Skip to main content

​
Overview

The /siaa/keywords/<nombre_doc> endpoint returns the TF-IDF (Term Frequency-Inverse Document Frequency) keywords that have been automatically extracted from a specified document. This is useful for debugging routing issues and understanding which terms the system considers most relevant for each document.

​
Endpoint

GET /siaa/keywords/<nombre_doc>

​
Parameters

​
nombre_doc
string
required
Document filename (case-insensitive). Spaces can be encoded as + or %20.Example: acuerdo_no._psaa16-10476.md

​
Response

​
documento
string
The normalized document name
​
keywords
array of strings
List of top TF-IDF keywords for this document, ordered by relevance score. The system extracts up to 20 keywords per document.

​
Error Response

If the document is not found:
​
error
string
Error message indicating the document was not found
Status code: 404

​
Example

​
Request

curl http://localhost:5000/siaa/keywords/acuerdo_no._psaa16-10476.md

​
Response

{
  "documento": "acuerdo_no._psaa16-10476.md",
  "keywords": [
    "sierju",
    "periodicidad",
    "reportar",
    "formulario",
    "recoleccion",
    "discrepancia",
    "roles",
    "funcionario",
    "responsable",
    "sancion",
    "incumplimiento",
    "disciplinario",
    "estadistica",
    "inventario",
    "magistrado",
    "administrador",
    "quinto",
    "habil",
    "carga",
    "diligenciar"
  ]
}

​
Use Cases

​
Debugging Document Routing

When a user query isn’t finding the expected document, check the keywords to understand what terms the routing algorithm associates with that document: 
curl http://localhost:5000/siaa/keywords/acuerdo_pcsja19-11207.md
If the expected keywords are missing, you may need to add manual keywords in the KEYWORDS_MANUALES configuration.

​
Understanding Document Content

Quickly understand the main topics covered in a document without reading the full text: 
curl http://localhost:5000/siaa/keywords/guia_civil_municipal.md

​
Validating TF-IDF Extraction

After updating documents or changing the tokenization rules, verify that important terms (including alphanumeric codes like “psaa16” or “10476”) are being extracted correctly.

​
Notes

  • Keywords are automatically generated during document loading using TF-IDF scoring
  • The system filters out stopwords and short tokens (less than 3 characters)
  • Alphanumeric terms with letters are always included (e.g., “psaa16”, “art5”)
  • Pure numeric terms are included only if they have 4+ digits (e.g., “10476”, “2016”)
  • Keywords are cached until documents are reloaded

Build docs developers (and LLMs) love

Get started for freeTalk to us