AMTTL: Adversarial Multi-task Transfer Learning

The AMTTL dataset is a monolingual Chinese text collection for token classification tasks, created by author gavinxing and last updated in January 2024. It is categorized as containing between 1,000 and 10,000 instances (1K<n<10K) and has crowdsourced annotations.

Use Cases

• Perform token classification on monolingual Chinese text using the provided annotations.
• Analyze text modality patterns within the 1K<n<10K instance range for the Chinese language.
• Utilize crowdsourced annotations for tasks in the token classification category.

Strengths

• Dataset size is specified within a known range of 1,000 to 10,000 instances (1K<n<10K).
• Annotations are crowdsourced, potentially providing diverse labeling perspectives.
• Dataset is recent, with a last update timestamp of January 2024.

Limitations

• Specific row count, column names, data fields, and sample data are unavailable, hindering preliminary assessment.
• The dataset description contains multiple '[More Information Needed]' sections, indicating incomplete documentation.
• The geographic scope is listed as 'us' but the language is Chinese (zh), which may indicate a specific, potentially narrow data origin.

Provenance

Source

Original source datasets, as indicated by tags.

Collection Method

Crowdsourced annotations.

Freshness

2024-01-09

Geography

Region tag indicates 'us', but primary language is Chinese (zh).