/Geometry-of-Compositionality

Word embedding based approach for textual compositionality

Primary LanguagePython

Geometry-of-Compositionality

This is Bi-Context datasets used in AAAI 2017 paper, Geometry of Compositionality. Files: (1) bicontext_English.txt: contains 104 English phrases and 208 sentences. (2) bicontext_Chinese.txt: contains 64 Chinese phrases and 128 sentences.

Data Format: In each line, the target phrase and its sentence are separated by tab, i.e., "phrase sentence". The phrases listed in the datasets are polysemous, and we provide two sentences containing it. The target phrase has literal meaning (compositional meaning) in one sentence, and has idiomatic meaning (non-compositional meaning) in the other sentence.