Semantic similarity for sequenced shingles
Fumito Konaka, Takao Miura · 2015
In this investigation, we introduce new kinds of sentence similarity, called Euclid similarity and Levenshtein similarity, to capture both word sequences and semantic aspects. This is especially useful for Semantic Textual Similarity (STS) so that we could retrieve SNS texts, short sentences or something including collocations. We show the usefulness of our approach by some experimental results.