least in part by comparing a distance threshold with the respective distances between the prior intents, wherein:
each of the plurality of similar intent groups is associated with a confidence determined based on the distance threshold and a strength of the prior engagements used to create the first bipartite graph,
the confidence associated with each similar intent group represents a degree of similarity between intents in the similar intent group,
the prior engagements are selected from the group consisting of search results, views, clicks, add-to-cart and purchases, which are ordered from weak engagements to strong engagements,
the confidence associated with each similar intent group becomes higher as the prior engagements are stronger engagements selected from the group,
the confidence associated with each similar intent group becomes higher as the distance threshold becomes smaller,
a number of the plurality of similar intent groups is reduced when the confidence becomes higher,
when the prior engagements are stronger engagements selected from the group, the distance threshold is determined to increase the number of the plurality of similar intent groups, and
a first intent and a second intent are grouped into a first similar intent group of the plurality of similar intent groups along with other intents;
receive, after associating the first intent to the second intent, an input query from a user;
detect the first intent based on the input query via a natural language processor;
retrieve the first intent, the second intent, and the other intents from the first similar intent group;
create, after receiving the input query and before querying the database, one single extended query containing all intents in the first similar intent group, by concatenating the first intent, the second intent, and the other intents in such sequential order;
query the database, with the one single extended query containing all intents in the first similar intent group, to generate extended query results; and
transmit the extended query results to the user in response to the input query.
|