Browse Prior Art Database

System for Visualizing and Influencing the Determining Factors for Answers from a Cognitive Question and Answer System

IP.com Disclosure Number: IPCOM000247267D
Publication Date: 2016-Aug-18
Document File: 6 page(s) / 623K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a system for visualizing and influencing the determining factors for answers from a cognitive question and answer system. The method is to map the feature list/set values provided by the cognitive question and answer system (retrieve feature list/scores) to the answer/document itself, such that the user can understand the factors that produced the output and then modify the queries to improve results.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 45% of the total text.

Page 01 of 6

System for Visualizing and Influencing the Determining Factors for Answers from a Cognitive Question and Answer System

When evaluating the results from a cognitive question and answer system, many factors influence the system's selection of an answer. The answer can be part of a document or the document itself. Investigating the determining factors behind the cognitive question and answer system's choice of answer can be a difficult and tedious task. Yet, the importance of understanding the determining factors is critical when trying to address incorrect answers.

For example:
A customer asks the cognitive question and answer system the following:
We are planning to connect to Hadoop (HDFS) server and write data into files. What could be the best way to connect to HDFS?

The cognitive question and answer system responds by suggesting a document with the following information as the answer:
Job stage fails with following error when attempting connection to remote hadoop file system: Message: Unable to connect to hdfs host myhost.domain.com on port 50111: Unknown error 255.

This example shows that the customer was asking for the "best way to connect", but the cognitive question and answer system provided an answer that addresses a "connection error". The user needs to understand the determining factors that caused cognitive question and answer system to choose this document as the answer, versus other documents that describe connection methods for Hadoop (HDFS) servers. Further, the user needs to know how to refine the contents of this document or a separate document to improve the chances for the cognitive question and answer system to return the correct answer.

The novel solution is a system for visualizing and influencing the determining factors for answers from a cognitive question and answer system. The method is to map the feature list/set values provided by the cognitive question and answer system (retrieve feature list/scores) to the answer/document itself. The mapping then either by highlights the selected text area in the document or enables the user to hover over a feature list value, which then reveals a highlight/overlay of a heat map to identify the area. The system can dynamically suggest relevant content or alternate mappings to improve the impact of a feature on the overall answer score. The system enables the user edit the contents of a feature and then save the changes directly back to the corpus. The system can then simulate how the changes impact the cognitive question and answer system results.

This novel solution optimizes the analysis of document information to the data scientist/analyst directly from a user interface (UI). It easily identifies how the features used by a cognitive question and answer system impact the score for an answer/document. This approach provides a simple interface to simulate how changes

1


Page 02 of 6

can impact the cognitive question and answer system results. It also simplifies sharing...