specter

# Dataset Card for Specter

This dataset is a collection of title-related-unrelated triplets from Scientific Publications on Specter. See [Specter](https://github.com/allenai/specter) for additional information.
This dataset can be used directly with Sentence Transformers to train embedding models.

## Dataset Subsets

### `triplet` subset

* Columns: "anchor", "positive", "negative"
* Column types: `str`, `str`, `str`
* Examples:
```python
{
'anchor': "Integrating children's contributions in the interaction design process",
'positive': 'Designing for or designing with? Informant design for interactive learning environments',
'negative': 'Power Operation in ISD: Technological Frames Perspectives.',
}
```
* Collection strategy: Reading the Specter dataset from [embedding-training-data](https://huggingface.co/datasets/sentence-transformers/embedding-training-data), followed by deduplication.
* Deduplified: Yes

### `pair` subset

* Columns: "anchor", "positive"
* Column types: `str`, `str`
* Examples:
```python
{
'anchor': 'Time-dependent trajectory regression on road networks via multi-task learning',
'positive': 'Convex multi-task feature learning',
}
```
* Collection strategy: Reading the Specter dataset from [embedding-training-data](https://huggingface.co/datasets/sentence-transformers/embedding-training-data), only taking the title and related title, and then performing deduplication.
* Deduplified: Yes

# 数据集卡片：Specter
本数据集是科学出版物中标题相关与不相关三元组的集合，更多信息可参见[Specter](https://github.com/allenai/specter)。该数据集可直接配合Sentence Transformers（句子变换器）用于嵌入模型的训练。

## 数据集子集

### `triplet` 子集
* 列名：锚样本（anchor）、正样本（positive）、负样本（negative）
* 列类型：均为字符串（str）
* 示例：
python
{
'anchor': "Integrating children's contributions in the interaction design process",
'positive': 'Designing for or designing with? Informant design for interactive learning environments',
'negative': 'Power Operation in ISD: Technological Frames Perspectives.',
}

* 采集策略：从[embedding-training-data](https://huggingface.co/datasets/sentence-transformers/embedding-training-data)数据集加载Specter数据集，随后执行去重操作。
* 已去重：是

### `pair` 子集
* 列名：锚样本（anchor）、正样本（positive）
* 列类型：均为字符串（str）
* 示例：
python
{
'anchor': 'Time-dependent trajectory regression on road networks via multi-task learning',
'positive': 'Convex multi-task feature learning',
}

* 采集策略：从[embedding-training-data](https://huggingface.co/datasets/sentence-transformers/embedding-training-data)数据集加载Specter数据集，仅提取标题及相关标题后进行去重。
* 已去重：是