Based on the assumption that semantic relatedness between documents is reflected in the distribution of the vocabulary, topic models are a widely used class of techniques for text analysis tasks. The application of topic models results in concepts, the so-called topics, and a high-dimensional description of the documents. For visualization tasks, they can be projected onto a lower-dimensional space using dimensionality reduction techniques. Though the quality of the resulting point layout mainly depends on the chosen topic model and dimensionality reduction technique, it is unclear which particular combinations are suitable for displaying the semantic relatedness between the documents. In this work, we propose a benchmark comprising various datasets, layout algorithms and their hyperparameters, and quality metrics for conducting an empirical study.
Paper
Full text
A Benchmark for the Use of Topic Models for Text Visualization Tasks
Semantic Scholar · Computer Science · 2022
Abstract
Based on the assumption that semantic relatedness between documents is reflected in the distribution of the vocabulary, topic models are a widely used class of techniques for text analysis tasks. The application of topic models results in concepts, the so-called topics, and a high-dimensional description of the documents. For visualization tasks, they can be projected onto a lower-dimensional space using dimensionality reduction techniques. Though the quality of the resulting point layout mainly depends on the chosen topic model and dimensionality reduction technique, it is unclear which particular combinations are suitable for displaying the semantic relatedness between the documents. In this work, we propose a benchmark comprising various datasets, layout algorithms and their hyperparameters, and quality metrics for conducting an empirical study.