#!/usr/bin/env python3 # ———————————————————————————————————————- # Script: classify_manager_sentiment.py # Description: Loads post-match manager quotes and rates overall sentiment using OpenAI’s GPT API. # Applies explicit guidelines to assign a score from -2 to 2 for each statement. # ———————————————————————————————————————-

import os import sys import time import json import logging import pandas as pd from dotenv import load_dotenv from openai import OpenAI from tqdm import tqdm

—————————————-

Configure logging

—————————————-

logging.basicConfig( level=logging.INFO, format=‘%(asctime)s %(levelname)s %(message)s’, handlers=[ logging.StreamHandler(sys.stdout), logging.FileHandler(‘classification.log’, mode=‘w’) ] ) logger = logging.getLogger(name)

Load environment variables

load_dotenv()

—————————————-

Initialize the OpenAI client (API key)

—————————————-

Directly assign your API key here; ensure it remains confidential

api_key = “key” client = OpenAI(api_key=api_key) logger.info(‘OpenAI client initialized’)

—————————————-

Define input CSV path

—————————————-

POST_MATCH_CSV = “data_match_texts.csv” FILE_ENCODING = ‘windows-1252’

—————————————-

Load data

—————————————-

def load_csv(path, encoding=None, kwargs): try: df = pd.read_csv(path, encoding=encoding, kwargs) logger.info(f”Loaded {len(df)} rows from {path}“) return df except Exception as e: logger.error(f”Failed to load {path}: {e}“) sys.exit(1)

posts_df = load_csv(POST_MATCH_CSV, encoding=FILE_ENCODING, on_bad_lines=‘skip’)

Validate required columns

required_cols = {‘text_id’, ‘game_id’, ‘manager_id’, ‘processed_tex’} missing = required_cols - set(posts_df.columns) if missing: logger.error(f”Missing columns in CSV: {missing}“) sys.exit(1)

—————————————-

Classification function

—————————————-

Change model= to either gpt-3.5-turbo or latest version of gpt-4.1-yyyy-mm-dd

def classify_text(text, model=‘model’, retries=3): ““” Uses GPT to rate sentiment from -2 (strong negative) to +2 (strong positive). Returns integer score or None on failure. ““” guidelines = ’’’## Task:

Please read each text carefully and rate the overall sentiment of the manager’s statement as positive or negative. Your rating should reflect the manager’s expressed tone, not your judgment of the match.

Rating Scale:

Score	Meaning
2	Strongly positive sentiment (clear optimism, satisfaction, praise).
1	Mildly positive sentiment (generally positive, slight reservations).
0	Neutral or unclear sentiment.
-1	Mildly negative sentiment (general disappointment, frustration).
-2	Strongly negative sentiment (clear criticism, significant disappointment).

Final Notes:

Use 0 if unsure or if sentiment is mixed without clear dominance. ’’’ prompt = f”““{guidelines} Now rate this manager statement:

"""{text}"""

Reply with only the integer score (e.g., 2 or -1).”“”

for attempt in range(1, retries+1):
    try:
        logger.info(f"API attempt {attempt}, prompt length={len(prompt)} chars")
        resp = client.chat.completions.create(
            model=model,
            messages=[{'role':'user','content':prompt}],
            temperature=0
        )
        out = resp.choices[0].message.content.strip()
        logger.info(f"[API] output: {out}")
        score = int(out)
        return score
    except ValueError:
        logger.error(f"Could not parse integer from response: {out}")
        return None
    except Exception as e:
        logger.error(f"API error on attempt {attempt}: {e}")
        if attempt == retries:
            logger.error("All API attempts failed; returning None")
            return None
        time.sleep(2**(attempt-1))
return None

—————————————-

Process all quotes

—————————————-

results = [] for _, row in tqdm(posts_df.iterrows(), total=len(posts_df), desc=‘Classifying’): score = classify_text(row[‘processed_tex’]) results.append({ ‘text_id’: row[‘text_id’], ‘game_id’: row[‘game_id’], ‘manager_id’: row[‘manager_id’], ‘score’: score })

results_df = pd.DataFrame(results)

—————————————-

Save results

—————————————-

output_file = ‘manager_sentiment_results.csv’ results_df.to_csv(output_file, index=False) logger.info(f”Saved {len(results_df)} records to {output_file}“) logger.info(”Score distribution:” + results_df[‘score’].value_counts(dropna=False).to_string())

#!/usr/bin/env python3 # ---------------------------------------------------------------------------------------------------------------------- # Script: classify_manager_sentiment.py # Description: Loads post-match manager quotes and rates overall sentiment using OpenAI's GPT API. # Applies explicit guidelines to assign a score from -2 to 2 for each statement. # ---------------------------------------------------------------------------------------------------------------------- import os import sys import time import json import logging import pandas as pd from dotenv import load_dotenv from openai import OpenAI from tqdm import tqdm # ---------------------------------------- # Configure logging # ---------------------------------------- logging.basicConfig( level=logging.INFO, format='%(asctime)s %(levelname)s %(message)s', handlers=[ logging.StreamHandler(sys.stdout), logging.FileHandler('classification.log', mode='w') ] ) logger = logging.getLogger(__name__) # Load environment variables load_dotenv() # ---------------------------------------- # Initialize the OpenAI client (API key) # ---------------------------------------- # Directly assign your API key here; ensure it remains confidential api_key = "key" client = OpenAI(api_key=api_key) logger.info('OpenAI client initialized') # ---------------------------------------- # Define input CSV path # ---------------------------------------- POST_MATCH_CSV = "data\post_match_texts.csv" FILE_ENCODING = 'windows-1252' # ---------------------------------------- # Load data # ---------------------------------------- def load_csv(path, encoding=None, **kwargs): try: df = pd.read_csv(path, encoding=encoding, **kwargs) logger.info(f"Loaded {len(df)} rows from {path}") return df except Exception as e: logger.error(f"Failed to load {path}: {e}") sys.exit(1) posts_df = load_csv(POST_MATCH_CSV, encoding=FILE_ENCODING, on_bad_lines='skip') # Validate required columns required_cols = {'text_id', 'game_id', 'manager_id', 'processed_tex'} missing = required_cols - set(posts_df.columns) if missing: logger.error(f"Missing columns in CSV: {missing}") sys.exit(1) # ---------------------------------------- # Classification function # ---------------------------------------- # Change model= to either gpt-3.5-turbo or latest version of gpt-4.1-yyyy-mm-dd def classify_text(text, model='model', retries=3): """ Uses GPT to rate sentiment from -2 (strong negative) to +2 (strong positive). Returns integer score or None on failure. """ guidelines = '''## Task: Please read each text carefully and rate the overall sentiment of the manager's statement as positive or negative. Your rating should reflect the manager’s expressed tone, not your judgment of the match. ## Rating Scale: | **Score** | **Meaning** | |----------|------------| | **2** | Strongly positive sentiment (clear optimism, satisfaction, praise). | | **1** | Mildly positive sentiment (generally positive, slight reservations). | | **0** | Neutral or unclear sentiment. | | **-1** | Mildly negative sentiment (general disappointment, frustration). | | **-2** | Strongly negative sentiment (clear criticism, significant disappointment). | ### Final Notes: - Use **0** if unsure or if sentiment is mixed without clear dominance. ''' prompt = f"""{guidelines} Now rate this manager statement: \"\"\"{text}\"\"\" Reply with only the integer score (e.g., 2 or -1).""" for attempt in range(1, retries+1): try: logger.info(f"API attempt {attempt}, prompt length={len(prompt)} chars") resp = client.chat.completions.create( model=model, messages=[{'role':'user','content':prompt}], temperature=0 ) out = resp.choices[0].message.content.strip() logger.info(f"[API] output: {out}") score = int(out) return score except ValueError: logger.error(f"Could not parse integer from response: {out}") return None except Exception as e: logger.error(f"API error on attempt {attempt}: {e}") if attempt == retries: logger.error("All API attempts failed; returning None") return None time.sleep(2**(attempt-1)) return None # ---------------------------------------- # Process all quotes # ---------------------------------------- results = [] for _, row in tqdm(posts_df.iterrows(), total=len(posts_df), desc='Classifying'): score = classify_text(row['processed_tex']) results.append({ 'text_id': row['text_id'], 'game_id': row['game_id'], 'manager_id': row['manager_id'], 'score': score }) results_df = pd.DataFrame(results) # ---------------------------------------- # Save results # ---------------------------------------- output_file = 'manager_sentiment_results.csv' results_df.to_csv(output_file, index=False) logger.info(f"Saved {len(results_df)} records to {output_file}") logger.info("Score distribution:\n" + results_df['score'].value_counts(dropna=False).to_string())