Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -5,7 +5,7 @@ import spacy
|
|
5 |
import subprocess
|
6 |
import nltk
|
7 |
from nltk.corpus import wordnet
|
8 |
-
import
|
9 |
from gensim import downloader as api
|
10 |
|
11 |
# Ensure necessary NLTK data is downloaded
|
@@ -28,76 +28,3 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
|
28 |
# Load AI Detector model and tokenizer from Hugging Face (DistilBERT)
|
29 |
tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english")
|
30 |
model = AutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english").to(device)
|
31 |
-
|
32 |
-
# Function to correct grammar using language-tool-python
|
33 |
-
def correct_grammar_with_language_tool(text):
|
34 |
-
tool = language_tool_python.LanguageTool('en-US')
|
35 |
-
matches = tool.check(text)
|
36 |
-
corrected_text = language_tool_python.utils.correct(text, matches)
|
37 |
-
return corrected_text
|
38 |
-
|
39 |
-
# AI detection function using DistilBERT
|
40 |
-
def detect_ai_generated(text):
|
41 |
-
inputs = tokenizer(text, return_tensors="pt", truncation=True, max_length=512).to(device)
|
42 |
-
with torch.no_grad():
|
43 |
-
outputs = model(**inputs)
|
44 |
-
probabilities = torch.softmax(outputs.logits, dim=1)
|
45 |
-
ai_probability = probabilities[0][1].item() # Probability of being AI-generated
|
46 |
-
return f"AI-Generated Content Probability: {ai_probability:.2f}%"
|
47 |
-
|
48 |
-
# Function to get synonyms using NLTK WordNet
|
49 |
-
def get_synonyms_nltk(word, pos):
|
50 |
-
synsets = wordnet.synsets(word, pos=pos)
|
51 |
-
if synsets:
|
52 |
-
lemmas = synsets[0].lemmas()
|
53 |
-
return [lemma.name() for lemma in lemmas]
|
54 |
-
return []
|
55 |
-
|
56 |
-
# Paraphrasing function using spaCy and NLTK with grammar correction
|
57 |
-
def paraphrase_with_spacy_nltk(text):
|
58 |
-
doc = nlp(text)
|
59 |
-
paraphrased_words = []
|
60 |
-
|
61 |
-
for token in doc:
|
62 |
-
# Map spaCy POS tags to WordNet POS tags
|
63 |
-
pos = None
|
64 |
-
if token.pos_ in {"NOUN"}:
|
65 |
-
pos = wordnet.NOUN
|
66 |
-
elif token.pos_ in {"VERB"}:
|
67 |
-
pos = wordnet.VERB
|
68 |
-
elif token.pos_ in {"ADJ"}:
|
69 |
-
pos = wordnet.ADJ
|
70 |
-
elif token.pos_ in {"ADV"}:
|
71 |
-
pos = wordnet.ADV
|
72 |
-
|
73 |
-
synonyms = get_synonyms_nltk(token.text.lower(), pos) if pos else []
|
74 |
-
|
75 |
-
# Replace with a synonym only if it makes sense
|
76 |
-
if synonyms and token.pos_ in {"NOUN", "VERB", "ADJ", "ADV"} and synonyms[0] != token.text.lower():
|
77 |
-
paraphrased_words.append(synonyms[0])
|
78 |
-
else:
|
79 |
-
paraphrased_words.append(token.text)
|
80 |
-
|
81 |
-
# Join the words back into a sentence
|
82 |
-
paraphrased_sentence = ' '.join(paraphrased_words)
|
83 |
-
|
84 |
-
# Correct the grammar of the paraphrased sentence using language-tool-python
|
85 |
-
corrected_sentence = correct_grammar_with_language_tool(paraphrased_sentence)
|
86 |
-
|
87 |
-
return corrected_sentence
|
88 |
-
|
89 |
-
# Gradio interface definition
|
90 |
-
with gr.Blocks() as interface:
|
91 |
-
with gr.Row():
|
92 |
-
with gr.Column():
|
93 |
-
text_input = gr.Textbox(lines=5, label="Input Text")
|
94 |
-
detect_button = gr.Button("AI Detection")
|
95 |
-
paraphrase_button = gr.Button("Paraphrase with spaCy & NLTK (Grammar Corrected)")
|
96 |
-
with gr.Column():
|
97 |
-
output_text = gr.Textbox(label="Output")
|
98 |
-
|
99 |
-
detect_button.click(detect_ai_generated, inputs=text_input, outputs=output_text)
|
100 |
-
paraphrase_button.click(paraphrase_with_spacy_nltk, inputs=text_input, outputs=output_text)
|
101 |
-
|
102 |
-
# Launch the Gradio app
|
103 |
-
interface.launch(debug=False)
|
|
|
5 |
import subprocess
|
6 |
import nltk
|
7 |
from nltk.corpus import wordnet
|
8 |
+
import language_check # Use language-check instead of language-tool-python
|
9 |
from gensim import downloader as api
|
10 |
|
11 |
# Ensure necessary NLTK data is downloaded
|
|
|
28 |
# Load AI Detector model and tokenizer from Hugging Face (DistilBERT)
|
29 |
tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english")
|
30 |
model = AutoModelForSequenceClassification.from_pretrained("distilbert-base-uncased-finetuned-sst-2-english").to(device)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|