On this page you will find STSS-65, which was the original Short Text Semantic Similarity dataset used in Li et Al. 2006 to evaluate the STASIS STSS measure.
On this page you will find STSS-131, the new benchmark dataset which is more representative of the English language and more demanding than STSS-65 and important for evaluating improved algorithms. This is an extract from a paper which is in press in ACM Transactions on Speech and Language Processing, estimated publication date December 2013. In due course the process to make it open access will be complete, for free download of the complete paper from the ACM TSLP website.
Arabic Word Semantic Similarity
I lead a research group which is extending Short Text Semantic Similarity to non-English languages. I have uploaded our Arabic Word Semantic Similarity benchmark dataset. This is intended to play the same role in Arabic as the Rubenstein & Goodenough dataset plays in English.
We hope you will find it useful for evaluating and comparing word semantic similarity measures for Arabic. The dataset is contained in the paper “Arabic Word Semantic Similarity” on my publications page.