Short Text Classification of Buyer-Initiated Questions in Online Auctions: A Score Assigning Method

Classification of short text (SMS, reviews, feedback, etc.) presents a unique set of challenges compared to classic text classification. Short texts are characterized by cryptic constructions, poor spelling, improper grammar, etc. that makes the application of traditional methods difficult. Proper classification enables us to use this information for further action. We study this problem in the context of online auctions. The paper presents a score assigning approach which outperforms traditional methods (e.g. Naive Bayes) in terms of accuracy.