Hierarchical topic model with an interpretable topic hierarchy

A system comprising:

• a memory component;
• and
• a processing device coupled to the memory component, the processing device to perform operations comprising:
• determining first-level topics in a hierarchical topic model (HTM) related to a corpus of documents, wherein a first-level topic of the first-level topics comprises multiple words;
• generating a number of clusters based on word embeddings of the multiple words;
• subdividing the multiple words into second-level topics comprising subtopics of the first-level topic, wherein a number of the second-level topics is equal to the number of clusters;
• assigning a document of the corpus of documents to the first-level topic and to a second-level topic of the second-level topics;
• receiving an indication of access to the document at a client computing device;
• and
• generating, using the HTM, custom content configured for the client computing device based on one or more other documents of the corpus of documents assigned to the first-level topic and the second-level topic.

A non-transitory computer-readable medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising:

• determining first-level topics in a hierarchical topic model (HTM) related to a corpus of documents;
• determining, based on hyperbolic word embeddings of words found in documents assigned to a first-level topic of the first-level topics, hierarchical relations between pairs of the words in the documents assigned to the first-level topic;
• determining statistical relations between the words and the documents assigned to the first-level topic;
• determining semantic relations between the pairs of the words in the documents assigned to the first-level topic;
• defining second-level topics as subtopics of the first-level topic in the HTM, based on the hierarchical relations, the statistical relations, and the semantic relations, wherein a number of the second-level topics is equal to a number of clusters determined using the hyperbolic word embeddings;
• and
• using the HTM to provide custom content for users based on access at a client computing device to documents in the corpus.