nfcorpus-fa / README.md
mehran-sarmadi's picture
Update README.md
44a1eac verified
metadata
configs:
  - config_name: default
    data_files:
      - split: train
        path: qrels/train.jsonl
      - split: dev
        path: qrels/dev.jsonl
      - split: test
        path: qrels/test.jsonl
  - config_name: corpus
    data_files:
      - split: corpus
        path: corpus.jsonl
  - config_name: queries
    data_files:
      - split: queries
        path: queries.jsonl

Dataset Summary

NFCorpus-Fa is a Persian (Farsi) dataset designed for the Retrieval task, with a focus on medical and nutrition information retrieval. It is a translated version of the original English NFCorpus (NutritionFacts Corpus), used in the BEIR benchmark, and is part of the FaMTEB (Farsi Massive Text Embedding Benchmark) under the BEIR-Fa collection.

  • Language(s): Persian (Farsi)
  • Task(s): Retrieval (Medical Information Retrieval, Nutrition Information Retrieval)
  • Source: Translated from the English NFCorpus using Google Translate
  • Part of FaMTEB: Yes — part of the BEIR-Fa collection

Supported Tasks and Leaderboards

This dataset evaluates models' ability to retrieve relevant scientific and nutritional content from a specialized corpus (PubMed articles) in response to health-related user queries. Performance can be compared on the Persian MTEB Leaderboard (filter by language: Persian).

Construction

  • Translated from the NFCorpus dataset using the Google Translate API
  • Original English dataset was curated using NutritionFacts.org queries and PubMed article annotations

Translation quality was validated using:

  • BM25 score comparisons
  • LLM-based assessment (GEMBA-DA framework)

Data Splits

As reported in the FaMTEB paper (Table 5):

  • Train: 114,208 samples
  • Dev: 0 samples
  • Test: 15,967 samples

Approximate total dataset size: 141k examples (user-provided figure)