tasks/text.py

from fastapi import APIRouter
from datetime import datetime
from datasets import load_dataset
from sklearn.metrics import accuracy_score
import random

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



#packages needed for inference
import pickle
import torch
import os

import nltk
from nltk.corpus import stopwords
import spacy

nltk.download('stopwords')
# Get the list of English stop words from NLTK
nltk_stop_words = stopwords.words('english')
# Load the spaCy model for English
nlp = spacy.load("en_core_web_sm")


def process_text(text):
    """
    Process text by:
    1. Lowercasing
    2. Removing punctuation and non-alphanumeric characters
    3. Removing stop words
    4. Lemmatization
    """
    # Step 1: Tokenization & Processing with spaCy
    doc = nlp(text.lower())  # Process text with spaCy

    # Step 2: Filter out stop words, non-alphanumeric characters, punctuation, and apply lemmatization
    processed_tokens = [
        re.sub(r'[^a-zA-Z0-9]', '', token.lemma_)  # Remove non-alphanumeric characters
        for token in doc 
        if token.text not in nltk_stop_words and token.text not in string.punctuation
    ]
    
    # Optional: Filter out empty strings resulting from the regex replacement
    processed_tokens = " ".join([word for word in processed_tokens if word])
    
    return processed_tokens

router = APIRouter()

DESCRIPTION = "TF-IDF + RF"
ROUTE = "/text"

@router.post(ROUTE, tags=["Text Task"], 
             description=DESCRIPTION)
async def evaluate_text(request: TextEvaluationRequest):
    """
    Evaluate text classification for climate disinformation detection.
    
    Current Model: Random Baseline
    - Makes random predictions from the label space (0-7)
    - Used as a baseline for comparison
    """
    
    # 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)
    test_dataset = train_test["test"]
    
    # Start tracking emissions
    tracker.start()
    tracker.start_task("inference")

    #--------------------------------------------------------------------------------------------
    # YOUR MODEL INFERENCE CODE HERE



    # Make random predictions (placeholder for actual model inference)
    true_labels = test_dataset["label"]

    current_file_path = os.path.abspath(__file__)
    current_dir = os.path.dirname(current_file_path)

    with open(os.path.join(current_dir,"tf-idf_vectorizer.pkl"), "rb") as tfidf_file:
        tfidf_vectorizer = pickle.load(tfidf_file)
    

    # Make predictions using the loaded model
    predictions = predict(test_dataset,tfidf_vectorizer,os.path.join(current_dir,"random_forest_model.pkl"))
    predictions = [LABEL_MAPPING[label] for label in predictions]

    #--------------------------------------------------------------------------------------------
    # YOUR MODEL INFERENCE STOPS HERE
    #--------------------------------------------------------------------------------------------   

    
    # Stop tracking emissions
    emissions_data = tracker.stop_task()
    
    # Calculate accuracy
    accuracy = accuracy_score(true_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