Topic Metric

Topic metrics quantify the prominence and coherence of themes within textual data, offering a powerful alternative to sentiment analysis for understanding complex information structures. Current research focuses on improving topic model coherence using techniques like BERTopic and LLMs, developing specialized evaluation metrics for topic-controlled tasks such as summarization, and applying these metrics to diverse fields including political science and astronomy to identify research priorities and trends. These advancements enhance the ability to analyze large text corpora, providing valuable insights for scientific planning and improving the accuracy of natural language processing applications.

Papers

October 24, 2023

Beyond Sentiment: Leveraging Topic Metrics for Political Stance Classification
Weihong Qi
Sentiment Analysis Implicit Sentiment Opinion Mining Stance Classification Sentiment Word Political Stance Topic Metric

May 23, 2023

Contextualized Topic Coherence Metrics
Hamed Rahimi, Jacob Louis Hoover, David Mimno, Hubert Naacke, Camelia Constantin, Bernd Amann
Neural Topic Model Coherence Evaluation Topic Metric

June 9, 2022

Topic-Controllable Summarization: Topic-Aware Evaluation and Transformer Methods
Tatiana Passali, Grigorios Tsoumakas
Text Summarization Topic Analysis Topic Embeddings Controllable Summarization 3D Transformer Topic Metric

March 1, 2022

Determining Research Priorities for Astronomy Using Machine Learning
Brian Thomas, Harley Thronson, Anthony Buonomo, Louis Barbier
Machine Learning Astronomical Data Latent Dirichlet Allocation Topic Metric

Topic Metric

Papers

Beyond Sentiment: Leveraging Topic Metrics for Political Stance Classification

Contextualized Topic Coherence Metrics

Topic-Controllable Summarization: Topic-Aware Evaluation and Transformer Methods

Determining Research Priorities for Astronomy Using Machine Learning