This is the repo for our paper Knowledge-Aware Bayesian Deep Topic Model in NeurIPS2022.
In this paper, we propose a novel Bayesian-aware framework for incorporating knowledge graph into deep generative model. Knowledge graph contains a lot of side information and common sense, and holds great potential to improve the performance and interpretability of most purely data-driven models. However, most existing topic models mainly focus on the word co-occurrence patterns, ignoring such easy-to-obtain prior domain knowledge, such as topic hierarchies in WordNet. Or several knowledge-based topic models only applicable to shallow hierarchies or sensitive to the quality of the provided prior knowledge. To this end, we develop a novel deep ETM that jointly models the documents and the given prior knowledge by embedding the words and topics into the same space. Guided by the provided knowledge, the proposed model tends to discover topic hierarchies that are organized into interpretable taxonomies. Besides, with a technique for adapting a given graph, our extended version allows the provided prior topic structure to be finetuned to match the target corpus.
code is coming.