DOI: http://dx.doi.org/10.26483/ijarcs.v9i2.5783
Volume 9, No. 2, March-April 2018
International Journal of Advanced Research in Computer Science
RESEARCH PAPER
Available Online at www.ijarcs.info
© 2015-19, IJARCS All Rights Reserved 714
ISSN No. 0976-5697
THEORETICAL ANSWER EVALUATION USING LSA, BLEU, WMD AND
FUZZY LOGIC
Praful Mishra
K.J Somaiya Institute Of Engineering And
Information Technology, Sion, Mumbai, India
Anmol Mishra
K.J Somaiya Institute Of Engineering And
Information Technology, Sion, Mumbai, India
Aniket Bharti
K.J Somaiya Institute Of Engineering And
Information Technology, Sion, Mumbai, India
Prof. Sarita Ambadekar
K.J Somaiya Institute Of Engineering And
Information Technology, Sion, Mumbai, India
Abstract : Assessment of student answers to grade their overall understanding of a subject is a critical task. However grading can be monotonous and
sometimes can be tedious task for the teachers. Automatic Grading can reduce tedium on teachers but it is complicated by free form student inputs.
The main task of automatic grading system is to assign ordinal scores to student answers, based on “model” or ideal answers. Here we introduce a
novel framework comprising of three building blocks Word Mover Distance(WMD)a statistical model Latent Semantic Analysis(LSA),Bilingual
Evaluation Understudy(BLEU) and Fuzzy logic, a model based on degree of truth to output scores. In other words LSA is used to identify the
semantic similarity between two concepts. Word Mover’s Distance (WMD), uses vector encoding of words to calculate the minimum cumulative
distance that words from a reference solution need to travel to match words from a student answer. This cumulative distance assess the distance
between two documents in a meaningful way, even when they have no words in common. Fuzzy logic is a primitive model in this system which is
used to output the final score based on inputs which are the outputs of LSA and WMD. This proposed method gives better precision, enhanced
dependability of results, thus saving the effort and time of staff.
Keywords: Latent Semantic Analysis, Singular Value Decomposition, N-gram, Word Mover Distance, Fuzzy Logic.
I. INTRODUCTION
Theory based examinations are held periodically to assess
students academically.The purpose of these assessment is to
gain insight of student understanding and knowledge
enhancement. However the manual evaluation of answers
sometimes can be monotonous ,bias errors and tiring. To
overcome these difficulties a faster and reliable method to
evaluate answer is required. Natural Language Processing
(NLP) is a technique in Artificial Intelligence that enables us
to analyze and synthesize natural language. NLP is further
divided into syntactic analysis and semantic analysis. Latter is
used for analysing grammar and arrangement of words in such
a manner that they show relationship among themselves.
While former is used to extract meaning from text. We publish
a paper that uses both syntactic and semantic methods to
evaluate student’s answer and allot them marks.
This paper proposes a algorithm to avoid this
gruesome manual answer evaluation. The assessment of
answers is done using a novel framework comprising of
generative probabilistic technique and degree of truth
techniques. Due to the freedom of input students write a
particular sentence in various form.LSA[1] measures the
semantic similarity of these answers with standard answer by
finding out important topics in both, BLEU[2]avoids
overrating of answer by LSA. WMD[4] measures the
similarity of students answer with standard answers even if
sentences in both are written different way but mean the same.
LSA,BLEU,WMD along with soft computing technique Fuzzy
Logic[3] gives the overall assessment of student and standard
answer.
II. EXISTING SYSTEM
Attali and Burstein devloped E-rater (Electronic Essay Grade),
that checks the writing style and the structure of the essays
rather than the specific content. Text features like vocabulary
level used, word occurring probability, correlation, word
length and essay length were extracted and parsed using
MSNLP (Microsoft Natural Language Processing) tool with
training essays. These features are allotted weightage. Already
graded essay are used for evaluating new essays, its features
are compared to already graded essays. The strength is the
agreement between the E-rater and human is above 97%. The
weakness of E-rater is that it requires a number of manually
scored training essays to score the answers.
C-rater (Concept rater) (Burstein et al., 2001; Yigal et
al., 2008) was also developed by ETS (Educational Testing
Service) and it is also called as content rater. The scoring is
based on the content and concept. It uses natural language