1602353775wzj commited on
Commit
aecc4f0
·
verified ·
1 Parent(s): e93ae36

Add POEMOGermanEmotionClassification (train=dual-annotator union≤2; test=XMLGold; anti-leakage)

Browse files
README.md CHANGED
@@ -1,3 +1,154 @@
1
- ---
2
- license: cc-by-nc-4.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: other
3
+ task_categories:
4
+ - text-classification
5
+ task_ids:
6
+ - multi-label-classification
7
+ language:
8
+ - de
9
+ multilinguality:
10
+ - monolingual
11
+ size_categories:
12
+ - 1K<n<10K
13
+ pretty_name: POEMOGermanEmotionClassification
14
+ tags:
15
+ - poetry
16
+ - german
17
+ - emotion-classification
18
+ - aesthetic-emotions
19
+ - multi-label-classification
20
+ - line-level
21
+ - mteb
22
+ - poetrymteb
23
+ - embedding-evaluation
24
+ annotations_creators:
25
+ - expert-generated
26
+ source_datasets:
27
+ - tnhaider/poetry-emotion
28
+ configs:
29
+ - config_name: default
30
+ data_files:
31
+ - split: train
32
+ path: data/train-*
33
+ - split: test
34
+ path: data/test-*
35
+ default: true
36
+ dataset_info:
37
+ - config_name: default
38
+ features:
39
+ - name: id
40
+ dtype: string
41
+ - name: title
42
+ dtype: string
43
+ - name: author
44
+ dtype: string
45
+ - name: year
46
+ dtype: string
47
+ - name: period
48
+ dtype: string
49
+ - name: poem
50
+ dtype: string
51
+ - name: labels
52
+ sequence: int64
53
+ - name: label_names
54
+ sequence: string
55
+ - name: source
56
+ dtype: string
57
+ splits:
58
+ - name: train
59
+ num_examples: 2366
60
+ - name: test
61
+ num_examples: 1198
62
+ ---
63
+
64
+ # POEMOGermanEmotionClassification
65
+
66
+ Line-level **multi-label aesthetic emotion classification** for German poetry (PoetryMTEB), derived from [PO-EMO](https://github.com/tnhaider/poetry-emotion) (Haider et al., LREC 2020).
67
+
68
+ ## Dataset Card
69
+
70
+ | Item | Description |
71
+ |------|-------------|
72
+ | **Source** | [tnhaider/poetry-emotion](https://github.com/tnhaider/poetry-emotion) |
73
+ | **Paper** | [PO-EMO: Conceptualization, Annotation, and Modeling of Aesthetic Emotions in German and English Poetry](http://www.romanklinger.de/publications/HaiderEgerKimKlingerMenninghaus2020LREC_PO-EMO.pdf) (LREC 2020) |
74
+ | **Languages** | German (`de`) |
75
+ | **Unit** | Verse **line** (`poem` field holds one line) |
76
+ | **Labels** | Up to **2** of 9 aesthetic emotions (multi-label) |
77
+ | **train** | Non-gold dual-annotator lines; label = **union** of both annotators, capped at ≤2 |
78
+ | **test** | `XMLGold` gold / adjudicated `emo1`/`emo2` |
79
+ | **Anti-leakage** | Gold poems (and overlapping line texts) are **excluded from train** |
80
+ | **Size** | train=2366; test=1198 |
81
+ | **Evaluation metrics** | Multi-label classification on embeddings: **macro/micro F1**, **Average Precision (AP)** |
82
+
83
+ ### Split anti-leakage checks
84
+
85
+ | Check | Result |
86
+ |-------|--------|
87
+ | Gold poems matched in complete TSV & removed from train | 47 poem_ids |
88
+ | Train ∩ test normalized line text | 0 |
89
+ | Train poems | 107 |
90
+ | Test poems | 48 |
91
+
92
+ ## Label taxonomy (9)
93
+
94
+ | id | label_name | train | test | total |
95
+ |---:|------------|------:|-----:|------:|
96
+ | 0 | `Beauty/Joy` | 957 | 346 | 1303 |
97
+ | 1 | `Sadness` | 598 | 185 | 783 |
98
+ | 2 | `Uneasiness` | 599 | 165 | 764 |
99
+ | 3 | `Vitality` | 414 | 191 | 605 |
100
+ | 4 | `Awe/Sublime` | 242 | 173 | 415 |
101
+ | 5 | `Suspense` | 265 | 151 | 416 |
102
+ | 6 | `Annoyance` | 120 | 168 | 288 |
103
+ | 7 | `Humor` | 184 | 74 | 258 |
104
+ | 8 | `Nostalgia` | 0 | 70 | 70 |
105
+
106
+ Codebook: `label_taxonomy.json`.
107
+
108
+ ## Features
109
+
110
+ | Field | Type | Description |
111
+ |-------|------|-------------|
112
+ | `id` | string | Example id |
113
+ | `title` | string | Poem title |
114
+ | `author` | string | Poet |
115
+ | `year` | string | Publication year |
116
+ | `period` | string | Literary period (when available) |
117
+ | `poem` | string | **Line** text (classification input) |
118
+ | `labels` | list[int64] | Emotion class indices (0–8), length 1–2 |
119
+ | `label_names` | list[string] | Canonical emotion names |
120
+ | `source` | string | `dual_annotator_union` (train) or `XMLGold` (test) |
121
+
122
+ ## Construction method
123
+
124
+ 1. **test**: parse `XMLGold/*.xml` TEI lines; take gold attributes `emo1` / `emo2` (normalize to the 9-class scheme).
125
+ 2. **train**: read `tsv/poemo.german.complete.prodody.emotion.tsv`; for each line take the union of `emotion{1,2}anno{1,2}`, drop `NONE`, keep at most 2 labels (frequency-preferring).
126
+ 3. **Exclude leakage**: remove any complete-TSV poem whose line fingerprint matches a gold poem (exact or Jaccard ≥ 0.7); additionally drop any train line whose normalized text appears in gold.
127
+ 4. Drop lines with no remaining emotion labels.
128
+
129
+ ## Citation
130
+
131
+ ```bibtex
132
+ @inproceedings{Haider2020,
133
+ author = {Thomas Haider and Steffen Eger and Evgeny Kim and Roman Klinger and Winfried Menninghaus},
134
+ title = {{PO-EMO}: Conceptualization, Annotation, and Modeling of Aesthetic Emotions in {German} and {English} Poetry},
135
+ booktitle = {Proceedings of the 12th International Conference on Language Resources
136
+ and Evaluation (LREC'20)},
137
+ year = {2020},
138
+ editor = {Nicoletta Calzolari and Khalid Choukri and Thierry Declerck and Hrafn
139
+ Loftsson and Bente Maegaard and Joseph Mariani and Asuncion Moreno
140
+ and Jan Odijk and Stelios Piperidis},
141
+ address = {Marseille, France},
142
+ month = {May},
143
+ publisher = {European Language Resources Association (ELRA)},
144
+ date = {11-16},
145
+ language = {english},
146
+ url = {http://www.romanklinger.de/publications/HaiderEgerKimKlingerMenninghaus2020LREC_PO-EMO.pdf}
147
+ }
148
+ ```
149
+
150
+ Upstream repository: https://github.com/tnhaider/poetry-emotion
151
+
152
+ ## License
153
+
154
+ Follow the upstream [PO-EMO](https://github.com/tnhaider/poetry-emotion) repository terms for redistribution of the annotations and texts.
data/test-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6599579c6e06ed899424653ff7a7c8bca4e1865522374fd4d4008d03c3da19f
3
+ size 51872
data/train-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa3e2945b414f72115db62385becef3356e2d3d91b37ca3669c3020b91ed67ea
3
+ size 95034
dataset_infos.yaml ADDED
@@ -0,0 +1,41 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ configs:
2
+ - config_name: default
3
+ data_files:
4
+ - path: data/train-*.parquet
5
+ split: train
6
+ - path: data/test-*.parquet
7
+ split: test
8
+ default: true
9
+ dataset_info:
10
+ configs:
11
+ - config_name: default
12
+ dataset_size: 146906
13
+ download_size: 146906
14
+ features:
15
+ - dtype: string
16
+ name: id
17
+ - dtype: string
18
+ name: title
19
+ - dtype: string
20
+ name: author
21
+ - dtype: string
22
+ name: year
23
+ - dtype: string
24
+ name: period
25
+ - dtype: string
26
+ name: poem
27
+ - sequence:
28
+ dtype: int64
29
+ name: labels
30
+ - sequence:
31
+ dtype: string
32
+ name: label_names
33
+ - dtype: string
34
+ name: source
35
+ splits:
36
+ - name: train
37
+ num_bytes: 95034
38
+ num_examples: 2366
39
+ - name: test
40
+ num_bytes: 51872
41
+ num_examples: 1198
label_taxonomy.json ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "n_classes": 9,
3
+ "labels": [
4
+ {
5
+ "id": 0,
6
+ "name": "Beauty/Joy"
7
+ },
8
+ {
9
+ "id": 1,
10
+ "name": "Sadness"
11
+ },
12
+ {
13
+ "id": 2,
14
+ "name": "Uneasiness"
15
+ },
16
+ {
17
+ "id": 3,
18
+ "name": "Vitality"
19
+ },
20
+ {
21
+ "id": 4,
22
+ "name": "Awe/Sublime"
23
+ },
24
+ {
25
+ "id": 5,
26
+ "name": "Suspense"
27
+ },
28
+ {
29
+ "id": 6,
30
+ "name": "Annoyance"
31
+ },
32
+ {
33
+ "id": 7,
34
+ "name": "Humor"
35
+ },
36
+ {
37
+ "id": 8,
38
+ "name": "Nostalgia"
39
+ }
40
+ ],
41
+ "train_label_rule": "union of emotion1/2 from annotator1 and annotator2, capped at 2 (freq-preferring)",
42
+ "test_label_rule": "XMLGold emo1/emo2",
43
+ "counts": {
44
+ "train": {
45
+ "Suspense": 265,
46
+ "Beauty/Joy": 957,
47
+ "Uneasiness": 599,
48
+ "Sadness": 598,
49
+ "Awe/Sublime": 242,
50
+ "Vitality": 414,
51
+ "Humor": 184,
52
+ "Annoyance": 120
53
+ },
54
+ "test": {
55
+ "Vitality": 191,
56
+ "Sadness": 185,
57
+ "Beauty/Joy": 346,
58
+ "Suspense": 151,
59
+ "Humor": 74,
60
+ "Annoyance": 168,
61
+ "Uneasiness": 165,
62
+ "Awe/Sublime": 173,
63
+ "Nostalgia": 70
64
+ },
65
+ "all": {
66
+ "Suspense": 416,
67
+ "Beauty/Joy": 1303,
68
+ "Uneasiness": 764,
69
+ "Sadness": 783,
70
+ "Awe/Sublime": 415,
71
+ "Vitality": 605,
72
+ "Humor": 258,
73
+ "Annoyance": 288,
74
+ "Nostalgia": 70
75
+ }
76
+ }
77
+ }