Complex and Precise Movie and Book Annotations in French Language for Aspect Based Sentiment Analysis

Complex and Precise Movie and Book Annotations in French Language for Aspect Based Sentiment Analysis Stefania Pecore, Jeanne Villaneau IRISA-D6 Media et Interactions Universit´e de Bretagne Sud, France {stefania.pecore,jeanne.villaneau}@univ-ubs.fr Abstract Aspect Based Sentiment Analysis (ABSA) aims at collecting detailed opinion information according to products and their features, via the recognition of targets of the opinions in text. Though some annotated data have been produced in challenges as SemEval, resources are still scarce, especially for languages other than English. We are interested in enhancing today’s mostly statistical text classification with the use of linguistics tools, in order to better define and analyze what has been written. The work presented in this paper focuses on two French datasets of movies and books online reviews. In reviews, text length is much higher compared to a tweet, giving us the opportunity to work on a challenging and linguistically interesting dataset. Moreover, movies and books are products that make classifying opinions into aspects quite complex. This article provides an analysis of the particularities of the two domains during the process of collecting and annotating data, a precise annotation scheme for each domain, examples and statistics issued from the annotation phase, and some perspectives on our future work. Keywords: French Aspect Based Sentiment Analysis, opinion mining, user reviews 1. Introduction The Web makes it relatively easy to collect a large amount of textual data. Researchers, as well as enterprises, politi- cians and governments are very interested in analyzing them, especially to learn more about people’s opinions. This is due to many reasons: to improve services and prod- ucts, to know the public opinion, to avoid terrorism acts. For researchers, opinion analysis is a multifaceted prob- lem, with different names according to the studied aspect: subjectivity analysis (Wiebe et al., 2004), opinion mining (Pang and Lee, 2008), sentiment extraction (Das and Chen, 2007). While the subject is mature, as proved by the many published surveys (Pang and Lee, 2008; Liu, 2012), there is still room for improvement, demonstrated by the interest for the yearly NLP conferences and workshops as SemEval or Wassa(Mohammad and Bravo-Marquez, 2017) and by the challenges that opinion analysis still offers (Breck and Cardie, 2016; Mohammad, 2016). Aspect Based Sentiment Analysis (ABSA) aims at ”deter- mining the orientation of sentiment expressed on each as- pect” (Liu, 2012). ABSA was introduced as a shared task for the first time in SemEval-2014 (Pontiki et al., 2014), with datasets in English language for two domains: laptops and restaurants. In SemEval-2015, the task was repeated and extended, adding the hotel domain with a dataset of whole reviews and not just isolated sentences (Pontiki et al., 2015). In SemEval-2016 (Pontiki et al., 2016), new and multilingual datasets were provided: restaurant reviews in six languages (English, French, Dutch, Russian, Spanish and Turkish), hotel reviews in Arabic, consumer electronics reviews in three languages (English, Dutch and Chinese), telecom reviews in Turkish and museum reviews in French. In this paper we propose an annotated French dataset for different domains compared with SemEval …

Content truncated to 500 words