Tonari-no-usagi commited on
Commit
8c2e363
·
verified ·
1 Parent(s): dd9cd39

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +124 -3
README.md CHANGED
@@ -1,3 +1,124 @@
1
- ---
2
- license: cc-by-nc-4.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - ja
4
+ license: cc-by-nc-4.0
5
+ task_categories:
6
+ - text-generation
7
+ - question-answering
8
+ tags:
9
+ - function-calling
10
+ - tool-use
11
+ - agent
12
+ - japanese
13
+ - synthetic
14
+ size_categories:
15
+ - 1k<n<10k
16
+ ---
17
+
18
+ # 🇯🇵 Japanese Function Calling Dataset (Synthetic)
19
+ **Robust & High-Quality Tool-Use Dataset for AI Agents**
20
+
21
+ <div align="center">
22
+
23
+ [![Buy Full Version](https://tonarinousagi.gumroad.com/l/japanese-function-calling)
24
+ [![License: CC BY-NC 4.0](https://img.shields.io/badge/License-CC_BY--NC_4.0-lightgrey.svg)](https://creativecommons.org/licenses/by-nc/4.0/)
25
+
26
+ </div>
27
+
28
+ ## Overview (概要)
29
+
30
+ This dataset contains Japanese instruction-response pairs designed to train LLMs for **Function Calling (Tool Use)** capabilities.
31
+ Unlike simple translation datasets, this dataset is generated to reflect **real-world Japanese usage**, including:
32
+ - **Casual speech** (e.g., Tameguchi)
33
+ - **Ambiguous instructions** (Subject omission)
34
+ - **Irrelevant noise** (Realistic conversational fillers)
35
+
36
+ 本データセットは、LLM(大規模言語モデル)に**日本語でのFunction Calling(ツール利用/関数呼び出し)能力**を学習させるために設計された合成データセットです。
37
+ 単なる翻訳ではなく、**「実用的なAIエージェント」**の開発を目的としており、以下のようなリアルな日本語の特徴を含んでいます。
38
+ - **多様な話し言葉:** 敬語だけでなく、タメ口など。
39
+ - **曖昧な指示:** 主語の省略や、倒置法。
40
+ - **ノイズ:** 文脈と関係のないフィラー(「えーと」など)や余計な情報。
41
+
42
+ ---
43
+
44
+ ## Why this dataset? (特徴)
45
+
46
+ | Feature | Description |
47
+ | :--- | :--- |
48
+ | **High Precision** | Generated by **Gemini 2.5 Flash** and strictly validated via Python scripts. Only valid JSON is included.<br>Gemini 2.5 Flashを用いて生成し、スクリプトでJSON構文を厳密にチェック済みです。 |
49
+ | **Robustness** | Contains various speech styles to prevent the model from overfitting to "textbook Japanese".<br>「教科書的な日本語」への過学習を防ぐため、崩した表現を多数収録しています。 |
50
+ | **Agent-Ready** | Covering 5+ essential domains: Calendar, Weather, Smart Home, Messaging, and Booking.<br>カレンダー、天気、家電操作、予約など、エージェント開発に必須のドメインを網羅。 |
51
+
52
+ ---
53
+
54
+ ## Dataset Structure (データ構造)
55
+
56
+ The data is provided in **JSONL** format, compatible with standard chat-completion pipelines.
57
+ データは一般的なチャット形式(JSONL)で提供されます。
58
+
59
+ ### Sample Data
60
+ ```json
61
+ {
62
+ "messages": [
63
+ {
64
+ "role": "user",
65
+ "content": "来週の火曜、大阪に出張なんだけど、天気どう?傘いる?"
66
+ },
67
+ {
68
+ "role": "assistant",
69
+ "tool_calls": [
70
+ {
71
+ "function": {
72
+ "name": "get_weather",
73
+ "arguments": "{\"location\": \"Osaka\", \"date\": \"next Tuesday\"}"
74
+ }
75
+ }
76
+ ]
77
+ }
78
+ ]
79
+ }
80
+
81
+ ### Defined Tools (収録ツール定義)
82
+ The dataset assumes the following tools are available:
83
+ 以下のツール定義に基づいています。
84
+
85
+ - `manage_schedule`: Add/Check/Delete calendar events.
86
+ - `get_weather`: Get weather forecast.
87
+ - `control_device`: Smart home control (Lights, AC, Locks).
88
+ - `search_restaurant`: Search and book restaurants.
89
+ - `send_message`: Send texts to contacts.
90
+
91
+ ---
92
+
93
+ ## Access & Licensing (ライセンスと入手方法)
94
+
95
+ ### 1. Free Preview (Hugging Face)
96
+ - **Content:** A subset of the dataset (e.g., 500 samples).
97
+ - **License:** [CC-BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/)
98
+ - **Usage:** Research and personal use only. **No Commercial Use.**
99
+ - **内容:** データセットの一部(サンプル)。
100
+ - **用途:** 研究、個人開発のみ。**商用利用不可。**
101
+
102
+ ### 2. Full Version (Gumroad)
103
+ - **Content:** Full dataset (10,000+ samples).
104
+ - **License:** **Commercial Use Allowed.** You can use this data to train your proprietary models and sell them.
105
+ - **Link:** [Available on Gumroad](ここにGumroadのリンクを貼る)
106
+ - **内容:** フルセット(10,000件以上)。
107
+ - **用途:** **商用利用可能。** 自社モデルの学習に使用し、そのモデルを販売することができます。
108
+
109
+ ---
110
+
111
+ ## 🛠 Generation Process (作成プロセス)
112
+
113
+ 1. **Schema Definition:** Defined typical API schemas for AI agents.
114
+ 2. **Generation:** Used **Gemini 2.5 Flash** to generate diverse Japanese prompts (Standard, Dialect, Casual) and corresponding JSON.
115
+ 3. **Validation:** Automatically parsed every JSON output. Invalid JSON or hallucinated parameters were discarded.
116
+
117
+ ---
118
+
119
+ ## Disclaimer (免責事項)
120
+ While we strive for high accuracy, this is synthetic data. Please verify important outputs in your specific application context.
121
+ 高精度を目指して作成されていますが、合成データであるた���、実際の運用環境での動作を保証するものではありません。
122
+ ```
123
+
124
+ ---