Spaces:
Runtime error
Runtime error
Upload app.py
Browse files
app.py
ADDED
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import numpy as np
|
3 |
+
import pandas as pd
|
4 |
+
import re
|
5 |
+
import shap
|
6 |
+
|
7 |
+
from transformers import (
|
8 |
+
AutoTokenizer,
|
9 |
+
AutoModelForSequenceClassification,
|
10 |
+
TextClassificationPipeline,
|
11 |
+
)
|
12 |
+
|
13 |
+
tokenizer = AutoTokenizer.from_pretrained("chinhon/fake_tweet_detect")
|
14 |
+
|
15 |
+
model = AutoModelForSequenceClassification.from_pretrained("chinhon/fake_tweet_detect")
|
16 |
+
|
17 |
+
tweet_detector = TextClassificationPipeline(model=model, tokenizer=tokenizer)
|
18 |
+
|
19 |
+
# tweak the extent of text cleaning as you wish
|
20 |
+
def clean_text(text):
|
21 |
+
text = re.sub(r"http\S+", "", text)
|
22 |
+
text = re.sub(r"\n", " ", text)
|
23 |
+
text = re.sub(r"\'t", " not", text) # Change 't to 'not'
|
24 |
+
text = re.sub(r"(@.*?)[\s]", " ", text) # Remove @name
|
25 |
+
text = re.sub(r"$\d+\W+|\b\d+\b|\W+\d+$", " ", text) # remove digits
|
26 |
+
text = re.sub(r"[^\w\s\#]", "", text) # remove special characters except hashtags
|
27 |
+
text = text.strip(" ")
|
28 |
+
text = re.sub(
|
29 |
+
" +", " ", text
|
30 |
+
).strip() # get rid of multiple spaces and replace with a single
|
31 |
+
return text
|
32 |
+
|
33 |
+
def tweet_detect(text):
|
34 |
+
data = [clean_text(text)]
|
35 |
+
prediction = tweet_detector(data)
|
36 |
+
|
37 |
+
pred_label = [x.get("label") for x in prediction]
|
38 |
+
|
39 |
+
if pred_label == ["LABEL_1"]:
|
40 |
+
return "Fake Tweet"
|
41 |
+
elif pred_label == ["LABEL_0"]:
|
42 |
+
return "Real Tweet"
|
43 |
+
|
44 |
+
#Define Gradio interface
|
45 |
+
gradio_ui = gr.Interface(
|
46 |
+
fn=tweet_detect,
|
47 |
+
title="Detect Fake Tweets",
|
48 |
+
description="Enter a tweet and see if the transformer model can identify if it was written by state-backed trolls.",
|
49 |
+
inputs=gr.inputs.Textbox(lines=10, label="Paste tweet text here [English Only]"),
|
50 |
+
outputs=gr.outputs.Label(type="auto", label="Prediction"),
|
51 |
+
interpretation="shap",
|
52 |
+
enable_queue=True
|
53 |
+
)
|
54 |
+
|
55 |
+
gradio_ui.launch()
|