6.domain extraction from research papers

www.jst.org.in 42 | Page
Journal of Science and Technology (JST)
Volume 2, Issue 4, April 2017, PP 42-50
www.jst.org.in ISSN: 2456 - 5660
Domain Extraction From Research Papers
Dr. R. Jayanthi1
, S. Sheela2
1
Dr. Mrs. R. Jayanthi, Assistant Professor, PG and Research Department of Computer Science, Quaid E –
Millath college for Women, Chennai, India..
2
Ms. S. Sheela, M.Phil Research Scholar, Computer Science, Quaid E – Millath college for Women, Chennai,
India.
Abstract: Automatically finding domain specific key terms from a given set of research paper is a challenging
task and research papers to a particular area of research is a concern for many people including students,
professors and researchers. A domain classification of papers facilitates that search process. That is, having a
list of domains in a research field, we try to find out to which domain(s) a given paper is more related. Besides,
processing the whole paper to read take a long time. In this paper, using domain knowledge requires much
human effort, e.g., manually composing a set of labeling a large corpus. In particular, we use the abstract and
keyword in research paper as the seeing terms to identify similar terms from a domain corpus which are then
filtered by checking their appearance in the research papers. Experiments show the TF –IDF measure and the
classification step make this method more precisely to domains. The results show that our approach can extract
the terms effectively, while being domain independent.
Keywords - Text Mining, Information Extraction, Domain keyword extraction, Term Frequency and Inverse
Document Frequency (TF – IDF)
I. INTRODUCTION
Domain – specific terms are term that have significant meaning (s) in a specific domain [1]. We extract
the term from the research papers. Here a “term” refers to a word or compound words representing a concept of
a specific domain, e.g., in chemistry, “alcohol” is a term that refers to an organic compound in which the
hydroxyl functional group is bound to a saturated carbon atom [2]. Terminology extraction are using rule based
techniques, supervised learning techniques or a combination of these two types of techniques all of which rely
on some domain knowledge. Acquiring such domain knowledge requires much human effort (e.g., manually
labeling a large corpus) [3].
To overcome this approach uses semantic extraction by building knowledge from a large corpus. The
semantic extraction refers to range of processing techniques that identify and extract entities, facts, attributes,
concepts, and events to populate metadata fields. The purpose of this is to enable the analysis of semi-
structured or unstructured content. Semantic extraction is usually based on three approaches,
1. Rule based: Matching similar to entity extraction, this approach requires the support of one or more
vocabularies.
2. Machine learning: A statistical analysis of the content, that potentially compute intensive application that
can benefit within the document corpus.
3. Hybrid solution: Statistically driven, but enhanced by a vocabulary. This is typically the best approach if the
content set is focused on a specific subject area.
The extracted information also should be machine understandable as well as human understandable in
term of research paper from set of domain corpus [4]. A statistical method is proposed in this paper and it is
based on, 3 steps.
First, extract the abstract and keyword in the collection of research paper using pdfifilter. Second,
terms which are similar to a certain domain occur frequently in research paper. Third, count word introduced
learning to Predict from Text (Weighted Scoring Method). The highly count value is the domain name of
particular research paper. The TF–IDF is introduced, the weighting method of measure the value of domain
corpus. User can easily find out the domain name from research paper as well as to separate folder for each
domain paper to save on your computer. Later users don’t waste the time for searching the research paper. We
can save the time for searching the research paper with domain name.
II. LITERATURE REVIEW
For text mining of domain extraction techniques we have studied few related papers. In this section we
describe the different techniques with different authors which are related to the domain extraction.
In this paper [2] existing terminology extraction approaches are mostly domain dependent. They use
domain specific linguistic rules, supervised machine learning techniques. In particular, we use the title words
and the keywords in research papers as the seeing terms and word2vec to identify similar terms from an open-
domain corpus as the candidate terms, which are the filtered by checking their occurrence in researchpapers.

Journal of Science and Technology
Rakhi Chakraborty, explains [5] it is extremely time consuming and difficult task to extract keyword or
feature manually. So an automated process that extracts keywords or features needs to be established. This paper
proposes a new domain keyword extraction technique that includes a new weighting method on the base of the
conventional TF - IDF. Term frequency-Inverse document frequency is widely used to express the documents
feature weight, which can’t reflect the division of terms in the document, and then can’t reflect the significance
degree and the difference between categories.
Hospice Houngbo, Robert E. Merer, Method Mention Extraction from Scientific Research Paper,
explains [6] scientific publications contain many references to method terminologies used during scientific
experiments. In this study we report our attempt to automatically extract such method terminologies from
scientific research papers, using rule-based and machine learning techniques. We first used some linguistic
features to extract fine-grained method sentences from a large biomedical corpus and then applied well
established methodologies to extract the method terminologies.
The author of [7] the computational linguistics community and its sub-fields have changed over the
years with respect to their foci, methods used, and domain problems. We extract these characteristics by
matching semantic extraction patterns, learned using bootstrapping, to the dependency trees of sentences in an
article’s abstract.
In this paper [8] the dynamics of a research community can be studied by extracting information from
its publications. Such information cannot be extracted using approaches that assume words are independent of
each other in a document. We use dependency trees, which give rich information about structure of a sentence,
and extract relevant information from them by matching semantic patterns.
III. DOMAIN KEYWORD EXTRACTION TECHNIQUES
Domain Corpus
In linguistics, a corpus (plural corpora) or text corpus is a large and structured set of texts (nowadays usually
electronically stored and processed). They are used to do statistical analysis and hypothesis testing, checking
occurrences or validating linguistic rules within a specific language territory. This paper, uses seven Domain
name with set of large corpus.
Figure 1: Domain Corpus in Research Paper
Figure1 shows the list for each domain some number of keywords is manually setting such as Text Mining 51,
Data Mining 66, Cryptography 56, Software Engineering 53, Big Data 57, Network Security 78, and Image
Processing 73.
PDF IFILTER
An IFilter is a plugin that allows Microsoft's search engines to index various file formats (as documents, email
attachments, database records, audio metadata etc.) So that they become searchable. Without an appropriate
IFilter, contents of a file cannot be parsed and indexed by the search engine. An IFilter acts as a plug-in for
extracting full-text and metadata for search engines.
A search engine usually works in two steps: The search engine goes through a designated place, e.g. a file folder
or a database, and indexes all documents or newly modified documents, including the various types documents,
in the background and creates internal data to store indexing result. A user specifies some keywords he would
like to search and the search engine answers the query immediately by looking up the indexing result and
responds to the user with that contains the keywords. During Step 1, the search engine itself doesn't understand
format of a document.

Therefore, it looks on Windows registry for an appropriate IFilter to extract the data from the document
format, filtering out embedded formatting and any other non-textual data. Figure 2 explain, one full research
paper (paper format pdf) read after using this tool to extract abstract and keywords in research paper.
Figure 2: Extract abstract and Index Terms from research paper TEXT
Similarity Mining
Text Vector must be generated before text similarity calculation between domain corpus and extraction
of research paper. A Chinese corpus converting economy, military, education, culture, and other fields, contains
approximately 100,000 documents [9]. Generating word figure3 sets from research paper such as domain
corpus. The process is relatively simple feature words were extracted from research paper and put in the domain
corpus.
Figure 3: Intersection of Domain Corpus and Extracted Research Paper
Figure 4, getting the result to calculate the total word for each domain. Total result of particular domain
name in given research paper, t is the maximum number appearance of the word.
t = count (Maximum Number of words appearance in domain corpus for each domain)
Figure 4: Count the words appearance in domain corpus for each domain

Finally, figure 5 explain the maximum number of count in the domain result of one research paper. For
each paper for calculating and get the result for 200 research paper.
Figure 5: Domain Result
EVALUATION MEASURE
Precision
In the field of information retrieval, precision is the fraction of retrieved documents that are relevant to
the query:
Precision = (relevant items retrieved) / (retrieved items) = P (relevant|retrieved)
Precision takes all retrieved documents into account, but it can also be evaluated at a given cut-off rank,
considering only the topmost results returned by the system. This measure is called precision. For example for a
text search on a set of documents precision is the number of correct results divided by the number of all returned
results. Precision is also used with recall, the percent of all relevant documents that is returned by the search.
The two measures are sometimes used together in the F1 Score (or f-measure) to provide a single measurement
for a system. Note that the meaning and usage of "precision" in the field of information retrieval differs from the
definition of accuracyand precision within other branches of science and technology.
Recall
Recall in information retrieval is the fraction of the documents that are relevant to the query that are
successfully retrieved.
Recall = (relevant items retrieved) / (relevant items) = P(retrieved|relevant)
For example for text search on a set of documents recall is the number of correct results divided by the number
of results that should have been returned. In binary classification, recall is called sensitivity. So it can be looked
at as the probability that a relevant document is retrieved by the query. It is trivial to achieve recall of 100% by
returning all documents in response to any query. Therefore, recall alone is not enough but one needs to measure
the number of non-relevant documents also, for example by computing the precision.
F –Measure
F1 measure is a derived effectiveness measurement. The resultant value is interpreted as a weighted average of
the precision and recall. The best value is 1 and the worst is 0.
F –Measure = 2((precision*recall) / (precision+recall))
TF – IDF
TF: Term Frequency, which measures how frequently a term occurs in a document. Since Table1, every
document is different in length it is possible that a term would appear much more times in long documents than
shorter ones. Thus, the term frequency is often divided by the document length (the total number of terms in the
document) as a way of classification [10].
TF(t) = (Number of times term t appears in a document) / (Total number of terms in the document)
The domain result is Data mining then value is 1 otherwise 0.
Total number of paper: 200
TF (t) = 40/56 = 0.7142857142

Where, text mining paper total count is 40 and domain corpus count is 56. The TF calculation number of
times t appear in the document value dividing the total number of terms in the document. Table 1, explain the
number of paper appears in the particular domain and the number of times appears, then the total calculation for
each domain.
Table 1: TF calculation
rm is. While computing TF, all terms are
considered equally important. However it is known that certain terms, such as "is", "of", and "that", may appear
a lot of times but have little importance. Thus we need to weigh down the frequent terms while scale up the rare
ones, by computing the following [10]:
IDF(t) = log_e(Total number of documents / Number of documents with term t in it)
IDF(t) = log_e (200 / 20)
Total no of paper =200
No of paper occur = 10
Finally,
tf-idf = tf *idf
= 0.7142857142 * 10
tf-idf = 7.142857142
Example, Consider a document containing 100 words wherein the word cat appears 3 times. The term frequency
(i.e., tf) for cat is then (3 / 100) = 0.03. Now, assume we have 10 million documents and the word cat appears in
one thousand of these. Then, the inverse document frequency (i.e., idf) is calculated as log(10,000,000 / 1,000) =
4. Thus, the Tf-idf weight is the product of these quantities: 0.03 * 4 = 0.12.
Domain name Paper1 Paper 2 Paper 3 Paper 4 Paper …. Paper
200
Total
Data mining 1 0 0 1 … 1 20
Text mining 0 1 1 1 … 1 40
Network
security
1 0 1 1 … 0 20
Cryptography 1 1 0 0 … 1 35
Software
engineering
1 0 1 0 … 0 25
Image
processing
1 0 1 1 ... 0 30
Big data 1 1 0 1 … 1 30
IDF: Inverse Document Frequency, which measures how important a te

IV. EXPERIMENTAL RESULT IN DOMAIN EXTRACTION
Step 1: Create Domain Corpus in Word document or notepad
Step 2: Read Research paper(pdf format) using Pdf IFilter Tool
Step 3: Extract Abstract and Index Terms from Research Paper

Step 4: Then count the term frequency from the collected term and rank them according to the high term
frequency (Using Term Frequency Similarity)
Step 5: Domain Extraction Result from Research Paper
Step 6: Precision, Recall and F – Measure calculation from Research Paper

Step 7: TF – IDF calculation
Step 8: Domain Extraction from Research paper for each paper domain name for the following image
Step 9: Graph Representation from Research paper for each Domain
V. CONCLUSION
The abstract and index terms extracting can be used for extracting domain name from specified research
paper and applied to document classification. This paper uses PDF IFilter is one of the best-known and most
commonly used research paper extractions currently in use. Term similarity mining approach is the term
frequency value counting from research paper. We use different performance measurements, including
precision, recall and F – measures with respect to individual human annotations and a weighted measure. Finally
TF –IDF measurement is calculated for each domain from overall research paper. This paper experimented with
the identification of individual research paper with domain name.

REFERENCES
[1] Su Nam Kim and Lawrence Cavedon, Classifying Domain-Specific Terms Using a Dictionary, In Proceedings of Australasian Language
Technology Association Workshop, 2011, 57−65.
[2] Birong Jiang, Edong Xun and Jianzhong Qi, A Domain Independent approach from Extracting Terms from Research Papers, Springer
International Publishing Switzerland 2015, 155- 166.
[3] P. Velardi, M. Missikoff, R. Basili, Identification of Relevant Terms to Support the Construction of Domain Ontologies, ACL workshop
on Human Language Technology and Knowledge Management, 2001, 5:1 – 5:8.
[4] Rishabh Upadhyay, Akihiro Fujii, Semantic Knowledge Extraction from Research Documents, Computer Science and Information System
(Fed CSIS) 2016.
[5] Rakhi Chakraborty, Domain Keyword Extraction Technique: A new weighting method based on frequency analysis, National Conference
on Advancement of Computing in Engineering Research, ACER 2013.
[6] Hospice Houngbo, Robert E. Merer, Method Mention Extraction from Scientific Research Paper, Proceeding of COLING 2012, 1211 –
1222.
[7] Sonal Gupta, Christopher D.Manning, Analyzing the dynamics of Research by Extracting key Aspects of Scientific Papers, International
Joint Conference on Natural Language Processing (IJCNLP), 2011.
[8] Sonal Gupta, Christopher D.Manning, Identifying Focus, Techniques and Domain of Scientific Papers, NIPS workshop on
Computational Social Science and the Wisdom of Crowds (NIPS-CSS), 2010.
[9] Gang Chen, Feng Liu, Mohammad Shojafer, Fuzzy System and Data Mining proceeding of FSDM 2015, Frontiers in Artificial
Intelligence and Applications 281, IOS Press 2016, ISBN 978-1-61499-618-7.
[10] H. Wu and R. Luk and K. Wong and K. Kwok, Interpreting TF-IDF term weights as making relevance decisions, ACM Transactions on
Information Systems, 2008.
BIBLIOGRAPHY OF AUTHORS
Dr. R. Jayanthi, MCA, M.Phil., Ph.D., working as an Assistant Professor in PG &
Research Department of Computer Science at Quaid-E-Millath Govt. College for
Women (Autonomous), Chennai. Her areas of interests are Data Mining, Text
Mining, Natural Language Processing, Information Extraction and Business
Intelligence. She has published articles in more than 10 InternationalJournals.
S. Sheela received her M.sc. Computer science in 2015 from Queen Mary’s College
for Women. She is pursuing her M.Phil. Computer Science under the supervision of
Dr. Mrs. R. Jayanthi, MCA., M.Phil., Ph.D., Assistant Professor in PG & Research
Department of Computer Science at Quaid-EMillath government College for
Women, Affiliated to university of Madras. She has presented papers in International
conferences and published papers in International Journals. Her area of interest is
Text Mining, Cryptography and Network Security.

6.domain extraction from research papers

More Related Content

What's hot (19)

Similar to 6.domain extraction from research papers (20)

Recently uploaded (20)

6.domain extraction from research papers