Towards a Universal Sentiment Classifier in Multiple languages

Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing, pages 511–520 Copenhagen, Denmark, September 7–11, 2017. c⃝2017 Association for Computational Linguistics Towards a Universal Sentiment Classifier in Multiple languages Kui Xu and Xiaojun Wan Institute of Computer Science and Technology, Peking University The MOE Key Laboratory of Computational Linguistics, Peking University {kuixu, wanxiaojun}@pku.edu.cn Abstract Existing sentiment classifiers usually work for only one specific language, and differ- ent classification models are used in dif- ferent languages. In this paper we aim to build a universal sentiment classifier with a single classification model in multiple different languages. In order to achieve this goal, we propose to learn multilingual sentiment-aware word embeddings simul- taneously based only on the labeled re- views in English and unlabeled parallel data available in a few language pairs. It is not required that the parallel data ex- ist between English and any other lan- guage, because the sentiment information can be transferred into any language vi- a pivot languages. We present the eval- uation results of our universal sentiment classifier in five languages, and the results are very promising even when the parallel data between English and the target lan- guages are not used. Furthermore, the u- niversal single classifier is compared with a few cross-language sentiment classifiers relying on direct parallel data between the source and target languages, and the re- sults show that the performance of our uni- versal sentiment classifier is very promis- ing compared to that of different cross- language classifiers in multiple target lan- guages. 1 Introduction Nowadays, a large amount of user-generated con- tent (UGC) appears online everyday, such as tweets, comments and product reviews. Sentiment classification on these data has become a popular research topic over the past few years (Pang et al., 2002; Blitzer et al., 2007; Agarwal et al., 2011; Li- u, 2012). Distributed representations of words or word embeddings have been widely explored, and have proved its great usability for the sentimen- t classification task (Tang et al., 2014; Zhou et al., 2015; Xu et al., 2015; Bollegala et al., 2016; Fer- reira et al., 2016). Most existing sentiment classifiers rely on la- beled training data and the data are usually language-dependent. In other words, a sentiment classifier is learned from a labeled dataset in a spe- cific language and this sentiment classifier can be used for sentiment classification in this language. However, labeled training data for sentiment clas- sification are not available or not easy to obtain in many languages in the world (e.g., Malaysian, Mongolian, Uighur). Without reliable labeled da- ta, it is hard to build a sentiment classifier in these resource-poor languages. Fortunately, there are a few studies investigating the task of cross-language sentiment classification (Banea et al., 2008; Wan, 2009; Meng et al., 2012; Xiao and Guo, 2013; Gao et al., 2015; Chen et al., 2015; Zhou et al., 2015; Li et al., 2017; Zhou et al., 2016a,b), which aims to make use of the labeled data in a source language …

Content truncated to 500 words