Corpus Manipulation Commands

Corpus Manipulation Commands

The crispt tool allows fine-grained control over the "Corpus" object—the central data structure in CRISP-T. You can use it to inspect, debug, and manually edit your data.

General Syntax:

crispt --inp <corpus_path> [action] --out <output_path>

Inspecting the Corpus

Print Structure

View a readable summary of your corpus.

crispt --inp ./corpus --print "documents 10"

• Formats:
  • --print documents: Show document summaries.
  • --print metadata: Show corpus-level metadata.
  • --print "documents 5": First 5 docs.
  • --print "dataframe stats": Stats of the numeric data.

Inspect DataFrame

View details about the structured numeric data (CSV).

• --df-cols: List all column names.
• --df-row-count: Total number of rows.
• --df-row <index>: View a specific row by its 0-based index.

Inspect Documents

• --doc-ids: List all document IDs.
• --doc-id <ID>: View full details (text + metadata) for a specific document.

Inspect Relationships

• --relationships: List all defined links between data points.
• --relationships-for-keyword <keyword>: Show links involving a specific word.

Export Metadata

Export the internal metadata of your corpus (e.g., from semantic search) as a DataFrame.

crispt --inp ./corpus --metadata-df --out ./exported

• --metadata-keys "key1,key2": Specific keys to include.

Editing the Corpus

Create New Corpus

Start a fresh corpus without importing files.

crispt --id my_new_corpus --name "My Study" --out ./new_corpus

Add/Remove Documents

Manually add or remove text.

• Add: crispt --doc "id|title|text_content"
  • Example: crispt --doc "101|Interview 1|The participant said..."
• Remove: crispt --remove-doc <id>

Edit Metadata

Add custom metadata fields to the corpus object.

crispt --meta "project_phase=pilot"

Relational Editing

• Add Link: crispt --add-rel "first|second|relation"
• Clear All Links: crispt --clear-rel

(See Data Linking Commands for automated linking)