Browse Prior Art Database

MACHINE LEARNING BASED PREDICTIVE MODEL FOR ANALYSING THE SENTIMENTS IN SHORT TEXT

IP.com Disclosure Number: IPCOM000239081D
Publication Date: 2014-Oct-10
Document File: 9 page(s) / 547K

Publishing Venue

The IP.com Prior Art Database

Related People

Kumar, Naveen: INVENTOR [+2]

Abstract

The present innovation is aimed at providing a machine learning based predictive model for analyzing the sentiments in short texts. Generally, sentiment analysis using machine learning techniques comprises the steps of pre – processing input content, extracting features from the input content and building a mathematical model for making predictions. The crux of the instant approach lies in the innovative feature extraction process which has manifested in building a good predictive model. More details on the technical and non – technical specifics such as objects, working & implementation, advantages etc. may be found in the sections that follow.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 27% of the total text.

Page 01 of 9

MACHINE LEARNING BASED PREDICTIVE MODEL FOR ANALYSING THE SENTIMENTS IN SHORT TEXT

Majumdar, Tapas; Kumar, Naveen;

GTO - Data Marketing Services, Cognizant Technology Solutions U.S. Corporation

The present innovation is aimed at providing a machine learning based predictive model for analyzing the sentiments in short texts. Generally, sentiment analysis using machine learning techniques comprises the steps of pre - processing input content, extracting features from the input content and building a mathematical model for making predictions. The crux of the instant approach lies in the innovative feature extraction process which has manifested in building a good predictive model. More details on the technical and non - technical specifics such as objects, working & implementation, advantages etc. may be found in the sections that follow.


2. Background

Sentiment is the attitude, opinion or feeling toward something, such as a person, organization, product or location. In the pre- web, pre-text analytics world, sentiment analysis was of very limited scope or limited to indirect measures. Interested entities had focus groups and/ or other forms of "qualitative research" which were applied in specific domains and operated upon limited data samples.

However, in today's world, because of the enormous popularity of web technologies and social media, the volume and relevance of information, opinion, interests and criticism, shared by individuals has increased manifolds. This shared information is usually indicative of the views and expressions of desirable masses on various subject - matter, e.g. reviews on different products/brands/services. Accordingly, over the last few years, sentiment analysis has drifted away from the realms of manual scrutiny and has moved to automated approaches.

Due to high business feasibility, automated sentiment analysis has been a subject of constant research and improvement. Various researchers have tried different approaches to improve the accuracy of sentiment detection of short text like tweets


1. Abstract



Page 02 of 9

and comments. One of the approach that has come across as a champion in the area of Sentiment Analysis and Opinion mining is that of Machine Learning. It is proven that Machine Learning has high accuracy w.r.t. predicting sentiments from short texts such as tweets.

A survey of the work done in relation to analyzing the sentiments in short texts would show that the work done in this area is scarce. For instance, twitter sentiment analysis is a relatively under constructed area. Prominent works relating to twitter sentiment analysis are illustrated as under -

 (Go et al., 2009)1 performed sentiment analysis on Twitter. They identified the tweet polarity using emoticons as noisy labels and collected a training dataset of 1.6 million tweets. They reported an accuracy of 81.34%.

 (Davidov et al., 2010)2 used 50 hashtags and 15 emoticons as noisy labels to create a dataset for Twitter sentime...