File size: 1,845 Bytes
95fc527
25b797a
20712aa
2ee1f34
25b797a
20712aa
25b797a
 
 
8aec507
 
1294d13
95fc527
20712aa
 
1294d13
 
 
2ee1f34
1294d13
 
2ee1f34
1294d13
 
 
2ee1f34
b147674
20712aa
 
 
 
 
b147674
 
 
 
 
25b797a
b147674
25b797a
1294d13
b147674
20712aa
 
 
 
 
b147674
 
 
 
25b797a
b147674
 
 
 
 
25b797a
b147674
 
 
25b797a
 
b147674
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
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
import os
from fastapi import FastAPI
from pydantic import BaseModel
from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification
from langdetect import detect, DetectorFactory

# Ensure consistent language detection results
DetectorFactory.seed = 0

# Set Hugging Face cache directory to a writable location
os.environ["HF_HOME"] = "/tmp/huggingface"
os.makedirs(os.environ["HF_HOME"], exist_ok=True)

app = FastAPI()

# Load the original tokenizer from the base model
original_tokenizer = AutoTokenizer.from_pretrained("tabularisai/multilingual-sentiment-analysis")

# Load the fine-tuned model (johndoee/sentiment) and pass the tokenizer explicitly
multilingual_model = pipeline(
    "sentiment-analysis",
    model="johndoee/sentiment",
    tokenizer=original_tokenizer
)

# Load the English sentiment model
english_model = pipeline("sentiment-analysis", model="siebert/sentiment-roberta-large-english")

class SentimentRequest(BaseModel):
    text: str

class SentimentResponse(BaseModel):
    original_text: str
    language_detected: str
    sentiment: str
    confidence_score: float

def detect_language(text):
    try:
        return detect(text)
    except Exception:
        return "unknown"

@app.get("/")
def home():
    return {"message": "Sentiment Analysis API is running!"}

@app.post("/analyze/", response_model=SentimentResponse)
def analyze_sentiment(request: SentimentRequest):
    text = request.text
    language = detect_language(text)

    # Choose the appropriate model based on language
    if language == "en":
        result = english_model(text)
    else:
        result = multilingual_model(text)

    return SentimentResponse(
        original_text=text,
        language_detected=language,
        sentiment=result[0]["label"].lower(),
        confidence_score=result[0]["score"],
    )