BabelEnconding at SemEval-2020 Task 3: Contextual Similarity as a Combination of Multilingualism and Language Models
This addresses the challenge of graded contextual similarity for NLP researchers, but it is incremental as it builds on existing multilingual and language model techniques.
The paper tackled the problem of predicting contextual word similarity by proposing an approach using translation and multilingual language models, achieving top-3 rankings in six out of eight task/language combinations and being the highest scorer three times.
This paper describes the system submitted by our team (BabelEnconding) to SemEval-2020 Task 3: Predicting the Graded Effect of Context in Word Similarity. We propose an approach that relies on translation and multilingual language models in order to compute the contextual similarity between pairs of words. Our hypothesis is that evidence from additional languages can leverage the correlation with the human generated scores. BabelEnconding was applied to both subtasks and ranked among the top-3 in six out of eight task/language combinations and was the highest scoring system three times.