You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

Dhivehi Retrieval Tasks

The Dhivehi Cross-Lingual Information Retrieval dataset is a collection of Dhivehi queries paired with three candidate documents (doc1, doc2, doc3). Each query has one relevant document indicated by doc_id.

Dataset Details

  • Languages: Dhivehi (dv)
  • Columns: "query", "doc1", "doc2", "doc3", "doc_id"
  • Column types: str, str, str, str, str
  • Number of rows: 1000

Examples:

{
  "query": "ދިވެހިރާއްޖޭގެ ސްކޫލްތަކުގައި ރަސްމީކޮށް ބޭނުންކުރަނީ ކޮން ބަހެއް ހެއްޔެވެ؟",
  "doc1": "ޔޫރޯ އަކީ ޔޫރަޕްގެ ގިނަ ގައުމުތަކުގެ ފައިސާ އެވެ.",
  "doc2": "ދިވެހި ބަސް ރަސްމީކޮށް؛ ސްކޫލްތަކުގައި އިނގިރޭސި ބަސް ވަރަށް ބޮޑަށް ބޭނުންކުރެއެވެ.",
  "doc3": "ލޯކަލް ފެރީތަކުން ކައިރީގައި ހުންނަ މީހުން ދިރިއުޅޭ ޖަޒީރާތައް ގުޅާލަނީ އެޓޯލްއެއްގެ ތެރެއިންނެވެ.",
  "doc_id": "doc2"
}

Collection strategy: For each query, three candidate documents were provided, with exactly one marked as the relevant document (doc_id).

Downloads last month
6