AIEnrichmentNodeEditor

Overview

The AIEnrichmentNodeEditor is a user interface component in the Vantage analytics and data platform designed to facilitate the integration of AI-driven data enrichment processes into workflows. This node allows users to leverage various AI tasks, such as classification, sentiment analysis, entity extraction, and summarization, enabling enhanced data insights.

Purpose

The primary goal of the AIEnrichmentNodeEditor is to configure the parameters of an AI enrichment task within a data processing workflow. Users can select a specific task type, customize prompts, choose output columns, and determine batch sizes for processing data rows.

Settings

1. Task

• Name: task
• Input Type: Dropdown
• Description: Selects the type of AI enrichment task to perform. Options include:
  • Classify: Assign a category to each row.
  • Sentiment: Detect the tone (positive, negative, neutral).
  • Extract Entities: Pull out key figures (people, places, dates).
  • Summarize: Condense text into a summary.
  • Custom: Allow users to write their own prompt.
• Default Value: classify
• Effect of Changes: Changing the selected task updates the prompt template to a predefined template relevant to the chosen task.

2. Prompt Template

• Name: promptTemplate
• Input Type: Textarea
• Description: This customizable field defines the prompt that will be passed to the AI model, incorporating contextual elements from the dataset. Placeholders such as {{row}} and {{column_name}} can be used for dynamic content.
• Default Value: Based on the selected task, e.g., for “classify,” it defaults to "Classify this record into a single category. Respond with ONLY the category name.\n\nRecord: {{row}}".
• Effect of Changes: Modifying this field alters how the AI model interprets the data for processing.

3. Output Column

• Name: outputColumn
• Input Type: Text
• Description: Specifies the name of the new column where results from the AI enrichment will be stored. This column is added to each row of the dataset after processing.
• Default Value: ai_result
• Effect of Changes: Changing the output column name changes where the results will be recorded in the dataset.

4. Batch Size

• Name: batchSize
• Input Type: Numeric (range slider)
• Description: Defines how many rows will be processed per AI enrichment call. A higher number may lead to faster processing but might reduce granularity or precision.
• Default Value: 5
• Effect of Changes: Adjusting the batch size affects the speed and quality of processing, with larger batches yielding quicker results but potentially less detailed outcomes.

Data Expectations

The AIEnrichmentNodeEditor expects the following data formats:

• Rows: The platform should provide a structured dataset where rows contain text data for processing.
• Columns: Users should have the ability to reference existing columns, as indicated by the upstreamColumns detected from the selected node and edges.
• Contextual Data: Any prompt should logically incorporate relevant attributes from the current dataset to efficiently deliver insightful output.

AI Integrations

The AIEnrichmentNodeEditor integrates with multiple AI services or models that offer capabilities like:

• Text classification
• Sentiment analysis
• Entity recognition
• Summarization

These integrations require backend support for invoking AI models, processing responses, and applying results to the dataset.

Billing Impacts

Using the AI enrichment features may incur additional fees based on:

• The number of API calls made to AI models
• The computational resources used for processing (e.g., processing time, data volume) Users are advised to check their service agreements and analytics service costs to understand how this might affect billing.

Use Cases & Examples

Use Cases

1. Customer Feedback Analysis:

   • A company wants to analyze customer feedback from surveys to identify sentiment trends over time. By creating an AI enrichment node, the company can apply sentiment analysis to automatically categorize feedback as positive, negative, or neutral.

2. Social Media Posts Categorization:

   • A marketing team needs to categorize social media mentions into topics such as product inquiries, customer complaints, or general comments. By using the classify option in AIEnrichmentNodeEditor, the team can effectively automate this process.

3. Data Cleaning through Entity Extraction:

   • An organization needs to extract key entities from unstructured text in their dataset to facilitate better data cleaning and organization for analytics. The extract entities feature can streamline this task by pulling relevant names and dates directly from entries.

Example Configuration

Scenario: A customer service team aims to classify customer inquiries as "Complaint," "Inquiry," or "Praise".

Configuration Steps:

• Select Task: Choose classify from the task dropdown.
• Prompt Template: Leave it at the default value which states: "Classify this record into a single category. Respond with ONLY the category name.\n\nRecord: {{row}}".
• Output Column: Set as customer_inquiry_category.
• Batch Size: Select 10 to process a set of inquiries at once for faster results.

Sample Configuration Data:

{
  "task": "classify",
  "promptTemplate": "Classify this record into a single category. Respond with ONLY the category name.\n\nRecord: {{row}}",
  "outputColumn": "customer_inquiry_category",
  "batchSize": 10
}

This configuration will allow the system to classify each inquiry and store the results in a new column for further analysis.