Sentiment Analysis

from Oyemi import Encoder

enc = Encoder()

# Check valence of individual words from Glassdoor reviews
test_words = ["great", "excellent", "supportive", "good",
              "stressful", "poor", "frustrated", "toxic"]

print("Word Valence Detection:")
for word in test_words:
    try:
        parsed = enc.encode_parsed(word)
        if parsed:
            valence = parsed[0].valence_name
            print(f"  {word:15} -> {valence}")
    except:
        print(f"  {word:15} -> unknown")

import re
from Oyemi import Encoder

def analyze_sentiment(text):
    """Analyze sentiment of text using Oyemi valence detection"""
    enc = Encoder()

    # Tokenize (simple word extraction)
    words = re.findall(r'\b[a-z]+\b', text.lower())

    valence_counts = {'positive': 0, 'negative': 0, 'neutral': 0}
    sentiment_words = {'positive': [], 'negative': []}

    for word in words:
        try:
            parsed = enc.encode_parsed(word, raise_on_unknown=False)
            if parsed:
                valence = parsed[0].valence_name
                valence_counts[valence] += 1
                if valence != 'neutral':
                    sentiment_words[valence].append(word)
        except:
            pass

    # Calculate sentiment score (-1 to +1)
    total = sum(valence_counts.values())
    if total == 0:
        score = 0
    else:
        score = (valence_counts['positive'] - valence_counts['negative']) / total

    return {
        'score': score,
        'label': 'positive' if score > 0.1 else 'negative' if score < -0.1 else 'neutral',
        'counts': valence_counts,
        'positive_words': sentiment_words['positive'],
        'negative_words': sentiment_words['negative']
    }

# Test on Glassdoor-style feedback
feedback = "Great learning opportunities and supportive colleagues with good training programs"
result = analyze_sentiment(feedback)

print(f"Text: {feedback}")
print(f"Sentiment: {result['label']} ({result['score']:.2f})")
print(f"Positive words: {result['positive_words']}")
print(f"Negative words: {result['negative_words']}")

# Glassdoor-style reviews
survey_data = [
    {"id": "001", "company": "KPMG", "feedback": "Great learning opportunities and supportive colleagues with good training"},
    {"id": "002", "company": "Primark", "feedback": "Stressful environment with poor management and low pay for work"},
    {"id": "003", "company": "J.P. Morgan", "feedback": "Excellent training and competitive salary with career development"},
    {"id": "004", "company": "McDonald's", "feedback": "Understaffed and stressful with low wages and frustrated workers"},
    {"id": "005", "company": "Vodafone", "feedback": "Good work balance but slow processes and average management"},
]

# Analyze all responses
results = []
for response in survey_data:
    sentiment = analyze_sentiment(response['feedback'])
    results.append({
        'id': response['id'],
        'company': response['company'],
        'score': sentiment['score'],
        'label': sentiment['label'],
        'key_words': sentiment['positive_words'] + sentiment['negative_words']
    })

# Display results
print("Glassdoor Sentiment Analysis:")
print("-" * 60)
for r in results:
    print(f"{r['id']} | {r['company']:12} | {r['label']:8} | {r['score']:+.2f} | {r['key_words'][:3]}")

from collections import defaultdict
import numpy as np

# Aggregate by department
dept_stats = defaultdict(lambda: {'scores': [], 'labels': []})

for r in results:
    dept_stats[r['dept']]['scores'].append(r['score'])
    dept_stats[r['dept']]['labels'].append(r['label'])

# Calculate department metrics
print("Department Sentiment Summary:")
print("=" * 50)

for dept, stats in sorted(dept_stats.items()):
    avg_score = np.mean(stats['scores'])
    positive_pct = stats['labels'].count('positive') / len(stats['labels']) * 100
    negative_pct = stats['labels'].count('negative') / len(stats['labels']) * 100

    # Determine health status
    if avg_score >= 0.2:
        health = "Healthy"
    elif avg_score >= 0:
        health = "Mixed"
    else:
        health = "At Risk"

    print(f"\n{dept}:")
    print(f"  Average Score: {avg_score:+.2f}")
    print(f"  Positive: {positive_pct:.0f}% | Negative: {negative_pct:.0f}%")
    print(f"  Status: {health}")

from collections import Counter

# Collect all negative words
all_negative_words = []
for response in survey_data:
    sentiment = analyze_sentiment(response['feedback'])
    all_negative_words.extend(sentiment['negative_words'])

# Count occurrences
negative_counts = Counter(all_negative_words)

print("Top Employee Concerns:")
print("-" * 30)
for word, count in negative_counts.most_common(10):
    print(f"  {word}: {count} mentions")

# Also track positive themes
all_positive_words = []
for response in survey_data:
    sentiment = analyze_sentiment(response['feedback'])
    all_positive_words.extend(sentiment['positive_words'])

positive_counts = Counter(all_positive_words)

print("\nTop Positive Themes:")
print("-" * 30)
for word, count in positive_counts.most_common(5):
    print(f"  {word}: {count} mentions")