Upload 4 files

README.md

@@ -1,9 +1,11 @@
 ---
 title: Ai Text Detector
-emoji: 📉
-colorFrom: yellow
-colorTo: yellow
-sdk: static
+emoji: 🐢
+colorFrom: blue
+colorTo: gray
+sdk: gradio
+sdk_version: 5.10.0
+app_file: app.py
 pinned: false
 license: apache-2.0
 ---

app.py

@@ -0,0 +1,32 @@
+import os
+
+import gradio as gr
+import torch
+from transformers import pipeline
+
+from utils import clean_text
+
+
+pipeline = pipeline(
+    task="text-classification",
+    model="fakespot-ai/roberta-base-ai-text-detection-v1",
+    device="cuda" if torch.cuda.is_available() else "cpu",
+    token=os.environ.get("ACCESS_TOKEN")
+)
+
+
+def predict(text):
+    cleaned_text = clean_text(text)
+    predictions = pipeline(cleaned_text, top_k=None)[0]
+    return {
+        p["label"]: p["score"] for p in predictions
+    }
+
+
+demo = gr.Interface(
+    predict,
+    inputs=gr.Textbox(),
+    outputs=gr.Label(num_top_classes=2),
+    title="AI Text Detector"
+)
+demo.launch()

requirements.txt

@@ -0,0 +1,2 @@
+transformers
+torch

utils.py

@@ -0,0 +1,42 @@
+import re
+from html import unescape
+
+
+def clean_text(t):
+    t = clean_markdown(t)
+    t = t.replace("\n"," ")
+    t = t.replace("\t"," ")
+    t = t.replace("^M"," ")
+    t = t.replace("\r"," ")
+    t = t.replace(" ,", ",")
+    t = re.sub(" +", " ", t)
+    return t
+
+
+def clean_markdown(md_text):
+    # Remove code blocks
+    md_text = re.sub(r'```.*?```', '', md_text, flags=re.DOTALL)
+    # Remove inline code
+    md_text = re.sub(r'`[^`]*`', '', md_text)
+    # Remove images
+    md_text = re.sub(r'!\[.*?\]\(.*?\)', '', md_text)
+    # Remove links but keep link text
+    md_text = re.sub(r'\[([^\]]+)\]\(.*?\)', r'\1', md_text)
+    # Remove bold and italic (groups of *, _)
+    md_text = re.sub(r'(\*\*|__)(.*?)\1', r'\2', md_text)
+    md_text = re.sub(r'(\*|_)(.*?)\1', r'\2', md_text)
+    # Remove headings
+    md_text = re.sub(r'#+ ', '', md_text)
+    # Remove blockquotes
+    md_text = re.sub(r'^>.*$', '', md_text, flags=re.MULTILINE)
+    # Remove list markers
+    md_text = re.sub(r'^(\s*[-*+]|\d+\.)\s+', '', md_text, flags=re.MULTILINE)
+    # Remove horizontal rules
+    md_text = re.sub(r'^\s*[-*_]{3,}\s*$', '', md_text, flags=re.MULTILINE)
+    # Remove tables
+    md_text = re.sub(r'\|.*?\|', '', md_text)
+    # Remove raw HTML tags
+    md_text = re.sub(r'<.*?>', '', md_text)
+    # Decode HTML entities
+    md_text = unescape(md_text)
+    return md_text