utils.py

import pandas as pd
from datetime import datetime
from sklearn.metrics.pairwise import cosine_similarity
from sklearn.feature_extraction.text import TfidfVectorizer
from nltk import word_tokenize, FreqDist, bigrams
from nltk.corpus import stopwords
import re

def get_match_times(team):
    fixtures = pd.read_csv('epl_fixtures.csv')
    team_fixtures = fixtures.loc[(fixtures['Home Team'] == team) | (fixtures['Away Team'] == team)]
    match_times = [datetime.strptime(x, '%d/%m/%Y %H:%M') for x in team_fixtures['Date']]
    return match_times

def is_valid_post(post, match_times):
    post_time = post["time"]

    for time in match_times:
        if abs((post_time - time).total_seconds()) <= 86400:
            return True
    return False

def generate_dataset(file, mode='train'):
    comments = pd.read_csv(file, encoding='utf-8')
    
    if mode == 'single':
        matches = comments.groupby(['Match_ID', 'Comment_Team'])
        return [matches.get_group(x) for x in matches.groups]
    
    elif mode == 'double':
        match_ids = comments.groupby(['Match_ID'])
        return_list = []
        for id in match_ids.groups:
            df = match_ids.get_group(id)
            return_list.append((df[df['Comment_Team'] == 'Home'], df[df['Comment_Team'] == 'Away']))
        return return_list
    
    elif mode == 'train':
        win = comments[comments["Result"] > 0].reset_index(drop=True)
        draw = comments[comments["Result"] == 0].reset_index(drop=True)
        lose = comments[comments["Result"] < 0].reset_index(drop=True)
        return win, draw, lose

def calculate_similarity(X, Y):
    return cosine_similarity(X, Y)

def tf_idf(comments):
    comments = comments["Comment"].to_list()
    vectorizer = TfidfVectorizer(tokenizer=word_tokenize, stop_words='english', lowercase=False)
    X = vectorizer.fit_transform(comments)
    return vectorizer, X

def create_freqdist(sents, mode='unigram'):
    stopws = stopwords.words('english')
    tokens = [token.lower() for comment in sents
              for token in word_tokenize(comment)
              if token not in stopws
              and not re.fullmatch('\W+', token)]
    if mode == 'unigram':    
        return FreqDist(tokens)
    elif mode == 'bigram':
        bg = bigrams(tokens)
        return FreqDist(bg)