Multi-News Dataset for Semantic Role Labeling via QA

A dataset for Semantic Role Labeling (SRL) constructed from the Multi-News summarization corpus. It was created by the author 'rubenwol' and uploaded to the Hugging Face platform in November 2021. The dataset applies the Question Answer driven Semantic Role Labeling (QA-SRL) framework to news articles.

Use Cases

• Training models to identify semantic roles (e.g., agent, patient) in sentences using the QA-SRL framework.
• Developing question-answering systems that generate and answer questions about 'who did what to whom' in news text.
• Evaluating the performance of SRL parsers on a dataset derived from multi-document news summaries.
• Fine-tuning language models for structured prediction tasks based on predicate-argument relationships.

Strengths

• Built upon the established Multi-News corpus, providing a foundation in news domain text.
• Applies the structured QA-SRL methodology for semantic role annotation.

Limitations

• Unknown dataset size, row count, and specific column structure limit reproducibility assessments.
• Data is from 2021 and may not reflect current events or language usage trends.
• Potential bias towards topics and writing styles present in the source Multi-News articles.

Provenance

Source

Derived from the Multi-News summarization dataset.

Collection Method

Annotations created using the QA-SRL framework.

Time Range

null

Freshness

Last updated on the platform in 2021; no update frequency specified.

Geography

null