README.md

---
license: mit
language:
- am
---

# AmQA: Amharic Question Answering Dataset


Amharic question and answer dataset in a prompt and completion format.

## Dataset Details

The dataset is formatted into the following formats.
```
"inputs": "ከዚህ በታች በተገለጸው አውድ ተከታዩን ጥያቄ ይመልሱ፡ {row['context']} {row['question']}",
"targets": "ከጥያቄው ጋር የሚስማማው ምላሽ {row['answer_text']} ነው።"

"inputs": "ከዚህ በታች ያለውን ዝርዝር መረጃ በመጠቀም ለሚከተለው ጥያቄ መልስ ይስጡ፡ {row['context']} {row['question']}",
"targets": "ከጥያቄው አንጻር ትክክለኛው መልስ {row['answer_text']} ነው።"

"inputs": "ከዚህ በታች ያለውን ጽሑፍ በማጣቀስ እባክዎን ለሚከተለው መልስ ይስጡ {row['context']} {row['question']}",
"targets": "ለጥያቄው መልስ {row['answer_text']} ነው።"

"inputs": "የተሰጠውን ጥያቄ ለመመለስ ከዚህ በታች የቀረበውን መረጃ ይመልከቱ፡ {row['context']} {row['question']}",
"targets": "ለተጠቀሰው ጥያቄ ትክክለኛው ምላሽ {row['answer_text']} ነው።"

"inputs": "ለሚከተለው ጥያቄ ምላሽ ለመስጠት ከዚህ በታች የቀረበውን አውድ ተጠቀም፡ {row['context']} {row['question']}",
"targets": "ለጥያቄው መልሱ {row['answer_text']} ነው።"

"inputs": "የተሰጠውን ጥያቄ ከዚህ በታች በተሰጠው አውድ መሰረት መልሱ፡ {row['context']} {row['question']}",
"targets": "ለጥያቄው ትክክለኛው ምላሽ {row['answer_text']} ነው።"
```

### Dataset Description

Abstract
Question Answering (QA) returns concise answers or answer lists from natural language text given a context document. To advance robust models' development, large amounts of resources go into curating QA datasets. There is a surge of QA datasets for languages like English, however, this is not the case for Amharic. Amharic, the official language of Ethiopia, is the second most spoken Semitic language in the world. There is no published or publicly available Amharic QA dataset. Hence, to foster the research in Amharic QA, we present the first Amharic QA (AmQA) dataset. We crowdsourced 2628 question-answer pairs over 378 Wikipedia articles. Additionally, we run an XLMRLarge-based baseline model to spark open- domain QA research interest. The best-performing baseline achieves an F-score of 69.58 and 71.74 in reader-retriever QA and reading comprehension settings respectively.

Dataset
In Amharic, interrogative sentences can be formulated using information-seeking pronouns like “ምን” (what), “መቼ” (when), “ማን” (who), “የት” (where), “የትኛው” (which), etc. and prepositional interrogative phrases like “ለምን” [ለ-ምን] (why), “በምን” [በ-ምን] (by what), etc. Besides, a verb phrase could be used to pose questions (Getahun 2013; Baye 2009). As shown bellow, the AmQA dataset contains context, question, and answer triplets. The contexts are articles collected from Amharic Wikipedia dump file. The question-answer pairs are created by crowdsourcing and annotated using the Haystack QA annotation tool. 2628 question and answer pairs are created from 378 documents. The whole AmQA dataset can be found here. We also split the datset into train, dev, and test with a size of 1728, 600, and 300 respectively.



- **Curated by:** [Tilahun Abedissa, Ricardo Usbeck, Yaregal Assabie]
- **Language(s) (NLP):** [Amharic]
- **License:** [MIT]

### Dataset Sources

<!-- Provide the basic links for the dataset. -->

- **Repository:** [https://github.com/semantic-systems/amharic-qa]
- **Paper [optional]:** [https://arxiv.org/abs/2303.03290]

## Dataset Structure

The dataset is restructured in a new JSON format of the following: 
```
formatted_json = {
            "prompt": f"ከዚህ በታች በተሰጠው አውድ/አንቀጽ መሰረት የተሰጠውን ጥያቄ መልሱ:\n{row['context']}\n\n{row['question']}",
            "completion": f"ለተሰጠው ጥያቄ መልሱ {row['answer_text']} ነው።"
        }

```

## Citation

**BibTeX:**
```
[@misc{abedissa2023amqa,
      title={AmQA: Amharic Question Answering Dataset}, 
      author={Tilahun Abedissa and Ricardo Usbeck and Yaregal Assabie},
      year={2023},
      eprint={2303.03290},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}]
```