Hugging Face's logo

Herelles
/
camembert-base-lupan

Text Classification
Transformers
Safetensors
PyTorch
French
camembert
text classification
urban planning
natural risks
risk management
geography
Model card Files Community
camembert-base-lupan
File size: 4,408 Bytes 
64a2a87
 
 
 
 
 
 
 
 
 
 
 
 
112f4e7
64a2a87
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
61de029
64a2a87
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
 
---
datasets:
- Herelles/lupan
language:
- fr
tags:
- text classification
- pytorch
- camembert
- urban planning
- natural risks
- risk management
- geography
inference: false
---
# CamemBERT LUPAN (Local Urban Plans And Natural risks)
## Overview

In France, urban planning and natural risk management operate the Local Land Plans (PLU – Plan Local d'Urbanisme) and the Natural risk prevention plans (PPRn – Plan de Prévention des Risques naturels) containing land use rules. To facilitate automatic extraction of the rules, we manually annotated a number of those documents concerning Montpellier, a rapidly evolving agglomeration exposed to natural risks, then fine-tuned a model.

This model classifies input text in French to determine if it contains an urban planning rule. It outputs one of 4 classes: Verifiable (indicating the possibility of verification with satellite images), Non-verifiable (indicating impossibility of verification with satellite images), Informative (containing non-strict rules in the form of recommendations), and Not pertinent (absence of any of the above rules). For better quality results, it is recommended to add a title and a subtitle to each textual input.

For more details please refer to our article:  https://www.nature.com/articles/s41597-023-02705-y

## Training and evaluation data 

The model is fine-tuned on top of CamemBERT using our corpus: https://huggingface.co/datasets/Herelles/lupan

This is the first corpus in the French language in the fields of urban planning and natural risk management. 

## Example of use

Attention: to run this code you need to have intalled `transformers`, `torch` and `numpy`. You can do it with `pip install transformers torch numpy`.

Load necessary libraries:
```
from transformers import CamembertTokenizer, CamembertForSequenceClassification

import torch

import numpy as np
```

Define tokenizer:
```
tokenizer = CamembertTokenizer.from_pretrained("camembert-base")
```

Define the model:
```
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')

model = CamembertForSequenceClassification.from_pretrained("herelles/camembert-base-lupan")

model.to(device)
```

Define segment to predict:
```
new_segment = '''Article 1 : Occupations ou utilisations du sol interdites
 
1) Dans l’ensemble de la zone sont interdits :
 
Les constructions destinées à l’habitation ne dépendant pas d’une exploitation agricole autres 
que celles visées à l’article 2 paragraphe 1).'''
```

Get the prediction:
```
test_ids = []
test_attention_mask = []

# Apply the tokenizer
encoding = tokenizer(new_segment, padding="longest", return_tensors="pt")

# Extract IDs and Attention Mask
test_ids.append(encoding['input_ids'])
test_attention_mask.append(encoding['attention_mask'])
test_ids = torch.cat(test_ids, dim = 0)
test_attention_mask = torch.cat(test_attention_mask, dim = 0)

# Forward pass, calculate logit predictions
with torch.no_grad():
  output = model(test_ids.to(device), token_type_ids = None, attention_mask = test_attention_mask.to(device))

prediction = np.argmax(output.logits.cpu().numpy()).flatten().item()

if prediction == 0:
  pred_label = 'Not pertinent'
elif prediction == 1:
  pred_label = 'Pertinent (Soft)'
elif prediction == 2:
  pred_label = 'Pertinent (Strict, Non-verifiable)'
elif prediction == 3:
  pred_label = 'Pertinent (Strict, Verifiable)'

print('Input text: ', new_segment)
print('\n\nPredicted Class: ', pred_label)
```

## Citation

To cite the data set please use:
```
@article{koptelov2023manually,
  title={A manually annotated corpus in French for the study of urbanization and the natural risk prevention},
  author={Koptelov, Maksim and Holveck, Margaux and Cremilleux, Bruno and Reynaud, Justine and Roche, Mathieu and Teisseire, Maguelonne},
  journal={Scientific Data},
  volume={10},
  number={1},
  pages={818},
  year={2023},
  publisher={Nature Publishing Group UK London}
}
 ```

To cite the code please use:
```
@inproceedings{koptelov2023towards,
  title={Towards a (Semi-) Automatic Urban Planning Rule Identification in the French Language},
  author={Koptelov, Maksim and Holveck, Margaux and Cremilleux, Bruno and Reynaud, Justine and Roche, Mathieu and Teisseire, Maguelonne},
  booktitle={2023 IEEE 10th International Conference on Data Science and Advanced Analytics (DSAA)},
  pages={1--10},
  year={2023},
  organization={IEEE}
}
```