Systems and methods for generating a knowledge graph including interaction information
In a computer-implemented method for generating a knowledge graph of interaction information, a data item including an interaction between at least two entities is received. A knowledge tuple for the data item is generated, the knowledge tuple including entities of the data item and a topic of the interaction of the data item. At least one thread to which the data item corresponds is identified, wherein a thread includes connected data items. An interaction tuple for the data item is generated, the interaction tuple including the entities of the data item, the thread corresponding to the data item, and a type of interaction of the data item. The knowledge graph is populated with the knowledge tuple and the interaction tuple, the knowledge graph including nodes associated with the entities, the topics, and the interactions, wherein the nodes associated with interactions include the type of interaction and thread of the data item.
1 . A computer-implemented method for generating a knowledge graph of interaction information, the method comprising:
receiving a data item comprising an interaction between at least two entities;
generating a knowledge tuple for the data item, the knowledge tuple comprising entities of the data item and a topic of the interaction of the data item, wherein the knowledge tuple generated for the data item is generated at a large language model, wherein the generating the knowledge tuple for the data item comprises:
extracting, at the large language model, the entities of the data item and the topic of the interaction of the data item; and
extracting, at the large language model, at least one relationship between the entities;
identifying at least one thread to which the data item corresponds, wherein a thread comprises connected data items;
generating an interaction tuple for the data item, the interaction tuple comprising the entities of the data item, the thread corresponding to the data item, and a type of interaction of the data item; and
populating the knowledge graph the knowledge tuple and the interaction tuple, the knowledge graph comprising nodes associated with the entities, the topics, and the interactions, wherein the nodes associated with interactions comprise the type of interaction and thread of the data item.
2 . The method of claim 1 , wherein the interaction tuple for the data item is generated at a large language model.
3 . The method of claim 2 , wherein the generating the interaction tuple for the data item comprises:
extracting, at the large language model, the entities of the data item;
receiving the at least one thread to which the data item corresponds; and
classifying the data item according to the type of interaction of the data item.
4 . The method of claim 3 , wherein the type of interaction of the data item is also based on the thread to which the data item corresponds.
5 . The method of claim 1 , wherein the interaction comprises at least one of an electronic communication and a calendar event.
6 . The method of claim 1 , wherein the type of interaction of the data item is selected from a library of interaction types.
7 . The method of claim 6 , wherein the library of interaction types is based on a domain of the knowledge graph.
8 . The method of claim 1 , wherein the knowledge graph is populated with a plurality of knowledge tuples and a plurality of interaction tuples corresponding to a domain.
9 . The method of claim 8 , further comprising:
receiving a plurality of data items comprising the data item;
determining at least one domain for the plurality of data items; and
filtering out data items of the plurality of data items that do not correspond to the domain of the knowledge graph, such that the data items not corresponding to the domain of the knowledge graph are disregarded.
10 . The method of claim 8 , wherein the domain comprises one of a scheduling domain, a sales domain, a planning domain, a forecasting domain, a strategic planning domain, a project management domain, a product management domain, and an engineering management domain.
11 . A non-transitory computer readable storage medium having computer readable program code stored thereon for causing a computer system to perform a method for generating a knowledge graph of interaction information, the method comprising:
receiving a data item comprising an interaction between at least two entities;
generating, at a first large language model, a knowledge tuple for the data item, the knowledge tuple comprising entities of the data item and a topic of the interaction of the data item, wherein the generating the knowledge tuple for the data item comprises:
extracting, at the first large language model, the entities of the data item and the topic of the interaction of the data item; and
extracting, at the first large language model, at least one relationship between the entities;
identifying at least one thread to which the data item corresponds, wherein a thread comprises connected data items;
generating, at a second large language model, an interaction tuple for the data item, the interaction tuple comprising the entities of the data item, the thread corresponding to the data item, and a type of interaction of the data item; and
populating the knowledge graph the knowledge tuple and the interaction tuple, the knowledge graph comprising nodes associated with the entities, the topics, and the interactions, wherein the nodes associated with interactions comprise the type of interaction and thread of the data item.
12 . The computer readable storage medium of claim 11 , wherein the type of interaction of the data item is also based on the thread to which the data item corresponds, and wherein the type of interaction of the data item is selected from a library of interaction types.
13 . The computer readable storage medium of claim 12 , wherein the library of interaction types is based on a domain of the knowledge graph.
14 . The computer readable storage medium of claim 11 , the method further comprising:
receiving a plurality of data items comprising the data item;
determining at least one domain for the plurality of data items; and
filtering out data items of the plurality of data items that do not correspond to the domain of the knowledge graph, such that the data items not corresponding to the domain of the knowledge graph are disregarded.
15 . A computer-implemented method for generating a knowledge graph of interaction information, the method comprising:
receiving a plurality of data items comprising a data item, wherein the data item of the plurality of data items comprises an interaction between at least two entities;
determining at least one domain for the data item;
provided the at least one domain for the data item corresponds to a domain of a knowledge graph:
generating, at a first large language model, a knowledge tuple for the data item, the knowledge tuple comprising entities of the data item and a topic of the interaction of the data item;
identifying at least one thread to which the data item corresponds, wherein a thread comprises connected data items;
generating, at a second large language model, an interaction tuple for the data item, the interaction tuple comprising the entities of the data item, the thread corresponding to the data item, and a type of interaction of the data item, wherein the type of interaction of the data item is selected from a library of interaction types, and, wherein the library of interaction types is based on a domain of the knowledge graph; and
populating the knowledge graph the knowledge tuple and the interaction tuple, the knowledge graph comprising nodes associated with the entities, the topics, and the interactions, wherein the nodes associated with interactions comprise the type of interaction and thread of the data item, and wherein the knowledge graph is populated with a plurality of knowledge tuples and a plurality of interaction tuples corresponding to a domain.