sofiaoliveira commited on
Commit
a0c1e67
1 Parent(s): 6f4632a

Create README.md

Browse files
Files changed (1) hide show
  1. README.md +97 -0
README.md ADDED
@@ -0,0 +1,97 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - multilingual
4
+ - pt
5
+ tags:
6
+ - bert-large-portuguese-cased
7
+ - semantic role labeling
8
+ - finetuned
9
+ - dependency parsing
10
+ license: Apache 2.0
11
+ datasets:
12
+ - PropBank.Br
13
+ - CoNLL-2012
14
+ - Universal Dependencies
15
+ metrics:
16
+ - F1 Measure
17
+ ---
18
+
19
+ # BERTimbau large fine-tune in Portuguese Universal Dependencies and semantic role labeling
20
+
21
+ ## Model description
22
+
23
+ This model is the [`neuralmind/bert-large-portuguese-cased`](https://huggingface.co/neuralmind/bert-large-portuguese-cased) fine-tuned first on the Universal Dependencies Portuguese dataset and then fine-tuned on the PropBank.Br data. This is part of a project from which resulted the following models:
24
+
25
+ * [liaad/srl-pt_bertimbau-base](https://huggingface.co/liaad/srl-pt_bertimbau-base)
26
+ * [liaad/srl-pt_bertimbau-large](https://huggingface.co/liaad/srl-pt_bertimbau-large)
27
+ * [liaad/srl-pt_xlmr-base](https://huggingface.co/liaad/srl-pt_xlmr-base)
28
+ * [liaad/srl-pt_xlmr-large](https://huggingface.co/liaad/srl-pt_xlmr-large)
29
+ * [liaad/srl-pt_mbert-base](https://huggingface.co/liaad/srl-pt_mbert-base)
30
+ * [liaad/srl-en_xlmr-base](https://huggingface.co/liaad/srl-en_xlmr-base)
31
+ * [liaad/srl-en_xlmr-large](https://huggingface.co/liaad/srl-en_xlmr-large)
32
+ * [liaad/srl-en_mbert-base](https://huggingface.co/liaad/srl-en_mbert-base)
33
+ * [liaad/srl-enpt_xlmr-base](https://huggingface.co/liaad/srl-enpt_xlmr-base)
34
+ * [liaad/srl-enpt_xlmr-large](https://huggingface.co/liaad/srl-enpt_xlmr-large)
35
+ * [liaad/srl-enpt_mbert-base](https://huggingface.co/liaad/srl-enpt_mbert-base)
36
+ * [liaad/ud_srl-pt_bertimbau-large](https://huggingface.co/liaad/ud_srl-pt_bertimbau-large)
37
+ * [liaad/ud_srl-pt_xlmr-large](https://huggingface.co/liaad/ud_srl-pt_xlmr-large)
38
+ * [liaad/ud_srl-enpt_xlmr-large](https://huggingface.co/liaad/ud_srl-enpt_xlmr-large)
39
+
40
+ 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).
41
+
42
+ ## Intended uses & limitations
43
+
44
+ #### How to use
45
+
46
+ To use the transformers portion of this model:
47
+ ```python
48
+ from transformers import AutoTokenizer, AutoModel
49
+
50
+ tokenizer = AutoTokenizer.from_pretrained("liaad/ud_srl-pt_bertimbau-large")
51
+ model = AutoModel.from_pretrained("liaad/ud_srl-pt_bertimbau-large")
52
+ ```
53
+
54
+ To use the full SRL model (transformers portion + a decoding layer), refer to the [project's github](https://github.com/asofiaoliveira/srl_bert_pt).
55
+
56
+
57
+ #### Limitations and bias
58
+
59
+ - The model was trained only for 10 epochs in the Universal Dependencies dataset.
60
+
61
+
62
+ ## Training procedure
63
+
64
+ The model was trained on the Universal Dependencies Portuguese dataset; then on Portuguese semantic role labeling data (PropBank.Br) using 10-fold Cross-Validation. The 10 resulting models 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).
65
+
66
+ ## Eval results
67
+
68
+ | Model Name | F<sub>1</sub> CV PropBank.Br (in domain) | F<sub>1</sub> Buscapé (out of domain) |
69
+ | --------------- | ------ | ----- |
70
+ | `srl-pt_bertimbau-base` | 76.30 | 73.33 |
71
+ | `srl-pt_bertimbau-large` | 77.42 | 74.85 |
72
+ | `srl-pt_xlmr-base` | 75.22 | 72.82 |
73
+ | `srl-pt_xlmr-large` | 77.59 | 73.84 |
74
+ | `srl-pt_mbert-base` | 72.76 | 66.89 |
75
+ | `srl-en_xlmr-base` | 66.59 | 65.24 |
76
+ | `srl-en_xlmr-large` | 67.60 | 64.94 |
77
+ | `srl-en_mbert-base` | 63.07 | 58.56 |
78
+ | `srl-enpt_xlmr-base` | 76.50 | 73.74 |
79
+ | `srl-enpt_xlmr-large` | **78.22** | 74.55 |
80
+ | `srl-enpt_mbert-base` | 74.88 | 69.19 |
81
+ | `ud_srl-pt_bertimbau-large` | 77.53 | 74.49 |
82
+ | `ud_srl-pt_xlmr-large` | 77.69 | 74.91 |
83
+ | `ud_srl-enpt_xlmr-large` | 77.97 | **75.05** |
84
+
85
+
86
+ ### BibTeX entry and citation info
87
+
88
+ ```bibtex
89
+ @misc{oliveira2021transformers,
90
+ title={Transformers and Transfer Learning for Improving Portuguese Semantic Role Labeling},
91
+ author={Sofia Oliveira and Daniel Loureiro and Alípio Jorge},
92
+ year={2021},
93
+ eprint={2101.01213},
94
+ archivePrefix={arXiv},
95
+ primaryClass={cs.CL}
96
+ }
97
+ ```