Using cohorts in a question answering system

Patent metadata

Field	Value
Publication number	US-9842296-B2
Application number	US-201414492661-A
Country	US
Kind code	B2
Filing date	Sep 22, 2014
Priority date	Sep 17, 2014
Publication date	Dec 12, 2017
Grant date	Dec 12, 2017

Abstract

A cohort analysis mechanism analyzes cohorts with similar attributes to extrapolate additional knowledge and answer a question in a question answering system. The cohort analysis mechanism identifies cohorts for an entity of the question and extracts relevant data concerning the cohorts. The cohort analysis mechanism aggregates the relevant information for evidence scoring and answer scoring to answer a question posed to the question answering system. The aggregating of the data includes combining and ranking answers from the cohorts, gathering evidence and then answering the question with the gathered evidence.

First claim

The invention claimed is:

1. A method for analyzing cohorts to answer a question in a question answering system, the method comprising the steps of:
identifying the cohorts for a person in the question, wherein the cohorts are persons meeting a threshold of similar attributes to the person of the question and available in a corpus of data to answer the question;
extracting answers and evidence from data of the cohorts;
combining and ranking the extracted answers and evidence from the cohorts;
using the combined and ranked answers and evidence to gather statistically significant evidence; and
answering the question with the gathered statistically significant evidence.

2. The method of claim 1 wherein the cohort analysis mechanism is part of a question answering application that answers a natural language question.

3. The method of claim 1 wherein the person in the question is a patient, the question is a medical question for the patient and the cohort analysis mechanism processes the medical history of the patient and finds the cohorts from medical histories of other patients to predict an answer from the cohorts.

4. The method of claim 1 wherein the cohort analysis mechanism determines an answer to the question using evidence gathered from a set of cohorts where case attributes in the set of cohorts does not match exactly attributes of the person in the question.

5. The method of claim 4 wherein the question is a medical question for a patient and the cohort analysis mechanism processes the medical history of the patient and finds the cohorts from medical histories of other patients to predict an answer from the cohorts.

6. The method of claim 5 wherein the cohort analysis mechanism combines the answer from the cohorts with answers using conventional question answering in an answer scoring block of a conventional question answering system to answer the question.

7. The method of claim 1 wherein the attributes include age, race and symptoms.

8. The method of claim 1 wherein the threshold is a reference percentage allowing the cohorts to be divided into relative strengths for different percentage thresholds and wherein the relative strengths of the cohorts are used to score the answers and evidence.

9. A method for analyzing cohorts to answer a question in a question answering system, the method comprising the steps of:
providing a cohort analysis mechanism that executes on at least one processor that analyzes cohorts with similar attributes, wherein the cohorts are persons meeting a threshold of similar attributes to a person in the question and available in a corpus of data, wherein the cohort analysis mechanism is part of a question answering application that answers a natural language question;
analyzing the cohorts to answer the question comprising:
identifying the cohorts for the person in the question;
extracting answers and evidence from data of the cohorts;
combining and ranking the extracted answers and evidence from the cohorts;
using the combined and ranted answers and evidence to gather statistically significant evidence;
answering the question with the gathered statistically significant evidence;

wherein the cohort analysis mechanism determines an answer to the question using evidence gathered from a set of cohorts where case attributes in the set of cohorts does not exactly match attributes of the question;
wherein the question is a medical question for a patient and the cohort analysis mechanism processes the medical history of the patient to find attributes of the patient and searches the medical histories of other patients to identify cohorts that are patients meeting the threshold of similar attributes where the attributes include age and symptoms; and
wherein the cohort analysis mechanism combines the answer from the cohorts with answers using conventional question answering in an answer scoring block of a conventional question answering system to answer the question.

10. A method for analyzing cohorts to answer a question in a question answering system, the method comprising the steps of:
providing a cohort analysis mechanism that executes on at least one processor that analyzes cohorts with similar attributes, wherein the cohorts are animals meeting a threshold of similar attributes to an animal in the question and available in a corpus of data, wherein the cohort analysis mechanism is part of a question answering application that answers a natural language question;
analyzing the cohorts to answer the question comprising:
identifying the cohorts for the animal in the question;
extracting answers and evidence from data of the cohorts;
combining and ranking the extracted answers and evidence from the cohorts;
using the combined and ranted data from the cohorts to gather statistically significant evidence;
answering the question with the gathered statistically significant evidence;

wherein the cohort analysis mechanism determines an answer to the question using evidence gathered from a set of cohorts where case attributes in the set of cohorts does not exactly match attributes of the question; and
wherein the cohort analysis mechanism combines the answer from the cohorts with answers using conventional question answering in an answer scoring block of a conventional question answering system to answer the question.

11. The method of claim 10 wherein the threshold is a reference percentage allowing the cohorts to be divided into relative strengths for different percentage thresholds and wherein the relative strengths of the cohorts are used to score the answers and evidence.

Assignees

• IBM

Inventors

• Clark Adam T
• Huebert Jeffrey K
• Payton Aspen L
• Petri John E
• Stevens Richard J
• Whaley Kathryn L

Classifications

• G06N5/022 Primary

  Knowledge engineering; Knowledge acquisition · CPC title

• G06F17/3053

  Physics · mapped topic

• G06N5/041 Primary

  Abduction · CPC title

• G06N5/046

  Forward inferencing; Production systems · CPC title

• G06N5/043

  Distributed expert systems; Blackboards · CPC title