Module task_mapping 

Task-Dataset-Backend Mapping System

This module provides a cohesive system for mapping:

• Tasks (NER, NED, Coreference, etc.) → Datasets
• Datasets → Backends that can evaluate them
• Backends → Tasks they support (via trait inspection)

Design Philosophy

• Trait-based capabilities: Backend capabilities are determined by trait implementations
• Many-to-many relationships: A dataset can support multiple tasks, a backend can support multiple tasks
• Explicit capabilities: Each backend declares what tasks it supports via traits
• Dataset metadata: Each dataset declares what tasks it can evaluate
• Task requirements: Each task declares what datasets are suitable

Trait-Based Capability Detection

Backends are queried for capabilities using trait bounds:

• Model → NER capability
• ZeroShotNER → Zero-shot NER capability
• RelationExtractor → Relation extraction capability
• DiscontinuousNER → Discontinuous NER capability
• CoreferenceResolver → Coreference resolution capability

Structs

TaskMapping

Comprehensive task-dataset-backend mapping.

Enums

Task

Information extraction and NLP tasks supported by anno.

Traits

CoreferenceResolverTrait

Trait for coreference resolution algorithms.

DiscontinuousNERTrait

Support for discontinuous entity spans.

RelationExtractorTrait

Joint entity and relation extraction.

ZeroShotNERTrait

Zero-shot NER for open entity types.

Functions

backend_tasks

Detect backend capabilities via trait inspection.

dataset_tasks

Tasks that a dataset supports for evaluation.

detect_backend_capabilities

Runtime capability detection for a backend instance.

detect_backend_capabilities_by_name

Capability detection using backend name (fallback when type_id isn’t available).

get_dataset_tasks

Get all tasks that a dataset supports.

get_task_backends

Get all backends that support a task.

get_task_datasets

Get all datasets suitable for a task.

task_datasets

Mapping from tasks to suitable datasets.