Learning Geometric Word Meta-Embeddings
This work addresses the challenge of integrating multiple word embedding sources for natural language processing tasks, but it is incremental as it builds on existing meta-embedding methods with geometric refinements.
The paper tackles the problem of combining different word embeddings by proposing a geometric framework that transforms embeddings into a common latent space using orthogonal rotations and Mahalanobis scaling, resulting in improved performance on word similarity and analogy benchmarks.
We propose a geometric framework for learning meta-embeddings of words from different embedding sources. Our framework transforms the embeddings into a common latent space, where, for example, simple averaging of different embeddings (of a given word) is more amenable. The proposed latent space arises from two particular geometric transformations - the orthogonal rotations and the Mahalanobis metric scaling. Empirical results on several word similarity and word analogy benchmarks illustrate the efficacy of the proposed framework.