A New Method to Identify Short-Text Authors Using Combinations of Machine Learning and Natural Language Processing Techniques

摘要

Identifying authors by their style of writing is a very challenging task. This problem has several applications, one of which is to identify fake online reviews written by spam accounts. The existence of such fake reviews degrades the credibility of the whole review collection, hence these fake reviews should be identified and removed. This process, however, needs to be automated since it is impossible to perform it manually in large review collections. Current authorship identification approaches identify authors based on large-scale texts such as documents. For this reason, these methods do not scale well to short texts such as online reviews that have limited features to learn from. This paper introduces a new method of author identification in short texts using combinations of machine learning algorithms and natural language processing techniques. The experiments we conducted on Yelp reviews gave promising results.