Question answering via tailored knowledge induction
Systems and techniques that facilitate question answering via tailored knowledge induction are provided. In various embodiments, a system can access a natural language question. In various aspects, the system can identify an answer for the natural language question, based on a tailored knowledge graph iteratively generated from a set of structured sub-queries associated with the natural language question.
1 . A system, comprising:
a memory configured to store computer-executable components; and
a processor that executes at least one of computer-executable components that:
receives, via a user interface, a natural language question from a user;
converts, via semantic parsing, the natural language question into a logical query comprising a set of structured sub-queries;
generates a tailored knowledge graph based on the set of structured sub-queries associated with the natural language question, wherein the generating comprises iteratively, for respective structured sub-queries of the set of structured sub-queries:
converting, via an inverse of semantic parsing, the respective structured sub-query into a respective unstructured text,
fetching, from a document corpus, one or more respective unstructured documents corresponding to the respective unstructured text,
generating, using triplet extraction, a respective sub-answer for the respective structured sub-query, based on the one or more respective unstructured documents corresponding to the respective unstructured text,
inserting, based on the respective sub-answer, at least one of respective nodes or respective edges into the tailored knowledge graph;
identifies, using the tailored knowledge graph, an answer for the natural language question; and
provides, via the user interface, the answer to the user.
2 . The system of claim 1 , wherein the iteratively generating the tailored knowledge graph further comprises:
updates, via slot-filling, the logical query based on the respective sub-answer.
3 . The system of claim 1 , wherein the at least one of the computer-executable components further:
generates the answer based at least in part on an ontology associated with the natural language question.
4 . The system of claim 1 , wherein the at least one of the computer-executable components further:
ranks the one or more respective unstructured documents based on respective semantic similarity scores between the respective unstructured text and the one or more respective unstructured documents.
5 . The system of claim 1 , wherein the at least one of the computer-executable components further:
prunes one or more nodes or edges from the tailored knowledge graph based on relevance to the natural language question.
6 . The system of claim 1 , wherein the at least one of the computer-executable components further:
receives, via the user interface, confirmation or correction of the identified answer from the user.
7 . The system of claim 1 , wherein the at least one of the computer-executable components further:
outputs an explanation describing how the tailored knowledge graph led to the identified answer.
8 . A computer-implemented method, comprising:
receiving, by a system operatively coupled to a processor, via a user interface, a natural language question from a user;
converting, by the system and via semantic parsing, the natural language question into a logical query comprising a set of structured sub-queries;
generating, by the system, a tailored knowledge graph based on the set of structured sub-queries, wherein the generating comprises iteratively, for respective structured sub-queries of the set of structured sub-queries:
converting, via an inverse of semantic parsing, the respective structured sub-query into a respective unstructured text,
fetching, from a document corpus, one or more respective unstructured documents corresponding to the respective unstructured text,
generates, using triplet extraction, a respective sub-answer for the respective structured sub-query, based on the one or more respective unstructured documents corresponding to the respective unstructured text,
inserts, based on the respective sub-answer, at least one of respective nodes or respective edges into the tailored knowledge graph;
identifying, by the system, using the tailored knowledge graph, an answer for the natural language question; and
providing, by the system, via the user interface, the answer to the user.
9 . The computer-implemented method of claim 8 , wherein the iteratively generating the tailored knowledge graph further comprises:
updating, by the system and via slot-filling, the logical query based on the respective sub-answer.
10 . The computer-implemented method of claim 8 , wherein the system generates the answer based at least in part on an ontology associated with the natural language question.
11 . The computer-implemented method of claim 8 , further comprising:
ranking, by the system, the one or more respective unstructured documents based on respective semantic similarity scores between the respective unstructured text and the one or more respective unstructured documents.
12 . The computer-implemented method of claim 8 , further comprising:
pruning, by the system, one or more nodes or edges from the tailored knowledge graph based on relevance to the natural language question.
13 . The computer-implemented method of claim 8 , further comprising:
receiving, by the system, via the user interface, confirmation or correction of the identified answer from the user.
14 . The computer-implemented method of claim 8 , further comprising:
outputting, by the system, an explanation describing how the tailored knowledge graph led to the identified answer.
15 . A computer program product for facilitating question answering via tailored knowledge induction, the computer program product comprising a non-transitory computer-readable memory having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
receive, via a user interface, a natural language question from a user;
convert, via semantic parsing, the natural language question into a logical query comprising a set of structured sub-queries;
generate a tailored knowledge graph based on the set of structured sub-queries, wherein the generating comprises iteratively, for respective structured sub-queries of the set of structured sub-queries:
converting, via an inverse of semantic parsing, the respective structured sub-query into a respective unstructured text,
fetching, from a document corpus, one or more respective unstructured documents corresponding to the respective unstructured text,
generates, using triplet extraction, a respective sub-answer for the respective structured sub-query, based on the one or more respective unstructured documents corresponding to the respective unstructured text,
inserts, based on the respective sub-answer, at least one of respective nodes or respective edges into the tailored knowledge graph;
identify, using the tailored knowledge graph, an answer for the natural language question; and
provide, via the user interface, the answer to the user.
16 . The computer program product of claim 15 , wherein the iteratively generating the tailored knowledge graph further comprises:
updating, via slot-filling, the logical query based on the respective sub-answer.
17 . The computer program product of claim 15 , wherein the program instructions are further executable to cause the processor to:
generate the answer based at least in part on an ontology associated with the natural language question.
18 . The computer program product of claim 15 , wherein the program instructions are further executable to cause the processor to:
rank the one or more respective unstructured documents based on respective semantic similarity scores between the respective unstructured text and the one or more respective unstructured documents.
19 . The computer program product of claim 15 , wherein the program instructions are further executable to cause the processor to:
prune one or more nodes or edges from the tailored knowledge graph based on relevance to the natural language question.