Question-answering system for answering relational questions by utilizing two paths where at least one path uses BERT model
A question-answering system that receive a natural-language question includes a database to provide a basis for that answer and a structured-query generator that constructs a structured query from the question and uses it to obtain an answer to the question from the database. The method further comprises generating a first answer to said first question, wherein generating said first answer comprises causing said question to traverse a first path through said question-answering system, wherein the first path comprises a knowledge base constructed from content of the document collection using a machine learning BERT model; and causing said question to traverse a second path through said question-answering system to generate a second answer to said question, said second answer differing from said first answer, wherein the second path comprises a structured query generator and the database, the database being distinct from the knowledge base.
1. A method comprising a computer-implemented question-answering system carrying out the steps of:
receiving a document collection;
extracting features from documents in said document collection;
using said features to build tables in a database;
receiving a question, said question being a natural language question;
constructing a query corresponding to said question, said query being a query in a database query language;
providing said query to the database, said database having been constructed from documents in a document collection;
receiving information from said database in response to having provided said query thereto;
based on said information, generating a first answer to said question, wherein generating said first answer comprises causing said question to traverse a first path through said question-answering system, wherein the first path comprises a knowledge base constructed from content of the document collection using a machine learning BERT model;
causing said question to traverse a second path through said question-answering system to generate a second answer to said question, said second answer differing from said first answer, wherein the second path comprises a structured query generator and the database, the database being distinct from the knowledge base;
generating a score for said first answer, and generating a score for said second answer; and
selecting one of said first and second answers as an output answer based on said scores thereof.
2. A method comprising a question-answering system carrying out the steps of:
receiving a document collection, extracting features from documents in said document collection, and using said features to build tables in a database;
constructing said database, comprises using said question-answering system to direct a prior question towards each of said documents and, based at least in part on answers to said prior question, constructing a table for inclusion in said database, including using a natural language processing procedure to produce an answer to the prior question based on natural language content of said document;
receiving a first question, said question being a natural language question;
constructing a query corresponding to said first question, said query being a query in a database query language;
providing said query to the database, said database having been constructed from documents in a document collection;
receiving information from said database in response to having provided said query thereto;
based on said information, generating a first answer to said first question, wherein generating said first answer comprises causing said question to traverse a first path through said question-answering system, wherein the first path comprises a knowledge base constructed from content of the document collection using a machine learning BERT model; and
causing said question to traverse a second path through said question-answering system to generate a second answer to said question, said second answer differing from said first answer, wherein the second path comprises a structured query generator and the database, the database being distinct from the knowledge base.
3. The method of claim 2 , wherein using the features to build tables in the database comprise building relational tables, and constructing the query corresponding to the first question comprises constructing a relational database query.
4. The method of claim 3 , further comprising processing the relational database query in the database to yield a database response, and forming the information provided from the databased from said database response.
5. The method of claim 2 , wherein the prior question is one of a plurality of prior questions that are directed to each of said documents and known prior to receiving the document collection.
6. A method comprising a question-answering system carrying out the steps of:
receiving a question, said question being a natural language question,
constructing a query corresponding to said question, said query being a query in a database query language,
providing said query to a database, said database having been constructed from documents in a document collection,
receiving information from said database in response to having provided said query thereto,
based on said information, generating a first answer to said question, wherein generating said first answer comprises causing said question to traverse a first path through said question-answering system, wherein the first path comprises a knowledge base constructed from content of the document collection using a machine learning BERT model, and
causing said question to traverse a second path through said question-answering system to generate a second answer to said question, said second answer differing from said first answer, wherein the second path comprises a structured query generator and the database, the database being distinct from the knowledge base.
7. The method of claim 6 , wherein said question-answering system carries out the further steps of causing said question to traverse a first path to generate said first answer and causing said question to traverse a second path to generate a second answer that differs from said first answer.
8. The method of claim 6 , wherein said documents are human-readable documents.
9. The method of claim 6 , wherein said documents are natural-language documents.
10. The method of claim 6 , wherein said question-answering system carries out the further step of constructing said database based on said documents, wherein constructing said database comprises constructing a relational database.
11. The method of claim 6 ,
wherein generating said first answer comprises causing said question to traverse a first path through said question-answering system and
wherein said question-answering system carries out the further steps of:
generating a score for said first question,
causing said question to traverse a second path through said question-answering system to generate a second answer to said question, said second answer differing from said first answer,
generating a score for said second question, and
selecting one of said first and second questions as an output answer based on said scores thereof.
12. The method of claim 6 , wherein said question-answering system carries out the further step of: constructing said database based on said documents, wherein constructing said database comprises constructing a schema for said database based at least in part on said documents.
13. The method of claim 6 , wherein said question-answering system carries out the further step of generating a second answer to said question, said second answer differing from said first answer.
14. The method of claim 6 , wherein said question-answering system carries out the further step of compiling a knowledge base based on said document collection and using said knowledge base to provide a second answer to said question.
15. The method of claim 6 , wherein said question-answering system carries out the further step of receiving a second answer to said question, said second answer being different from said first answer and rejecting one of said first and second answers.
16. The method of claim 6 , wherein said question-answering system carries out the further step of receiving said document collection, extracting features from documents in said document collection, and using said features to build tables in said database.
17. The method of claim 6 , wherein said question-answering system carries out the further step of: constructing said database, wherein constructing said database comprises using said question-answering system to direct a question towards each of said documents and, based at least in part on answers to said questions, constructing a table for inclusion in said database.
18. The method of claim 6 , wherein said question-answering system carries out the further step of constructing said database, wherein constructing said database comprises identifying a table in said documents and incorporating said table into said database.
19. The method of claim 6 , wherein generating said first answer comprises causing said question to traverse a first path through said question-answering system, and wherein said question-answering system carries out the further step of generating a score for said first question, causing said question to traverse a second path through said question-answering system to generate a second answer to said question, said second answer differing from said first answer, generating a score for said second question, and selecting one of said first and second questions as an output answer based on said scores thereof.