Generating a superset of question/answer action paths based on dynamically generated type sets

Patent metadata

Field	Value
Publication number	US-9971967-B2
Application number	US-201314103981-A
Country	US
Kind code	B2
Filing date	Dec 12, 2013
Priority date	Dec 12, 2013
Publication date	May 15, 2018
Grant date	May 15, 2018

Abstract

An approach is provided for generating supersets of Q/A action paths based on dynamically generated type sets. In the approach, a corpus of knowledge that is used by the QA system is analyzed. The analyzing is performed according to a natural language processing (NLP) of a number of key words that have been found to exceed an expected frequency, such as a frequency found in general language usage. Runtime code is generated by utilizing the key words. The generated runtime code is able to provide answers from the corpus of knowledge, such as by being in a natural language question or in a structured query language expression, with the provided answers being related to one or more predicted questions.

First claim

What is claimed is:

1. A method, in an information handling system comprising a processor and a memory, for generating question-answer paths for a Question/Answer (QA) system, the method comprising:
analyzing a corpus of knowledge used by the QA system, wherein the analyzing comprises identifying a plurality of key words included in the corpus of knowledge, wherein the key words occur in the corpus of knowledge at a higher than expected frequency;
generating a runtime code by utilizing the key words, wherein the generated runtime code comprises an initial structured query language expression including a plurality of search fields, and wherein the generated runtime code is adapted to provide answers from the corpus of knowledge, and wherein the provided answers are related to one or more predicted questions;
in response to generating the runtime code, receiving a first user input from a user, the first user input comprising a selected one of the plurality of search fields included in the initial structured query language expression;
in response to receiving the first user input, generating an edited runtime code, wherein the edited runtime code comprises a complete structured query language expression for the selected search field;
generating a natural language question based upon the edited runtime code;
in response to generating the natural language question, receiving a second user input from the user, the second user input comprising one or more edits to the natural language question;
updating the natural language question based on the second user input;
storing the updated natural language question and one or more of the provided answers in a format compatible for processing by the QA system; and
training the QA system using the updated natural language question and the one or more of the provided answers, wherein the stored updated natural language question and the one or more of the provided answers comprises a question-answer path utilized by the QA system.

2. The method of claim 1 further comprising:
accessing a structured database from the corpus of knowledge, wherein the structured database is included in the corpus of knowledge and facilitates searching for one or more possible answers to the one or more predicted questions using the runtime code.

3. The method of claim 2 further comprising:
querying the structured database to retrieve responsive information;
utilizing the retrieved responsive information to formulate the one or more possible answers to the one or more predicted questions; and
presenting the formulated possible answers to the user.

4. The method of claim 2 further comprising:
selecting a set of data sources included in the corpus of knowledge, wherein the selected set of data sources are related to a common domain of information;
extracting a plurality of words from the set of data sources;
calculating a domain-usage frequency for each of the plurality of words, wherein the domain-usage frequency is the frequency that each of the plurality of words occurs in the set of data sources;
retrieving a general-usage frequency for each of the plurality of words, wherein the general-usage frequency is the frequency that each of the plurality of words occurs in general language usage;
for each of the plurality of words, comparing the domain-usage frequency of the word with the general-usage frequency of the word; and
identifying a set of the plurality of words from the plurality of words as the plurality of key words, the identifying based on the comparison revealing that the domain-usage frequency corresponding to each word in the set of the plurality of words is statistically significantly higher than the general-usage frequency corresponding to the respective word.

5. The method of claim 4 further comprising:
searching a plurality of schemas corresponding to the structured database for the plurality of key words;
finding a selected one of the key words in a selected one of the schemas, wherein the selected schema corresponds to a selected database table included in the structured database; and
building the initial structured query language expression to search for the selected key word in the selected database table.

6. The method of claim 1 further comprising:
presenting the generated natural language question to the user at a user interface (UI); and
receiving the one or more edits, at the UI, from the user to the generated natural language question.

7. The method of claim 1 wherein the corpus of knowledge includes a first corpus that relates to a first domain of knowledge and a second corpus that relates to a second domain of knowledge, wherein the key words include a first set of key words related to the first corpus and second set of key words related to the second corpus, and wherein the first and second sets of key words differ between the two domains.

8. An information handling system comprising:
one or more processors;
a memory coupled to at least one of the processors;
a display; and
a set of instructions stored in the memory and executed by at least one of the processors to generate question-answer paths for a Question/Answer (QA) system, wherein the set of instructions perform actions of:
analyzing a corpus of knowledge used by the QA system, wherein the analyzing comprises identifying a plurality of key words included in the corpus of knowledge, wherein the key words occur in the corpus of knowledge at a higher than expected frequency;
generating a runtime code by utilizing the key words, wherein the generated runtime code comprises an initial structured query language expression including a plurality of search fields, and wherein the generated runtime code is adapted to provide answers from the corpus of knowledge, and wherein the provided answers are related to one or more predicted questions;
in response to generating the runtime code, receiving a first user input from a user, the first user input comprising a selected one of the plurality of search fields included in the initial structured query language expression;
in response to receiving the first user input, generating an edited runtime code, wherein the edited runtime code comprises a complete structured query language expression for the selected search field;
generating a natural language question based upon the edited runtime code;
in response to generating the natural language question, receiving a second user input from the user, the second user input comprising one or more edits to the natural language question;
updating the natural language question based on the second user input;
storing the updated natural language question and one or more of the provided answers in a format compatible for processing by the QA system; and
training the QA system using the updated natural language question and the one or more of the provided answers, wherein the stored updated natural language question and the one or more of the provided answers comprises a question-answer path utilized by the QA system.

9. The information handling system of claim 8 wherein the actions further comprise:
accessing a structured database from the corpus of knowledge, wherein the structured database is included in the corpus of knowledge and facilitates searching for one or more possible answers to the one or more predicted questions using the runtime code.

10. The information handling system of claim 9 wherein the actions further comprise:
querying the structured database to retrieve responsive information;
utilizing the retrieved responsive information to formulate the one or more possible answers to the one or more predicted questions; and
presenting the formulated possible answers to the us

Assignees

• IBM

Inventors

• Bufe Iii John P
• Winkler Timothy P

Classifications

• G06N5/02 Primary

  Knowledge representation; Symbolic representation · CPC title