skip to main content
US FlagAn official website of the United States government
dot gov icon
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
https lock icon
Secure .gov websites use HTTPS
A lock ( lock ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Explore Research Products in the PAR
It may take a few hours for recently added research products to appear in PAR search results.


Title: Improving Neural Topic Models using Knowledge Distillation
Abstract: Topic models are often used to identify human-interpretable topics to help make sense of large document collections. We use knowledge distillation to combine the best attributes of probabilistic topic models and pretrained transformers. Our modular method can be straightforwardly applied with any neural topic model to improve topic quality, which we demonstrate using two models having disparate architectures, obtaining state-of-the-art topic coherence. We show that our adaptable framework not only improves performance in the aggregate over all estimated topics, as is commonly reported, but also in head-to-head comparisons of aligned topics.  more » « less
Award ID(s):
2031736
PAR ID:
10225167
Author(s) / Creator(s):
; ;
Date Published:
Journal Name:
Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)
Page Range / eLocation ID:
1752 to 1771
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ; ; ; (, WWW '22: Proceedings of the ACM Web Conference 2022)
    Researchers using social media data want to understand the discussions occurring in and about their respective fields. These domain experts often turn to topic models to help them see the entire landscape of the conversation, but unsupervised topic models often produce topic sets that miss topics experts expect or want to see. To solve this problem, we propose Guided Topic-Noise Model (GTM), a semi-supervised topic model designed with large domain-specific social media data sets in mind. The input to GTM is a set of topics that are of interest to the user and a small number of words or phrases that belong to those topics. These seed topics are used to guide the topic generation process, and can be augmented interactively, expanding the seed word list as the model provides new relevant words for different topics. GTM uses a novel initialization and a new sampling algorithm called Generalized Polya Urn (GPU) seed word sampling to produce a topic set that includes expanded seed topics, as well as new unsupervised topics. We demonstrate the robustness of GTM on open-ended responses from a public opinion survey and four domain-specific Twitter data sets. 
    more » « less
  2. ; ; (, Proceedings of the AAAI Conference on Artificial Intelligence)
    Dynamic topic models (DTM) are commonly used for mining latent topics in evolving web corpora. In this paper, we note that a major limitation of the conventional DTM based models is that they assume a predetermined and fixed scale of topics. In reality, however, topics may have varying spans and topics of multiple scales can co-exist in a single web or social media data stream. Therefore, DTMs that assume a fixed epoch length may not be able to effectively capture latent topics and thus negatively affect accuracy. In this paper, we propose a Multi-Scale Dynamic Topic Model (MS-DTM) and a complementary Incremental Multi-Scale Dynamic Topic Model (IMS-DTM) inference method that can be used to capture latent topics and their dynamics simultaneously, at different scales. In this model, topic specific feature distributions are generated based on a multi-scale feature distribution of the previous epochs; moreover, multiple scales of the current epoch are analyzed together through a novel multi-scale incremental Gibbs sampling technique. We show that the proposed model significantly improves efficiency and effectiveness compared to the single scale dynamic DTMs and prior models that consider only multiple scales of the past. 
    more » « less
  3. ; (, COLING)
    Most previous work in unsupervised semantic modeling in the presence of metadata has assumed that our goal is to make latent dimensions more correlated with metadata, but in practice the exact opposite is often true. Some users want topic models that highlight differences between, for example, authors, but others seek more subtle connections across authors. We introduce three metrics for identifying topics that are highly correlated with metadata, and demonstrate that this problem affects between 30 and 50% of the topics in models trained on two real-world collections, regardless of the size of the model. We find that we can predict which words cause this phenomenon and that by selectively subsampling these words we dramatically reduce topic-metadata correlation, improve topic stability, and maintain or even improve model quality 
    more » « less
  4. ; ; ; (, Advances in Quantitative Ethnography: Second International Conference, ICQE 2020, Malibu, CA, USA, February 1-3, 2021, Proceedings)
    Ruis, Andrew; Lee, Seung B. (Ed.)
    When text datasets are very large, manually coding line by line becomes impractical. As a result, researchers sometimes try to use machine learning algorithms to automatically code text data. One of the most popular algorithms is topic modeling. For a given text dataset, a topic model provides probability distributions of words for a set of “topics” in the data, which researchers then use to interpret meaning of the topics. A topic model also gives each document in the dataset a score for each topic, which can be used as a non-binary coding for what proportion of a topic is in the document. Unfortunately, it is often difficult to interpret what the topics mean in a defensible way, or to validate document topic proportion scores as meaningful codes. In this study, we examine how keywords from codes developed by human experts were distributed in topics generated from topic modeling. The results show that (1) top keywords of a single topic often contain words from multiple human-generated codes; and conversely, (2) words from human-generated codes appear as high-probability keywords in multiple topic. These results explain why directly using topics from topic models as codes is problematic. However, they also imply that topic modeling makes it possible for researchers to discover codes from short word lists. 
    more » « less
  5. ; (, Proceedings of the Fourteenth International AAAI Conference on Web and Social Media)
    Understanding how political attention is divided and over what subjects is crucial for research on areas such as agenda setting, framing, and political rhetoric. Existing methods for measuring attention, such as manual labeling according to established codebooks, are expensive and can be restrictive. We describe two computational models that automatically distinguish topics in politicians' social media content. Our models---one supervised classifier and one unsupervised topic model---provide different benefits. The supervised classifier reduces the labor required to classify content according to pre-determined topic list. However, tweets do more than communicate policy positions. Our unsupervised model uncovers both political topics and other Twitter uses (e.g., constituent service). These models are effective, inexpensive computational tools for political communication and social media research. We demonstrate their utility and discuss the different analyses they afford by applying both models to the tweets posted by members of the 115th U.S. Congress. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email