HYUKI5's picture
WhereIsAI/github-issue-similarity@9d058e4cfbc1 を複製し、カードを日本語で書く
3324bb9 verified
|
Raw History Blame Contribute Delete
5.96 kB
---
license: mit
pretty_name: WhereIsAI/github-issue-similarity (mirror)
source_datasets:
- WhereIsAI/github-issue-similarity
tags:
- laya
- laya-finetune
- mirror
- domain-code
- use-permissive
- genre-other
- label-binary
- annot-unknown
- lang-en
language:
- en
task_categories:
- sentence-similarity
size_categories:
- 10K<n<100K
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: valid
path: data/valid-*
- split: test
path: data/test-*
- config_name: positive
data_files:
- split: train
path: positive/train-*
- split: valid
path: positive/valid-*
- split: test
path: positive/test-*
dataset_info:
- config_name: default
features:
- name: text1
dtype: string
- name: text2
dtype: string
- name: label
dtype: int64
splits:
- name: train
num_bytes: 181474610
num_examples: 18565
- name: valid
num_bytes: 14656141
num_examples: 1547
- name: test
num_bytes: 13135402
num_examples: 1548
download_size: 58129604
dataset_size: 209266153
- config_name: positive
features:
- name: anchor
dtype: string
- name: positive
dtype: string
splits:
- name: train
num_bytes: 79405713
num_examples: 9457
- name: valid
num_bytes: 6160932
num_examples: 774
- name: test
num_bytes: 5782206
num_examples: 807
download_size: 25212890
dataset_size: 91348851
---
# WhereIsAI/github-issue-similarity(複製)
## データの内容
GitHub の issue どうしの類似性を判定するためのデータ。1 行は 2 つの文章 `text1` と `text2` の組と、その類似性を表す整数の `label` から成る。元のラベルを決まった規則で変換して作られており、出典は論文 arXiv:2309.12871 である。
## データの概要
| 項目 | 値 |
|---|---|
| 行数 | 32,698(2 構成・6 分割の合計) |
| 列数 | 2〜3(構成による) |
| 容量 | 83.35 MB(ファイル 8 本) |
| ファイルの形式 | parquet 6 |
| 言語 | en |
| 公開 | public |
### 構成と分割
| 構成(config) | 分割(split)ごとの行数 | 列数 |
|---|---|---:|
| `default` | `train` 18,565、`valid` 1,547、`test` 1,548 | 3 |
| `positive` | `train` 9,457、`valid` 774、`test` 807 | 2 |
### 列(構成 `default`・分割 `train`。構成ごとに列が違う)
| 列 | 型 | 先頭の行の値 |
|---|---|---|
| `text1` | `string` | `<h3 dir="auto">Feature request</h3> <p dir="auto…` |
| `text2` | `string` | `<p dir="auto">Hello,<br> I tried to import this:…` |
| `label` | `int64` | `0` |
行数と列は HF のデータセットビューアの集計(元のデータセットの現在の版で数えたもの)。全データセットの一覧は [HYUKI5/dataset-catalog](https://huggingface.co/datasets/HYUKI5/dataset-catalog) にある。
## 複製について
[`WhereIsAI/github-issue-similarity`](https://huggingface.co/datasets/WhereIsAI/github-issue-similarity) をリビジョン `9d058e4cfbc11d782186a1dbb5031a6509b895e1` の時点でそのまま複製したもの。中身は変えていない。判断モデル Laya(`convaiinnovations/laya`)の微調整に使う候補として、調査メモ(`Laya 学習用データセット調査`)に載せたデータを集めている。
## 出典
- 元のデータセット: https://huggingface.co/datasets/WhereIsAI/github-issue-similarity
- 複製したリビジョン: `9d058e4cfbc11d782186a1dbb5031a6509b895e1`(2024-05-03 08:32:33+00:00)
- 元のライセンス表記: `mit`
## 論文
- https://arxiv.org/abs/2309.12871
## アノテーションの方法
元のカードにアノテーションの節が無かった。下の「調査メモの記述」と「元のカード」を参照。
## 微調整に使うときの注意
- 調査メモの台帳での区分: A(許諾型のライセンスだけで、疑義の注記が無い)。区分は記載の文言で機械的に付けたもので、法的な判定ではない。学習に入れる前に元のカードとライセンスの原文で確かめる。
- Laya の入力は 2,048 token まで(ModernBERT-large / mmBERT)。長い本文は先頭から黙って切られるので、切れる行は事前に除くか、要約・切り出しをしてから入れる。
- 分割は記事・文書単位で行い、同じ本文から作った別の問いを学習側と較正側に分けない。温度と閾値は学習に使っていない分割で別々に当てる。
## 調査メモの記述
| メモ | 節 | ライセンス(記載) | 確認 | 注記 |
|---|---|---|---|---|
| `jev_sources_tasksource_bekko.csv` | CSV(tasksource-jev-typed-decisions) | mit / license_use=commercial | 確認済 | license_use=commercial / [Jev 元データ source] source_annotation_converted(deterministic recast; no model r |
## 元のカード(日本語訳)
以下は元のデータセットのカードの本文を日本語に訳したもの(機械翻訳)。訳に疑問があるときは、末尾の原文を確かめる。
---
# GIS: Github Issue Similarity データセット
このデータセットは次の論文で公開された: https://arxiv.org/abs/2309.12871
# 引用
本データセットを研究で使う場合は、次のように引用されたい。
```bibtex
@article{li2023angle,
title={AnglE-optimized Text Embeddings},
author={Li, Xianming and Li, Jing},
journal={arXiv preprint arXiv:2309.12871},
year={2023}
}
```
<details>
<summary>原文(英語)</summary>
# GIS: Github Issue Similarity Dataset
This dataset was released from the paper: https://arxiv.org/abs/2309.12871
# Citation
If you use our dataset in your research, welcome to cite us as follows:
```bibtex
@article{li2023angle,
title={AnglE-optimized Text Embeddings},
author={Li, Xianming and Li, Jing},
journal={arXiv preprint arXiv:2309.12871},
year={2023}
}
```
</details>