Embedding Words in Non-Vector Space with Unsupervised Graph Learning
This work addresses the limitation of standard vector-based word embeddings for natural language processing by proposing a graph-based alternative, offering potential improvements in tasks like similarity and analogy, though it is incremental as it builds on existing methods like GloVe.
The paper tackles the problem of representing words in a non-vector space by introducing GraphGlove, an unsupervised graph-based method that learns word embeddings as nodes in a weighted graph, using shortest path distances. It shows that this approach substantially outperforms vector-based methods like GloVe on word similarity and analogy tasks, with analysis revealing hierarchical structures similar to WordNet.
It has become a de-facto standard to represent words as elements of a vector space (word2vec, GloVe). While this approach is convenient, it is unnatural for language: words form a graph with a latent hierarchical structure, and this structure has to be revealed and encoded by word embeddings. We introduce GraphGlove: unsupervised graph word representations which are learned end-to-end. In our setting, each word is a node in a weighted graph and the distance between words is the shortest path distance between the corresponding nodes. We adopt a recent method learning a representation of data in the form of a differentiable weighted graph and use it to modify the GloVe training algorithm. We show that our graph-based representations substantially outperform vector-based methods on word similarity and analogy tasks. Our analysis reveals that the structure of the learned graphs is hierarchical and similar to that of WordNet, the geometry is highly non-trivial and contains subgraphs with different local topology.