esa-sceva/satcom-synth-qa

---
dataset_name: satcom-synth-qa
tags:
- satellite-communications
- synthetic-data
- question-answering
- esa-sceva
language: en
license: apache-2.0
task_categories:
- question-answering
size_categories:
- 100K<n<1M
---

# esa-sceva/satcom-synth-qa

## Summary
Synthetic dataset of question-answer pairs on satellite communications, created to support model fine-tuning and evaluation in the SatCom domain.

## Description
Generated from SatCom documents using large models (LLaMA 70B 3.3 Instruct and Qwen 2 72B Instruct).
Two single-hop generation strategies were applied:
1. Joint QA generation from full documents.
2. Two-step process with separate question and answer generation for improved quality.

The dataset covers a wide range of SatCom topics, providing diverse factual and conceptual questions.

## Composition
- Around 1.1 million QAs before filtering
- JSONL format
- Fields: `question`, `answer`
- Language: English

## Intended use
Training or evaluating models for factual question answering and domain adaptation in satellite communications.

## Quality control
- Automatic structure validation
- Filtering for coherence and domain relevance
- Spot-checking by SatCom experts

## Example
```python
from datasets import load_dataset
ds = load_dataset("esa-sceva/satcom-synth-qa", split="train")
print(ds[0]["question"])
print(ds[0]["answer"])

数据集名称：satcom-synth-qa
标签：卫星通信（satellite-communications）、合成数据（synthetic-data）、问答（question-answering）、esa-sceva
语言：英语
许可协议：Apache 2.0
任务类别：问答任务
样本规模：10万<样本量<100万

# esa-sceva/satcom-synth-qa

## 数据集概述
本数据集为卫星通信（satellite communications）领域的问答对合成数据集，旨在支撑卫星通信（SatCom）领域的模型微调与评估工作。

## 数据集详情
本数据集基于大语言模型（Large Language Model，LLM）——LLaMA 70B 3.3 Instruct与Qwen 2 72B Instruct，从卫星通信文档中生成。采用两种单跳生成策略：
1. 基于完整文档联合生成问答对；
2. 分两步分别生成问题与答案，以提升生成质量。
本数据集覆盖卫星通信领域的广泛主题，提供多样化的事实性与概念性问答内容。

## 数据组成
- 过滤前共计约110万条问答对；
- 采用JSONL格式存储；
- 数据字段包含`question`（问题）与`answer`（答案）；
- 语言为英语。

## 适用场景
可用于训练或评估卫星通信领域的事实性问答模型，以及开展领域自适应相关研究。

## 质量管控
- 自动化结构校验；
- 针对内容连贯性与领域相关性进行筛选；
- 由卫星通信领域专家进行抽样复核。

## 使用示例
python
from datasets import load_dataset
ds = load_dataset("esa-sceva/satcom-synth-qa", split="train")
print(ds[0]["question"])
print(ds[0]["answer"])