ISP Classifier

The ISP Classifier automatically categorizes customer complaints into diagnostic codes for faster resolution.

Classification Workflow

flowchart TD
    A[Customer Complaint] --> B[Extract Keywords]
    B --> C{Keyword Match?}
    C -->|Yes| D[Return ISP Code]
    C -->|No| E[LLM Fallback]
    E --> F[Generate Classification]
    F --> G[Return Code + Reason]
    D --> H[Log & Route]
    G --> H

    style A fill:#e3f2fd
    style D fill:#c8e6c9
    style G fill:#c8e6c9

Multi-Stage Classification

flowchart LR
    A[Input Text] --> B[Preprocessing]
    B --> C{Exact Match?}
    C -->|Yes| D[ISP-XXX]
    C -->|No| E{Partial Match?}
    E -->|Yes| F[ISP-XXX]
    E -->|No| G[LLM Analysis]
    G --> H[ISP-XXX + Confidence]

Architecture Diagram

graph TD
    A[Complaint] --> B[Rule Engine]
    A --> C[LLM Engine]
    B --> D{Confidence > 80%}
    C --> D
    D -->|Yes| E[Return Code]
    D -->|No| F[Fallback Logic]
    F --> E

    style B fill:#e8eaf6
    style C fill:#fff3e0
    style E fill:#c8e6c9

ISP Codes Reference

Code	Description	Example
ISP-001	ONT/Fiber Issue	Red light, no fiber
ISP-002	Router Problem	WiFi not working
ISP-003	DNS Issue	Can't resolve sites
ISP-004	Speed Problem	Slow connection
ISP-005	Billing Issue	Payment error
ISP-006	Outage	Area-wide down

Demo Scripts

# Rule-based classifier
python app-baseline-class.py

# LLM-enhanced classifier
python app-classifier1.py

How It Works

sequenceDiagram
    participant U as User
    participant A as App
    participant R as Rules
    participant L as LLM

    U->>A: "Red light on ONT"
    A->>R: Check keywords
    R-->>A: Match: ISP-001
    A-->>U: "ONT Issue Detected"

    Note over U,A: If rules fail, LLM is used