Beyond text qa multimedia answer generation by harvesting web information
Introduction to a Question Answering System
1.
2. INTRODUCTION
Question Answer system is a man machine communication device. The
basic idea of QA system is to provide correct responses to the questions
in a human like manner giving short and accurate answers.
With the advent of Internet and the World Wide Web (WWW) massive
amount of textual information are available to general public. However, it
is difficult to find specific information with them.
Question answering system has deduction capability with the help of
which users can retrieve the exact answer to a question instead of a set of
documents.
QA systems typically require a great deal of human effort, in order to
create linguistic rules that can cope with the vast variety of questions that
can be asked, and the many different ways in which they can be
formulated.
3. OUR PROJECT
Our project on question answering system deals with three different
stages:
1. The naïve bayesian classifier stage which classifies answers in terms of a
simple yes/no as per the question asked.
2. Exact answer extraction stage which gives the exact answer of a given
question asked by the user
For eg: Who is the president of America?
Answer: Barack Obama
3. The Passage Retrieval stage which gives the the relevant passage from
the database from where the answer to a given question might be found.
Our QA system facilitates the user to ask a question in any form as user
wishes and provides them with the exact answer or passage containing
the answer.
5. TECNIQUES (contd..)
QUESTION CLASSIFICATION: The task of the Question Classification
module is to assign a semantic category
to a given question. Every question can be
classified into following taxonomy
6. TECNIQUES (contd..)
PASSAGE RETRIEVAL: The goal of the passage retrieval module is to find
relevant documents from a given collection. A
document is deemed as relevant if it can potentially
contain the correct answer for a given question.
ANSWER EXTRACTION: The answer extraction module is the last module
in the QA pipeline and, therefore, its goal is to
extract an answer from the relevant passages
returned by the passage retrieval module, and
present the answer.
8. APPLICATION
(Zhang & Lee, 2003) used a naive Bayesian classifier for task of
question classification, trained on the standard data set for question
classification of Li & Roth.
Question Answering System can be applied to prove other algorithms
like Support Vector Machine (SVM) and K Nearest Neighbor (KNN)
algorithm.
Exact answer to a given question can be extracted using surface text
patterns that give the different forms in which answer can appear.
(Ravichandran & Hovy, 2001) developed a method to learn patterns,
using bootstrapping.
The technique works in two-phases, where the first (Algorithm 1) is
used to acquire patterns from a set of seed examples, and the second
validates the learnt patterns by calculating their precision.
9. FUTURE ENHANCEMENT
The Question Answering System (QA) can be used to apply Support
Vector Machine (SVM) and K Nearest Neighbor (KNN) algorithms.
The system can be further extended to the field of Machine Learning.
Machine Learning is the field of study that is concerned with the question
of how to construct computer programs that automatically improve with
experience (Mitchell, 1997). Within this view, a computer program
is said to learn from an experience with respect to some task, if the
program’s performance at the task improves with the experience.
The field of machine learning is divided into three broad categories :
1. Supervised learning
2. Unsupervised learning
3. Reinforced learning
These can be stepwise implemented by enhancing the current system.
10. LITERATURE SURVEY
[1] AnswerBus Question Answering System Zhiping Zheng, School of
Information University of Michigan.
[2] Steven Abney, Michael Collins, and Amit Singhal. Answer
Extraction. Proceedings of ANLP 2000. Seattle, WA. April 29 - May 3,
2000.
[3] From search engines to question answering system-The problems
of World Knowledge, Relevance,Precision and
Deduction, Lotfi.A.Zadeh, University of California.
[4] Peter Clark, John Thompson, and Bruce Porter. A knowledge-
based approach to question answering. AAAI’99 Fall Symposium on
Question-Answering Systems. Orlando, Florida. 1999.
11. CONCLUSION
Existing search engines, with Google at the top, have many truly
remarkable capabilities. But there is a basic limitation – search engines do
not have deduction capability – a capability which a question-answering
system is expected to have. Nevertheless, search engines are extremely
useful because a skilled human user can get around search engine
limitations. In this perspective, a search engine may be viewed as a semi-
mechanized question-answering system.
Our project on Question Answering system aims at providing exact
answers to user’s question irrespective of the form in which the question is
asked. Apart from this another stage of the project also provides the user
with the entire passage containing data related to user’s question wherein
the correct answer can be extracted.
We have also succesfully applied the Naïve Bayesian approach of
retrieving answers to a particular question.