Skip to content

Data Analysis Commands

Jump to bottom
Bell Eapen edited this page Jan 29, 2026 · 1 revision

Data Analysis Commands

CRISP-T offers two main types of analysis: Text Analysis (NLP) for unstructured documents and Numeric/Structural Analysis for quantitative data and corpus structure.

1. Text Analysis (NLP)

These commands are run using crisp and operate on the documents in your corpus.

Topic Modeling (LDA)

Discover hidden topics in your text.

crisp --inp ./corpus --topics --num 5 --out ./analyzed
  • --topics: Run Latent Dirichlet Allocation (LDA).
  • --num <N>: Number of topics to find (default: 3).

Topic Assignment

Assign each document to its most relevant topic.

crisp --inp ./corpus --assign --out ./analyzed
  • --assign: Adds 'dominant_topic' metadata to each document.

Sentiment Analysis

Analyze the emotional tone of your text using VADER.

crisp --inp ./corpus --sentiment --sentence --out ./analyzed
  • --sentiment: Document-level sentiment scores (pos, neg, neu, compound).
  • --sentence: Sentence-level sentiment breakdown.

Summarization

Generate an extractive summary of your corpus.

crisp --inp ./corpus --summary --num 5
  • --summary: Extract key sentences.
  • --num <N>: Number of sentences to include in the summary.

Categories & Coding

Extract potential qualitative codes or themes.

crisp --inp ./corpus --cat --num 10
crisp --inp ./corpus --codedict --num 10
  • --cat: List common categories/themes.
  • --codedict: Generate a preliminary coding dictionary (Category, Property, Dimension).

Run Everything

Run all available NLP tasks at once.

crisp --inp ./corpus --nlp --out ./analyzed

2. Numeric & Structural Analysis

These commands analyze the relationships and patterns in your data over time or structure.

Topological Data Analysis (TDA)

Analyze the "shape" of your data using the Mapper algorithm (via crispt).

crispt --inp ./corpus --tdabm "outcome:var1,var2:radius"
  • Format: y_variable:x_variables:radius
  • Example: crispt --tdabm "satisfaction:age,income:0.3"
  • (Visualized with crispviz --tdabm)

Temporal Analysis

Analyze how your data changes over time (via crispt).

  • Summary: crispt --temporal-summary W (Weekly summary of event counts).
  • Sentiment Trends: crispt --temporal-sentiment M:mean (Monthly mean sentiment).
  • Topic Trends: crispt --temporal-topics W:5 (Top 5 topics per week).
  • Subgraphs: crispt --temporal-subgraphs M (Graph snapshots per month).

Period Codes: D (Day), W (Week), M (Month), Y (Year).

Clone this wiki locally