README.md

---
configs:
  - config_name: default
    data_files:
      - 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

**DBPedia-Fa** is a Persian (Farsi) dataset tailored for the **Retrieval** task, focusing on **entity retrieval**. It is a translated version of the English DBpedia dataset used in the **BEIR benchmark**, and a key component of the [FaMTEB (Farsi Massive Text Embedding Benchmark)](https://huggingface.co/spaces/mteb/leaderboard), under the **BEIR-Fa** collection.

- **Language(s):** Persian (Farsi)  
- **Task(s):** Retrieval (Entity Retrieval)  
- **Source:** Translated from the English DBpedia dataset used in BEIR  
- **Part of FaMTEB:** Yes — under BEIR-Fa

## Supported Tasks and Leaderboards

The dataset is used to evaluate **text embedding models** on their ability to retrieve **structured entity information** in response to **heterogeneous queries**. Results are benchmarked on the **Persian MTEB Leaderboard** on Hugging Face Spaces (language filter: Persian).

## Construction

The dataset was created by:

- Translating an English DBpedia entity retrieval dataset into Persian using the **Google Translate API**  
- Retaining the structure of the original **knowledge base entity retrieval task**, where the goal is to return accurate entities from DBpedia based on various query types

According to the *FaMTEB* paper, the **BEIR-Fa datasets** underwent:

- **BM25 retrieval score comparisons** with English counterparts  
- **LLM-based evaluation** using the **GEMBA-DA framework** to confirm **high translation quality** for retrieval tasks

## Data Splits

As reported in the FaMTEB paper (Table 5):

- **Train:** 0 samples  
- **Dev:** 0 samples  
- **Test:** 4,651,208 samples  

**Total:** ~4.69 million examples