Lexical Entailment is the process of deriving relationships between a pair of given words that satisfy certain properties. It aims to identify the semantic relation- ships between words, which can be broadly explained as a hyponym-hypernym relationship. This project achieves the following:
- A thorough exploration of the various techniques present in monolingual lexical entailment.
- An explanation of the technique that we use for preprocessing word vectors to emphasize lexical entailment.
- A expansion of the above technique to be able to handle cross lingual lexical entailment by pre processing bi lingual word spaces.
- The code can perform cross lingual lexical entailment across 4 different languages as of May 2020 (English, Russian, Italian, German)
An important point to note here is that depending on which languages you want to perform cross lingual lexical entailment on, you need to ensure that you have the relevant constraint condition files (Dictionary translations etc.) set up. The report that details the entire project can be found here