sofiaoliveira commited on
Commit
bcbe60f
1 Parent(s): cd6283f

Added README.md

Browse files
Files changed (1) hide show
  1. README.md +101 -0
README.md ADDED
@@ -0,0 +1,101 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - multilingual
4
+ - pt
5
+ tags:
6
+ - xlm-roberta-base
7
+ - semantic role labeling
8
+ - finetuned
9
+ license: Apache 2.0
10
+ datasets:
11
+ - PropBank.Br
12
+ metrics:
13
+ - F1 Measure
14
+ ---
15
+
16
+ # MyModel
17
+
18
+ ## Model description
19
+
20
+ This model is the [`xlm-roberta-base`](https://huggingface.co/xlm-roberta-base) fine-tuned on Portuguese semantic role labeling data. This is part of a project from which resulted the following models:
21
+
22
+ * [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
23
+ * [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
24
+ * [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
25
+ * [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
26
+ * [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
27
+ * [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
28
+ * [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
29
+ * [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
30
+ * [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
31
+ * [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
32
+ * [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
33
+ * [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
34
+ * [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
35
+ * [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
36
+
37
+ For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
38
+
39
+
40
+ ## Intended uses & limitations
41
+
42
+ #### How to use
43
+
44
+ To use the transformers portion of this model:
45
+ ```python
46
+ from transformers import AutoTokenizer, AutoModel
47
+
48
+ tokenizer = AutoTokenizer.from_pretrained("liaad/srl-pt_xlmr-base")
49
+ model = AutoModel.from_pretrained("liaad/srl-pt_xlmr-base")
50
+ ```
51
+
52
+ To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
53
+
54
+
55
+ #### Limitations and bias
56
+
57
+ - This model does not include a Tensorflow version. This is because the "type_vocab_size" in this model was changed (from 1 to 2) and, therefore, it cannot be easily converted to Tensorflow.
58
+
59
+
60
+ ## Training data
61
+
62
+ Pretrained weights were left identical to the original model [`xlm-roberta-base`](https://huggingface.co/xlm-roberta-base). A randomly initialized embeddings layer for "token_type_ids" was added.
63
+
64
+
65
+ ## Training procedure
66
+
67
+ The models were trained on the PropBank.Br datasets, using Cross-Validation. They were tested on the folds as well as on a smaller opinion dataset "Buscapé". For more information, please see the accompanying article (See BibTeX entry and citation info below) and the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
68
+
69
+ ## Eval results
70
+
71
+
72
+ | Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
73
+ | --------------- | ------ | ----- |
74
+ | `srl-pt_bertimbau-base` | 76.30 | 73.33 |
75
+ | `srl-pt_bertimbau-large` | 77.42 | 74.85 |
76
+ | `srl-pt_xlmr-base` | 75.22 | 72.82 |
77
+ | `srl-pt_xlmr-large` | 77.59 | 73.84 |
78
+ | `srl-pt_mbert-base` | 72.76 | 66.89 |
79
+ | `srl-en_xlmr-base` | 66.59 | 65.24 |
80
+ | `srl-en_xlmr-large` | 67.60 | 64.94 |
81
+ | `srl-en_mbert-base` | 63.07 | 58.56 |
82
+ | `srl-enpt_xlmr-base` | 76.50 | 73.74 |
83
+ | `srl-enpt_xlmr-large` | **78.22** | 74.55 |
84
+ | `srl-enpt_mbert-base` | 74.88 | 69.19 |
85
+ | `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
86
+ | `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
87
+ | `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
88
+
89
+
90
+ ### BibTeX entry and citation info
91
+
92
+ ```bibtex
93
+ @misc{oliveira2021transformers,
94
+ title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
95
+ author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
96
+ year={2021},
97
+ eprint={2101.01213},
98
+ archivePrefix={arXiv},
99
+ primaryClass={cs.CL}
100
+ }
101
+ ```