Neural Vector Conceptualization for Word Vector Space Interpretation
This addresses the interpretability challenge in NLP models for researchers and practitioners, though it appears incremental as it builds on prior approaches by operating in the original space and handling non-linearities.
The paper tackles the problem of interpreting distributed word vector spaces in NLP by introducing a neural model that conceptualizes word vectors, activating higher-order concepts and learning non-linear relations, resulting in less entropic concept activation profiles compared to cosine similarity.
Distributed word vector spaces are considered hard to interpret which hinders the understanding of natural language processing (NLP) models. In this work, we introduce a new method to interpret arbitrary samples from a word vector space. To this end, we train a neural model to conceptualize word vectors, which means that it activates higher order concepts it recognizes in a given vector. Contrary to prior approaches, our model operates in the original vector space and is capable of learning non-linear relations between word vectors and concepts. Furthermore, we show that it produces considerably less entropic concept activation profiles than the popular cosine similarity.