What is Sentiment Analysis?

What it is

Sentiment analysis (also called opinion mining) uses natural language processing (NLP) to analyze large volumes of text—tweets, emails, reviews, support tickets, etc.—to determine whether the expressed sentiment is positive, negative, neutral, or somewhere in between. Its goal is to help businesses understand customers, improve experience, and manage brand reputation.

---

Core technology and approaches

Sentiment analysis is built on NLP and is commonly implemented using one of three approaches:

Rule-based (lexicon-driven)

• Uses predefined word lists/lexicons (examples: positive — affordable, well-made, fast; negative — expensive, poorly made, slow) and simple scoring based on word matches and frequency.
• Strength: simple and interpretable.
• Weaknesses: struggles with sarcasm, negation, idioms and other contextual language.
• Example failure cases:
  • Sarcasm: “a pair of shoes so wellmade they lasted me one week”
  • Negation: “I wouldn’t say the shoes were inexpensive”
  • Idiom: “at this price the shoes are a steal”

Machine-learning based

• Trains classifiers on labeled text to learn patterns and context.
• Typical features: word frequencies, emotive phrases, review length, and other text features.
• Common algorithms:
  1. Linear regression — predicting a sentiment score from features
  2. Naive Bayes — probabilistic classification using word occurrence likelihoods
  3. Support Vector Machines (SVM) — finds an optimal boundary between classes; often used for binary classification
• Strength: can better handle nuance, sarcasm and negation if trained with appropriate data.
• Requirement: sizable labeled datasets for good performance.

Hybrid

• Combines lexicons and machine learning to improve robustness and accuracy.

---

Types of sentiment analysis and outputs

• Polarity scoring: basic positive / neutral / negative (sometimes with finer levels such as extremely positive/negative).
• Fine-grained (graded) sentiment: numeric scale (for example, 0–100 where 0 is neutral and 100 is extreme sentiment).
• Aspect-based sentiment analysis (ABSA): evaluates sentiment toward specific product or service aspects (e.g., sentiment about a travel app’s new chatbot).
• Emotion detection: identifies specific emotions or psychological states (e.g., frustration) beyond simple polarity.

---

Applications and use cases

• Customer experience optimization
• Support triage (e.g., escalate angry or frustrated tickets)
• Product and feature feedback analysis
• Market research and competitive intelligence
• Spotting trends and opportunities from large volumes of textual feedback

Sentiment analysis converts noisy textual feedback (including tricky cases) into actionable insights for business decisions.

---

Practical notes and examples

Sample review phrases used to illustrate concepts:

“these shoes are affordable and shipping was fast.” “a pair of shoes so wellmade they lasted me one full week.” “I wouldn’t say the shoes were inexpensive.” “at this price the shoes are a steal.”

Key pitfalls to watch for:

• Sarcasm
• Negation
• Idioms
• Context dependence

Model choice and the quality and variety of training data determine how well these issues are handled.

---

Format of results

Models may return:

• Categorical labels (positive / negative / neutral)
• Graded numeric scores
• Aspect-specific sentiments
• Explicit emotion labels

---

Main speaker / source

The summary is based on an unnamed video narrator/presenter; no specific author or organization is cited in the subtitles.