nv-3mei commited on
Commit
af6c18f
·
1 Parent(s): 44f11d9

update README

Browse files
README.md CHANGED
@@ -22,7 +22,7 @@ Nemotron-Personas-Japan: 現実世界の分布に基づいた合成ペルソナ
22
  # データセット概要 (Dataset Overview)
23
  Nemotron-Personas-Japan は、日本における人口の多様性と豊かさを捉えることを目的とし、実世界の人口統計、地理的分布、性格特性の分布に基づいて合成的に生成されたペルソナのオープンソースデータセットです。名前、性別、年齢、背景、婚姻状況、学歴、職業、居住地などの統計に基づいて生成した初のデータセットされた [Nemotron-Personas](https://huggingface.co/datasets/nvidia/Nemotron-Personas) の日本語版です。本バージョンでは、日本語における多様なモデリングユースケースに適した高品質のペルソナを提供します
24
 
25
- Nemotron-Personas-Japan は、日本のモデル開発者が重要な地域固有の人口統計や文化的背景を取り入れたソブリンAIシステムを開発することを支援します。本データセットは、日本の地理的・人口統計的な実分布を反映することで、合成データの多様性を高め、バイアスを軽減し、[model collapse](https://en.wikipedia.org/wiki/Model_collapse)(他モデルの出力を無条件に学習することで生じる劣化)を防ぎます。特に、過去のペルソナデータセットと比較して、年齢(例:高齢者ペルソナ)、地理(例:地方在住ペルソナ)、学歴、職業など、複数の軸で人口統計的分布をより代表するように設計されています。例えば、実際の名前、年齢、職業、文化的・教育的背景を含んだ高品質なマルチターン会話データを生成でき、そのデータにユニークな視点や切り口をもたらします。
26
 
27
  本データセットは、合成データ生成のためのエンタープライズ向け複合AIシステム [NeMo Data Designer](https://docs.nvidia.com/nemo/microservices/latest/generate-synthetic-data/index.html) を用いて作成されました。独自の確率的グラフィカルモデル(PGM)、Apache-2.0 ライセンスの GPT-OSS-120B モデル、さらに Data Designer に組み込まれた拡張可能なバリデータや評価器群を活用しています。拡張版の Nemotron-Personas-Japan は、まもなく NeMo Data Designer 上で直接利用可能になります。
28
 
@@ -118,8 +118,8 @@ These extensions build on a strong foundation of public data, helping create per
118
 
119
  ## シードデータ (Seed Data)
120
  日本の人口における社会人口学的および地理的な多様性と複雑性を捉えるために、Nemotron-Personas-Japan は以下の情報を利用しました:
121
- * 総務省統計局が公開している国勢調査
122
- * 名字由来net が提供した日本人の名字・名前データ
123
 
124
  In order to capture the socio-demographic and geographic diversity and complexity of Japan's population, Nemotron-Personas-Japan leveraged
125
  * [population census data](https://www.stat.go.jp/data/jinsui/2024np/index.html) published by the Statistics Bureau of Japan (e-Stat)
@@ -149,17 +149,17 @@ The dataset includes 22 fields: 6 persona fields and 16 contextual fields shown
149
 
150
  The analysis below provides a breakdown across various axes of the dataset to emphasize the built-in diversity and pattern complexity of data.
151
 
152
- ## Names
153
- 本データセットはペルソナに焦点を当てているため、氏名は専用のフィールドとしては提供されていません。しかし、ペルソナ合成には、名字由来net によって提供された 20,000 件のユニークな名前と 97,000 件のユニークな姓が組み込まれています。
154
 
155
  Since the focus of this dataset is on personas, names aren’t provided as dedicated fields. However, infused into persona generation are 20,000 unique first names, 97,000 unique last names provided by [Myoji-Yurai.net](http://Myoji-Yurai.net).
156
 
157
  ## 年齢分布 (Age Distribution)
158
- 日本のペルソナにおける年齢分布は、国全体の実際の人口構造を反映しており、中高年層が大きな割合を占め、若年層に向かって徐々に減少していくという特徴があります。この分布はつぼ型をしており、若者が少なく、第一次および第二次ベビーブーム世代に大きな膨らみがあります。さらに、日本の女性は世界的に見ても非常に長寿であり、高齢者に占める女性の割合は高くなっています。
159
 
160
  なお、このデータセットには 18 歳未満の未成年は含まれていません。
161
 
162
- The age distribution of our Japanese personas mirrors the country’s real demographic structure, characterized by a large proportion of people in the middle to older age groups and a gradual decline toward the younger cohorts. The distribution is shaped like a “vase-shaped,” with fewer young people and a significant expansion in the generations born during the baby boom and the second baby boom. Furthermore, Japanese women are exceptionally long-lived by global standards, and the proportion of women among the elderly is high.
163
 
164
  Note that minors under 18 are excluded from this dataset.
165
  <center>
 
22
  # データセット概要 (Dataset Overview)
23
  Nemotron-Personas-Japan は、日本における人口の多様性と豊かさを捉えることを目的とし、実世界の人口統計、地理的分布、性格特性の分布に基づいて合成的に生成されたペルソナのオープンソースデータセットです。名前、性別、年齢、背景、婚姻状況、学歴、職業、居住地などの統計に基づいて生成した初のデータセットされた [Nemotron-Personas](https://huggingface.co/datasets/nvidia/Nemotron-Personas) の日本語版です。本バージョンでは、日本語における多様なモデリングユースケースに適した高品質のペルソナを提供します
24
 
25
+ Nemotron-Personas-Japan は、日本のモデル開発者が重要な地域固有の人口統計や文化的背景を取り入れた[ソブリンAI](https://www.nvidia.com/en-us/lp/industries/global-public-sector/sovereign-ai-technical-overview/)システムを開発することを支援します。本データセットは、日本の地理的・人口統計的な実分布を反映することで、合成データの多様性を高め、バイアスを軽減し、[model collapse](https://en.wikipedia.org/wiki/Model_collapse)(他モデルの出力を無条件に学習することで生じる劣化)を防ぎます。特に、過去のペルソナデータセットと比較して、年齢(例:高齢者ペルソナ)、地理(例:地方在住ペルソナ)、学歴、職業など、複数の軸で人口統計的分布をより代表するように設計されています。例えば、実際の名前、年齢、職業、文化的・教育的背景を含んだ高品質なマルチターン会話データを生成でき、そのデータにユニークな視点や切り口をもたらします。
26
 
27
  本データセットは、合成データ生成のためのエンタープライズ向け複合AIシステム [NeMo Data Designer](https://docs.nvidia.com/nemo/microservices/latest/generate-synthetic-data/index.html) を用いて作成されました。独自の確率的グラフィカルモデル(PGM)、Apache-2.0 ライセンスの GPT-OSS-120B モデル、さらに Data Designer に組み込まれた拡張可能なバリデータや評価器群を活用しています。拡張版の Nemotron-Personas-Japan は、まもなく NeMo Data Designer 上で直接利用可能になります。
28
 
 
118
 
119
  ## シードデータ (Seed Data)
120
  日本の人口における社会人口学的および地理的な多様性と複雑性を捉えるために、Nemotron-Personas-Japan は以下の情報を利用しました:
121
+ * 総務省統計局が公開している国[勢調査](https://www.stat.go.jp/data/jinsui/2024np/index.html)
122
+ * [名字由来net](http://Myoji-Yurai.net) が提供した日本人の名字・名前データ
123
 
124
  In order to capture the socio-demographic and geographic diversity and complexity of Japan's population, Nemotron-Personas-Japan leveraged
125
  * [population census data](https://www.stat.go.jp/data/jinsui/2024np/index.html) published by the Statistics Bureau of Japan (e-Stat)
 
149
 
150
  The analysis below provides a breakdown across various axes of the dataset to emphasize the built-in diversity and pattern complexity of data.
151
 
152
+ 氏名 (Names)
153
+ 本データセットはペルソナに焦点を当てているため、氏名は専用のフィールドとしては提供されていません。しかし、ペルソナ合成には、[名字由来net](https://myoji-yurai.net/) によって提供された 20,000 件のユニークな名前と 97,000 件のユニークな姓が組み込まれています。
154
 
155
  Since the focus of this dataset is on personas, names aren’t provided as dedicated fields. However, infused into persona generation are 20,000 unique first names, 97,000 unique last names provided by [Myoji-Yurai.net](http://Myoji-Yurai.net).
156
 
157
  ## 年齢分布 (Age Distribution)
158
+ 日本のペルソナにおける年齢分布は、国全体の実際の人口構造を反映しており、中高年層が大きな割合を占め、若年層に向かって徐々に減少していくという特徴があります。この分布では若者が少なく、第一次および第二次ベビーブーム世代に大きな膨らみがあります。さらに、日本の女性は世界的に見ても非常に長寿であり、高齢者に占める女性の割合は高くなっています。
159
 
160
  なお、このデータセットには 18 歳未満の未成年は含まれていません。
161
 
162
+ The age distribution of our Japanese personas mirrors the country’s real demographic structure, characterized by a large proportion of people in the middle to older age groups and a gradual decline toward the younger cohorts. The distribution shows fewer young people and a significant expansion in the generations born during the baby boom and the second baby boom. Furthermore, Japanese women are exceptionally long-lived by global standards, and the proportion of women among the elderly is high.
163
 
164
  Note that minors under 18 are excluded from this dataset.
165
  <center>
images/nemotron_personas_japan_field_stats.png CHANGED

Git LFS Details

  • SHA256: 4af1885cf46bdbd76d16bd54a18aaadb8d5d84be6d185b3f1163e747b9eb6e3a
  • Pointer size: 131 Bytes
  • Size of remote file: 215 kB

Git LFS Details

  • SHA256: 7bdfd683675c7f60ad227dee42195b0e3005e78f46ead7323070fe244af6fb10
  • Pointer size: 131 Bytes
  • Size of remote file: 218 kB