System and method for hybrid question answering over knowledge graph

A device, comprising:

• a processing system including a processor;
• and
• a memory that stores executable instructions that, when executed by the processing system, facilitate performance of operations, the operations comprising:
• identifying, without human intervention, a main entity of a natural language question;
• locating, within a knowledge graph, a focal node corresponding to the main entity;
• identifying, within the knowledge graph, a candidate answer set comprising a group of other entities within a predetermined proximity to the focal node, wherein the predetermined proximity comprises two hops from the focal node;
• extracting, from the knowledge graph, contextual information for the group of other entities of the candidate answer set, wherein the contextual information comprises a respective answer type of each of the other entities of the candidate answer set;
• encoding the natural language question and the contextual information of the group of other entities of the candidate answer set to obtain an encoded vectorial representation of the natural language question and a plurality of encoded vectorial representations of the candidate answer set, wherein the encoding uses pre-trained language model embeddings obtained via a bidirectional encoder representations from transformer (BERT) encoding process;
• and
• selecting one of the other entities of the candidate answer set to obtain a selected one of the candidate answer set as an answer to the natural language question, wherein the selecting is based on scoring of the encoded vectorial representation of the natural language question under an influence of the contextual information, and wherein the scoring involves evaluations of cross-attention between the natural language question and the respective answer type of each of the other entities of the candidate answer set.

A non-transitory, machine-readable medium, comprising executable instructions that, when executed by a processing system including a processor, facilitate performance of operations, the operations comprising:

• identifying an entity of a natural language question;
• locating, within a knowledge graph, a node corresponding to the entity;
• generating a candidate answer set comprising a group of other entities located a predetermined proximity to the node within the knowledge graph, wherein the predetermined proximity comprises three hops from the node;
• determining, from the knowledge graph, contextual information for the group of other entities of the candidate answer set, wherein the contextual information comprises a respective answer type of each of the other entities of the candidate answer set;
• encoding the natural language question and the contextual information of the group of other entities of the candidate answer set to obtain an encoded vectorial representation of the natural language question and an encoded vectorial representation of the candidate answer set, wherein the encoding uses pre-trained language model embeddings obtained via a bidirectional encoder representations from transformer (BERT) encoding process;
• and
• selecting one of the other entities of the candidate answer set to obtain a selected one of the candidate answer set as an answer to the natural language question, wherein the selecting is based on scoring of the encoded vectorial representation of the natural language question under an influence of the contextual information, and wherein the scoring involves evaluations of cross-attention between the natural language question and the respective answer type of each of the other entities of the candidate answer set.