Tags

Organizing and categorizing content with hierarchical tags

Tags in Clarifeye provide a hierarchical categorization system for your document content. They enable structured classification of text chunks with reasoning and evidence.

Tag can either be at:

• The document level
• The chunk level

Overview

Tags allow you to:

• Create hierarchical taxonomies for document classification
• Tag content with explanations and supporting quotes
• Build searchable category trees
• Organize knowledge by domain-specific classifications

Chunk tags

Tag structure

A chunk tag is defined as a tagging tree. For example you can create a tree like this:

{
  "id": "technical-positions",
  "name": "Technical Positions",
  "children": [
    {
      "id": "infra",
      "name": "Infrastructure",
      "children": [
        {
          "id": "sre",
          "name": "SRE",
          "children": [],
          "description": "Job focus on SRE"
        }
      ],
      "description": "Jobs related to the management of infrastructure"
    }
  ],
  "description": "Categories of technical job positions in the software industry"
}

Thanks to that structure, clarifeye will be able to process recursively and classify chunks into the appropriate tags.

UI creation

You can create tags through the UI, to do so go to the extractors tab and create a new tag extractor. On the extractor you will be able to configure:

• the tagging tree
• the LLM model to use
• custom extraction instructions for the LLM.
• the version of chunks to use.

API creation

You can also create tags through the API, to do so you will need to leverage the table apis to retrieve data from the chunks table and then write data to the tag table.

Document tags

Document tags are a way to tag documents with a hierarchical categorization system.

Tag structure.

Document tag have the following structure:

{
  "tags": [
    {
      "name": "DocumentType",
      "description": "The type or category of the document.",
      "possible_values": [
        "Article",
        "Report",
        "Contract",
        "Policy",
        "Procedure",
        "Guideline",
        "Manual",
        "Instruction",
        "Study",
        "Memo",
        "Proposal",
        "Presentation",
        "Other"
      ],
      "allow_multiple_values": false,
      "allow_llm_to_infer": true
    },
    {
      "name": "Language",
      "description": "The primary language used in the document.",
      "possible_values": [
        "English",
        "Spanish",
        "French",
        "German",
        "Chinese",
        "Other"
      ],
      "allow_multiple_values": false,
      "allow_llm_to_infer": false
    },
    {
      "name": "Topic",
      "description": "Subject of the document.",
      "possible_values": null,
      "allow_multiple_values": true,
      "allow_llm_to_infer": true
    }
  ]
}

UI creation

You can create document tags through the UI, to do so go to the tasks tab and create a new document tagging task. On the document tagging task you will be able to configure:

• the tagging tree
• the version of documents to use.
• the LLM model to use
• custom extraction instructions for the LLM.
• the version of chunks to use.

API creation

You can also create document tags through the API, to do so you will need to leverage the table apis to retrieve data from the documents table and then write data to the tag table.