Telugu Dialogue Dataset for Instruction Tuning

NLPT compiled a Telugu text dataset from two instruction-tuning sources: 'gpteacher_cleaned_filtered_telugu_romanized' and 'yahma_alpaca_cleaned_telugu_filtered_and_romanized'. The dataset contains an unspecified number of dialogue or instruction-response pairs. It was last updated on the Hugging Face platform in February 2024.

Use Cases

• Fine-tune a Telugu language model on instruction-response pairs from the 'yahma_alpaca_cleaned_telugu_filtered_and_romanized' source.
• Train a dialogue generation model using conversational data from the 'gpteacher_cleaned_filtered_telugu_romanized' component.
• Evaluate the performance of text generation models on romanized Telugu text across two distinct curated datasets.
• Perform comparative analysis of dialogue styles and instruction formats between the two merged source datasets.

Strengths

• Combines data from two established instruction-tuning sources, 'gpteacher' and 'alpaca'.
• Data has undergone cleaning and filtering processes as indicated by source names.
• Text is romanized, providing a standardized character set for processing.

Limitations

• Exact row count, column names, and dataset size are unknown.
• The dataset's composition and balance between the two source components are unspecified.
• Potential for overlap or duplication between the two merged sources is not addressed.

Provenance

Source

Combination of 'gpteacher_cleaned_filtered_telugu_romanized' and 'yahma_alpaca_cleaned_telugu_filtered_and_romanized' datasets.

Collection Method

Aggregated and hosted by NLPT on Hugging Face.

Freshness

Last updated on 2024-02-26.