LSIS at SemEval-2016 Task 7: Using Web Search Engines for English and Arabic Unsupervised Sentiment Intensity Prediction

Proceedings of SemEval-2016, pages 469–473, San Diego, California, June 16-17, 2016. c⃝2016 Association for Computational Linguistics LSIS at SemEval-2016 Task 7: Using Web Search Engines for English and Arabic Unsupervised Sentiment Intensity Prediction Amal Htait *,+ amal.htait@lsis.org *Aix-Marseille University, CNRS LSIS UMR 7296 13397, Marseille France Sebastien Fournier *,+ sebatien.fournier@lsis.org Patrice Bellot *,+ patrice.bellot@lsis.org +Aix-Marseille University, CNRS CLEO OpenEdition UMS 3287 13451, Marseille France Abstract In this paper, we present our contribution in SemEval2016 task71: Determining Senti- ment Intensity of English and Arabic Phrases, where we use web search engines for English and Arabic unsupervised sentiment intensity prediction. Our work is based, first, on a group of classic sentiment lexicons (e.g. Sen- timent140 Lexicon, SentiWordNet). Second, on web search engines’ ability to find the co- occurrence of sentences with predefined nega- tive and positive words. The use of web search engines (e.g. Google Search API) enhance the results on phrases built from opposite polarity terms. 1 Introduction A sentiment lexicon is a list of words and phrases, such as ”excellent”, ”awful” and ”not bad”, each is being assigned with a positive or negative score reflecting its sentiment polarity and strength. Sen- timent lexicon is crucial for sentiment analysis (or opining mining) as it provides rich sentiment infor- mation and forms the foundation of many sentiment analysis systems (Pang and Lee, 2008; Liu, 2012). Sentence intensity is essential when we need to compare sentences having the same polarity orien- tation. It is expressed by words or phrases with dif- ferent strengths. For example, the word ”excellent” is stronger than ”good”. The sentiment words, like ”good” and ”bad”, are used to express positive and negative sentiments. But also intensifier and dimin- isher words can change the degree of the expressed sentiment, an intensifier increases the intensity of a 1http://alt.qcri.org/semeval2016/task7/ positive or negative word like ”very” and a dimin- isher decrease its intensity like ”barely”. However, sentence intensity prediction of short sentences faces several challenges: 1. Due to the nature of the short sentence itself; the limited size of the sentences, the informal language of the content that may contain slang words and non-standard expressions (e.g. LOL instead of laughing out loud, greaaaaaat etc.), and the high level of noise due to the absence of spell checker tools. 2. Due to the sentiment lexicons not including all of the vocabulary needed, or may not be to- tally balanced between positive and negative sentences. Our proposal is to: 1. Calculate the probability of positivity for the phrase in the sentiment lexicons, using the point-wise mutual information (PMI) (Cover et al., 1994). 2. When the phrase is not included in the senti- ment lexicons, we use the web search engine to find probability of positivity for the phrase based on its co-occurrence near the word ”ex- cellent” and near the word ”poor”. 2 Related work The sentiment words are the main factor for sen- timent classification, by consequence sentiment words and phrases can be used for sentiment clas- sification in an unsupervised method, a method …

Content truncated to 500 words