Automated Arabic Antonym Extraction Using a Corpus Analysis Tool

The automatic extraction of semantic relations between words from textual corpora is an extremely challenging task. The increasing need for language resources supporting Natural language processing (NLP) applications has encouraged the development of automated methods for the extraction of semantic relations between words. The use of corpus statistical and similarity distribution methods can help in the task of semantic relation extraction between pairs of words. In this paper, we present a pattern-based bootstrapping approach using Arabic language corpora and a corpus analysis tool (Sketch Engine) to extract the semantic relations (antonyms) between word pairs. The algorithm uses LogDice and pattern cooccurrence to classify the extracted pairs into antonyms. Results of evaluation show that our approach is able to extract the antonym relations with a precision of 76%.