Rajesh Piryani

Sentiment analysis of textual reviews; Evaluating machine learning, unsupervised and SentiWordNet approaches

This paper presents our experimental results on performance evaluation of all the three approaches for document-level sentiment classification. We have implemented two Machine Learning based classifiers (Naïve Bayes and SVM), the Unsupervised Semantic Orientation approach (SO-PMI-IR algorithm) and the SentiWordNet approaches for sentiment classification of movie reviews. We used two pre-existing large datasets and collected one of moderate size on our own. The paper primarily makes two useful contributions: (a) it presents a comprehensive evaluative account of performance of all the three available approaches on use with movie reviews, and (b) it presents a new modified Adjective+Adverb combine scheme of SentiWordNet approach.

A Content-Based eResource Recommender System to Augment eBook-Based Learning

This paper presents our experimental work to design a content-based recommendation system for eBook readers. The system automatically identifies a set of relevant eResources for a reader, reading a particular eBook, and presents them to the user through an integrated interface. The system involves two different phases. In the first phase, we parse the textual content of the eBook currently read by the user to identify learning concepts being pursued. This requires analysing the text of relevant parts of the eBook to extract concepts and subsequently filter them to identify learning concepts of interest to Computer Science domain. In the second phase, we identify a set of relevant eResources from the World Wide Web. This involves invoking publicly available APIs from Slideshare, LinkedIn, YouTube etc. to retrieve relevant eResources for the learning concepts identified in the first part. The system is evaluated through a multi-faceted process involving tasks like sentiment analysis of user reviews of the retrieved set of eResources for recommendations. We strive to obtain an additional wisdom-of-crowd kind of evaluation of our system by hosting it on a public Web platform.

Information and Relation Extraction for Semantic Annotation of eBook Texts

This paper presents our algorithmic approach for information and relation extraction from unstructured texts (such as from eBook sections or webpages), performing other useful analytics on the text, and automatically generating a semantically meaningful structure (RDF schema). Our algorithmic formulation parses the unstructured text from eBooks and identifies key concepts described in the eBook along with relationship between the concepts. The extracted information is then used for four purposes: (a) for generating some computed metadata about the text source (such as readability of an eBook), (b) generate a concept profile for each distinct part of text, (c) identifying and plotting relationship between key concepts described in the text, and (d) to generate RDF representation for the text source. We have done our experiments on eBook texts from Computer Science domain; however, the approach can be applied to work on different forms of text in other domains as well. The results are not only useful for concept based tagging and navigation of unstructured text documents (such as eBook) but can also be used to design a comprehensive and sophisticated learning recommendation system.

Lexicon Ensemble and Lexicon Pooling for Sentiment Polarity Detection

ABSTRACT This paper presents our experimental work towards detecting sentiment polarity of free-form texts: first by using an ensemble of sentiment lexicons and then through a lexicon pooled machine learning classifier. In the ensemble design, we combined four different sentiment lexicons in different ways to determine sentiment polarities of different text data. The ensemble approach, however, did not achieve superior performance as initially thought. Therefore, in the second design, we tried to pool the sentiment lexicon knowledge into the machine learning classification process itself of a multinomial naive Bayes classifier. The experimental designs are evaluated on three document and two sentence datasets. The lexicon pooled approach obtains superior accuracy levels as compared to standard naive Bayes classifier as well as lexicon-based methods. Furthermore, as the amount of training data decreases, the accuracy levels of lexicon pooled machine learning classifier decays slowly as compared to standalone naive Bayes classifier. The framework presented proves useful and robust and can be extended to any classification task.

A Linguistic Rule-Based Approach for Aspect-Level Sentiment Analysis of Movie Reviews

Aspect-level sentiment analysis refers to sentiment polarity detection from unstructured text at a fine-grained feature or aspect level. This paper presents our experimental work on aspect-level sentiment analysis of movie reviews. Movie reviews generally contain user opinion about different aspects such as acting, direction, choreography, cinematography, etc. We have devised a linguistic rule-based approach which identifies the aspects from movie reviews, locates opinion about that aspect and computes the sentiment polarity of that opinion using linguistic approaches. The system generates an aspect-level opinion summary. The experimental design is evaluated on datasets of two movies. The results achieved good accuracy and shows promise for deployment in an integrated opinion profiling system.

A Lexicon Pooled Machine Learning Classifier for Opinion Mining from Course Feedbacks

This paper presents our algorithmic design for a lexicon pooled approach for opinion mining from course feedbacks. The proposed method tries to incorporate lexicon knowledge into the machine learning classification process through a multinomial process. The algorithmic formulations have been evaluated on three datasets obtained from ratemyprofessor.com. The results have also been compared with standalone machine learning and lexicon based approaches. The experimental results show that the lexicon pooled approach obtains higher accuracy than both the standalone implementations. The paper, thus proposes and demonstrates how a lexicon pooled hybrid approach may be a preferred technique for opinion mining from course feedbacks and hence suitable for develpment in a practical caurse feedback mining system.

An Algorithmic Formulation for Extracting Learning Concepts and Their Relatedness in eBook Texts

In this paper, we present an algorithmic formulation to automatically extract learning concepts and their relationships from eBook texts and to generate an RDF data that can be used for a number of purposes. Our algorithmic approach first extracts various parts of an eBook (such as chapters and sections) and then through a sentence-level parsing scheme identifies learning concepts described in the eBook text. We have programmed for the identification and extraction of relationships between different learning concepts occurring in a section. We have also been able to extract some general data about the eBooks such as author, price, and reviews (through eBook content mining and web crawling). The learning concepts, their relationships and other useful information extracted from the eBooks; is then programmatically transformed into a machine readable RDF data. The automated process of concept and relation extraction and their subsequent storage into RDF data, makes our effort important and useful for tasks like Information Extraction, Concept-based Search and Machine Reading.