Integrating Context Knowledge With Concepts Ontology For Handling Lexical Semantic Ambiguity Of Natural Language Interface In Question Answering (QA) System
Question Answering(QA) Systems allow the user to ask questions in a natural language and obtain an exact answers. In general, the QA system is composed of four modules, they are namely: question processing, documents processing, passages processing, and answer processing. Question processing modu...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | en_US |
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Question Answering(QA) Systems allow the user to ask questions in a natural
language and obtain an exact answers. In general, the QA system is composed of four
modules, they are namely: question processing, documents processing, passages
processing, and answer processing. Question processing module is considered the
most fundamental component in the natural language interface of the QA system, and
its quality impacts the performance of the overall QA system. This module receives
natural language questions (NLQs). The most difficult problem in developing a QA
system is so hard to find an exact answer to the NLQ. One of the most challenging
problems in returning answers is how to resolve lexical semantic ambiguity in the
NLQs. Lexical semantic ambiguity may occurs when a user's NLQ contains words
that have more than one meaning. As a result, QA system performance can be
negatively affected by these ambiguous words. In this research, we aim to resolve
this problem by introducing CKCO (Context Knowledge & Concepts Ontology)
approach. This approach integrates context knowledge and concepts ontology of the
proposed domain, into a shallow natural language processing (SNLP) technique.
Concepts ontology represents real world facts that describe the proposed domain,
while context knowledge contains a set of words with their senses obtained from
WordNet Domain and a group of words within the proposed domain serve as context
labels, and it is determined based on neighborhood words in the NLQ. SNLP
technique includes shallow syntactic analysis based on chunking method, and shallow
semantic analysis using semantic role labeling method. Experimental results show
that CKCO approach has 78.2% accuracy on a test set of 150 NLQs. We applied
CKCO approach to a QA prototype in a university domain for new students to
examine the impact of our approach in retrieving correct answers. The QA prototype
was evaluated based on two aspects: prototype performance and acceptance of users.
Experimental results show that the CKCO approach together with other components
of our QA system yield a result which is 77.3% for accuracy which is 4.7% above the
accuracy yielded by the same prototype without disambiguation. Acceptance aspect
was evaluated based on its user's perspective, to determine the perceived ease of use
(PEU), perceived satisfaction (PS), and perceived usefulness (PU) of the QA system
that we developed. The results from the respondents reflect that our QA prototype was
acceptable. |
|---|