Skip to content

Latest commit

 

History

History
62 lines (35 loc) · 3.21 KB

File metadata and controls

62 lines (35 loc) · 3.21 KB
graph LR
    AbstractResolver["AbstractResolver"]
    Resolver["Resolver"]
    WordAligner["WordAligner"]
    langextract_data["langextract.data"]
    langextract_tokenizer["langextract.tokenizer"]
    Resolver -- "implements" --> AbstractResolver
    Resolver -- "delegates to" --> WordAligner
    Resolver -- "uses" --> langextract_data
    Resolver -- "uses" --> langextract_tokenizer
Loading

CodeBoardingDemoContact

Details

One paragraph explaining the functionality which is represented by this graph. What the main flow is and what is its purpose.

AbstractResolver

This component defines the contract for any class responsible for parsing LLM output and aligning extractions. It establishes the resolve and align methods as the fundamental operations for this subsystem, ensuring extensibility and adherence to a common interface.

Related Classes/Methods:

  • langextract.resolver.AbstractResolver:resolve
  • langextract.resolver.AbstractResolver:align

Resolver

The concrete implementation of AbstractResolver, this component is responsible for parsing LLM-generated YAML or JSON into data.Extraction objects. It orchestrates the entire resolution and alignment process, handling parsing logic, error management, and the ordering of extracted information. It acts as the primary orchestrator within this subsystem.

Related Classes/Methods:

  • langextract.data.Extraction:44-106

WordAligner

A specialized utility component focused on the precise alignment of extracted text spans with the original source text. It employs string comparison algorithms to identify exact, partial, or fuzzy matches, calculating token and character offsets. This component handles the granular details of text alignment.

Related Classes/Methods:

  • langextract.resolver.WordAligner:520-882

langextract.data

Module responsible for defining data structures, such as Extraction objects, used across the system.

Related Classes/Methods: None

langextract.tokenizer

Module responsible for tokenizing text for processing and alignment.

Related Classes/Methods: None