Automated generation of dialogue flow from documents

Patent metadata

Field	Value
Publication number	US-12505301-B2
Application number	US-202117452624-A
Country	US
Kind code	B2
Filing date	Oct 28, 2021
Priority date	Oct 28, 2021
Publication date	Dec 23, 2025
Grant date	Dec 23, 2025

Abstract

A computerized method, system and computer program product for building a dialogue flow. One embodiment of the method may comprise receiving an input document, the input document comprising content, and generating, by a question-answer pipeline, a plurality of question-answer pairs from the content of the input document. For each question-answer pair, the method may further comprise feeding the question of the question-answer pair into an intent of a dialogue flow structure, and feeding the answer of the question-answer pair as one response of the intent. The method may further comprise tagging each of the plurality of question-answer pairs with a corresponding document section index, reading, by a conversational agent, the input document to a user, pausing the reading when the conversational agent reaches one of the document section indices in the input document, and in response, reading the question corresponding to the document section indicia to the user.

First claim

What is claimed is:

1 . A computerized method for building a dialogue flow, comprising:
receiving a plurality of training inputs; storing the plurality of training inputs into a data store; using the plurality of training inputs to train a question-answer pipeline; receiving an input document, the input document comprising content; generating, by the question-answer pipeline, a plurality of question-answer pairs from the content of the input document, wherein the plurality of question-answer pairs includes a plurality of questions generated by the question-answer pipeline and a plurality of corresponding answers generated by the question-answer pipeline; for at least one question-answer pair: feeding a respective question of the at least one question-answer pair into an intent of a dialogue flow structure; feeding a respective answer of the at least one question-answer pair as one response of the intent; tagging the at least one question-answer pair with a corresponding document section index; reading, by a conversational agent, the input document to a user; and pausing the reading when the conversational agent reaches the corresponding document section index in the input document, and in response, reading the respective question tagged to the corresponding document section index to the user.

2 . The computerized method of claim 1 , further comprising:
by the conversational agent, receiving a user answer to the respective question read to the user; and responding to the received user answer according to the dialogue flow structure.

3 . The computerized method of claim 1 , wherein the generating the plurality of question-answer pairs from the content of the input document comprises:
generating, by an answer generation subsystem of the question-answer pipeline from the input document, a plurality of answer candidates based on the content of the input document; generating, by a question generation subsystem of the question-answer pipeline from the input document, a question for each of the plurality of answer candidates to form a plurality of question-answer candidate pairs; ranking, by a ranking subsystem of the question-answer pipeline, the plurality of question-answer candidate pairs; and selecting a predetermined number of highest ranked question-answer candidate pairs as the plurality of question-answer pairs.

4 . The computerized method of claim 1 , further comprising:
for at least one question-answer pair, generating, by a machine learning model, one or more additional answers with similar semantic meanings; and associating the additional answers with the at least one question-answer pair to form a <question, correct answers> tuple.

5 . The computerized method of claim 4 , further comprising:
generating, by the machine learning model, one or more contradictory answers with opposing semantic meanings; and further associating the contradictory answers with the at least one question-answer pair to form <question, correct answers, incorrect answers> tuple.

6 . The computerized method of claim 5 , further comprising feeding the incorrect answers as another response option of the intent.

7 . The computerized method of claim 4 , wherein the similar semantic meanings comprise one or more paraphrases of the respective answer.

8 . A computer program product for building a dialogue flow, comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to:
collect user input for a plurality of input document texts via a graphical web user interface; use the user input to train a first deep learning model; feed a new input document text to the first deep learning model; extract, by the first deep learning model, a plurality of question-answer pairs from the input text; tag the plurality of question-answer pairs with a corresponding document section index; generate, by a second deep-learning model for at least some of the plurality of question-answer pairs, at least one additional answer with a similar semantic meaning to a respective answer and at least one contradictory additional answer with an opposite semantic meanings to the respective answer; form a plurality of <question, correct answers, incorrect answers> tuples from the at least some of the plurality of question-answer pairs, the additional answers, and the contradictory answers; and for at least one <question, correct answers, incorrect answers> tuple: feed the question into a new intent of a dialogue flow structure; feed the correct answers as one response of the intent; and feed the incorrect answers as another response option of the intent.

9 . The computer program product of claim 8 , further comprising program instructions to repeat the extracting, tagging, generating, forming, and feeding operations for a plurality of new input documents.

10 . The computer program product of claim 8 , further comprising program instructions to:
read, by a conversational agent, the new input document to a user; and pause the reading when the conversational agent reaches the corresponding document section index in the input document, and in response, read a respective question tagged to the corresponding document section index to the user.

11 . A system for building a dialogue flow, comprising:
a processing unit; a memory coupled to the processing unit, wherein the memory contains program instructions executable by the processing unit to cause the processing unit to: receive a plurality of training inputs; store the plurality of training inputs into a data store; use the plurality of training inputs to train a question-answer pipeline; receive an input document, the input document comprising content; generate, by the question-answer pipeline, a plurality of question-answer pairs from the content of the input document, wherein the plurality of question-answer pairs includes a plurality of questions generated by the question-answer pipeline and a plurality of corresponding answers generated by the question-answer pipeline; for at least one question-answer pair: feed a respective question of the at least one question-answer pair into an intent of a dialogue flow structure; feed a respective answer of the at least one question-answer pair as one response of the intent; tag the at least one question-answer pair with a corresponding document section index; read, by a conversational agent, the input document to a user; and pause the reading when the conversational agent reaches the corresponding document section index in the input document, and in response, read the respective question tagged to the corresponding document section index to the user.

12 . The system of claim 11 , further comprising program instructions to:
by the conversational agent, receive a user answer to the respective question read to the user; and respond to the received user answer according to the dialogue flow structure.

13 . The system of claim 11 , wherein the program instructions to generate the plurality of question-answer pairs from the content of the input document further comprises program instructions to:
generate, by an answer generation subsystem of the question-answer pipeline from the input document, a plurality of answer candidates based on the content of the input document; generate, by a question generation subsystem of the question-answer pipeline from the input document, a question for each of the plurality of answer candidates to form a plurality of question-answer candidate pairs; rank, by a ranking subsystem of the question-an

Assignees

• IBM

Inventors

• Wang Dakuo
• Xu Anbang
• Yu Mo
• Gan Chuang
• Liu Xiaotong
• Liu Haibin

Classifications

• G06F18/2178

  based on feedback of a supervisor · CPC title

• G06F18/214

  Generating training patterns; Bootstrap methods, e.g. bagging or boosting · CPC title

• G06N5/04

  Inference or reasoning models · CPC title

• G06F40/295

  Named entity recognition · CPC title

• G06N3/006

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