Skip to main content

ZQ Classify

Configure and run ZettaQuant SEAL (Secure, Efficient, Accurate and Low-Cost) to analyze financial text directly inside Snowflake.

This page is where you define what to analyze, run the pipeline, and review results.

Overview​

ZQ Classify is ZettaQuant’s core analysis engine. It processes financial documents and assigns structured labels to sentences using AI models running natively inside Snowflake.

It enables you to:

  • Identify relevant content
  • Classify policy, sentiment, and risk signals
  • Generate structured outputs for downstream analytics

Pipeline Features​

  • Advanced AI Capabilities
    High-precision relevancy detection and structured financial text classification.

  • Batch Processing
    Efficiently process large document collections across your configured dataset.

  • Incremental Analysis
    Run analysis on newly ingested documents without reprocessing historical data.

  • Structured Outputs
    Results are written directly to Snowflake tables for reporting and metrics.

Workflow​

Before running ZQ Classify, ensure the following prerequisites are completed:

Prerequisites​

  1. Configure your dataset in Data Configuration
  2. Apply required Access Grants

Run Analysis​

  1. Create analysis topics
  2. Click Run to start the pipeline
  3. Review results and proceed to Metrics

Creating Topics​

Each pipeline run is driven by one or more topics.

A topic includes:

  • Topic name
  • Analysis mode
  • Description
  • Optional labels (for classification)

Topics define how documents are interpreted during analysis.

Analysis Modes​

Relevancy​

Determines whether each sentence is related to a topic.

Output:

  • RELEVANT
  • IRRELEVANT

Use for:

  • Event detection
  • Topic discovery
  • Content filtering
Relevancy Mode Configuration

Classification​

Assigns structured labels to sentences.

Labels are defined when creating the topic.

Use for:

  • Sentiment analysis
  • Policy stance detection
  • Risk categorization
Classification Mode Configuration

Running the Pipeline​

After configuring topics:

  1. Choose the topics to run
  2. Click Run

The system provisions compute resources and processes documents automatically.

Execution status is visible in the interface, including:

  • Pending
  • Running
  • Completed
Pipeline Run Status

Results​

After a pipeline run completes, ZQ Classify displays:

Summary​

  • Total Topics
  • Completed runs
  • Pending / Running jobs

Results Table​

Displays per-topic outputs:

  • Analysis mode
  • Total sentences
  • Relevant / labeled counts

All results are stored in Snowflake and used by the Metrics module.

Pipeline Results and Summary

Best Practices​

  • Start with a limited number of topics before scaling
  • Use precise topic descriptions
  • Validate access grants before running large jobs
  • Monitor execution in Telemetry for long-running workloads

For large datasets, ensure sufficient warehouse size and compute pool capacity.

Troubleshooting​

  • Pipeline failures → Check execution details in Telemetry & Logs

  • Performance issues → Review warehouse size and compute pool status

  • Permission issues → Confirm required data access grants are applied

  • No results generated → Ensure the dataset is configured correctly and contains valid documents

  • Unexpected labels → Refine topic definitions and classification structure

Next Steps​

After running ZQ Classify:

  1. Metrics - Create document-level indicators
  2. Telemetry - Monitor execution and logs
  3. Data Configuration - Change dataset