Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,69 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
# BART-perspectives Model
|
2 |
+
|
3 |
+
## Overview
|
4 |
+
|
5 |
+
The BART-perspectives model is a sequence-to-sequence Transformer model hosted on Hugging Face Model Hub. Built on top of Facebook's BART-large (specifically the `philschmid/bart-large-cnn-samsum` finetune), it is specifically designed to extract perspectives from textual data at scale. The model provides an in-depth analysis of the speaker's identity, their emotions, the object of these emotions, and the reason behind these emotions.
|
6 |
+
|
7 |
+
## Usage
|
8 |
+
|
9 |
+
It is designed to be used with the `perspectives` library:
|
10 |
+
|
11 |
+
```python
|
12 |
+
from perspectives import DataFrame
|
13 |
+
|
14 |
+
# Load DataFrame
|
15 |
+
df = DataFrame(texts = [list of sentences])
|
16 |
+
|
17 |
+
# Load model
|
18 |
+
df.load_model()
|
19 |
+
|
20 |
+
# Get perspectives
|
21 |
+
df.get_perspectives()
|
22 |
+
|
23 |
+
# Search
|
24 |
+
df.search(speaker='...', emotion='...')
|
25 |
+
```
|
26 |
+
|
27 |
+
You can use also this model directly with a pipeline for text generation:
|
28 |
+
|
29 |
+
```python
|
30 |
+
from transformers import pipeline
|
31 |
+
|
32 |
+
# Load the model
|
33 |
+
generator = pipeline('text-generation', model='helliun/bart-perspectives')
|
34 |
+
|
35 |
+
# Get perspective
|
36 |
+
perspective = generator("Describe the perspective of this text: <your text>", max_length=1024, do_sample=False)
|
37 |
+
print(perspective)
|
38 |
+
```
|
39 |
+
You can also use it with `transformers.AutoTokenizer` and `transformers.AutoModelForSeq2SeqLM`:
|
40 |
+
|
41 |
+
```python
|
42 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
43 |
+
|
44 |
+
# Load the model
|
45 |
+
tokenizer = AutoTokenizer.from_pretrained("helliun/bart-perspectives")
|
46 |
+
model = AutoModelForSeq2SeqLM.from_pretrained("helliun/bart-perspectives")
|
47 |
+
|
48 |
+
# Tokenize the sentence
|
49 |
+
inputs = tokenizer.encode("Describe the perspective for this sentence: <your text>", return_tensors='pt')
|
50 |
+
|
51 |
+
# Pass the tensor through the model
|
52 |
+
results = model.generate(inputs)
|
53 |
+
|
54 |
+
# Decode the results
|
55 |
+
decoded = tokenizer.decode(results[:,0])
|
56 |
+
print(decoded)
|
57 |
+
```
|
58 |
+
|
59 |
+
## Training and Evaluation
|
60 |
+
|
61 |
+
The model was fine-tuned on a diverse corpus encompassing multiple subjects and emotion ranges. Please refer to the model's card on Hugging Face for detailed information about specific datasets, metrics, training, and evaluation processes.
|
62 |
+
|
63 |
+
## Contributing and Support
|
64 |
+
|
65 |
+
Please raise an issue on the [Hugging Face Model card](https://huggingface.co/helliun/bart-perspectives) if you encounter any problems using the model. Contributions like fine-tuning on additional data or improving the model architecture are always welcome!
|
66 |
+
|
67 |
+
## License
|
68 |
+
|
69 |
+
The model is open source and free to use under the MIT license.
|