Sentiment analysis for Excel and CSV files
Find out which reviews and comments are positive, neutral or negative, for every row of your spreadsheet. Free, no sign-up.
or drop it here
CSV, Excel, Parquet or JSON. Up to 100 MB or 100,000 rows per file.
No file handy?
1. Columns to read
Pick the columns that help decide. They're read in the order you click them.
Numbers and dates can help, but they're less useful than text: they're read as words, so there's no math or date comparison. Include a text column for the best results.
2. What to decide for each row
3. Review
Preview first to check the results. Previews are free and don't count toward your daily limit.
Not quite right? Reword the question, describe the categories more clearly, or choose different columns, then preview again.
Preparing your file…
Your file is ready
Always free. For your privacy, we don't keep anything: your file and results are deleted automatically within an hour.
Whether it's product reviews, app store comments or open-ended survey answers, reading every row to judge the mood doesn't scale. Formulas that look for words like "great" or "bad" miss sarcasm, mixed opinions and anything phrased indirectly.
Here each row is read as a whole and marked positive, neutral or negative, with a confidence score. You can also use a 1 to 5 scale instead of three labels, which works well for estimating star ratings.
How to do it
- Upload your Excel or CSV file and pick the column with the text, such as the review or comment.
- Use the ready-made Positive / Neutral / Negative categories, or switch to "Rate on a scale" for a 1 to 5 score.
- Preview 10 rows to check the results, then download your file with a Sentiment column added.
Example
Real results from the sample file (customer reviews). The last row shows a case the model wasn't sure about, flagged for you to check:
| Review | Sentiment | Confidence | Needs review |
|---|---|---|---|
| Best tacos I've had in years. The al pastor was perfect and the staff remembered our names. | Positive | 100% | |
| Waited 45 minutes for a cold burger. The manager didn't seem to care. Won't be back. | Negative | 100% | |
| Decent coffee, nothing special. Seating is limited but the wifi works. | Neutral | 100% | |
| Good, not great. I'd come back if I was in the area. | Positive | 38% | Yes |
Tips for better results
- Describe what each label means for your business. "Negative" might mean any complaint, or only angry customers.
- For mixed reviews ("great food, slow service") decide whether you want them as Neutral or judged by the overall tone.
- Use the "needs review" flag to pull out the borderline comments; those are often the most useful to read.
Questions
How accurate is the sentiment analysis?
In our tests it told positive from negative reviews correctly about 98% of the time. Three-way and 1 to 5 ratings are harder because people disagree on them too; it was within one star of the real rating 97% of the time.
Does it work for languages other than English?
It works best in English. Other languages often work but are less reliable, so check the "needs review" rows carefully.
Can I get a score instead of labels?
Yes. Choose "Rate on a scale" and define the levels, for example 1 to 5 stars. You get both the level and a numeric score.
More things you can do
- Categorize bank transactions
- Classify support tickets
- Categorize survey responses
- Score sales leads
- Detect spam submissions
- Categorize products
Or classify your spreadsheet any other way you like.