{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/source-lda-enhancing-probabilistic-topic","title":"Source-LDA: Enhancing probabilistic topic models using prior knowledge sources","arxiv_id":"1606.00577","date":"2016-06-02","proceeding":null,"authors":["Justin Wood","Patrick Tan","Wei Wang","Corey Arnold"],"abstract":"A popular approach to topic modeling involves extracting co-occurring n-grams\nof a corpus into semantic themes. The set of n-grams in a theme represents an\nunderlying topic, but most topic modeling approaches are not able to label\nthese sets of words with a single n-gram. Such labels are useful for topic\nidentification in summarization systems. This paper introduces a novel approach\nto labeling a group of n-grams comprising an individual topic. The approach\ntaken is to complement the existing topic distributions over words with a known\ndistribution based on a predefined set of topics. This is done by integrating\nexisting labeled knowledge sources representing known potential topics into the\nprobabilistic topic model. These knowledge sources are translated into a\ndistribution and used to set the hyperparameters of the Dirichlet generated\ndistribution over words. In the inference these modified distributions guide\nthe convergence of the latent topics to conform with the complementary\ndistributions. This approach ensures that the topic inference process is\nconsistent with existing knowledge. The label assignment from the complementary\nknowledge sources are then transferred to the latent topics of the corpus. The\nresults show both accurate label assignment to topics as well as improved topic\ngeneration than those obtained using various labeling approaches based off\nLatent Dirichlet allocation (LDA).","url_abs":"http://arxiv.org/abs/1606.00577v3","url_pdf":"http://arxiv.org/pdf/1606.00577v3.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"source-lda-enhancing-probabilistic-topic","repo_url":"https://github.com/ucla-scai/Source-LDA","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":0,"framework":"none","reach":null}],"tasks":[{"task_slug":"topic-models","task_name":"Topic Models"}],"methods":[],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}