The impact of summarisation on textual entailment - a case study on global warming arguments

The challenging task of recognizing textual entailment aims to check whether the meaning of a smaller text - the hypothesis h - can be inferred from another text T. Current methods interleave natural language processing, machine learning, search and lexical resources. All these instruments pose computational challenges that make textual entailment unfeasible for large texts. Hence, we investigate how textual entailment is affected by text summarization. By summarising the text T we expect a decrease of accuracy, but an increase of computation speed. We aim to assess the expected decrease in accuracy caused by summarisation against time benefits due to smaller text given to entailment machinery. Our results show that the time needed for computing entailment is decreased four times, while the accuracy decreases with two percentages.