tasks/text.py

from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.svm import SVC
from sklearn.metrics import accuracy_score

from .utils.evaluation import TextEvaluationRequest
from .utils.emissions import tracker, clean_emissions_data, get_space_info

router = APIRouter()

DESCRIPTION = "SVM Text Classifier with TF-IDF"
ROUTE = "/text_svm"

@router.post(ROUTE, tags=["Text Task"], 
             description=DESCRIPTION)
async def evaluate_text_svm(request: TextEvaluationRequest):
    """
    Evaluate text classification for climate disinformation detection.
    
    Current Model: SVM Classifier
    - Uses TF-IDF for text vectorization
    - Trains and evaluates a Support Vector Machine (SVM) model
    """
    # Get space info
    username, space_url = get_space_info()

    # Define the label mapping
    LABEL_MAPPING = {
        "0_not_relevant": 0,
        "1_not_happening": 1,
        "2_not_human": 2,
        "3_not_bad": 3,
        "4_solutions_harmful_unnecessary": 4,
        "5_science_unreliable": 5,
        "6_proponents_biased": 6,
        "7_fossil_fuels_needed": 7
    }

    # Load and prepare the dataset
    dataset = load_dataset(request.dataset_name)

    # Convert string labels to integers
    dataset = dataset.map(lambda x: {"label": LABEL_MAPPING[x["label"]]})

    # Split dataset
    train_test = dataset["train"].train_test_split(test_size=request.test_size, seed=request.test_seed)
    train_dataset = train_test["train"]
    test_dataset = train_test["test"]

    # Extract text and labels
    train_texts = [x["text"] for x in train_dataset]
    train_labels = [x["label"] for x in train_dataset]
    test_texts = [x["text"] for x in test_dataset]
    test_labels = [x["label"] for x in test_dataset]

    # TF-IDF Vectorization
    vectorizer = TfidfVectorizer(max_features=5000)
    train_vectors = vectorizer.fit_transform(train_texts)
    test_vectors = vectorizer.transform(test_texts)

    # Train SVM Classifier
    model = SVC(kernel="linear", probability=True)
    model.fit(train_vectors, train_labels)

    # Start tracking emissions
    tracker.start()
    tracker.start_task("inference")

    # Inference
    predictions = model.predict(test_vectors)

    # Stop tracking emissions
    emissions_data = tracker.stop_task()

    # Calculate accuracy
    accuracy = accuracy_score(test_labels, predictions)

    # Prepare results dictionary
    results = {
        "username": username,
        "space_url": space_url,
        "submission_timestamp": datetime.now().isoformat(),
        "model_description": DESCRIPTION,
        "accuracy": float(accuracy),
        "energy_consumed_wh": emissions_data.energy_consumed * 1000,
        "emissions_gco2eq": emissions_data.emissions * 1000,
        "emissions_data": clean_emissions_data(emissions_data),
        "api_route": ROUTE,
        "dataset_config": {
            "dataset_name": request.dataset_name,
            "test_size": request.test_size,
            "test_seed": request.test_seed
        }
    }
    
    return results