Identifying Sentiments in Algerian Code-switched User-generated Comments

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 2698–2705 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 2698 Identifying Sentiments in Algerian Code-switched User-generated Comments Wafia Adouane1, Samia Touileb2, and Jean-Philippe Bernardy1 1 Department of Philosophy, Linguistics and Theory of Science (FLoV), Centre for Linguistic Theory and Studies in Probability (CLASP), University of Gothenburg 2 Language Technology Group, Department of Informatics, University of Oslo 1 firstname.lastname@gu.se, 2samiat@ifi.uio.no Abstract We present in this paper our work on Algerian language, an under-resourced North African colloquial Arabic variety, for which we built a comparably large corpus of more than 36,000 code-switched user-generated comments annotated for sentiments. We opted for this data domain because Algerian is a colloquial language with no existing freely available corpora. Moreover, we compiled sentiment lexicons of positive and negative unigrams and bigrams reflecting the code-switches present in the language. We compare the performance of four models on the task of identifying sentiments, and the results indicate that a CNN model trained end-to-end fits better our unedited code-switched and unbalanced data across the predefined sentiment classes. Additionally, injecting the lexicons as background knowledge to the model boosts its performance on the minority class with a gain of 10.54 points on the F-score. The results of our experiments can be used as a baseline for future research for Algerian sentiment analysis. Keywords: Algerian Arabic, code-switching, user-generated data, sentiment analysis, under-resourced colloquial languages 1. Introduction Sentiment Analysis (SA) is a well-established NLP task which is commonly framed as a text classification prob- lem. SA includes various related applications, depending on the domain and its real-world use, such as product re- viewing and opinion mining. It has been applied to several domains, mostly for curated monolingual data. Recently, however, the focus has been extended to new domains and settings, namely to user-generated data which reflects real- world use cases. In this paper, we focus on identifying and classifying senti- ments by analysing user-generated comments on YouTube videos. We work with comments written in Algerian Ara- bic, a non-standardised Arabic variety characterised by heavy code-switching between co-existing languages and language varieties mainly Modern Standard Arabic (MSA), Berber, French, and local Arabic variants. There is a large body of work on SA for Arabic, largely for MSA and Middle Eastern Arabic varieties (for exam- ple (Rushdi-Saleh et al., 2011; Abdul-Mageed and Diab, 2012; Zbib et al., 2012; Aly and Atiya, 2013; Nabil et al., 2015; ElSahar and El-Beltagy, 2015; Salameh et al., 2015)). Nevertheless, there has been less work done for Northern African Arabic varieties, which are indeed colloquial lan- guages, due primarily to the scarcity of written linguistic resources. However, new social media platforms made it possible to obtain user-generated data reflecting real use of such lan- guages. In interactive communication channels, users use speech-like languages to express themselves with sponta- neous spelling for non-standardised languages. Hence this domain is potentially a useful resource for analysing the linguistic properties of this kind of unedited …

Content truncated to 500 words