Professional Documents
Culture Documents
NC - Updated Thesis
NC - Updated Thesis
Master of Technology
(Computer Science & Engineering)
Submitted By:
Nilam Chourasiya
(Enrollment No. – -0837CS20MT03)
Dipshikha Sharma
Supervisor SIMS, Indore
RECOMMENDATION
Dipshikha
Sharma
Supervisor
SIMS, Indore
FORWARDED BY:
i
Sanghvi Institute of Management & Science, Indore (M.P.)
Department of Computer Science and Engineering
CERTIFICATE
Date: Date:
xiv
Sanghvi Institute of Management & Science, Indore (M.P.)
Department of Computer Science and Engineering
Dipshikha
Sharma
Supervisor
SIMS, Indore
xiv
Sanghvi Institute of Management & Science, Indore
Department of Computer Science and Engineering
I further declare that to the best of my knowledge the present work does not contain
any part of the work which has been submitted for the award of any degree either in
this University or in other University/ Deemed University without proper citation.
Nilam Chourasiya
(0837CS20MT03)
UNDERTAKING
The work entitled “Question & Answering System Using Natural Language
Processing” our own work. A check for plagiarism has been carried out on the
thesis/project/ dissertation is found within the acceptable limit.
xiv
xiv
ACKNOWLEDGEMENT
Finally, I am very much thankful to my parents for their personal attention and
care.
Nilam Chourasiya
(0837CS20MT03)
xiv
Abstract
Question and Answering System is one of the major research are in Natural Language
Processing and Information Retrieval. Main challenges of Question and Answer system
gives exact answer of question which give by user. Question and Answering system can
be classified into three category are open domain, closed domain and restricted domain.
Using advanced Natural Language Processing tool we will be developed a framework for
question answering system. In this paper we work on restricted domain question
answering system. Proposed system work on keyword and question matching and return
precise answer of question.
VII
TABLE OF CONTENTS
Page No.
Certificate of Recommendation i
Certificate of Approval ii
Acknowledgement iv
Abstract v
Table of Content vi
List of Figures viii
List of Tables ix
1.1 Background 1
1.2 Motivation 2
1.7 Summary 7
vi
CHAPTER 3: PROBLEM IDENTIFICATION 13-20
3.4 Summary 20
4.7 Summary 32
5.3 Summary 40
6.1 Conclusion 41
References 42
List of Publication 44
Plagiarism Report 48
Vii
LIST OF FIGURES
viii
LIST OF TABLES
CHAPTER 1
INTRODUCTION
1.1 Background
In Computer Science, Question Answering System always helps to user in easier way to find the
answers of given question. Information area is new area for human. With help of web crawlers,
we can get any data readily available. We are only a tick far from getting to a page at remote
corner of the world. We have constantly needed PCs to act savvy [1]. To achieve this undertaking
the field of Artificial Intelligence appeared. One of the key obstructions in making PCs clever is
comprehension of Natural Language. Normal dialect handling which manages comprehension of
dialects is sub division of Artificial Intelligence. Square outline of Question Answering System is
appears in beneath figure1.1. Question Answering is a great NLP application. Assignment that an
inquiry noting framework acknowledges is given an inquiry and gathering of archives, finds the
correct response for the question. It incorporates two integral objectives: first comprehend the
different issues in normal dialect comprehension and portrayal and the second to plan regular
dialect interface to PCs [2].
1.2 Motivation
We are always in a quest of Information However there is distinction in data and information.
Data Retrieval or web seek is develop and we can get applicable data readily available. Question
Answering is a specific type of Information Access, which looks for knowledge. We are occupied
with getting the important pages as well as we are keen on finding particular solution to inquiries.
Each module contains several sub modules and these modules use several Natural Language
Although the set of documents, which are retrieved by the search engine, contain a lot of
information about the search topic but it may or may not contain exactly that information which
the user is looking for [6].The basic idea behind the question answering system is that the users
just have to enter the question and the system will retrieve the most appropriate and precise
answer for that question and return it to the user. Hence in those cases where the user is looking
for a short and precise answer, question answering System plays a great role rather than Search
Engines, which usually provide a large set of links of those web pages which might contain the
answer of that question. Below figure 1.2 shows basic process of extraction answer.
Pre-processing
Question template matching
Answering.
Question Answering Systems can be classified on the basis of the domains over which it has been
constructed.
Open domain question answering systems are domain independent. It relies on general ontology
and world knowledge. Usually these systems have a large collection of data from where the
required answer is to be found out. Since in case of Open Domain question answering information
content is not of particular domain it can answer questions of various fields however here deep
reasoning is not possible [3].
Close domain question answering systems deal with questions in a specific domain [3]. LUNAR
and BASEBALL are the example of close domain QA systems .In this case the data set contains a
very limited amount of focused and structured information. Hence in case of close domain
question answering systems deep reasoning is possible but the problem with these systems was
that due to the very small size of data set they are not more than a 'Toy Systems"[4].
Chapter-1 Introduction: This chapter deals with all the introductory requirements for
understanding the domain area. It gives the details, which are necessary to understand the work,
and measures its outcomes. It provides the motivations, Background, problems understanding and
a view of proposed solution. This is very first and essential part of the report, which contains the
brief details about the Question Answering System.
Chapter-2 Literature Review: It presents a survey on technologies available with the domains.
In this a wide variety of existing mechanism, algorithms and architectures is studied for
identifying the issues removed and remains in Question Answering area.
Chapter-4Proposed Work: After studying the different existing mechanism this identifies the
System Preliminary. It gives a clear understanding the Algorithm with its steps. It will help the
solution to provide better resolution of the current situations of security.This chapter also gives
implementation plan and Testing Strategy of above security problems by suggesting an
architectural solution. Here in this chapter the implementation of our proposed system will be
done. The implementation is working on which platform, what kind of theme and approach is
followed is referred in this section.
Chapter-6Conclusion and Future Work: This chapter gives concluding remarks on the
dissertations and gives a final analysis and comparisons along with some future directions of the
work. The future scope and the short summary will be discussed. It gave an idea how we can
expand the work in future which we have performed in this report.
1.7 Summary
This chapter allocates with all the introductory requirements for understanding the domain
area. It gives the details, which are necessary to understand the work, and measures its outcomes.
It provides the motivations, background, problems understanding and a view of proposed
solution.
CHAPTER 2
LIRERATURE SURVEY
2.1 Literature Review
1. Jinzhong Xu et al.,[1] had worked on Research of Automatic Question Answering Sysytem in
Network Teaching . This method offer students and teachers to exchange and answer the Question
based on Natural Language.
This paper advances a model of automatic question answering system which is based on
Natural Language Processing. This paper introduces theory of semantic representation and
Ontology. This paper has researched the key technologies of Question answering system
based on Natural language Processing. This paper does not gives an accurate answer for
complex questions.
2. Shouning Qu et al.,[2] had worked on Research and design of Intelligent Question Answering System.
This model proposes a model which supports natural language and finds answer from the intelligent
Question answering system.
This system proposes an improved text classification algorithm to classify the database question
accurately. The classification algorithm with improved TFIDF method gives more accurate answer
than the traditional method. After user inputs text, the system analyses the data using natural
processing, positions the target relevant to category , matches the answer and provides efficient
answer.
3. Wael Salloum et al.,[3] had worked on A Question Answering System based on Conceptual Graph
Formalism . This paper proposes a new text based question answering system which converts
knowledge into documents and question into Conceptual Graph formalism. For every question type
there is a different conceptual grah formalism, thus for each question many CG’s are generated. In
this paper projection operator is used to compare questions CG to a sentences CG , and then the exact
answer is extracted from it. More reaserch can be done on extracting answers from simple sentences
and combining sentences of similar meaning.
4. Tilani Gunawardena et al.,[4] had worked on An Automated Answering System with template
Matching for Natural Language Questions. This system uses Closed Domain Question answering
system to find the answers. Therefore answers are stored in a database by domain experts.
The final answer extracted in this paper has the ability to answer the question asked in SMS or
English language. For this answer extraction process a Template Matching technique is applied in this
paper. This paper does not guarantees that the sytems gives an accurate answer since the system deals
with the i) lack of understanding of problem domain ii) handling SMS abbreviations iii) Handling
Spelling mistakes.
5. Erfan Najmi et al.,[5] had worked on Intelligent Semantic Question Answering System. This paper
introduces an approach for Question answering system using Semantic Technologies.
This paper converts the Query into Resource Description Framework (RDF) triples and then
searches the answer in the RDF files. The advantage of having this system is that it has less
Computation time. The system also has some disadvantages that it does not provide answers to
Descriptive or Long answer type Question.
6. PayaJ Biswas et al.,[6] had worked on the A Framework for Restricted Domain Question Answering
System. This paper proposes a framework for restricted domain question answering system.
This model makes use of Information Extraction than that of Information Retrieval Process used
by search engines. This framework can be used to develop a system which provides exact and precise
answer. Also this model provides a proper flow of data for answer extraction. The major issues exited
in the proposed model is that the performance depends on the search engine and the NLP tools used.
7. Varsha Bhoir et al.,[7] has proposed Question Answering System : A Heuristic Approach. The
proposed model works for specific domain of tourism, which is a restricted domain model.
The main aim of this restricted domain Question answering system is to improve the accuracy of the
extracting answer. The system returns precise answers related to the tourism domain. This system
uses an integrated answer retrieval technique which combines of web crawler and Keyword oriented
procedure.
8. Sangdo Han et al.,[8] has worked on Keyword Question Answering System with Report Generation
for Linked Data. This paper introduces a Question answering system that extracts answer from
Linked Data and generate report in Natural Language.
This system uses entity disambiguation and distributed word similarity to match each keywords to
property in Linked Data. To extract Keyword related Information, this model uses SPARQL query.
This system returned the correct answer for 95% of the questions.
9. Sreelakshmi V et al.,[9] has worked on the Open domain Question answering System Using Semantic
Role Labeling. In this paper the system finds an answer using online search and Semantic Role
Labeling.
The goal of the Semantic Role Labeling is to identify all the constituents that fill a
semantic role i.e. to determine the roles like Agent , patient, Location, etc in a sentence. The result of
this system was compared to that of the system using pattern matching to find the answers. This
system can be enhanced in future for Complex Questions.
10. Unmesh Sasikumar et al.,[10] has done “A Survey of Natural language Question Answering System”
. This survey paper describes the different method of natural language question answering system.
The different types of Question answering system described in this paper is as: i) Web based
Question answering system ii) Information Extraction based question answering system iii)Restricted
Domain Question answering system iv) Rule based question answering system v) Classification of
Questioners Level vi) Question answering system based on Information Retrieval .
transaction by one id
of one specific role
which could be the
loop hole of this
Architecture.
2.4 Summary
During the analysis and taking the experiment results of the system finds that data are more
accurate than the classical process so that it gives better results from the old process. In this a
wide variety of existing mechanism, algorithms and architectures is studied for identifying the
issues removed and remains in Question Answer area. Later on, this gives a brief categorization
of various approaches, which has been suggested over the last few years on Question Answer.
CHAPTER 3
PROBLEM IDENTIFICATION
3.1 Problem Domain
In the last two decades dozens of question answering systems have been developed using some
new concepts and techniques. Hyo-Jung et al. [16] presented Chinese question classification
based on mining association rules, which extract word and bi-gram from questions as classic
features.
Kepei Zhang and Jieyupresented a Chinese Question• Answering System with Question
Classification, which uses word, named entity, part of speech (POS) and semantics as a classic
feature to classify the question.
Santosh Kumar Ray discussed some of the existing approaches for question classification and
proposed a new method based on the usage of the Word Net. Svetlana Stoyanchev [6] presented a
document retrieval experiment on a question answering system, and evaluated the use of named
entities and of noun, verb, and prepositional phrases as exact match phrases in a document
retrieval query, while Kangavari, Samira
Ghandchi, and Manak presented simplest approach to improve the accuracy of a question
answering system. ChiyoungSeoa, Sang-Won Leeb, and Hyoung-JooKima [20] showed in their
performance study that RDBMS implementation using inverted index technique almost always
outperforms the IR implementations.
Paloma Moreda Hector Llorens offered two proposals for using semantic information in QAS,
specifically in the answer extraction step. Its aim is to determine the improvement in performance
of current QA systems, especially when dealing with common noun questions.
Liang Yunjuan&Ma Lijuan discussed the design of dynamic knowledge-based full-text retrieval
system, inverted index technology research and analysis, given some of indexing code, in order to
improve the retrieval accuracy and to achieve a reasonable.
Many architecture have also been proposed for developing Question Answering Systems.
Kolomiyets and Moneshave proposed a model based on the translation of question statement and
document into a computer readable format which is a little bit sophisticated and expensive.
Mohammad Reza Kangavari et al. have also proposed an architecture using dynamic patterns and
semantic relations among words verb and keywords. Both these architectures may perform well
but they are very complex, containing a large number of modules which is difficult to implement.
The usual Question Answering system is designed to answer simple wh-questions like “who”,
“what”, “when”, “where”, etc. Neverthless the recent QA research targets on extending the
system to answer complex questions, synopsis questions, view questions etc.The paper proposes a
Question Answering system that answers simple factoid, questions by using a technique called
Semantic Role Labeling.
1. Question classes
2. Question processing
3. Context and QA
5. Answer extraction
6. Answer formulation
A question may belong to dissimilar type and depending on its category we require different
strategies to answer the question. We all might have a range of attack for a category of factoid
questions, which on the other hand will not work for questions that need much deeper
understanding of facts. We need a profound understanding of what category a question belongs.
Question: Who is also known as “chacha”, and was born on 14th November?
Requires understanding of not only facts but we must have knowledge of Scattering of Light.
Even aswill discuss later, questions can be classified depending on its form in its structure. We
discuss such categories in Chapter dedicated to bloom’s taxonomy.
Same question may be asked in different forms. We may ask it in interrogative way or the
assertive way. We need to understand the semantics of the question. We need to recognize what
the question is asking for before proceeding to answer the question itself. The practice to
understand the question is termed as Question Processing.
Example 1.4: We are seeking same information using different forms of the question:
Interrogative:
Assertive:
Questions are always asked in a context. Questions are rarely asked in universal context. We
could required to have a knowledge of context before proceeding to resolve a question.
For a person who just reached Mumbai: He is interested in finding Taj Hotel.
Earlier than we can answer questions we need sources which are significant and exhaustive. We
all require a data source which will behave as the base for all the information required for
addressing the questions. It may be collection documents. It can be the whole web which we can
search for. It can be a database from where we can get the answers for structured queries.
Depending on the question we may want to extract specific type of information from the data
sources. Reliant on the complexity of the question we would like to know the expectation of user.
Extraction of Place:
Simple extraction may be enough for certain questions. We may want the partial answers to be
extracted from various sources and combine them. At the same time, we want the results of the
QA System to be as natural as possible. For generating answers we require to generate answers
which are termed as answer formulation.
We must answer questions, even the complex question must clarify in few seconds. Persons
would not like to wait for hours in front of computer to get answers to questions. Watson for
instance when played out Jeopardy was able to reply to in average of 3 seconds. We need to
develop architecture such that the end product is a real time system.
Question Answering machine to look for. But other languages like Hindi have a loss of such
resources. Therefore we translate a given query to more formative language and get the answer,
which is translated, back to original language.
3.2.9 Interactive QA
We do now not need a tedious query answering machine that just solutions questions that we pose
to it. The QA system to be interactive and clear doubt in case it finds the question ambiguous.
Example 1.8:
We desire the QA scheme to not only reproduce what is there in the text gathering. We want it to
do more. We want it learn facts and apply reason to generate new facts which will be helpful in
answering the posed question.
We need to arrange data as per its sort so that the appropriate response look ends up noticeably
productive.After we know the category of question we will only search for the document
categories, which are relevant for our question.
We would like to know intention of the user by means of analyzing his/her previous queries. For
accomplishing the task we need to build a profile of user. We would like to get the answers based
upon the taste of the user.
Example 1.10:
The question expansion emphasize on the comparison of multiple Tokens for a particular
keyword. The result shows that the expansions are feasible and efficient to find more similar
documents and sentences, as well as retrieving the answers from web resources to the user
questions.
A straightforward Question Answering structure was implemented using the technique called
Semantic Role Labeling. The system consisted of three phases called Query Processing,
Document Processing, Answer Processing. In each of these levels, numerous language processing
devices were used. The system is web based and therefore uses the search engine to extract
information from the web.
A basic Question Answering framework was actualized utilizing the method called Semantic Role
Labeling. The system consisted of three phases called Query Processing, Document Processing,
and Answer Processing. In each of these phases, several language-processing components were
used. The system is web based and hence uses the search engine to extract information from the
web.
3.4 Summary:
In this chapter we identify problem in existing system. Later on, this will give a brief
categorization of various approaches, which has been suggested over the last few years on
Question Answering System using Data mining approaches.
CHAPTER 4
PROPOSED WORK
Identification) are chosen as a Head Word. This Head word will be useful in the next
phase of Clustering.
6. Clustering: - Here Clustering technique is applied on Wikipedia data set for answer
extraction process. Clustering is the task of grouping a set of objects in such a way that
objects in the same group are more similar to each other than to those in other group. The
Clustering Technique used in this paper is K-Means. K-means Clustering is a method of
Vector Quantization. It aims to partition n Observations into k clusters in which each
observation belongs to the cluster with the nearest mean.
The Head Word Selected in the above phase is used in this phase to form clusters. To
apply K-Means on the data set TF-IDF has been used.Three clusters are formed using K-
means algorithm, from where and the answers will be retrieved.
7. Templates Matching :- . Templates are the predefined format of the answer which will be
presentedin front of the user. Templates are formed by the Head word , selected in the 5th
phase and by the Question Format decided in the Question analysis Phase.
The Answer Extraction process will match the following Templates in its Database to give
an exact answer. Templates such as:
i. “Head Word” is . . . . . . .
ii. “Head Word” means . . . . . . .
iii. “Head Word” is known as . . . . . .
iv. “Head Word” is called. . . . . . . .
v. “Head Word” can be defined as . . . . . . .
8. Answer Extraction: - In this Phase the templates generated in the above phase will be
matched to that in the Clusters formed in the clustering phase.
Proposed System show in below figure 4.1.Proposed system performs following operations:
1. Question Processing: In this module the given Question is processed to get some important
information from it. Steps through which question Processing Module passes and their
descriptions are given below. Steps in Question Processing Module:
a. Find the Type of given question using Wh word.
b. Find out the expected type of answer.
c. Get the Keywords from the Question.
he first step in the QA System is the Question Processing or Question Classification module.
Various information, which we will get through this module, are the Type of Question, Expected
Answer Type, Focus or Head Word of the Question and the Question Keywords.
2. Document Processing: Once the question has been processed we will move towards the
document processing module. In this module the documents which are relevant to the given
question are retrieved and processed. Following steps used in document processing.
a. Get the question in hand and search relevant documents using a reliable search
engine.
b. Take top relevant documents.
c. Extract the content from these documents.
d. Save these contents in to file
3. Answer processing: This module presents algorithms for extracting the potential answer
for all the three categories of questions that is Definition Type of Question, Descriptive
Type of Question and Factoid Type of Question.
4. Dataset Clustering: cluster dataset using fuzzy c-mean algorithm then process for
question and answer processing.
The architecture of the proposed system is shown below. The question asked by the user
goes through different stages from preprocessing to Question identification, and then from
clustering the data sets and finally template matching is performed to get the required
answer.
Below shown is the flow chart of the proposed system. The flow chart describes the way in which
user gets an answer for its Query. If the question entered by the user is not valid, then the system
returns an error. If the question is valid it performs the steps shown to get a exact answer.
type_of_question=questionIdentification();
//generate template
print question type
//select HeadWord from tokens;
K-mean(wikipedia_database)
ans[]= search wikipedia (HeadWord);
for i=1 to ans.length
matchTemplate(template, ans[i])
add to answer
return answer;
}
Netbeans 8.0.1
Fast & Smart Code Editing-An IDE is much more than a text editor. The Net Beans Editor
indents lines, matches text and brackets, and best parts source code syntactically and
semantically. What's more, it provides code templates, coding tips, and refactoring methods. The
editor supports many languages PHP, Java, C/C++, HTML, Groovy, JavaScript and JSP and
Servlet. Since the editor is extensible, many plugin are available for many other languages.
Keeping an obvious overview regarding large software, with many folders as well as files, and
millions of lines regarding code, can be a daunting process. NetBeans IDE delivers different
views of one's data, from many project glass windows to helpful tools for putting together your
software and handling them successfully, letting a person drill into your data efficiently, while
providing you version instruments via Subversion, Mercurial, and Git integration out from the
box.
Design GUIs for Java SE, Java EE,HTML, PHP and Java ME applications quickly and smoothly
by using editors and drag-and-drop tools in the IDE. Regarding Java SE applications, the
NetBeans GUI Builder automatically manages correct spacing and alignment, whilst supporting
in-place enhancing, as effectively. The GUI builder is so convenient to use and spontaneous that
many experts have used for you to prototype GUIs dwell at buyer presentations.
The expense of buggy code enhances the longer the item remains unfixed. NetBeans gives static
research tools, especially integration using the widely used FindBugs software, for determining
and fixing common problems in Coffee code. Additionally, the NetBeans Debugger lets you place
breakpoints in your source signal, add discipline watches, stage through your own code, come
upon methods, take pic and keep an eye on execution since it occurs. NetBeans Profiler delivers
expert support for optimizing your application's pace and ram usage, and helps it be easier to
develop reliable and also scalable Java SE, JavaFX and also Java EE applications. NetBeans IDE
has a visual debugger with regard to Java SE applications, permitting you to debug person
interfaces without researching source signal. Take GUI snapshots of your applications and select
user software elements for you to jump on the similar source signal.
Java
JAVA is easy to program- Java was designed to be easy to use and is therefore easy to
write, compile, debug, and learn than other programming languages.
Java is pure object oriented language.
Java is platform-independent-One of the most significant advantages of Java is its ability
to move easily from one computer system to another. The ability to run the program on
many different operating systems is crucial to WWW, and Java succeeds at this by being
platform-independent at both the source and binary levels.
JDK 8
JDK 8 namely Java Development Kit 8 is the latest version of Java and JDK 8 is released as part
of Java Standard Edition.
4.4Tools Used
Considering the each feature of real implementation and checking the results of existing
approaches some improvements has to be performed in later stages of this work. It requires some
tools, which facilitates the development and proves the results as required. The category is
divided into two basic areas:
JDK 1.7
NetBeans 8.0
The implementation of the proposed concept of machine learning based weather prediction
system need to incorporate the following third party class libraries for supporting the entire
implementation.
S. Classes Description
No.
1 java. util.ArrayList The array list provide a high level data structure
that consumes data to intermediately store and
use during program execution
2 java.sql Provides the API for retrieving and processing
data kept in a data source (usually a relational
database) using the Java TM programming
language.
3 javax.swing.filechooser That is swing GUI used to select a file and
directory for user input
The given section provides the description of classes that are implemented for developing the
components of the proposed working model.
This section provides the details of the implemented user interface design and their navigational
model for user working. The figure 4.3 shows the initial project screen.
4.7 Summary
After studying the different existing mechanism this identifies the System Preliminary. It gives a
clear understanding the Algorithm with its steps. It will help the solution to provide better
resolution of the current situations of security.This chapter also gives implementation plan and
Testing Strategy of above security problems by suggesting an architectural solution. Here in this
chapter the implementation of our proposed system will be done. The implementation is working
on which platform, what kind of theme and approach is followed is referred in this section.
CHAPTER 5
RESULT ANALYSIS
5.1 Result Analysis
Question and Answering System is developed in this research with help Java (JDK1.8) and Net
Beans IDE8.02 on window operating sytem7. All forms of Question Answering System design in
Swing. Graph plotted for computation time, type of question and memory management using
JFree Chart Library. In Result Analysis compare Proposed Question Answering system with
existing Question Answering system in term of computation time and memory.
In Question Answering System took each type of questions for experiment like Factoid Question,
Descriptive and Definition. Wikipedia used as dataset for search Question answers. Below figure
shows that home screen of project.
1 Definition Type 6
2 Description Type 2
3 Factoid Type 2
We calculate computation time for Exiting Question Answering system and Proposed
Question Answering system. And results shown with help of graph. From experiments
found that Proposed Question Answering system less computation time compare to
Existing Question Answering system.
We calculate computation memory for Exiting Question Answering system and Proposed
Question Answering system. And results shown with help of graph. From experiments
found that Proposed Question Answering system less computation memory compare to
Existing Question Answering system.
5.3 Summary
Developing a solution is an approach proving mechanism but to prove its results is a complicated
task because it measures each and every step of the solution and let it compare with the existing
mechanisms. So as to do that effectively this chapter gives a detailed result analysis to prove
effectiveness of the suggested mechanism.
CHAPTER 6
CONCLUSION & FUTURE WORK
6.1 Conclusion
In this thesis we have proposed a structure for space question Answering System utilizing Data
mining devices and programming. This structure can be utilized to build up a Question Answering
System for removing definite and exact answer from limited area printed informational collection.
The proposed structure not just gives a straightforward and implementable structure for creating
question Answering System yet in addition gives an appropriate stream of information for answer
extraction. Since the proposed show works over watchwords and headword and is autonomous of
the inquiry or sentence structure, it has decreased the overhead of question standardization.
Besides since the structure is given for confined area, it likewise handles the issue of word sense
disambiguation. The significant issue, which exists with the proposed structure, is that its
execution is reliant on the execution of the web crawler and the utilized Data mining apparatuses.
REFERENCES
[1]. Sreelakshmi V, Sangeetha Jamal, Survey Paper : Question Answering Systems, in National Conference on
Computing and Communication - (NCCC ), March 2014, GEC Idukki.
[2]. M Ramprasath, S Hariharan Improved Question Answering System by semantic reformulation, IEEE- Fourth
International Conference on Advanced Computing, 2012.
[3]. Ali Mohamed Nabil Allam, and Mohamed Hassan Haggag, The Question Answering Systems: A Survey,
International Journal of Research and Reviews in Information Sciences (IJRRIS), September 2012 Science Academy
Publisher, United Kingdom
[4] Molla D., and Vicedo J., "Question answering in restricted domains: An overview", Computer Linguist, ppAI-6 1,
2007
[5] Moreda P., Llorens H., Saquete E., & Palomar M., "Combining semantic information in question answering
systems", Information Processing & Management, pp.870-885, 20 1 1.
[6] Svetlana Stoyanchev, and Young Chol Song, and William Lahti, "Exact Phrases in Information Retrieval for
Question Answering", Coling 2008: Proceedings of the 2nd workshop on Information Retrieval for Question
Answering (IR4QA), pp. 9- 16 Manchester,UK. August 2008".
[7] Woods W.A, Kaplan R.A, Nash-Webber.B, "The lunar sciences naturallanguage information system" , Final
report: BBN Report #2378. Technical report, Bolt Beranek and Newman Inc.,Cambridge, MA., June 1972.
[8] Green RF, Wolf A.K., Chomsky, K. Laughery, "BASEBALL: An automatic question answerer", in: Proceedings
of Western Computing Conference, vol.19, pp. 2 19-224, 196 1.
[9] Ittycheriah A, Franz M, Zhu WJ, Ratnaparkhi A and Mammone RJ. IBM‟s statistical question answering system.
In Proceedings of the Text Retrieval Conference TREC-9, 2000.
[10] Athira P. M., Sreeja M. and P. C. Reghuraj
Department of Computer Science and Engineering, Government
Engineering College, Sreekrishnapuram, Palakkad, Kerala, India, 678633. Architecture of an Ontology-Based
Domain-Specific Natural Language Question Answering System.
[12] Sanjay K Dwivedi, Vaishali Singh. Research and reviews in question answering system Department of
Computer Science, B. B. A. University (A Central University) Luck now, Uttar Pradesh, 226025, India.
[13] Poonam Gupta, Vishal Gupta
Assistant Professor, Computer Science & Engineering Department University
Institute of Engineering & Technology Panjab University, Chandigarh.
[14] Kolomiyets, Oleksander. And Moens, Marie-Francine. “A survey on question answering technology from an
information retrieval perspective”. Journal of Information Sciences 181, 2011.5412-5434. DOI:
10.1016/j.ins.2011.07.047. Elsevier.
[15] Moreda, Paloma., Llorens Hector., Saquete, Estela. And Palomar, Manuel. “Combining semantic information in
question answering systems” Journal of Information Processing and Management 47, 2011. 870- 885. DOI:
10.1016/j.ipm.2010.03.008. Elsevier.
[16] Ko, Jeongwoo., Si, Luo., and Nyberg Eric. “Combining evidence with a probabilistic framework for answer
ranking and answer merging in question answering” Journal: Information Processing and Management 46,
2010
541-554. DOI: 10.1016/j.ipm.2009.11.004. Elsevier.
[17] Pachpind Priyanka P, Bornare Harshita N, Kshirsagar Rutumbhara B, Malve Ashish D BE Comp S.N.D COE &
RC, YEOLA,” An Automatic Answering System Using Template Matching For Natural Language Questions”.
PLAGIARISM
Self Certificate
This is to certify that on “Question & Answering System Using
Natural Language Processing” for the partial fulfillment of M.Tech
(C.S) YDC program is original and well checked by plagiarism detector
software (http://smallseotools.com/plagiarism-checker).
Chapter % of Unique
Number Plagiarism Checked In Content
1 Introduction 100 %
2 Literature Review 93 %
3 Problem Identification 89%
4 Proposed Work 95 %
5 Result Analysis 100%
6 Conclusion & Future work 100%
"NC.docx"
Licensed to: Sachin Choudhary
Relation chart:
Distribution graph:
% 0.9 wrds: 16
% 0.5 wrds: 9
347 - Ok / 42 - Failed
Mtech Scholar
Abstract:
Question and Answering System is one of the major module and Answer Processing module. Each Processing
research are in Natural Language. Main challenges of and Information Retrieval module contains several sub
Question and Answer system gives exact answer of modules and these modules use several Natural Language
question which give by user. Question and Answering Processing Techniques in order to extract the proper
system can be classified into three category are open answer. The usual Question Answering system is designed
domain, closed domain and restricted domain. Using to answer simple wh-questions like “who”, “what”,
advanced Natural Language Processing tool we will be “when”, “where”, etc. But the recent QA research focuses
developed a framework for question answering system. on extending the system to answer complex questions,
In this paper we work on restricted domain question summary questions, opinion questions etc. The paper
answering system. Proposed system work on keyword and proposes a Question Answering system that answers
question matching and return precise answer of simple factoid, wh-questions by using a technique called
question. Semantic Role Labeling.
1. INTRODUCTION:
Although the set of documents which are retrieved by the
search engine contain a lot of information about the search
topic but it may or may not contain exactly that
information which the user is looking for [1].The basic
idea behind the question answering system is that the users Figure 1. Block Diagram Question Answering System
just have to enter the question and the system will retrieve
the most appropriate and precise answer for that question The rest of the paper is organized as follows. The next
and return it to the user. Hence in those cases where the section describes the general architecture of a Question
user is looking for a short and precise answer, question Answering System. Section 3 discusses some of the
answering System plays a great role rather than Search related works in this area. The proposed system
Engines, which usually provide a large set of links of those architecture is described in section 4. The paper concludes
web pages which might contain the answer of that with the experimental setup and results.
question. A typical Question Answering system can be
divided into 3 modules namely: Question Processing 2. ARCHITECTURE OF A QUESTION
module, Document Processing or Information Retrieval ANSWERING
1
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
described in detail in the following subsections. Another system developed by Pragisha K. Et.al [11],
described about the. It receives Malayalam natural
Question Answering Systems can be classified on the basis language questions from the user and extracts most
of the domains over which it has been constructed. appropriate response by analyzing a collection of
Malayalam documents. The system handles four each
Open Domain Question Answering question.
Close Domain Question Answering Research and reviews in question answering system
Restricted Domain Question Answering developed by Sanjay K Dwivedi Et.al[12] propose
taxonomy for characterizing Question Answer (QA)
Open domain question answering systems are domain systems, survey of major QA systems described in
independent. It relies on general ontology and world literature and provide a qualitative analysis of them.
knowledge. Usually these systems have a large collection
of data from where the required answer is to be found out. Table [I] presents comparison about different types of
Since in case of Open Domain question answering question answering system [22].
information content is not of particular domain it can
answer questions of various fields however here deep S. No Type of Question Question and
reasoning is not possible [3]. and Answering Answering System
System Methods
Close domain question answering systems deal with 1 Multilingual Tokenization and
questions in a specific domain [3]. LUNAR and Question/Answering pos tagging., Word
BASEBALL are the example of close domain QA systems sense
.In this case the data set contains a very limited amount of disambiguation,
focused and structured information . hence in case of close Answer type
domain question answering systems deep reasoning is identification,
possible but the problem with these systems was that due Keywords
to the very small size of data set they are not more than a expansion,
'Toy Systems"[4]. Semantic
Disambiguation
Research in restricted-domain question answering 2 Analysis of the Query
(RDQA) addresses problems related to the incorporation Asks Question- Reformulation,
of domain- specific information into current state-of-the- Answering System NGram Mining, N-
art QA technology with the hope of achieving deep Gram Filtering, N-
reasoning capabilities and reliable accuracy performance Gram Tiling.
in real world applications. In fact, as a not too-long-term 3 Multilingnality, Answering
vision, Spatial- temporal architecture
context awareness,
3. LITERATURE SURVEY: Textual entailment
4 A Question Expected Answer
In most of the research papers [4, 5, 6] LUNAR [7] and Answering System Type, Named
BASEBALL [8] have been discussed as the earlier based on Entities Presence,
developed question answering systems. However there are Information
various question answering systems which have been Retrieval and
developed with different concepts since the idea of QA Validation
System has been coined 5 A Hybrid Question Module,
Answering System Hypothesis
In a system developed Athira P. M, Et.al [10], presented based on Generation
an architecture of ontology-based domain-specific natural Information Module, Document
language question answering that applies semantics and Retrieval and Processing and
domain knowledge to improve both query construction and Answer Validation Indexing
answer extraction. 6 A specifiable Answering
2
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
domain multilingual architecture and their descriptions are given below. Steps
Question in Question Processing Module:
In a System developed by Poonam Gupta Et.al [13] A a.
Find the Type of given
Survey of Text Question Answering Techniques. Question
answering is a difficult form of information retrieval question using Wh word.
characterized by information needs that are at least b. Find out the expected type of
somewhat expressed as natural language Template answer.
Matching Automatic Answering System For natural c. Get the Keywords from the
languages questions proposed by Pachpind Priyanka Et.al Question.
[17], Frequently Asked QA System that replies with pre-
stored answers to user questions asked in regular English, d. Find out the Focus of the
rather than keyword or sentence structure based retrieval question.
mechanisms.
he first step in the QA System is the Question Processing
4. Proposed System: or Question Classification module. Various information,
which we will get through this module, are the Type of
Since both the Open Domain QA System and Close Question, Expected Answer Type, Focus or Head Word of
Domain QA System have their own pros and cons a new the Question and the Question Keywords.
concept of Question Answering has been coined by
Molla& Vice do [4] called RESTRICTED DOMAIN QA Question Type
SYSTEM, which is the midway of these two domains.
WH Factoid Definition Descriptive
We are convinced that research in restricted domains will word Type Type Type
drive the convergence between structured knowledge-
based and free text-based question answering. Question Who How What
When What
What Why
Where
Which
3
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
Type of Question, Descriptive Type of [6] Svetlana Stoyanchev, and Young Chol Song, and
Question and Factoid Type of Question. William Lahti, "Exact Phrases in Information Retrieval for
Question Answering", Coling 2008: Proceedings of the
4. Dataset Clustering: cluster dataset using 2nd workshop on Information Retrieval for Question
fuzzy c-mean algorithm then process for Answering (IR4QA), pp. 9- 16 Manchester,UK. August
question and answer processing. 2008".
[7] Woods W.A, Kaplan R.A, Nash-Webber.B, "The lunar
5. CONCLUSION: sciences naturallanguage information system" , Final
report: BBN Report #2378. Technical report, Bolt Beranek
In this paper we have proposed a framework for restricted and Newman Inc.,Cambridge, MA., June 1972.
domain question Answering System using advanced NLP [8] Green RF, Wolf A.K., Chomsky, K. Laughery,
tools and software. This framework can be used to develop "BASEBALL: An automatic question answerer", in:
a Question Answering System for extracting exact and Proceedings of Western Computing Conference, vol.19,
precise answer from restricted domain textual data set. The pp. 2 19-224, 196 1.
proposed framework not only provides a simple and [9] Ittycheriah A, Franz M, Zhu WJ, Ratnaparkhi A and
implementable framework for developing question Mammone RJ. IBM‟s statistical question answering
Answering System but also provides a proper flow of data system. In Proceedings of the Text Retrieval Conference
for answer extraction. TREC-9, 2000.
[10] Athira P. M., Sreeja M. and P. C. Reghuraj
Since the proposed model works over keywords and
Department of Computer Science and Engineering,
headword and is independent of the question or sentence Government Engineering College, Sreekrishnapuram,
structure, it has reduced the overhead of question Palakkad, Kerala, India, 678633. Architecture of an
normalization. Moreover since the framework is given for Ontology-Based Domain-Specific Natural Language
restricted domain, it also handles the issue of word sense Question Answering System.
disambiguation. The major problem which exists with the [11] Pragisha K. “design and implementation of a QA
proposed framework is that it's performance is dependent system in Malayalam”.
on the performance of the search engine and the used NLP [12] Sanjay K Dwivedi, Vaishali Singh. Research and
tools. reviews in question answering system Department of
Computer Science, B. B. A. University (A Central
6. REFERENCES: University) Luck now, Uttar Pradesh, 226025, India.
[13] Poonam Gupta, Vishal Gupta
Assistant Professor,
[1]. Sreelakshmi V, Sangeetha Jamal, Survey Paper : Computer Science & Engineering Department University
Question Answering Systems, in National Conference on Institute of Engineering & Technology Panjab University,
Computing and Communication - (NCCC ), March 2014, Chandigarh.
GEC Idukki. [14] Kolomiyets, Oleksander. And Moens, Marie-
[2]. M Ramprasath, S Hariharan Improved Question Francine. “A survey on question answering technology
Answering System by semantic reformulation, IEEE- from an information retrieval perspective”. Journal of
Fourth International Conference on Advanced Computing, Information Sciences 181, 2011.5412-5434. DOI:
2012. 10.1016/j.ins.2011.07.047. Elsevier.
[3]. Ali Mohamed Nabil Allam, and Mohamed Hassan [15] Moreda, Paloma., Llorens Hector., Saquete, Estela.
Haggag, The Question Answering Systems: A Survey, And Palomar, Manuel.“Combining semantic information
International Journal of Research and Reviews in in question answering systems” Journal of Information
Information Sciences (IJRRIS), September 2012 Science Processing and Management 47, 2011. 870- 885. DOI:
Academy Publisher, United Kingdom 10.1016/j.ipm.2010.03.008. Elsevier.
[4] Molla D., and Vicedo J., "Question answering in [16] Ko, Jeongwoo., Si, Luo., and Nyberg Eric.
restricted domains: An overview", Computer Linguist, “Combining evidence with a probabilistic framework for
ppAI-6 1, 2007 answer ranking and answer merging in question
[5] Moreda P., Llorens H., Saquete E., & Palomar M., answering” Journal: Information Processing and
"Combining semantic information in question answering Management 46,
2010 541-554. DOI:
systems", Information Processing & Management, pp.870- 10.1016/j.ipm.2009.11.004. Elsevier.
885, 20 1 1. [17] Pachpind Priyanka P, BornareHarshita N,
4
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
KshirsagarRutumbhara B, Malve Ashish D BE Comp Answering System. Section 3 discusses some of the
S.N.D COE & RC, YEOLA,” An Automatic Answering related works in this area. The proposed system
System Using Template Matching For Natural Language architecture is described in section 4. The paper concludes
Questions”. with the experimental setup and results.
1.
I Introduction
II Architecture of a Question Answering
Although the set of documents which are retrieved by the
search engine contain a lot of information about the search In this section we describe the architecture of our system.
topic but it may or may not contain exactly that The overall architecture of the system can be subdivided
information which the user is looking for [1].The basic into three main modules: (1) pre-processing, (2) question
idea behind the question answering system is that the users template matching, and (3) answering. Each module is
just have to enter the question and the system will retrieve described in detail in the following subsections.
the most appropriate and precise answer for that question
and return it to the user. Hence in those cases where the Question Answering Systems can be classified on the basis
user is looking for a short and precise answer, question of the domains over which it has been constructed.
answering System plays a great role rather than Search
Engines, which usually provide a large set of links of those • Open Domain Question Answering
web pages which might contain the answer of that
question. A typical Question Answering system can be • Close Domain Question Answering
divided into 3 modules namely: Question Processing
module, Document Processing or Information Retrieval • Restricted Domain Question Answering
5
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
Another system developed by Pragisha K. Et.al [11], In a System developed by Poonam Gupta Et.al [13] A
described about the. It receives Malayalam natural Survey of Text Question Answering Techniques. Question
language questions from the user and extracts most answering is a difficult form of information retrieval
appropriate response by analyzing a collection of characterized by information needs that are at least
Malayalam documents. The system handles four each somewhat expressed as natural language Template
question. Matching Automatic Answering System For natural
languages questions proposed by Pachpind Priyanka Et.al
Research and reviews in question answering system [17], Frequently Asked QA System that replies with pre-
developed by Sanjay K Dwivedi Et.al[12] propose stored answers to user questions asked in regular English,
taxonomy for characterizing Question Answer (QA) rather than keyword or sentence structure based retrieval
systems, survey of major QA systems described in mechanisms.
literature and provide a qualitative analysis of them.
IV Proposed System
S. No Type of Question Question and
and Answering Answering System Since both the Open Domain QA System and Close
System Methods Domain QA System have their own pros and cons a new
1 Multilingual Tokenization and concept of Question Answering has been coined by
Question/Answering pos tagging., Word Molla& Vice do [4] called RESTRICTED DOMAIN QA
sense SYSTEM, which is the midway of these two domains.
disambiguation,
Answer type We are convinced that research in restricted domains will
identification, drive the convergence between structured knowledge-
Keywords based and free text-based question answering.
expansion,
Semantic
Disambiguation
2 Analysis of the Query
Asks Question- Reformulation,
Answering System NGram Mining, N-
Gram Filtering, N-
Gram Tiling.
3 Multilingnality, Answering
Spatial- temporal architecture
context awareness,
Textual entailment
4 A Question Expected Answer
6
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
Figure 5.2 Question Types Figure 5.3 Computation time for Existing and Proposed
System.
S.No. Question Type No. Of Questions Table 5.3 Computation time for Existing and Proposed
System.
1 Definition Type 6 Computation Computation
Time of Time of
2 Description Type 2
Question Existing Proposed
S.No.
3 Factoid Type 2 Number Question Question
Answering Answering
System System
8
International Journal of Research in Science & Engineering
Vol X Issue IX Jan 2023
P-ISSN:2394-8280 E-ISSN:2394-8299
(MS) (MS)
Question
1 13203 7215
Number 1
Question
2 9853 4978
Number 2
Question
3 10340 9734
Number 3
Question
4 22565 5123
Number 4
Question
5 11287 7460 Figure 5.5 Initial screens for Existing and Proposed
Number 5
System.
Computation Memory
Output Screen
9
International Journal of Research in Science & Engineering
Vol IV Issue IV Aug 2018
P-ISSN:2394-8280 E-ISSN:2394-8299
Question
4 2164 1946
Number 4
Question
5 1798 1816
Number 5
VI Conclusion
In this paper we have proposed a framework for restricted
domain question Answering System using advanced NLP
tools and software. This framework can be used to develop
a Question Answering System for extracting exact and
precise answer from restricted domain textual data set. The
Figure 5.8 Select Head Word of Question proposed framework not only provides a simple and
implementable framework for developing question
Answering System but also provides a proper flow of data
for answer extraction.
Figure 5.6 Answer of Proposed Question Ans
Since the proposed model works over keywords and
headword and is independent of the question or sentence
structure, it has reduced the overhead of question
normalization. Moreover since the framework is given for
restricted domain, it also handles the issue of word sense
disambiguation. The major problem which exists with the
proposed framework is that it's performance is dependent
on the performance of the search engine and the used NLP
tools.
References
[1]. Sreelakshmi V, Sangeetha Jamal, Survey Paper :
Question Answering Systems, in National Conference on
Computing and Communication - (NCCC ), March 2014,
GEC Idukki.
Table 5.4 Computation memory for Existing and Proposed [2]. M Ramprasath, S Hariharan Improved Question
System Answering System by semantic reformulation, IEEE-
Computation Computation Fourth International Conference on Advanced Computing,
Memory of Memory of 2012.
Existing Proposed [3]. Ali Mohamed Nabil Allam, and Mohamed Hassan
Question Haggag, The Question Answering Systems: A Survey,
S.No. Question Question
Number International Journal of Research and Reviews in
Answering Answering
System System Information Sciences (IJRRIS), September 2012 Science
(MB) (MB) Academy Publisher, United Kingdom
Question [4] Molla D., and Vicedo J., "Question answering in
1 2183 1940 restricted domains: An overview", Computer Linguist,
Number 1
Question ppAI-6 1, 2007
2 2434 2015 [5] Moreda P., Llorens H., Saquete E., & Palomar M.,
Number 2
"Combining semantic information in question answering
Question
3 2507 2473 systems", Information Processing & Management, pp.870-
Number 3
885, 20 1 1.
10
International Journal of Research in Science & Engineering
Vol IV Issue IV Aug 2018
P-ISSN:2394-8280 E-ISSN:2394-8299
[6] Svetlana Stoyanchev, and Young Chol Song, and [16] Ko, Jeongwoo., Si, Luo., and Nyberg Eric.
William Lahti, "Exact Phrases in Information Retrieval for “Combining evidence with a probabilistic framework for
Question Answering", Coling 2008: Proceedings of the answer ranking and answer merging in question
2nd workshop on Information Retrieval for Question answering” Journal: Information Processing and
Answering (IR4QA), pp. 9- 16 Manchester,UK. August Management 46,
2008". 2010 541-554. DOI: 10.1016/j.ipm.2009.11.004. Elsevier.
[7] Woods W.A, Kaplan R.A, Nash-Webber.B, "The lunar [17] Pachpind Priyanka P, BornareHarshita N,
sciences naturallanguage information system" , Final KshirsagarRutumbhara B, Malve Ashish D BE Comp
report: BBN Report #2378. Technical report, Bolt Beranek S.N.D COE & RC, YEOLA,” An Automatic Answering
and Newman Inc.,Cambridge, MA., June 1972. System Using Template Matching For Natural Language
[8] Green RF, Wolf A.K., Chomsky, K. Laughery, Questions”.
"BASEBALL: An automatic question answerer", in:
Proceedings of Western Computing Conference, vol.19,
pp. 2 19-224, 196 1.
[9] Ittycheriah A, Franz M, Zhu WJ, Ratnaparkhi A and
Mammone RJ. IBM‟s statistical question answering
system. In Proceedings of the Text Retrieval Conference
TREC-9, 2000.
[10] Athira P. M., Sreeja M. and P. C. Reghuraj
Department of Computer Science and Engineering,
Government Engineering College, Sreekrishnapuram,
Palakkad, Kerala, India, 678633. Architecture of an
Ontology-Based Domain-Specific Natural Language
Question Answering System.
11
International Journal for Rapid Research in Engineering Technology & Applied Science
Vol X Issue X Feb 2023
ISSN:2455-4723
I Introduction
Although the set of documents which are retrieved by the
search engine contain a lot of information about the search Figure 1. Block Diagram Question Answering System
topic but it may or may not contain exactly that
information which the user is looking for [1].The basic The rest of the paper is organized as follows. The next
idea behind the question answering system is that the users section describes the general architecture of a Question
just have to enter the question and the system will retrieve Answering System. Section 3 discusses some of the
the most appropriate and precise answer for that question related works in this area. The proposed system
and return it to the user. Hence in those cases where the architecture is described in section 4. The paper concludes
user is looking for a short and precise answer, question with the experimental setup and results.
answering System plays a great role rather than Search
Engines, which usually provide a large set of links of those
web pages which might contain the answer of that II Architecture of a Question Answering
question. A typical Question Answering system can be
divided into 3 modules namely: Question Processing In this section we describe the architecture of our system.
module, Document Processing or Information Retrieval The overall architecture of the system can be subdivided
into three main modules: (1) pre-processing, (2) question
template matching, and (3) answering. Each module is
described in detail in the following subsections.
Question Answering Systems can be classified on the basis Another system developed by Pragisha K. Et.al [11],
of the domains over which it has been constructed. described about the. It receives Malayalam natural
language questions from the user and extracts most
• Open Domain Question Answering appropriate response by analyzing a collection of
Malayalam documents. The system handles four each
• Close Domain Question Answering question.
• Restricted Domain Question Answering Research and reviews in question answering system
developed by Sanjay K Dwivedi Et.al[12] propose
Open domain question answering systems are domain taxonomy for characterizing Question Answer (QA)
independent. It relies on general ontology and world systems, survey of major QA systems described in
knowledge. Usually these systems have a large collection literature and provide a qualitative analysis of them.
of data from where the required answer is to be found out.
Since in case of Open Domain question answering S. No Type of Question Question and
information content is not of particular domain it can and Answering Answering System
answer questions of various fields however here deep System Methods
reasoning is not possible [3]. 1 Multilingual Tokenization and
Question/Answering pos tagging., Word
Close domain question answering systems deal with sense
questions in a specific domain [3]. LUNAR and disambiguation,
BASEBALL are the example of close domain QA systems Answer type
.In this case the data set contains a very limited amount of identification,
focused and structured information . hence in case of close Keywords
domain question answering systems deep reasoning is expansion,
possible but the problem with these systems was that due Semantic
to the very small size of data set they are not more than a Disambiguation
'Toy Systems"[4]. 2 Analysis of the Query
Asks Question- Reformulation,
Research in restricted-domain question answering Answering System NGram Mining, N-
(RDQA) addresses problems related to the incorporation Gram Filtering, N-
of domain• specific information into current state-of-the- Gram Tiling.
art QA technology with the hope of achieving deep 3 Multilingnality, Answering
reasoning capabilities and reliable accuracy performance Spatial- temporal architecture
in real world applications. In fact, as a not too-long-term context awareness,
vision. Textual entailment
4 A Question Expected Answer
III Literature Survey Answering System Type, Named
based on Entities Presence,
In most of the research papers [4, 5, 6] LUNAR [7] and Information
BASEBALL [8] have been discussed as the earlier Retrieval and
developed question answering systems. However there are Validation
various question answering systems which have been 5 A Hybrid Question Module,
developed with different concepts since the idea of QA Answering System Hypothesis
System has been coined based on Generation
Information Module, Document
In a system developed Athira P. M, Et.al [10], presented Retrieval and Processing and
an architecture of ontology-based domain-specific natural Answer Validation Indexing
language question answering that applies semantics and 6 A specifiable Answering
domain knowledge to improve both query construction and domain multilingual architecture
answer extraction. Question
V. Result Analysis
Question and Answering System is developed in this
research with help Java (JDK1.8) and Net Beans IDE8.02
on window operating sytem7. All forms of Question Figure 5.2 Question Types
Answering System design in Swing. Graph plotted for
computation time, type of question and memory
management using JFree Chart Library. In Result Analysis
compare Proposed Question Answering system with Table 5.1 Number of Question in Types
existing Question Answering system in term of
computation time and memory. S.No. Question Type No. Of Questions
Figure 5.3 Computation time for Existing and Proposed Figure 5.3 Computation Memories for Existing and
System. Proposed System.
Output Screen
Table 5.3 Computation time for Existing and Proposed
System.
Computation Computation
Time of Time of
Existing Proposed
Question
S.No. Question Question
Number
Answering Answering
System System
(MS) (MS)
Question
1 13203 7215
Number 1
Question
2 9853 4978
Number 2
Question
3 10340 9734 Figure 5.5 Initial screens for Existing and Proposed
Number 3
System.
Question
4 22565 5123
Number 4
Question
5 11287 7460
Number 5
Computation Memory
VI Conclusion
In this paper we have proposed a framework for restricted
domain question Answering System using advanced NLP
tools and software. This framework can be used to develop
a Question Answering System for extracting exact and
precise answer from restricted domain textual data set. The
proposed framework not only provides a simple and
Figure 5.8 Select Head Word of Question implementable framework for developing question
Answering System but also provides a proper flow of data
for answer extraction.
Figure 5.6 Answer of Proposed Question Ans Since the proposed model works over keywords and
headword and is independent of the question or sentence
structure, it has reduced the overhead of question
normalization. Moreover since the framework is given for
restricted domain, it also handles the issue of word sense
disambiguation. The major problem which exists with the
proposed framework is that it's performance is dependent
on the performance of the search engine and the used NLP
tools.
References
[1]. Sreelakshmi V, Sangeetha Jamal, Survey Paper :
Question Answering Systems, in National Conference on
Computing and Communication - (NCCC ), March 2014,
GEC Idukki.
[2]. M Ramprasath, S Hariharan Improved Question Department University Institute of Engineering &
Answering System by semantic reformulation, IEEE- Technology Panjab University, Chandigarh.
Fourth International Conference on Advanced Computing,
2012. [14] Kolomiyets, Oleksander. And Moens, Marie-
[3]. Ali Mohamed Nabil Allam, and Mohamed Hassan Francine. “A survey on question answering technology
Haggag, The Question Answering Systems: A Survey, from an information retrieval perspective”. Journal of
International Journal of Research and Reviews in Information Sciences 181, 2011.5412-5434. DOI:
Information Sciences (IJRRIS), September 2012 Science 10.1016/j.ins.2011.07.047. Elsevier.
Academy Publisher, United Kingdom
[4] Molla D., and Vicedo J., "Question answering in [15] Moreda, Paloma., Llorens Hector., Saquete, Estela.
restricted domains: An overview", Computer Linguist, And Palomar, Manuel.“Combining semantic information
ppAI-6 1, 2007 in question answering systems” Journal of Information
[5] Moreda P., Llorens H., Saquete E., & Palomar M., Processing and Management 47, 2011. 870- 885. DOI:
"Combining semantic information in question answering 10.1016/j.ipm.2010.03.008. Elsevier.
systems", Information Processing & Management, pp.870-
885, 20 1 1. [16] Ko, Jeongwoo., Si, Luo., and Nyberg Eric.
[6] Svetlana Stoyanchev, and Young Chol Song, and “Combining evidence with a probabilistic framework for
William Lahti, "Exact Phrases in Information Retrieval for answer ranking and answer merging in question
Question Answering", Coling 2008: Proceedings of the answering” Journal: Information Processing and
2nd workshop on Information Retrieval for Question Management 46,
Answering (IR4QA), pp. 9- 16 Manchester,UK. August 2010 541-554. DOI: 10.1016/j.ipm.2009.11.004. Elsevier.
2008". [17] Pachpind Priyanka P, BornareHarshita N,
[7] Woods W.A, Kaplan R.A, Nash-Webber.B, "The lunar KshirsagarRutumbhara B, Malve Ashish D BE Comp
sciences naturallanguage information system" , Final S.N.D COE & RC, YEOLA,” An Automatic Answering
report: BBN Report #2378. Technical report, Bolt Beranek System Using Template Matching For Natural Language
and Newman Inc.,Cambridge, MA., June 1972. Questions”.
[8] Green RF, Wolf A.K., Chomsky, K. Laughery,
"BASEBALL: An automatic question answerer", in:
Proceedings of Western Computing Conference, vol.19,
pp. 2 19-224, 196 1.
[9] Ittycheriah A, Franz M, Zhu WJ, Ratnaparkhi A and
Mammone RJ. IBM‟s statistical question answering
system. In Proceedings of the Text Retrieval Conference
TREC-9, 2000.
[10] Athira P. M., Sreeja M. and P. C. Reghuraj
Department of Computer Science and Engineering,
Government Engineering College, Sreekrishnapuram,
Palakkad, Kerala, India, 678633. Architecture of an
Ontology-Based Domain-Specific Natural Language
Question Answering System.