Ingeniero

Sentiment Analysis in Microblogging: A Practical Implementation
Mauro Cohen, Pablo Damiani, Sebastian Durandeu, Renzo Navas, Hernán Merlino, Enrique Fernández
Departamento de Computación, Facultad de Ingeniería, Universidad De Buenos Aires, Paseo Colón 850, Buenos Aires, Argentina {litodam, maurocohen, renzoe, sebastiandurandeu}@gmail.com, hmerlino@fi.uba.ar, enfernan@itba.edu.ar

Abstract.This paper presents a system that can take short messages relevant to a particular topic from a microblogging service such as Twitter or Facebook, analyze the messages for the sentiments they carry on, and classify them. In particular, the system addresses this problem by retrieving raw data from Twitter - one of the most popular microblogging platforms - pre-processing on that raw data, andfinally analyzing it using machine learning techniques to classify them by sentiment as either positive or negative. Keywords: Microblogging, sentiment analysis, sentiment classification, opinion mining, information retrieval, text mining, social web, twitter, python, nltk

1

Introduction

In the past few years, social networks have increased their popularity to become the mainstream platforms ofthe Internet world. An average Internet user nowadays spend more time on social networks than on search engines and e-mail [1]. Among the different social networks types, microblogging networks have gained a strong importance in recent years [2]. Templeton [3] defines microblogging as a small-scale form of blogging made up from short, succinct messages, used by both consumers and businesses toshare news, post status updates and carry on conversations. Millions of Internet users use microblogging to talk about their daily activities and to seek or share information. These published messages might also include real-time opinions and feelings on certain topics, for example likes or dislikes statements. There are different microblogging platforms available today: Twitter [4], Jaiku [5],Tumblr [6] to name just a subset. Among them, Twitter [4] has become the prevalent platform. Since Twitter’s inception in 2006, it has grown at an unprecedented rate. In just four years, the service has grown to approximately 20 million unique visitors each month with users sending short 140-character messages (known as “tweets”) approximately 40 million times a day [7]. Twitter is a public datasource that has been proven a valuable source of information [8].

CACIC 2011 - XVII CONGRESO ARGENTINO DE CIENCIAS DE LA COMPUTACIÓN

191

As Twitter gains popularity, it becomes more useful to analyze trends and sentiment of its users towards various topics. Determining the general attitude of users towards a product or service, for example, can help a business measure overall consumerattitudes, overall satisfaction and feeling about the brand. It can also provide a warning when there is a sudden change in sentiment [9]. As a whole, applying automated tools that attempt to classify tweets into either positive, negative or neutral categories automatically could be quite useful for companies and marketers. In this context, with the population of social networks and microblogging, newresearch fields on sentiment analysis - also known as sentiment extraction or opinion mining - have grown considerably and gained special attention lately. Sentiment analysis, grounded on machine learning techniques, is the task of identifying positive and negative opinions, emotions, and evaluations [10]. It aims to identify the sentiment or feeling in the users to something such as a product,company, place, person and others based on the content published in the web. In the view of the requester, it is possible to obtain a summary about what people are feeling about a topic, without the need of finding and reading all opinions and other news related to it. Sentiment analysis is mainly a text categorization problem which desires to detect favorable and unfavorable opinions related to a...