File size: 8,818 Bytes
10762db
676f964
5acf869
 
 
676f964
5acf869
 
 
 
 
 
 
 
8a5de30
5acf869
 
 
 
 
 
 
8a5de30
 
 
 
5acf869
 
0e74ffd
5acf869
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
5adc59f
5acf869
 
 
 
 
 
 
 
 
5adc59f
 
5acf869
 
 
 
 
 
 
 
 
 
 
 
 
 
676f964
5acf869
 
 
 
 
 
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
128
129
130
131
132
133
134
135
136
137
138
139
140
---
pipeline_tag: sentence-similarity
language: fr
datasets:
- stsb_multi_mt
tags:
- Text
- Sentence Similarity
- Sentence-Embedding
- camembert-large
license: apache-2.0
model-index:
- name: sentence-camembert-large by Van Tuan DANG
  results:
  - task:
      name: Sentence-Embedding
      type: Text Similarity
    dataset:
      name: Text Similarity fr
      type: stsb_multi_mt
      args: fr
    metrics:
    - name: Test Pearson correlation coefficient
      type: Pearson_correlation_coefficient
      value: 88.63
library_name: sentence-transformers
---
## Description:
This [**Sentence-CamemBERT-Large**](https://huggingface.co/Lajavaness/sentence-camembert-large) Model is an Embedding Model for French developed by [La Javaness](https://www.lajavaness.com/).  The purpose of this embedding model is to represent the content and semantics of a French sentence as a mathematical vector, allowing it to understand the meaning of the text beyond individual words in queries and documents. It offers powerful semantic search capabilities.
## Pre-trained sentence embedding models are state-of-the-art of Sentence Embeddings for French.

The [Lajavaness/sentence-camembert-large](https://huggingface.co/Lajavaness/sentence-camembert-large) model is an improvement over the [dangvantuan/sentence-camembert-base](https://huggingface.co/dangvantuan/sentence-camembert-large)  offering greater robustness and better performance on all STS benchmark datasets. It has been fine-tuned using the pre-trained [facebook/camembert-large](https://huggingface.co/camembert/camembert-large) and
[Siamese BERT-Networks with 'sentences-transformers'](https://www.sbert.net/) on dataset [stsb](https://huggingface.co/datasets/stsb_multi_mt/viewer/fr/train).  Additionally, it has been combined with [Augmented SBERT](https://aclanthology.org/2021.naacl-main.28.pdf) on dataset [stsb](https://huggingface.co/datasets/stsb_multi_mt/viewer/fr/train). The model benefits from Pair Sampling Strategies using two models: [CrossEncoder-camembert-large](https://huggingface.co/dangvantuan/CrossEncoder-camembert-large) and [dangvantuan/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large)

## Usage
The model can be used directly (without a language model) as follows:

```python
from sentence_transformers import SentenceTransformer
model =  SentenceTransformer("Lajavaness/sentence-camembert-large")

sentences = ["Un avion est en train de décoller.",
          "Un homme joue d'une grande flûte.",
          "Un homme étale du fromage râpé sur une pizza.",
          "Une personne jette un chat au plafond.",
          "Une personne est en train de plier un morceau de papier.",
          ]

embeddings = model.encode(sentences)
```

## Evaluation
The model can be evaluated as follows on the French test data of stsb.

```python
from sentence_transformers import SentenceTransformer
from sentence_transformers.readers import InputExample
from datasets import load_dataset
def convert_dataset(dataset):
    dataset_samples=[]
    for df in dataset:
        score = float(df['similarity_score'])/5.0  # Normalize score to range 0 ... 1
        inp_example = InputExample(texts=[df['sentence1'], 
                                    df['sentence2']], label=score)
        dataset_samples.append(inp_example)
    return dataset_samples

# Loading the dataset for evaluation
df_dev = load_dataset("stsb_multi_mt", name="fr", split="dev")
df_test = load_dataset("stsb_multi_mt", name="fr", split="test")

# Convert the dataset for evaluation

# For Dev set:
dev_samples = convert_dataset(df_dev)
val_evaluator = EmbeddingSimilarityEvaluator.from_input_examples(dev_samples, name='sts-dev')
val_evaluator(model, output_path="./")

# For Test set:
test_samples = convert_dataset(df_test)
test_evaluator = EmbeddingSimilarityEvaluator.from_input_examples(test_samples, name='sts-test')
test_evaluator(model, output_path="./")
```

**Test Result**: 
The performance is measured using Pearson and Spearman correlation:
- On dev


| Model  | Pearson correlation | Spearman correlation  | #params  |
| ------------- | ------------- | ------------- |------------- |
| [Lajavaness/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large)| **88.63** |**88.46** | 336M| 
| [dangvantuan/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large)| 88.2 |88.02 | 336M| 
| [Sahajtomar/french_semanti](https://huggingface.co/Sahajtomar/french_semantic)| 87.44 |87.30 | 336M| 
| [Lajavaness/sentence-flaubert-base](https://huggingface.co/Lajavaness/sentence-flaubert-base)| 87.14 |87.10 | 137M |
| [GPT-3 (text-davinci-003)](https://platform.openai.com/docs/models) | 85 | NaN|175B |
| [GPT-(text-embedding-ada-002)](https://platform.openai.com/docs/models) | 79.75 | 80.44|NaN |


- On test, Pearson and Spearman correlation are evaluated on many different benchmark datasets:


**Pearson score**
| Model                                                      | [STS-B](https://huggingface.co/datasets/stsb_multi_mt/viewer/fr/train)   | [STS12-fr ](https://huggingface.co/datasets/Lajavaness/STS12-fr)| [STS13-fr](https://huggingface.co/datasets/Lajavaness/STS13-fr) | [STS14-fr](https://huggingface.co/datasets/Lajavaness/STS14-fr) | [STS15-fr](https://huggingface.co/datasets/Lajavaness/STS15-fr) | [STS16-fr](https://huggingface.co/datasets/Lajavaness/STS16-fr) | [SICK-fr](https://huggingface.co/datasets/Lajavaness/SICK-fr) | params |
|------------------------------------------|-------|----------|----------|----------|----------|----------|---------|--------|
| [Lajavaness/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large)      | **86.26** | **87.42**    | **89.34**    | **88.05**    | **88.91**    | 77.15    | 83.13   | 336M   |
| [dangvantuan/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large)     | 85.88 | 87.28    | 89.25    | 87.91    | 88.54    | 76.90    | 83.26   | 336M   |
| [Sahajtomar/french_semantic](https://huggingface.co/Sahajtomar/french_semantic)               | 85.80 | 86.05    | 88.50    | 86.57    | 87.49    | 77.85    | 83.27   | 336M    |
| [Lajavaness/sentence-flaubert-base](https://huggingface.co/Lajavaness/sentence-flaubert-base)        | 85.39 | 86.64    | 87.24    | 85.68    | 87.99    | 75.78    | 82.84   | 137M   |
| [GPT3 (text-embedding-ada-002)](https://platform.openai.com/docs/models)            | 79.03 | 66.16    | 75.48    | 70.69    | 77.88    | 65.18    | -       | -      |


**Spearman score**
| Model                                                      | [STS-B](https://huggingface.co/datasets/stsb_multi_mt/viewer/fr/train)   | [STS12-fr ](https://huggingface.co/datasets/Lajavaness/STS12-fr)| [STS13-fr](https://huggingface.co/datasets/Lajavaness/STS13-fr) | [STS14-fr](https://huggingface.co/datasets/Lajavaness/STS14-fr) | [STS15-fr](https://huggingface.co/datasets/Lajavaness/STS15-fr) | [STS16-fr](https://huggingface.co/datasets/Lajavaness/STS16-fr) | [SICK-fr](https://huggingface.co/datasets/Lajavaness/SICK-fr) | params |
|:-------------------------------------|-------:|---------:|---------:|---------:|---------:|---------:|--------:|:-------|
| [Lajavaness/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large)   | **86.14** | **81.22**    | 88.61    | **86.28**    | **89.01**    | 78.65    | **77.71**   | 336M   |
| [dangvantuan/sentence-camembert-large](https://huggingface.co/dangvantuan/sentence-camembert-large) | 85.78 | 81.09    | 88.68    | 85.81    | 88.56    | 78.49    | 77.70   | 336M   |
| [Sahajtomar/french_semantic](https://huggingface.co/Sahajtomar/french_semantic)          | 85.55 | 77.92    | 87.85    | 83.96    | 87.63    | 79.07    | 77.14   | 336M   |
| [Lajavaness/sentence-flaubert-base](https://huggingface.co/Lajavaness/sentence-flaubert-base)    | 85.67 | 79.97    | 86.91    | 84.57    | 88.10    | 77.84    | 77.55   | 137M    |
| [GPT3 (text-embedding-ada-002)](https://platform.openai.com/docs/models)         | 77.53 | 64.27    | 76.41    | 69.63    | 78.65    | 75.30    | -       | -      |



## Citation


	@article{reimers2019sentence,
	   title={Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks},
	   author={Nils Reimers, Iryna Gurevych},
	   journal={https://arxiv.org/abs/1908.10084},
	   year={2019}
	}


	@article{martin2020camembert,
	   title={CamemBERT: a Tasty French Language Mode},
	   author={Martin, Louis and Muller, Benjamin and Su{\'a}rez, Pedro Javier Ortiz and Dupont, Yoann and Romary, Laurent and de la Clergerie, {\'E}ric Villemonte and Seddah, Djam{\'e} and Sagot, Beno{\^\i}t},
	   journal={Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics},
	   year={2020}
	}