Skip to main content
Back to Library
Prompt Engineering Guide

Mastering Analyze sentiment
on Perplexity Online 70B

Stop guessing. See how professional prompt engineering transforms Perplexity Online 70B's output for specific technical tasks.

The "Vibe" Prompt

"Analyze the sentiment of this text: [TEXT]"
Low specificity, inconsistent output

Optimized Version

STABLE
You are an expert sentiment analysis AI. Your task is to determine the sentiment of the provided text, categorizing it as 'Positive', 'Negative', or 'Neutral'. Follow these steps: 1. **Understand the Core Subject**: Identify the main topic or entity being discussed. 2. **Extract Key Opinion Phrases/Words**: Locate adjectives, adverbs, verbs, and nouns that convey emotion or judgment related to the core subject. 3. **Consider Context and Modifiers**: Account for intensifiers (e.g., 'very', 'extremely'), negations (e.g., 'not', 'never'), and conditional language that may alter the sentiment of a phrase. 4. **Identify Implicit Sentiment**: Look for subtle cues, sarcasm, or irony that might contradict explicit wording. 5. **Synthesize and Conclude**: Combine all extracted information to make a final, overall sentiment judgment. Provide your sentiment classification and a brief, step-by-step reasoning for your conclusion. **TEXT**: [TEXT] **Sentiment Analysis Breakdown:** 1. **Core Subject**: 2. **Key Opinion Phrases/Words**: 3. **Context/Modifiers Considered**: 4. **Implicit Sentiment Notes**: 5. **Overall Sentiment**:
Structured, task-focused, reduced hallucinations

Engineering Rationale

The optimized prompt provides a clear persona ('expert sentiment analysis AI'), specific instructions, and a structured chain-of-thought process. It breaks down the complex task of sentiment analysis into manageable steps, guiding the model to systematically evaluate the text. This reduces ambiguity, encourages deeper processing, and helps the model arrive at more accurate and consistent results by explicitly asking it to consider factors like context, modifiers, and implicit sentiment. The output format is also predefined, making it easier for the model to generate the desired response.

0%
Token Efficiency Gain
Optimized prompt consistently produces more detailed and logical sentiment breakdown than naive prompt.
Optimized prompt is less prone to misinterpreting subtle or nuanced sentiment (e.g., sarcasm, irony).
Naive prompt often gives just a label without justification, optimized always provides reasoning.

Ready to stop burning tokens?

Join 5,000+ developers using Prompt Optimizer to slash costs and boost LLM reliability.

Optimize My Prompts