Evaluating Neural Word Representations in Tensor-Based Compositional Settings
This work provides empirical guidance for NLP practitioners on when to use neural vs. traditional word representations in compositional settings, though it is incremental as it compares existing methods rather than introducing new ones.
The study compared neural word embeddings against traditional co-occurrence vectors across multiple compositional tasks, finding that co-occurrence vectors remained competitive in constrained tasks like verb disambiguation and sentence similarity, while neural embeddings outperformed them in larger-scale tasks such as paraphrase detection and dialogue act tagging, showing robust performance.
We provide a comparative study between neural word representations and traditional vector spaces based on co-occurrence counts, in a number of compositional tasks. We use three different semantic spaces and implement seven tensor-based compositional models, which we then test (together with simpler additive and multiplicative approaches) in tasks involving verb disambiguation and sentence similarity. To check their scalability, we additionally evaluate the spaces using simple compositional methods on larger-scale tasks with less constrained language: paraphrase detection and dialogue act tagging. In the more constrained tasks, co-occurrence vectors are competitive, although choice of compositional method is important; on the larger-scale tasks, they are outperformed by neural word embeddings, which show robust, stable performance across the tasks.