Adaptive cross-contextual word embedding for word polysemy with unsupervised topic modeling

Because of its efficiency, word embedding has been widely used in many natural language processing and text modeling tasks. It aims to represent each by a vector so such that the geometry between these vectors can capture semantic correlations words. An ambiguous often have diverse meanings different contexts, quality which is called polysemy. The bulk studies aimed generate only one single for word, whereas few made small number embeddings present word. However, it hard determine exact senses as depend on contexts. To address this problem, paper proposes novel adaptive cross-contextual (ACWE) method capturing polysemy contexts based topic modeling, defined over latent interpretable space. proposed ACWE consists two main parts, first an unsupervised probabilistic model designed obtain global embeddings, represented unified Based process then devised second part learn local polysemous In fact, adaptively adjusted updated with respect tailored corresponding validated datasets collected from Wikipedia IMDb tasks including similarity, induction, interpretability, classification. Experimental results indicate does not outperform established methods, consider six popular benchmark datasets, but also yields competitive performance compared state-of-the-art deep learning-based approaches without considering Moreover, significantly improves performances classification both precision F1, visualizations semantics words demonstrate feasibility advantage