playbooks

home / skills / dkyazzentwatwa / chatgpt-skills / topic-modeler

topic-modeler skill

/topic-modeler

This skill helps you extract topics from text collections using LDA, generate keywords, classify documents, and visualize results for insights.

npx playbooks add skill dkyazzentwatwa/chatgpt-skills --skill topic-modeler

Review the files below or copy the command above to add this skill to your agents.

Files (3)
SKILL.md
749 B
---
name: topic-modeler
description: Extract topics from text collections using LDA (Latent Dirichlet Allocation) with keyword extraction and topic visualization.
---

# Topic Modeler

Extract topics from text collections using LDA.

## Features

- **LDA Topic Modeling**: Latent Dirichlet Allocation
- **Topic Keywords**: Extract representative keywords per topic
- **Document Classification**: Assign documents to topics
- **Visualization**: Topic word clouds and distributions
- **Coherence Scores**: Evaluate topic quality

## CLI Usage

```bash
python topic_modeler.py --input documents.csv --column text --topics 5 --output topics.json
```

## Dependencies

- gensim>=4.3.0
- nltk>=3.8.0
- pandas>=2.0.0
- matplotlib>=3.7.0
- wordcloud>=1.9.0

Overview

This skill extracts coherent topics from collections of text using Latent Dirichlet Allocation (LDA) and provides keyword summaries, document assignments, and visualizations. It bundles preprocessing, keyword extraction, coherence scoring, and plot-ready outputs so you can turn raw text into actionable topic insights quickly. The tool is implemented in Python and designed for reproducible, auditable topic analysis.

How this skill works

The skill tokenizes and preprocesses documents, builds a dictionary and corpus, then fits an LDA model to discover latent topics. It returns top keywords per topic, assigns topic probabilities to each document, and computes coherence scores to help you evaluate model quality. Optional visualization modules produce word clouds and topic distribution plots for interpretation and reporting.

When to use it

  • Exploring themes in large document collections (news, reviews, support tickets).
  • Summarizing unlabeled text to guide annotation or taxonomy creation.
  • Clustering documents by dominant topics for routing or prioritization.
  • Monitoring topic drift over time in feeds or social media.
  • Evaluating different topic counts and preprocessing choices using coherence.

Best practices

  • Clean and normalize text (lowercase, remove stopwords, lemmatize) before modeling.
  • Try several topic counts and compare coherence scores to select the best K.
  • Inspect top keywords and sample documents per topic to validate interpretability.
  • Filter out extremely rare and extremely common tokens from the dictionary.
  • Seed the random state for reproducible LDA results when iterating.

Example use cases

  • Analyze customer support tickets to surface common complaint categories and route them to teams.
  • Process product reviews to identify recurring praise and pain points for prioritization.
  • Summarize weekly news articles into topic buckets for an editorial briefing.
  • Group internal documents by theme to accelerate knowledge discovery during onboarding.
  • Track evolving discussion topics on social channels to inform marketing campaigns.

FAQ

What inputs does the skill expect?

A CSV or tabular file with a text column; you specify the column name. The script also accepts plain lists of documents when used as a library.

How do I choose the number of topics?

Run models across a range of topic counts and compare coherence scores and human interpretability to pick the best balance.

What visualizations are provided?

Word clouds for each topic and topic distribution plots across documents; outputs are matplotlib and wordcloud objects or image files.