Improving Gender Classification of Blog Authors

Proceedings of the 2010 Conference on Empirical Methods in Natural Language Processing, pages 207–217, MIT, Massachusetts, USA, 9-11 October 2010. c⃝2010 Association for Computational Linguistics Improving Gender Classification of Blog Authors Arjun Mukherjee Bing Liu Department of Computer Science University of Illinois at Chicago 851 South Morgan Street Chicago, IL 60607, USA amukherj@cs.uic.edu Department of Computer Science University of Illinois at Chicago 851 South Morgan Street Chicago, IL 60607, USA liub@cs.uic.edu Abstract The problem of automatically classifying the gender of a blog author has important appli- cations in many commercial domains. Exist- ing systems mainly use features such as words, word classes, and POS (part-of- speech) n-grams, for classification learning. In this paper, we propose two new techniques to improve the current result. The first tech- nique introduces a new class of features which are variable length POS sequence pat- terns mined from the training data using a se- quence pattern mining algorithm. The second technique is a new feature selection method which is based on an ensemble of several fea- ture selection criteria and approaches. Empir- ical evaluation using a real-life blog data set shows that these two techniques improve the classification accuracy of the current state-of- the-art methods significantly. 1 Introduction Weblogs, commonly known as blogs, refer to on- line personal diaries which generally contain in- formal writings. With the rapid growth of blogs, their value as an important source of information is increasing. A large amount of research work has been devoted to blogs in the natural language processing (NLP) and other communities. There are also many commercial companies that exploit information in blogs to provide value-added ser- vices, e.g., blog search, blog topic tracking, and sentiment analysis of people’s opinions on prod- ucts and services. Gender classification of blog authors is one such study, which also has many commercial applications. For example, it can help the user find what topics or products are most talked about by males and females, and what products and services are liked or disliked by men and women. Knowing this information is crucial for market intelligence because the information can be exploited in targeted advertising and also product development. In the past few years, several authors have stu- died the problem of gender classification in the natural language processing and linguistic com- munities. However, most existing works deal with formal writings, e.g., essays of people, the Reuters news corpus and the British National Corpus (BNC). Blog posts differ from such text in many ways. For instance, blog posts are typically short and unstructured, and consist of mostly informal sentences, which can contain spurious information and are full of grammar errors, abbreviations, slang words and phrases, and wrong spellings. Due to these reasons, gender classification of blog posts is a harder problem than gender classifica- tion of traditional formal text. Recent work has also attempted gender classi- fication of blog authors using features such as content words, dictionary based content analysis results, POS (part-of-speech) tags and feature se- lection along with a …

Content truncated to 500 words