The HTML parser records the start and end character positions of every token and existing link within a single normalized text stream, allowing LinkBERT predictions and classifier flags to align precisely on a client-side UI slider without re-fetching.

flowchart LR
    A[Raw HTML] --> B[HTML Parser]
    B --> C[Flattened Text Stream]
    B --> D[Recorded Link Offsets]
    C --> E[LinkBERT Service]
    E --> F[Token Spans & Probs]
    D --> G[UI Interactive Slider]
    F --> G