Bag-of-Concepts Document Representation for Textual News Classification

Automatic classification of news articles is a relevant problem due to the large amount of news generated every day, so it is crucial that these news are classified to allow for users to access to information of interest quickly and effectively. Traditional classification systems represent documents as bag-of-words (BoW), which are oblivious to two problems of language: synonymy and polysemy. This paper shows the advantages of using a bag-of-concepts (BoC) representation of documents, which tackles synonymy and polysemy, in text news classification – using a Support Vector Machines algorithm. In order to create BoC representations, a Wikipedia-based semantic annotator is used. To evaluate the proposal we used a purpose-built corpus and the Reuters 21578 corpus. Results show that the efficiency of the BoC approach is very dependent on the performance of the semantic annotator in extracting concepts, which depends heavily on the characteristics of particular corpora, reaching performance increases up to 29.65%.