Spaces:
Sleeping
Sleeping
vkovacs
commited on
Commit
·
fb05782
1
Parent(s):
158b5a1
sentence split logic added
Browse files
app.py
CHANGED
@@ -1,5 +1,6 @@
|
|
1 |
import os
|
2 |
import torch
|
|
|
3 |
import numpy as np
|
4 |
from transformers import AutoModelForSequenceClassification
|
5 |
from transformers import AutoTokenizer
|
@@ -16,6 +17,23 @@ HF_TOKEN = os.environ["hf_read"]
|
|
16 |
SENTIMENT_LABEL_NAMES = {0: "Negative", 1: "No sentiment or Neutral sentiment", 2: "Positive"}
|
17 |
LANGUAGES = ["Czech", "English", "French", "German", "Hungarian", "Polish", "Slovakian"]
|
18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
|
20 |
def build_huggingface_path(language: str):
|
21 |
if language == "Czech" or language == "Slovakian":
|
@@ -39,22 +57,48 @@ def predict(text, model_id, tokenizer_id):
|
|
39 |
logits = model(**inputs).logits
|
40 |
|
41 |
probs = torch.nn.functional.softmax(logits, dim=1).cpu().numpy().flatten()
|
42 |
-
|
43 |
-
|
44 |
-
return
|
|
|
45 |
|
46 |
def predict_wrapper(text, language):
|
47 |
model_id = build_huggingface_path(language)
|
48 |
tokenizer_id = "xlm-roberta-large"
|
49 |
-
return predict(text, model_id, tokenizer_id)
|
50 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
|
52 |
with gr.Blocks() as demo:
|
53 |
-
gr.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
fn=predict_wrapper,
|
55 |
-
inputs=[
|
56 |
-
|
57 |
-
|
58 |
|
59 |
if __name__ == "__main__":
|
60 |
demo.launch()
|
|
|
|
1 |
import os
|
2 |
import torch
|
3 |
+
import spacy
|
4 |
import numpy as np
|
5 |
from transformers import AutoModelForSequenceClassification
|
6 |
from transformers import AutoTokenizer
|
|
|
17 |
SENTIMENT_LABEL_NAMES = {0: "Negative", 1: "No sentiment or Neutral sentiment", 2: "Positive"}
|
18 |
LANGUAGES = ["Czech", "English", "French", "German", "Hungarian", "Polish", "Slovakian"]
|
19 |
|
20 |
+
def load_spacy_model(model_name="xx_sent_ud_sm"):
|
21 |
+
try:
|
22 |
+
model = spacy.load(model_name)
|
23 |
+
except OSError:
|
24 |
+
spacy.cli.download(model_name)
|
25 |
+
model = spacy.load(model_name)
|
26 |
+
return model
|
27 |
+
|
28 |
+
def split_sentences(text, model):
|
29 |
+
# disable pipeline components not necessary for splitting
|
30 |
+
model.disable_pipes(model.pipe_names) # first disable all the pipes
|
31 |
+
model.enable_pipe("senter") # then enable the sentence splitter only
|
32 |
+
|
33 |
+
doc = model(text)
|
34 |
+
sentences = [sent.text for sent in doc.sents]
|
35 |
+
|
36 |
+
return sentences
|
37 |
|
38 |
def build_huggingface_path(language: str):
|
39 |
if language == "Czech" or language == "Slovakian":
|
|
|
57 |
logits = model(**inputs).logits
|
58 |
|
59 |
probs = torch.nn.functional.softmax(logits, dim=1).cpu().numpy().flatten()
|
60 |
+
label_pred = model.config.id2label[probs.argmax()]
|
61 |
+
probability_pred = f"{100*probs.max()}%"
|
62 |
+
return label_pred, probability_pred
|
63 |
+
|
64 |
|
65 |
def predict_wrapper(text, language):
|
66 |
model_id = build_huggingface_path(language)
|
67 |
tokenizer_id = "xlm-roberta-large"
|
|
|
68 |
|
69 |
+
spacy_model = load_spacy_model()
|
70 |
+
sentences = split_sentences(text, spacy_model)
|
71 |
+
|
72 |
+
results = []
|
73 |
+
for sentence in sentences:
|
74 |
+
label, probability = predict(sentence, model_id, tokenizer_id)
|
75 |
+
results.append({
|
76 |
+
"Sentence": sentence,
|
77 |
+
"Prediction": label,
|
78 |
+
"Probability": probability
|
79 |
+
})
|
80 |
+
|
81 |
+
output_info = f'Prediction made using the <a href="https://huggingface.co/{model_id}">{model_id}</a> model.'
|
82 |
+
return results, output_info
|
83 |
|
84 |
with gr.Blocks() as demo:
|
85 |
+
with gr.Row():
|
86 |
+
with gr.Column():
|
87 |
+
input_text = gr.Textbox(lines=6, label="Input Text", placeholder="Enter your text here...")
|
88 |
+
language_choice = gr.Dropdown(choices=LANGUAGES, label="Language", value="English")
|
89 |
+
predict_button = gr.Button("Submit")
|
90 |
+
|
91 |
+
with gr.Column():
|
92 |
+
result_table = gr.Dataframe(headers=["Sentence", "Prediction", "Probability"],
|
93 |
+
label="Sentence-level Predictions")
|
94 |
+
model_info = gr.Markdown()
|
95 |
+
|
96 |
+
predict_button.click(
|
97 |
fn=predict_wrapper,
|
98 |
+
inputs=[input_text, language_choice],
|
99 |
+
outputs=[result_table, model_info]
|
100 |
+
)
|
101 |
|
102 |
if __name__ == "__main__":
|
103 |
demo.launch()
|
104 |
+
|