README.md
3.4 KB · 93 lines · markdown Raw
1 ---
2 language: pt
3 license: apache-2.0
4
5 widget:
6 - text: "O futuro de DI caiu 20 bps nesta manhã"
7 example_title: "Example 1"
8 - text: "O Nubank decidiu cortar a faixa de preço da oferta pública inicial (IPO) após revés no humor dos mercados internacionais com as fintechs."
9 example_title: "Example 2"
10 - text: "O Ibovespa acompanha correção do mercado e fecha com alta moderada"
11 example_title: "Example 3"
12 ---
13
14 # FinBERT-PT-BR : Financial BERT PT BR
15
16 FinBERT-PT-BR is a pre-trained NLP model to analyze sentiment of Brazilian Portuguese financial texts.
17
18 The model was trained in two main stages: language modeling and sentiment modeling. In the first stage, a language model was trained with more than 1.4 million texts of financial news in Portuguese.
19 From this first training, it was possible to build a sentiment classifier with few labeled texts (500) that presented a satisfactory convergence.
20
21 At the end of the work, a comparative analysis with other models and the possible applications of the developed model are presented.
22 In the comparative analysis, it was possible to observe that the developed model presented better results than the current models in the state of the art.
23 Among the applications, it was demonstrated that the model can be used to build sentiment indices, investment strategies and macroeconomic data analysis, such as inflation.
24
25 ## Applications
26
27 ### Sentiment Index
28
29 ![Sentiment Index](sentiment_index_and_economy.png)
30
31
32 ## Usage
33
34 #### BertForSequenceClassification
35
36 ```python
37 from transformers import AutoTokenizer, BertForSequenceClassification
38 import numpy as np
39
40 pred_mapper = {
41 0: "POSITIVE",
42 1: "NEGATIVE",
43 2: "NEUTRAL"
44 }
45
46 tokenizer = AutoTokenizer.from_pretrained("lucas-leme/FinBERT-PT-BR")
47 finbertptbr = BertForSequenceClassification.from_pretrained("lucas-leme/FinBERT-PT-BR")
48
49 tokens = tokenizer(["Hoje a bolsa caiu", "Hoje a bolsa subiu"], return_tensors="pt",
50 padding=True, truncation=True, max_length=512)
51 finbertptbr_outputs = finbertptbr(**tokens)
52
53 preds = [pred_mapper[np.argmax(pred)] for pred in finbertptbr_outputs.logits.cpu().detach().numpy()]
54 ```
55
56 #### Pipeline
57
58 ```python
59 from transformers import (
60 AutoTokenizer,
61 BertForSequenceClassification,
62 pipeline,
63 )
64
65 finbert_pt_br_tokenizer = AutoTokenizer.from_pretrained("lucas-leme/FinBERT-PT-BR")
66 finbert_pt_br_model = BertForSequenceClassification.from_pretrained("lucas-leme/FinBERT-PT-BR")
67
68 finbert_pt_br_pipeline = pipeline(task='text-classification', model=finbert_pt_br_model, tokenizer=finbert_pt_br_tokenizer)
69 finbert_pt_br_pipeline(['Hoje a bolsa caiu', 'Hoje a bolsa subiu'])
70 ```
71
72 ## Author
73
74 - [Lucas Leme](https://www.linkedin.com/in/lucas-leme-santos/) - lucaslssantos99@gmail.com
75
76 ## Citation
77
78 ```latex
79 @inproceedings{santos2023finbert,
80 title={FinBERT-PT-BR: An{\'a}lise de Sentimentos de Textos em Portugu{\^e}s do Mercado Financeiro},
81 author={Santos, Lucas L and Bianchi, Reinaldo AC and Costa, Anna HR},
82 booktitle={Anais do II Brazilian Workshop on Artificial Intelligence in Finance},
83 pages={144--155},
84 year={2023},
85 organization={SBC}
86 }
87 ```
88
89 ## Paper
90
91 - Paper: [FinBERT-PT-BR: Sentiment Analysis of Texts in Portuguese from the Financial Market](https://sol.sbc.org.br/index.php/bwaif/article/view/24960)
92 - Undergraduate thesis: [FinBERT-PT-BR: Análise de sentimentos de textos em português referentes ao mercado financeiro](https://pcs.usp.br/pcspf/wp-content/uploads/sites/8/2022/12/Monografia_PCS3860_COOP_2022_Grupo_C12.pdf)
93