backend.py

import spacy
import pandas as pd
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
from youtube_transcript_api import YouTubeTranscriptApi, TranscriptsDisabled, VideoUnavailable
from googleapiclient.discovery import build
from fpdf import FPDF
import re
from wordcloud import WordCloud

# Initialize Spacy and VADER
nlp = spacy.load("en_core_web_sm")
sia = SentimentIntensityAnalyzer()

# YouTube Data API key
YOUTUBE_API_KEY = "YOUR_API_KEY"

def fetch_video_metadata(video_url):
    video_id = video_url.split('v=')[-1]
    youtube = build("youtube", "v3", developerKey=YOUTUBE_API_KEY)

    try:
        request = youtube.videos().list(part="snippet,statistics", id=video_id)
        response = request.execute()

        video_data = response['items'][0]

        metadata = {
            "channel_name": video_data['snippet']['channelTitle'],
            "video_title": video_data['snippet']['title'],
            "views": video_data['statistics']['viewCount'],
            "likes": video_data['statistics'].get('likeCount', 'N/A'),
            "dislikes": video_data['statistics'].get('dislikeCount', 'N/A'),
            "posted_date": video_data['snippet']['publishedAt']
        }

        return metadata, None

    except VideoUnavailable:
        return None, "Video is unavailable."
    except Exception as e:
        return None, str(e)

def fetch_transcript(video_url):
    video_id = video_url.split('v=')[-1]

    try:
        transcript = YouTubeTranscriptApi.get_transcript(video_id)
        text = " ".join([t['text'] for t in transcript])
        return text, None

    except (TranscriptsDisabled, VideoUnavailable):
        return None, "Transcript not available for this video."
    except Exception as e:
        return None, str(e)

def split_long_sentences(text):
    doc = nlp(text)
    sentences = []
    for sent in doc.sents:
        if len(sent.text.split()) > 25:
            sub_sentences = []
            current_chunk = []
            for token in sent:
                current_chunk.append(token.text)
                if token.is_punct and token.text in {".", "!", "?"}:
                    sub_sentences.append(" ".join(current_chunk).strip())
                    current_chunk = []
                elif token.text.lower() in {"and", "but", "because", "so"}:
                    if len(current_chunk) > 3:
                        sub_sentences.append(" ".join(current_chunk).strip())
                        current_chunk = []

            if current_chunk:
                sub_sentences.append(" ".join(current_chunk).strip())

            sentences.extend(sub_sentences)
        else:
            sentences.append(sent.text.strip())

    return sentences

def read_keywords(file_path):
    df = pd.read_excel(file_path.name)  # Use file_path.name since it's a Gradio file object
    attributes = df.columns.tolist()
    keywords = {attribute: df[attribute].dropna().tolist() for attribute in attributes}
    return keywords, attributes

def match_keywords_in_sentences(sentences, keywords):
    matched_keywords = {attribute: [] for attribute in keywords}
    for sentence in sentences:
        for attribute, sub_keywords in keywords.items():
            for keyword in sub_keywords:
                if keyword.lower() in sentence.lower():
                    matched_keywords[attribute].append(sentence)
    return matched_keywords

def analyze_sentiment_for_keywords(matched_keywords, sentences):
    sentiment_results = {attribute: [] for attribute in matched_keywords}
    for attribute, matched_sentences in matched_keywords.items():
        for sentence in matched_sentences:
            sentiment_score = sia.polarity_scores(sentence)["compound"]
            sentiment_results[attribute].append({"sentence": sentence, "score": sentiment_score})
    return sentiment_results

def generate_word_clouds(matched_keywords):
    wordclouds = {attribute: WordCloud().generate(" ".join(sentences)) for attribute, sentences in matched_keywords.items()}
    return wordclouds

def generate_pdf_with_sections(metadata, sentiment_results, wordclouds):
    pdf = FPDF()
    pdf.add_page()
    pdf.set_font("Arial", size=12)

    # Add metadata to PDF
    pdf.cell(200, 10, txt=f"Video Title: {metadata['video_title']}", ln=True)
    pdf.cell(200, 10, txt=f"Channel: {metadata['channel_name']}", ln=True)
    pdf.cell(200, 10, txt=f"Posted Date: {metadata['posted_date']}", ln=True)
    pdf.cell(200, 10, txt=f"Views: {metadata['views']}", ln=True)

    # Add Sentiment Analysis Results
    for attribute, sentiments in sentiment_results.items():
        pdf.cell(200, 10, txt=f"\nSentiments for {attribute}:", ln=True)
        for sentiment in sentiments:
            pdf.cell(200, 10, txt=f"  - {sentiment['sentence']} [Score: {sentiment['score']}]", ln=True)

    # Generate Wordclouds
    for attribute, wordcloud in wordclouds.items():
        wordcloud_image_path = f"{attribute}_wordcloud.png"
        wordcloud.to_file(wordcloud_image_path)
        pdf.add_page()
        pdf.image(wordcloud_image_path, x=10, y=10, w=180)

    output_pdf_path = "sentiment_report.pdf"
    pdf.output(output_pdf_path)
    return output_pdf_path