Improve TCCP Dataset Card + schema (poem/label) + dataset_infos.yaml

README.md

@@ -1,46 +1,157 @@
----
-license: cc-by-4.0
-task_categories:
-- text-classification
-language:
-- zh
-tags:
-- art
-pretty_name: TCCP - Tang Chinese Classical Poetry Dataset
-size_categories:
-- 1K<n<10K
----
-
-# TCCP - 唐代中文古典诗歌数据集
-
-## 数据概况
-
-TCCP数据集包含唐代古典诗歌，每首诗歌标注了主题和关键词。
-
-**总计**: 3206 首诗歌
-
-### 数据集统计
-| 数据集 | 诗歌数量 | 主题数量 | 关键词数量 |
-|-------|----------|----------|------------|
-| train | 2804 | 9 | 24529 |
-| test | 243 | 9 | 2107 |
-| validation | 159 | 9 | 3064 |
-
-
-### 数据格式
-python
-{
-"text": "诗歌正文",
-"title": "诗歌标题",
-"theme_label": "诗歌主题",
-"keywords": ["关键词1", "关键词2"]
-}
-
-### 使用示例
-python
-from datasets import load_dataset
-dataset = load_dataset("PoetryMTEB/TCCP")
-print(dataset['train'][0])
-## 数据来源
-唐代古典诗歌数据集:https://github.com/shuizhonghaitong/classification_GAT  
-刘昱彤,吴斌,白婷.古诗词图谱的构建及分析研究[J].计算机研究与发展,2020,57(06):1252-1268.
+---
+license: cc-by-4.0
+task_categories:
+- text-classification
+task_ids:
+- multi-class-classification
+language:
+- zh
+multilinguality:
+- monolingual
+size_categories:
+- 1K<n<10K
+pretty_name: TCCP (Tang Chinese Classical Poetry Theme Classification)
+tags:
+- poetry
+- chinese
+- classical-chinese
+- tang-poetry
+- theme-classification
+- multi-class-classification
+- mteb
+- poetrymteb
+- embedding-evaluation
+annotations_creators:
+- expert-generated
+source_datasets:
+- shuizhonghaitong/classification_GAT
+configs:
+- config_name: default
+  data_files:
+  - split: train
+    path: data/train-*
+  - split: validation
+    path: data/validation-*
+  - split: test
+    path: data/test-*
+  default: true
+dataset_info:
+- config_name: default
+  features:
+  - name: id
+    dtype: string
+  - name: title
+    dtype: string
+  - name: poem
+    dtype: string
+  - name: label
+    dtype: int64
+  - name: label_name
+    dtype: string
+  - name: keywords
+    sequence: string
+  splits:
+  - name: train
+    num_examples: 2804
+  - name: validation
+    num_examples: 159
+  - name: test
+    num_examples: 243
+---
+
+# TCCP — Tang Chinese Classical Poetry Theme Classification
+
+Multi-class **theme classification** of **Tang-dynasty classical Chinese poetry** for PoetryMTEB embedding evaluation.
+
+Upstream TCCP provides poem texts with expert/curated **theme** labels and keyword lists, originally used for graph-attention theme classification ([classification_GAT](https://github.com/shuizhonghaitong/classification_GAT)).
+
+## Dataset Card
+
+| Item | Description |
+|------|-------------|
+| **Source** | [shuizhonghaitong/classification_GAT](https://github.com/shuizhonghaitong/classification_GAT) (TCCP) |
+| **Paper** | 刘昱彤, 吴斌, 白婷. 古诗词图谱的构建及分析研究[J]. *计算机研究与发展*, 2020, 57(6): 1252-1268 |
+| **Languages** | Classical Chinese / Chinese (`zh`) |
+| **Unit** | Full poem text (`poem`) |
+| **Labels** | Single-label **9** themes (思乡 / 送别 / 战争 / …) |
+| **Size** | train=2804; validation=159; test=243; **total=3206** |
+| **Splits** | Preserved from the previous PoetryMTEB/TCCP release (train / validation / test) |
+| **License** | [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/) (PoetryMTEB redistribution; cite original authors) |
+| **Evaluation metrics** | Embedding classification: **accuracy**, **macro/micro F1**. Upstream GAT paper reports theme-classification accuracy. |
+
+> **Note on size:** The upstream GitHub README mentions ~3,247 poems; this Hub packaging contains **3206** labeled poems after cleaning/split alignment of the prior PoetryMTEB release.
+
+## Label taxonomy (9)
+
+| id | label_name | name_en | train | validation | test | total |
+|---:|------------|---------|------:|-----------:|-----:|------:|
+| 0 | `思乡` | Homesickness | 596 | 28 | 31 | 655 |
+| 1 | `送别` | Farewell | 565 | 30 | 42 | 637 |
+| 2 | `战争` | War | 546 | 26 | 34 | 606 |
+| 3 | `怀人` | Longing for someone | 447 | 15 | 7 | 469 |
+| 4 | `田园` | Pastoral | 168 | 12 | 19 | 199 |
+| 5 | `山水` | Landscape | 136 | 8 | 14 | 158 |
+| 6 | `咏物` | Object chanting | 119 | 11 | 28 | 158 |
+| 7 | `闺怨` | Boudoir lament | 114 | 12 | 29 | 155 |
+| 8 | `怀古` | Historical reflection | 113 | 17 | 39 | 169 |
+
+Codebook: `label_taxonomy.json`.
+
+## Features
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `id` | string | Example id (`tccp-{split}-{idx}`) |
+| `title` | string | Poem title |
+| `poem` | string | **Classification input**: full poem body (formerly `text`) |
+| `label` | int64 | Theme class id (0–8) |
+| `label_name` | string | Theme name in Chinese (formerly `theme_label`) |
+| `keywords` | list[string] | Keyword list associated with the poem |
+
+Schema aligns with other PoetryMTEB classification datasets (`id`, `poem`, `label`, `label_name`).
+
+## Construction method
+
+1. Load the existing `PoetryMTEB/TCCP` parquet splits.
+2. Rename/normalize fields: `text`→`poem`, `theme_label`→`label_name` + integer `label`.
+3. Keep upstream `title` and `keywords`; assign stable ids.
+4. Write PoetryMTEB parquet shards + `dataset_infos.yaml` + Dataset Card.
+
+No re-split was performed in this packaging pass (split sizes preserved for continuity with prior PoetryMTEB experiments).
+
+## How to load
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("PoetryMTEB/TCCP")
+print(ds["train"][0]["poem"][:80])
+print(ds["train"][0]["label_name"], ds["train"][0]["label"])
+```
+
+## Intended use
+
+- PoetryMTEB / MTEB-style **multi-class classification** probing of classical Chinese poem embeddings (theme recognition).
+- Literary / DH analyses of Tang poetry themes.
+
+## Citation
+
+```bibtex
+@article{liu2020gushi,
+  title   = {古诗词图谱的构建及分析研究},
+  author  = {刘昱彤 and 吴斌 and 白婷},
+  journal = {计算机研究与发展},
+  volume  = {57},
+  number  = {6},
+  pages   = {1252--1268},
+  year    = {2020}
+}
+```
+
+Upstream code/data release: [https://github.com/shuizhonghaitong/classification_GAT](https://github.com/shuizhonghaitong/classification_GAT)
+
+This Hub packaging: `PoetryMTEB/TCCP`
+
+## License
+
+Distributed under **CC BY 4.0** for this PoetryMTEB packaging. Please cite the original paper and GitHub repository when using the data.

data/test-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:48876c7c985526889841a1a0453477df89e5700d9e640c94b1be05ba2fc90e0e
-size 49954
+oid sha256:18d155ba9266dac7154c9e0af9f7bc9f3adf3867eb35b71ca9c15df1dcf63877
+size 52423

data/train-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a605bd725d13df890c3f987c1dae054423d1f9cbb1d7f92ab3556c71da4c7ba8
-size 1134852
+oid sha256:c0829a72fa585cc2bdafcd50e5ffc2cb6f7578e79682e7b7dd8cf53dc471450d
+size 1074015

data/validation-00000-of-00001.parquet

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:a49d4c0cd7260d449fb963e0b767ab94e50ad9b411833042307062c651fb6aba
-size 70605
+oid sha256:89a3443b46170ada457aaf62244a4253681a636cad67b385ce3fdd03390cbd60
+size 72391

dataset_infos.yaml

@@ -0,0 +1,39 @@
+configs:
+- config_name: default
+  data_files:
+  - path: data/train-*.parquet
+    split: train
+  - path: data/validation-*.parquet
+    split: validation
+  - path: data/test-*.parquet
+    split: test
+  default: true
+dataset_info:
+  configs:
+  - config_name: default
+    dataset_size: 1198829
+    download_size: 1198829
+    features:
+    - dtype: string
+      name: id
+    - dtype: string
+      name: title
+    - dtype: string
+      name: poem
+    - dtype: int64
+      name: label
+    - dtype: string
+      name: label_name
+    - sequence:
+        dtype: string
+      name: keywords
+    splits:
+    - name: train
+      num_bytes: 1074015
+      num_examples: 2804
+    - name: validation
+      num_bytes: 72391
+      num_examples: 159
+    - name: test
+      num_bytes: 52423
+      num_examples: 243

label_taxonomy.json

@@ -0,0 +1,98 @@
+{
+  "n_classes": 9,
+  "labels": [
+    {
+      "id": 0,
+      "name": "思乡",
+      "name_en": "Homesickness",
+      "name_zh": "思乡"
+    },
+    {
+      "id": 1,
+      "name": "送别",
+      "name_en": "Farewell",
+      "name_zh": "送别"
+    },
+    {
+      "id": 2,
+      "name": "战争",
+      "name_en": "War",
+      "name_zh": "战争"
+    },
+    {
+      "id": 3,
+      "name": "怀人",
+      "name_en": "Longing for someone",
+      "name_zh": "怀人"
+    },
+    {
+      "id": 4,
+      "name": "田园",
+      "name_en": "Pastoral",
+      "name_zh": "田园"
+    },
+    {
+      "id": 5,
+      "name": "山水",
+      "name_en": "Landscape",
+      "name_zh": "山水"
+    },
+    {
+      "id": 6,
+      "name": "咏物",
+      "name_en": "Object chanting",
+      "name_zh": "咏物"
+    },
+    {
+      "id": 7,
+      "name": "闺怨",
+      "name_en": "Boudoir lament",
+      "name_zh": "闺怨"
+    },
+    {
+      "id": 8,
+      "name": "怀古",
+      "name_en": "Historical reflection",
+      "name_zh": "怀古"
+    }
+  ],
+  "counts": {
+    "train": {
+      "战争": 546,
+      "怀古": 113,
+      "送别": 565,
+      "思乡": 596,
+      "咏物": 119,
+      "怀人": 447,
+      "闺怨": 114,
+      "田园": 168,
+      "山水": 136
+    },
+    "validation": {
+      "战争": 26,
+      "送别": 30,
+      "闺怨": 12,
+      "咏物": 11,
+      "怀古": 17,
+      "山水": 8,
+      "思乡": 28,
+      "田园": 12,
+      "怀人": 15
+    },
+    "test": {
+      "战争": 34,
+      "送别": 42,
+      "闺怨": 29,
+      "咏物": 28,
+      "怀古": 39,
+      "山水": 14,
+      "思乡": 31,
+      "田园": 19,
+      "怀人": 7
+    }
+  },
+  "source": {
+    "github": "https://github.com/shuizhonghaitong/classification_GAT",
+    "paper": "古诗词图谱的构建及分析研究"
+  }
+}