README.md

---
language:
- en
multilinguality:
- monolingual
size_categories:
- 100K<n<1M
task_categories:
- feature-extraction
- sentence-similarity
pretty_name: Specter
tags:
- sentence-transformers
dataset_info:
- config_name: pair
  features:
  - name: anchor
    dtype: string
  - name: positive
    dtype: string
  splits:
  - name: train
    num_bytes: 55252049
    num_examples: 380142
  download_size: 24217449
  dataset_size: 55252049
- config_name: triplet
  features:
  - name: anchor
    dtype: string
  - name: positive
    dtype: string
  - name: negative
    dtype: string
  splits:
  - name: train
    num_bytes: 152814049
    num_examples: 684098
  download_size: 62182004
  dataset_size: 152814049
configs:
- config_name: pair
  data_files:
  - split: train
    path: pair/train-*
- config_name: triplet
  data_files:
  - split: train
    path: triplet/train-*
---

# 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