Geometry of Compositionality
This provides a computationally efficient tool for natural language processing tasks, though it is incremental as it builds on existing word embedding methods.
The paper tackles the problem of detecting context-specific compositionality of words or phrases using a simple geometric test based on word embeddings, achieving competitive state-of-the-art accuracy across multiple languages and phenomena like idiomaticity and sarcasm.
This paper proposes a simple test for compositionality (i.e., literal usage) of a word or phrase in a context-specific way. The test is computationally simple, relying on no external resources and only uses a set of trained word vectors. Experiments show that the proposed method is competitive with state of the art and displays high accuracy in context-specific compositionality detection of a variety of natural language phenomena (idiomaticity, sarcasm, metaphor) for different datasets in multiple languages. The key insight is to connect compositionality to a curious geometric property of word embeddings, which is of independent interest.