Tamil and Sinhala Short Sentence Pairs for Similarity Measurement

This dataset supports research on deep learning techniques for measuring short sentence similarity in the low-resource languages Tamil and Sinhala. It was created by NLPC-UOM and is associated with a 2021 research paper on monolingual similarity measurement using Siamese networks. The dataset was last updated in March 2025.

Use Cases

• Train a Siamese network model on Tamil and Sinhala short sentence pairs for similarity scoring.
• Benchmark unsupervised techniques adapted from English for low-resource language similarity tasks.
• Evaluate deep learning models on short sentence similarity for the specific languages Tamil and Sinhala.

Strengths

• Focuses on two specific low-resource languages, Tamil and Sinhala, addressing a gap in NLP resources.
• Associated with a peer-reviewed research publication from 2021, providing methodological context.
• Dataset was updated on the Hugging Face platform in March 2025, indicating recent maintenance.

Limitations

• The dataset size, number of rows, and specific column structure are not provided, limiting reproducibility assessments.
• As a dataset for low-resource languages, the available data volume for training complex models may be constrained.
• The description does not specify the annotation method or potential label noise for the similarity scores.

Provenance

Source

NLPC-UOM (University of Moratuwa, Sri Lanka).

Collection Method

Created for research on deep learning-based sentence similarity measurement, utilizing existing unsupervised techniques for English.

Freshness

Last updated on the platform in March 2025.

Geography

Primarily relevant to the Tamil and Sinhala languages.