Skip to content

Latest commit

 

History

History
87 lines (60 loc) · 3.44 KB

llm-prompt-guide.md

File metadata and controls

87 lines (60 loc) · 3.44 KB

AI Column Extraction Feature

Documentation for AI-powered text extraction and transformation functionality

Overview

The AI Column Extraction feature allows users to create new columns by processing existing text data using AI language models. Users can extract specific information, transform content, or generate new insights based on source text.

Demo Video

AI-FeatureDemo-480p.mov

Configuration

Basic Settings

  1. New Column Name: Define the name for the column that will store the AI-generated results
  2. LLM Provider: Select the configured AI provider with pre-defined parameters - Model, Server URL, Temperature, Max Tokens
  3. Response Format: Choose the output structure
Response format Description
`Text Plain text response
`JSON Schema Structured response following a defined schema
`JSON Object Free-form JSON response

Input Configuration

  1. Description: Detailed instructions for the AI model about the desired extraction or transformation
  2. JSON Schema (Optional): Required when Response Format is set to JSON Schema. Defines the structure and validation rules for the output

JSON Schema generator tools

  1. Liquid Technologies JSON to Schema converter
  2. jsonschema.net
  3. easy-json-schema.github.io

Note: There are few manual edits that may be required. Refer to the JSON schema in the example below.

Preview Function

Shows sample processing using the first row of data Displays:

  1. Input Value: Original text from source column
  2. Response: AI-generated result based on configuration

Allows validation before processing entire dataset

Examples

Text Response Examples

Use Case Instruction
Basic Summary Generate a one-sentence summary of the input text
Language Translation Translate the text to German

JSON Schema Examples

Use Case Instruction JSON Schema
Person information extraction Extract in JSON format details of the person mentioned in the content { "name": "individual_schema", "schema": { "type": "object", "properties": { "name": { "type": "string", "description": "The name of the individual." }, "dateofbirth": { "type": "string", "description": "The date of birth of the individual." }, "placeofbirth": { "type": "string", "description": "The place where the individual was born." }, "dateofdeath": { "type": "string", "description": "The date of death of the individual." }, "placeofdeath": { "type": "string", "description": "The place where the individual died." } }, "required": [ "name", "dateofbirth", "placeofbirth", "dateofdeath", "placeofdeath" ], "additionalProperties": false }, "strict": true }

JSON object examples

Use Case Instruction
Content classification Classify the content and extract key entities, the response should be in JSON format

Use Cases

Content Transformation

  1. Summarization
  2. Translation
  3. Style conversion
  4. Format standardization

Information Extraction

  1. Entity recognition
  2. Key fact extraction
  3. Timeline creation
  4. Relationship mapping

Content Analysis

  1. Sentiment analysis
  2. Theme identification
  3. Category classification
  4. Key concept extraction