Abstract
Natural Language Inference (NLI) is a hot topic research in natural languageprocessing, contradiction detection between sentences is a special case of NLI.This is considered a difficult NLP task which has a big influence when added asa component in many NLP applications, such as Question Answering Systems, textSummarization. Arabic Language is one of the most challenging low-resourceslanguages in detecting contradictions due to its rich lexical, semanticsambiguity. We have created a data set of more than 12k sentences and namedArNLI, that will be publicly available. Moreover, we have applied a new modelinspired by Stanford contradiction detection proposed solutions on Englishlanguage. We proposed an approach to detect contradictions between pairs ofsentences in Arabic language using contradiction vector combined with languagemodel vector as an input to machine learning model. We analyzed results ofdifferent traditional machine learning classifiers and compared their resultson our created data set (ArNLI) and on an automatic translation of both PHEME,SICK English data sets. Best results achieved using Random Forest classifierwith an accuracy of 99%, 60%, 75% on PHEME, SICK and ArNLI respectively.