COMFO: Multilingual Corpus for Opinion Mining

The use of Machine Learning (ML) algorithms in opinion mining, particularly supervised learning algorithms, requires an annotated corpus to train the classification model in order to predict results that are close to reality. Unfortunately, there are still no resources for the automatic processing of textual data expressed in the Senegalese urban language. The objective of this paper is to build a multilingual corpus for opinion mining (COMFO). The process of building theCOMFOcorpus is composed of three steps: presentation of the data source, data collection and preparation, and annotation by lexical approach. The particularity of COMFO lies in the integration of foreign languages (French and English) and local languages, particularly urbanWolof, in order to reflect the collective opinion of Senegalese readers.