UWB@RuShiftEval Measuring Semantic Difference as per-word Variation in Aligned Semantic Spaces

Abstract

We present our system for measuring the lexical semantic change between corpora pairs, developed for the RuShiftEval competition. We measure semantic changes in a list of test words between three different corpora (for each pair), pre-Soviet, Soviet, and post-Soviet, each from a different time period. For each corpus, we train word embeddings, and we obtain linear transformations between the embedding spaces. Finally, we measure the similarity between the transformed vectors for each test word.