Neural Networks for Open Domain Targeted Sentiment

Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing, pages 612–621, Lisbon, Portugal, 17-21 September 2015. c⃝2015 Association for Computational Linguistics. Neural Networks for Open Domain Targeted Sentiment Meishan Zhang and Yue Zhang and Duy-Tin Vo Singapore University of Technology and Design {meishan zhang, yue zhang}@sutd.edu.sg, duytin vo@mymail.sutd.edu.sg Abstract Open domain targeted sentiment is the joint information extraction task that finds target mentions together with the senti- ment towards each mention from a text corpus. The task is typically modeled as a sequence labeling problem, and solved us- ing state-of-the-art labelers such as CRF. We empirically study the effect of word embeddings and automatic feature combi- nations on the task by extending a CRF baseline using neural networks, which have demonstrated large potentials for sentiment analysis. Results show that the neural model can give better results by significantly increasing the recall. In ad- dition, we propose a novel integration of neural and discrete features, which com- bines their relative advantages, leading to significantly higher results compared to both baselines. 1 Introduction Targeted sentiment analysis has drawn growing re- search interests over the past few years. Compared with traditional sentiment analysis tasks, which extract the overall sentiment of a document, a sen- tence or a tweet, targeted sentiment analysis ex- tracts the sentiment over given targeted entities from a text, and therefore is practically more infor- mative. An example is shown in Figure 1. There are at least two practical scenarios: (1) Certain entities of concern are specified, and the requirement is to extract the sentiment to- wards their mentions in a text. For exam- ple, one can be interested in the sentiment towards Google Inc., Microsoft and Face- book in financial news texts, or the sentiment towards Manchester United, Liverpool and Chelsea in tweets. So excited to meet my [baby Farah]+ !!! [Baseball Warehouse]+ : easy to under- stand information. The [#Afghan #Parlaiment Speaker]− should Resign . Saw [Erykah Badu]− last night , vile venue unfortunately . [AW service]0 will be back at work . Figure 1: Targeted sentiment analysis. (2) No specified target is given, and the require- ment is to find sentiments towards entities in the open domain. For example, one might be interested extracting the mentions to all per- sons and organizations, together with the sen- timents towards each mention, from a news archive or a collection of novels. There are two sub tasks in targeted sentiment analysis, namely entity recognition and sentiment classification for each entity mention which ap- ply to both scenarios above. In scenario (1), en- tity recognition is relatively trivial, and can typ- ically be achieved by pattern matching. Partly due to this reason, most previous work has ad- dressed targeted sentiment analysis as a pure clas- sification task, assuming that target mentions have been given (Jiang et al., 2011; Chen et al., 2012; Dong et al., 2014; Vo and Zhang, 2015). For scenario (2), a named entity recognition (NER) system can be used to extract targets, before the same …

Content truncated to 500 words