Update README.md
Browse files
README.md
CHANGED
@@ -6,3 +6,62 @@ tags:
|
|
6 |
language: de
|
7 |
---
|
8 |
# REDEWIEDERGABE Tagger: free indirect STWR
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
language: de
|
7 |
---
|
8 |
# REDEWIEDERGABE Tagger: free indirect STWR
|
9 |
+
|
10 |
+
This model is part of an ensemble of binary taggers that recognize German speech, thought and writing representation. They can be used to automatically detect and annotate the following 4 types of speech, thought and writing representation in German texts:
|
11 |
+
|
12 |
+
| STWR type | Example | Translation |
|
13 |
+
|--------------------------------|-------------------------------------------------------------------------|----------------------------------------------------------|
|
14 |
+
| direct | Dann sagte er: **"Ich habe Hunger."** | Then he said: **"I'm hungry."** |
|
15 |
+
| free indirect ('erlebte Rede', **this tagger**) | Er war ratlos. **Woher sollte er denn hier bloß ein Mittagessen bekommen?** | He was at a loss. **Where should he ever find lunch here?** |
|
16 |
+
| indirect | Sie fragte, **wo das Essen sei.** | She asked **where the food was.** |
|
17 |
+
| reported | **Sie sprachen über das Mittagessen.** | **They talked about lunch.** |
|
18 |
+
|
19 |
+
The ensemble is trained on the [REDEWIEDERGABE corpus](https://github.com/redewiedergabe/corpus) ([Annotation guidelines](http://redewiedergabe.de/richtlinien/richtlinien.html)), fine-tuning each tagger on the domain-adapted [lkonle/fiction-gbert-large](https://huggingface.co/lkonle/fiction-gbert-large). ([Training Code](https://github.com/cophi-wue/LLpro/blob/main/contrib/train_redewiedergabe.py))
|
20 |
+
|
21 |
+
**F1-Scores:**
|
22 |
+
|
23 |
+
| STWR type | F1-Score |
|
24 |
+
|-----------|-----------|
|
25 |
+
| direct | 90.76 |
|
26 |
+
| indirect | 79.16 |
|
27 |
+
| **free indirect (this tagger)** | **58.00** |
|
28 |
+
| reported | 70.47 |
|
29 |
+
|
30 |
+
----
|
31 |
+
|
32 |
+
**Demo Usage:**
|
33 |
+
|
34 |
+
```python
|
35 |
+
from flair.data import Sentence
|
36 |
+
from flair.models import SequenceTagger
|
37 |
+
|
38 |
+
|
39 |
+
sentence = Sentence('Sie sprachen über das Mittagessen. Sie fragte, wo das Essen sei. Woher sollte er das wissen? Dann sagte er: "Ich habe Hunger."')
|
40 |
+
|
41 |
+
rwtypes = ['direct', 'indirect', 'freeindirect', 'reported']
|
42 |
+
for rwtype in rwtypes:
|
43 |
+
model = SequenceTagger.load(f'aehrm/redewiedergabe-{rwtype}')
|
44 |
+
model.predict(sentence)
|
45 |
+
print(rwtype, [ x.data_point.text for x in sentence.get_labels() ])
|
46 |
+
# >>> direct ['"', 'Ich', 'habe', 'Hunger', '.', '"']
|
47 |
+
# >>> indirect ['wo', 'das', 'Essen', 'sei', '.']
|
48 |
+
# >>> freeindirect ['Woher', 'sollte', 'er', 'das', 'wissen', '?']
|
49 |
+
# >>> reported ['Sie', 'sprachen', 'über', 'das', 'Mittagessen', '.', 'Woher', 'sollte', 'er', 'das', 'wissen', '?']
|
50 |
+
```
|
51 |
+
|
52 |
+
**Cite**:
|
53 |
+
|
54 |
+
Please cite the following paper when using this model.
|
55 |
+
|
56 |
+
```
|
57 |
+
@inproceedings{ehrmanntraut-et-al-llpro-2023,
|
58 |
+
address = {Ingolstadt, Germany},
|
59 |
+
title = {{LLpro}: A Literary Language Processing Pipeline for {German} Narrative Text},
|
60 |
+
booktitle = {Proceedings of the 10th Conference on Natural Language Processing ({KONVENS} 2022)},
|
61 |
+
publisher = {{KONVENS} 2023 Organizers},
|
62 |
+
author = {Ehrmanntraut, Anton and Konle, Leonard and Jannidis, Fotis},
|
63 |
+
year = {2023},
|
64 |
+
}
|
65 |
+
```
|
66 |
+
|
67 |
+
|