🧠

Module 3 - Lecture - Natural Language Processing 2: Sentiment and Semantic Analysis

Jul 8, 2025

Overview

This lecture introduces sentiment analysis and semantic analysis, explains their importance in business, and discusses techniques and challenges in analyzing text data for emotional tone and meaning.

Sentiment Analysis: Definition and Applications

  • Sentiment analysis uses natural language processing (NLP) to determine if text expresses positive, negative, or neutral opinions.
  • It quantifies the overall sentiment in customer reviews, social media, and other text data automatically.
  • Businesses use sentiment analysis to understand customer opinions about products, services, or promotions.
  • Automated sentiment analysis eliminates the need for manual review of large volumes of text.

Techniques and Challenges in Sentiment Analysis

  • A basic approach is dictionary-based, where positive and negative words are tallied to determine sentiment.
  • Challenges include detecting subjectivity, handling conditional sentences, sarcasm, contextual word meanings, and pronoun resolution.
  • Lexicons may need to be domain-specific, as word sentiment can vary by context.

Decision Support Systems Using Sentiment Analysis

  • Sentiment analysis can be integrated into systems to compare customer opinions across competitors and product categories.
  • Example: A system for retailers like Costco analyzes online customer feedback and summarizes competitive sentiment by product category.
  • This helps businesses identify strengths and weaknesses in their offerings relative to competitors.

Semantic Analysis: Beyond Emotional Tone

  • Semantic analysis focuses on understanding the meaning of text, not just the sentiment.
  • Categorical approaches use dictionaries of related words (e.g., university, health, money) to identify topics in documents.
  • LIWC (Linguistic Inquiry and Word Count) categorizes words across over 100 topics and provides proportions for each in a text.

Hierarchical Semantic Networks

  • WordNet is a system that organizes all words into a hierarchical network based on their meanings and relationships.
  • Words are classified from broad categories (e.g., physical object) down to specific items (e.g., types of dogs).
  • This hierarchy enables finding synonyms and understanding word relationships for deeper semantic analysis.

Key Terms & Definitions

  • Sentiment Analysis — Automated detection of positive, negative, or neutral opinions in text.
  • Semantic Analysis — Process of determining the meaning or topics present in a text.
  • Lexicon — A dictionary of words labeled by sentiment or topic.
  • LIWC — Software tool for analyzing text by categorizing words into psychological and topical categories.
  • WordNet — Hierarchical lexical database that relates words by meaning and organizes them into a semantic network.

Action Items / Next Steps

  • Review how to build and use lexicons for sentiment analysis.
  • Explore the use of tools like LIWC and WordNet for semantic analysis.
  • Prepare to discuss how themes are extracted from online postings in the next session.