# US 2019/0171646 A1 · Providing answers to questions using hypothesis pruning

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2019/0171646 A1
- Document ID: US20190171646A1
- Status: Published application
- Kind: A1
- Application number: 16/267,688
- Filed: 2019-02-05
- Published: 2019-06-06
- Earliest claimed: 2010-09-28
- Assignees: International Business Machines Corporation
- Applicants: International Business Machines Corporation
- Inventors: Jennifer Chu-Carroll; David A. Ferrucci; David C. Gondek; Adam P. Lally; James W. Murdock, IV
- Web page: https://patentgrep.com/patent/US20190171646A1

## CPC

- G06F 16/24578: {using ranking}
- G06F 16/24575: {using context}
- G06F 16/248: Presentation of query results
- G06F 16/3329: Natural language query formulation
- G06F 16/9535: Search customisation based on user profiles and personalisation
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

A method, system and computer program product for generating answers to questions. In one embodiment, the method comprises receiving a query, conducting a search through one or more data sources to identify candidate answers to the query, and providing each of the candidate answers with a preliminary score. The method further comprises filtering out any of the candidate answers with a preliminary score that does not satisfy a defined condition. The candidate answers having preliminary scores that satisfy this condition form a subset of the candidate answers. Each of the candidate answers in this subset is processed to produce further scores. A ranking function is applied to these further scores to determine a ranking for each of the candidate answers in the subset; and after this ranking function is applied, one or more of the candidate answers are selected as one or more final answers to the query.

## Claims

### Claim 1

A system for generating answers to questions, comprising:
a computer device comprising at least one distinct software module, each distinct software module being embodied on a tangible computer-readable medium; a memory; and at least one processor coupled to the memory and operative for:
receiving an input query;
obtaining a plurality of candidate answers from text data, wherein each of the candidate answers is found in associated data content;
passing the candidate answers in sequence through a pre-filter module and a filtering module, including
the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers, and
the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module to direct a first group of the candidate answers to a post-filter module and to direct a second group of the candidate answers through a supporting passage retrieval process;
the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers;
the supporting passage retrieval process retrieving supporting passages for the second group of the candidate answers;
providing content dependent scores for the second group of the candidate answers based on the retrieved supporting passages for the second group of the candidate answers; and
returning one or more of the candidate answers to the input query based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

### Claim 2 (depends on 1)

The system according to claim 1, wherein the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers includes the pre-filter module using the scoring function to provide a pre-filter score for each of the candidate answers independent of the data content in which said each candidate answer is found.

### Claim 3 (depends on 1)

The system according to claim 1, wherein the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers includes the post-filter module providing the post-filter score for each candidate answer of the first group of candidate answers independent of the data content in which said each candidate answer of the first group of candidate answers is found.

### Claim 4 (depends on 1)

The system according to claim 1, wherein the returning one or more of the candidate answers to the input query includes using a merging/ranking module to rank the first and second groups of the candidate answers based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

### Claim 5 (depends on 4)

The system according to claim 4, wherein the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module includes the hypothesis filtering module filter the candidate answers to direct a third group of the candidate answers directly to the merging/ranking module.

### Claim 6 (depends on 5)

The system according to claim 5, wherein the first group of the candidate answers have a pre-filter score above a specified high score.

### Claim 7 (depends on 6)

The system according to claim 6, wherein the third group of the candidate answers have a pre-filter score below a specified low score.

### Claim 8

An article of manufacture, comprising:
at least one tangible computer readable hardware medium having computer readable program code logic to generate answers to questions, the computer readable program code logic, when executing, performing the following:
receiving an input query at a computer processor system;
obtaining, at the computer processor system, a plurality of candidate answers from text data, wherein each of the candidate answers is found in associated data content;
passing, at the computer processor system, the candidate answers in sequence through a pre-filter module and a filtering module, including
the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers, and
the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module to direct a first group of the candidate answers to a post-filter module and to direct a second group of the candidate answers through a supporting passage retrieval process;
the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers;
the supporting passage retrieval process retrieving supporting passages for the second group of the candidate answers;
providing, at the computer processor system, content dependent scores for the second group of the candidate answers based on the retrieved supporting passages for the second group of the candidate answers; and
returning, at the computer processor system, one or more of the candidate answers to the input query based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

### Claim 9 (depends on 8)

The article of manufacture according to claim 8, wherein the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers includes the pre-filter module using the scoring function to provide a pre-filter score for each of the candidate answers independent of the data content in which said each candidate answer is found.

### Claim 10 (depends on 8)

The article of manufacture according to claim 8, wherein the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers includes the post-filter module providing the post-filter score for each candidate answer of the first group of candidate answers independent of the data content in which said each candidate answer of the first group of candidate answers is found.

### Claim 11 (depends on 8)

The article of manufacture according to claim 8, wherein the returning one or more of the candidate answers to the input query includes using a merging/ranking module to rank the first and second groups of the candidate answers based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

### Claim 12 (depends on 11)

The article of manufacture according to claim 11, wherein the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module includes the hypothesis filtering module filter the candidate answers to direct a third group of the candidate answers directly to the merging/ranking module.

### Claim 13 (depends on 12)

The article of manufacture according to claim 12, wherein the first group of the candidate answers have a pre-filter score above a specified high score.

### Claim 14 (depends on 13)

The article of manufacture according to claim 13, wherein the third group of the candidate answers have a pre-filter score below a specified low score.

### Claim 15

A computer-implemented method of generating answers to questions, comprising:
receiving an input query at a computer processor system;
obtaining, at the computer processor system, a plurality of candidate answers from text data, wherein each of the candidate answers is found in associated data content;
passing, at the computer processor system, the candidate answers in sequence through a pre-filter module and a filtering module, including
the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers, and
the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module to direct a first group of the candidate answers to a post-filter module and to direct a second group of the candidate answers through a supporting passage retrieval process;
the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers;
the supporting passage retrieval process retrieving supporting passages for the second group of the candidate answers;
providing, at the computer processor system, content dependent scores for the second group of the candidate answers based on the retrieved supporting passages for the second group of the candidate answers; and
returning, at the computer processor system, one or more of the candidate answers to the input query based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

### Claim 16 (depends on 15)

The method according to claim 15, wherein the pre-filter module using a scoring function to provide a pre-filter score for each of the candidate answers includes the pre-filter module using the scoring function to provide a pre-filter score for each of the candidate answers independent of the data content in which said each candidate answer is found.

### Claim 17 (depends on 15)

The method according to claim 15, wherein the post-filter module providing a post-filter score for each candidate answer of the first group of candidate answers includes the post-filter module providing the post-filter score for each candidate answer of the first group of candidate answers independent of the data content in which said each candidate answer of the first group of candidate answers is found.

### Claim 18 (depends on 15)

The method according to claim 15, wherein the returning one or more of the candidate answers to the input query includes using a merging/ranking module to rank the first and second groups of the candidate answers based on the post-filter scores provided for the first group of the candidate answers and the content dependent scores provided for the second group of the candidate answers.

### Claim 19 (depends on 18)

The method according to claim 18, wherein the hypothesis filtering module filtering the candidate answers based on the scores provided in the pre-filter module includes the hypothesis filtering module filter the candidate answers to direct a third group of the candidate answers directly to the merging/ranking module.

### Claim 20 (depends on 19)

The method according to claim 19, wherein:
the first group of the candidate answers have a pre-filter score above a specified high score; and
the third group of the candidate answers have a pre-filter score below a specified low score.
