File size: 1,245 Bytes
6633877
 
e5e31f2
6633877
9a6d3bb
e5e31f2
 
6633877
e5e31f2
 
a03498c
e5e31f2
 
9a6d3bb
a03498c
9a6d3bb
 
 
 
 
 
e5e31f2
 
9a6d3bb
 
a03498c
6633877
e5e31f2
a03498c
6633877
 
 
e5e31f2
a03498c
 
 
9a6d3bb
6633877
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
from transformers import pipeline
import gradio as gr
import re

# Load models once for speed
spam_pipe = pipeline("text-classification", model="Titeiiko/OTIS-Official-Spam-Model")
zero_shot = pipeline("zero-shot-classification", model="facebook/bart-large-mnli")

def is_gibberish(text: str) -> bool:
    letters = len(re.findall(r"[a-zA-Z]", text))
    return len(text) == 0 or (letters / len(text) < 0.6)

def detect(text: str) -> dict:
    # 1️⃣ Ad/spam check
    spam_flag = spam_pipe(text)[0]["label"] != "LABEL_0"

    # 2️⃣ Relevance check (anything not a complaint counts as spam)
    zero_result = zero_shot(text, candidate_labels=["complaint", "not complaint"])
    not_complaint_flag = zero_result["labels"][0] == "not complaint"

    # 3️⃣ Gibberish check
    gibberish_flag = is_gibberish(text)

    # ✅ Final decision
    spam = spam_flag or not_complaint_flag or gibberish_flag

    return {
        "input": text,
        "spam": spam
    }

demo = gr.Interface(
    fn=detect,
    inputs=gr.Textbox(label="Enter complaint text"),
    outputs=gr.JSON(label="Result"),
    title="Spam Detector",
    description="Returns only the input and spam status (True/False)."
)

if __name__ == "__main__":
    demo.launch()