Experiments with language combinatorics in text classification: lessons learned and future implications

This paper presents a meta-analysis of experiments performed with language combinatorics (LC), a novel language model generation and feature extraction method based on combinatorial manipulations of sentence elements (e.g., words). Along recent years LC has been applied to a number of text classification tasks, such as affect analysis, cyberbullying detection or future reference extraction. We summarize two of the most extensive experiments and discuss general implications for future implementations of combinatorial language model.