Upload text data and explore how lexicon-based sentiment analysis scores each record. This lab uses a VADER-style approach and walks through both overall scores and a detailed, token-by-token breakdown for a sample sentence.
Sentiment analysis is a text analytics technique that automatically determines whether a piece of text expresses a positive, negative, or neutral opinion. It's widely used in marketing to analyze customer reviews, social media posts, survey responses, and any other text-based feedback at scale.
Instead of reading thousands of comments manually, sentiment analysis lets you quickly answer questions like: "Are customers happy with our new product?" or "How does sentiment differ between our brand and competitors?"
This lab uses VADER (Valence Aware Dictionary and sEntiment Reasoner), a rule-based sentiment analysis tool specifically designed for social media and informal text. Unlike machine learning approaches that require training data, VADER works "out of the box" using a curated dictionary of words with pre-assigned sentiment scores.
The VADER process:
VADER produces four sentiment scores for each text record:
| Score | Range | What It Means |
|---|---|---|
| Compound | -1 to +1 | The overall sentiment. This is the most useful single metric. Scores ≥ 0.05 are positive, ≤ -0.05 are negative, and between is neutral. |
| Positive | 0 to 1 | Proportion of text that is positive. Example: 0.45 means 45% of the emotional content is positive. |
| Neutral | 0 to 1 | Proportion of text that is neutral (factual statements, filler words). |
| Negative | 0 to 1 | Proportion of text that is negative. Example: 0.20 means 20% of the emotional content is negative. |
Note: Positive + Neutral + Negative always sum to 1.0 (100%), while Compound is calculated separately.
💡 Which score should I use?
For most marketing analyses, focus on the Compound score—it's the overall sentiment. The pos/neu/neg proportions are useful for understanding why the compound is what it is. For example, a compound of 0.15 with pos=0.35, neu=0.60, neg=0.05 tells you "moderately positive, but mostly factual text with little emotion."
Great use cases:
Limitations to keep in mind:
Always spot-check results and use sentiment analysis as a starting point for deeper investigation, not a final answer.
Per-record sentiment: Each row of your data receives all four sentiment scores, plus a categorical label (positive / neutral / negative) based on the compound value.
Summary statistics: See the overall distribution of sentiment across all records, including averages, standard deviations, and the proportion falling into each category.
Group comparisons: If your data has a grouping column (like brand or category), compare sentiment across groups to identify leaders and laggards.
Worked examples: For randomly chosen positive and negative samples, the lab shows a token-by-token breakdown so you can see exactly how VADER computed the final scores.
Use these case studies to practice reading sentiment outputs before uploading your own data. Presets include simulated Reddit-style posts about a university’s online enrollment system and detailed reviews of a new influencer swimwear brand.
Choose which column contains the text you want to analyze.
Drop a CSV/TSV file here or click to browse.
The first row should contain column names; text should be in one of the columns.
If you paste text here, the tool will ignore any uploaded file and use these lines instead.
Run the analysis to see overall sentiment across your text records.
This bar chart shows how many records are classified as positive, neutral, or negative based on VADER's compound score thresholds.
Download the sentiment scores for all analyzed records as a CSV file.
Run the analysis to see how each token in a relatively positive record contributes to its final sentiment score.
Run the analysis to see how each token in a relatively negative record contributes to its final sentiment score.
The analyzer first tokenizes the sentence and looks up each token in a sentiment lexicon. Tokens with known sentiment (e.g., "great", "terrible") get a base score; others are neutral.
Next, local rules adjust those base scores. Intensifiers like "very" or "extremely" boost nearby sentiment words; negations like "not" flip or attenuate scores; and all-caps or repeated punctuation can add emphasis.
Finally, the adjusted token scores are combined into overall positive, neutral, and negative proportions and a single compound score in \([-1, 1]\). The example above shows each token's contribution so you can see how the final numbers arise from the text.
🔗 Connecting Examples to Patterns
These examples show how VADER processes individual texts token-by-token. To understand patterns across ALL your records, scroll down to the Sentiment Distribution (Histogram) and Box Plot—they reveal whether your entire dataset skews positive/negative or shows mixed opinions across many records.
Run the analysis to generate an APA-style report of your sentiment results.
Run the analysis to generate a managerial interpretation of your sentiment results.
| Measure | Estimate | Std. Dev. | Min | Max |
|---|
📊 How to read bar colors: Each bar represents a range of compound scores (e.g., 0.10 to 0.21). Bars are colored based on where the bin center falls: green for positive territory (≥0.05), red for negative (≤−0.05), and gray for neutral. The bar's height shows how many records fall within that score range.
Run the analysis to see how compound sentiment scores are distributed across your text records.