Generating a chatbot from an FAQ

Patent metadata

Field	Value
Publication number	US-11226997-B2
Application number	US-201715832192-A
Country	US
Kind code	B2
Filing date	Dec 5, 2017
Priority date	Dec 5, 2017
Publication date	Jan 18, 2022
Grant date	Jan 18, 2022

Abstract

A chatbot answer database can be generated from a question and answer document. A question/answer processor can receive a document that includes questions and answers. A set of entities can be determined from the answers in the document. A set of triplets can be constructed from the answers, and stored in a chatbot database. For each answer, the question/answer processor can determine text for slots of the triplet based on the entities extracted from the answer. The triplet can be stored as a tree structure in a database of tree structures associated with the document. A chatbot can receive questions, and provide answers based on the database of tree structures associated with the document.

First claim

What is claimed is:

1. A method for generating a chatbot from a question and answer document, the method comprising:
receiving, by one or more processors, at least one document comprising a plurality of questions and a plurality of answers to the plurality of questions;
determining a plurality of entities in the plurality of answers to the plurality of questions;
for an answer of the plurality of answers:
determining text slots of a triplet based, at least in part, on one or more entities extracted from the answer,
determining that the triplet comprises a focus entity of the plurality of entities,
determining that a first slot of the text slots does not comprise a focus entity, and
in response to determining that the first slot does not comprise a focus entity, rewriting the triplet so that the first slot of the text slots of the triplet is the focus entity;
storing the rewritten triplet as a tree structure in a database of tree structures associated with the at least one document; and

providing, by the chatbot, an answer response based on the rewritten triplet.

2. The method of claim 1 , further comprising:
determining that the triplet does not include a focus entity;
and
in response to determining that the triplet does not contain a focus entity, discarding the triplet.

3. The method of claim 1 , further comprising:
disambiguating an entity to produce text for a slot of the triplet.

4. The method of claim 1 , further comprising:
flattening the answer to create a plurality of concepts, wherein a first concept is used for the triplet and a second concept is used for at least one additional triplet.

5. The method of claim 1 , wherein storing the triplet as the tree structure in the database of tree structures includes storing the focus entity as a top level node of the tree structure.

6. The method of claim 1 , wherein providing the answer comprises:
receiving a question;
creating a question triplet from the question;
determining if an answer triplet in the database of tree structures is a match to the question triplet; and
providing the answer response to the question in response to determining that the answer triplet matches the question triplet.

7. The method of claim 6 , further comprising:
determining that two slots of the question triplet match two slots of the answer triplet; and
providing answer text associated with the answer triplet in response determining that the two slots of the question triplet match the two slots of the answer triplet.

8. The method of claim 6 , further comprising:
determining that one slot of the question triplet matches one slot a plurality of answer triplets;
presenting the plurality of answer triplets to a user;
receiving an indication of a user selection of the plurality of answer triplets; and
providing answer text associated with the selected answer triplet.

9. A computer readable non-transitory hardware storage medium including computer executable instructions that, when executed by one or more processors, cause the one or more processors to perform operations for generating a chatbot from a question and answer document, the operations comprising:
instructions to receive at least one document comprising a plurality of questions and a plurality of answers to the plurality of questions;
instructions to determine a plurality of entities in the plurality of answers to the plurality of questions;
for an answer of the plurality of answers:
instructions to determine text slots of a triplet based, at least in part, on one or more entities extracted from the answer,
instructions to determine that the triplet comprises a focus entity of the plurality of entities,
instructions to determine that a first slot of the text slots does not comprise a focus entity, and
instructions to, in response to determining that the first slot does not comprise a focus entity, rewrite the triplet so that the first slot of the text slots of the triplet is the focus entity;
instructions to store the rewritten triplet as a tree structure in a database of tree structures associated with the at least one document; and

instructions to provide an answer response by the chatbot based on the rewritten triplet.

10. The machine readable storage medium of claim 9 , wherein the operations further comprise:
instructions to determine that the triplet does not include a focus entity; and
instructions to discard the triplet in response to determining that the triplet does not contain a focus entity.

11. The machine readable storage medium of claim 9 , wherein the operations further comprise:
store answer text associated with the answer used to create the triplet in the database; and
associate the answer text with the triplet in the database.

12. The machine readable storage medium of claim 9 , wherein the operations to store the triplet as the tree structure in the database of tree structures includes operations to store the focus entity as a top level node of the tree structure.

13. The machine readable storage medium of claim 9 , wherein the operations that provide the answer response comprise:
instructions to receive a question;
instructions to create a question triplet from the question;
instructions to determine if an answer triplet in the database of tree structures is a match to the question triplet; and
instructions to provide the answer response to the question in response to determining that the answer triplet matches the question triplet.

14. A system for generating a chatbot from a question and answer document, the system comprising:
one or more processors; and
a machine readable storage medium including a computer program product having computer executable instructions that, when executed by the one or more processors, perform operations comprising:
instructions to receive at least one document comprising a plurality of questions and a plurality of answers to the plurality of questions;
instructions to determine a plurality of entities in the plurality of answers to the plurality of questions;
for an answer of the plurality of answers:
instructions to determine text slots of a triplet based, at least in part, on one or more entities extracted from the answer,
instructions to determine that the triplet comprises a focus entity of the plurality of entities,
instructions to determine that a first slot of the text slots does not comprise a focus entity, and
instructions to, in response to determining that the first slot does not comprise a focus entity, rewrite the triplet so that the first slot of the text slots of the triplet is the focus entity;
instructions to store the rewritten triplet as a tree structure in a database of tree structures associated with the at least one document; and

instructions to providing an answer response by the chatbot based on the rewritten triplet.

15. The system of claim 14 , wherein the operations further comprise:
instructions to determine that the triplet does not include a focus entity of the set of one or more focus entities; and
instructions to discard the triplet in response to determining that the triplet does not contain a focus entity.

16. The system of claim 14 , wherein the operations to store the triplet as the tree structure in the database of tree structures include operations to store the focus entity as a top level node of the tree structure.

17. The system of claim 14 , wherein the operations that provide the answer response comprise:
instructions to receive a question;
instructions to create a question triplet from the question;

Assignees

• IBM

Inventors

• Pinel Florian
• Byron Donna K
• Dimascio Carmine
• Johnson Benjamin L

Classifications

• H04L51/02 Primary

  using automatic reactions or user delegation, e.g. automatic replies or chatbot-generated messages · CPC title

• G06F16/3329 Primary

  Natural language query formulation · CPC title

• G06N5/02

  Knowledge representation; Symbolic representation · CPC title

• G06F16/2246

  Trees, e.g. B+trees · CPC title

• G06N3/006

  based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO] · CPC title