by ReviewMiner (LX AI)
ReviewMiner turns a pile of customer reviews into sentiment and theme summaries — aspect coverage on one screen for product and CX teams who need signal without reading every line.
Decision-support only — confirm before you publish or act.
I built ReviewMiner to help teams move from raw text to clear signals without spending hours reading each comment. The first step is to create an account on the ReviewMiner dashboard. After you verify your email, you will see a prompt to generate an API key. This key lets you upload files programmatically or use the web uploader.
With the key in hand you can start uploading review data either through the drag‑and‑drop area or via a simple POST request to the /upload endpoint.
ReviewMiner accepts CSV, JSON Lines, or plain text files. The most reliable format is CSV because it lets you control column names and avoid parsing surprises. I recommend exporting reviews from your support platform, app store, or survey tool as a CSV that contains at least two columns: a unique identifier and the review text.
Your CSV should look like this:
id,review 1001,"The battery lasts all day, but the screen flickers sometimes." 1002,"Setup was easy, I love the camera quality." 1003,"Customer service replied fast, the device feels solid."
If you prefer JSON Lines, each line must be a JSON object with the same keys:
{"id":1001,"review":"The battery lasts all day, but the screen flickers sometimes."}
{"id":1002,"review":"Setup was easy, I love the camera quality."}
{"id":1003,"review":"Customer service replied fast, the device feels solid."}
Make sure the file is UTF‑8 encoded and does not contain a byte‑order mark. Files larger than 200 MB should be split into chunks of 50 MB to avoid timeout during upload.
Once the file is ready, you can start a job from the dashboard or via the API. I usually begin with a small sample of 500 reviews to verify that the aspect model works for my domain before scaling up.
ReviewMiner comes with a preset list of common aspects (price, usability, support, performance, features). You can edit this list to match your product vocabulary. For a SaaS tool I often add “integration”, “documentation”, and “security”.
If you prefer to use the API, the request looks like this:
POST https://api.reviewminer.com/v1/analyze
Headers: Authorization: Bearer <your_api_key>
Body: { "file_url": "https://your‑storage.com/reviews.csv", "aspects": ["price","usability","support","performance","features","integration","documentation","security"] }
The response includes a job ID that you can poll for results.
The core output of ReviewMiner is a table that shows each aspect, its aggregate sentiment score, the number of mentions, and a representative quote. I find this view most useful when I need to decide where to focus effort.
| Aspect | Sentiment Score (−1 to +1) | Mentions | Example Quote |
|---|---|---|---|
| price | ‑0.32 | 142 | "The subscription feels high compared to similar tools." |
| usability | +0.58 | 210 | "I could start a project in under five minutes, the UI is clear." |
| support | +0.12 | 87 | "Help desk answered my question within two hours." |
| performance | ‑0.05 | 165 | "Occasional lag when loading large reports, but generally fine." |
| features | +0.41 | 190 | "The new export option saves me a lot of time." |
| integration | +0.03 | 54 | "Webhook setup was straightforward, documentation could be clearer." |
| documentation | ‑0.18 | 63 | "I struggled to find the API reference; it is buried in the FAQ." |
| security | +0.25 | 41 | "GDPR compliance is clearly stated, which reassures our legal team." |
How to read the table:
I often sort the table by absolute sentiment score to see which aspects have the strongest polarity, then look at the mention count to gauge impact. For the sample above, usability shows the strongest positive signal with high volume, while price shows a negative signal with decent volume, suggesting a pricing review may be worthwhile.
Seeing the numbers is only the first step. I use the output to create a short‑term backlog and a longer‑term roadmap. The process below has worked for my team across several product lines.
By repeating this loop every two weeks you keep the product team grounded in what customers actually say, rather than relying on anecdotal feedback.
It is important to be clear about the limits of ReviewMiner so you do not mistake it for something it is not. I have seen teams over‑rely on automated summaries and miss nuance that only a human can catch.
The model provides a best‑effort estimate of sentiment and theme coverage. It can misclassify sarcasm or domain‑specific jargon. Always spot‑check a random sample of reviews, especially when the sentiment score is near 0.
If a review contains a legal complaint, a security disclosure, or a detailed feature request, you should read the full text. The summary is meant to highlight patterns, not to replace deep dives on individual items.
ReviewMiner looks at historical data only. It does not forecast how sentiment will change after a upcoming release or a market shift. Use it as a diagnostic tool, not a crystal ball.
For safety‑critical applications you should still follow guidance from sources such as the OWASP LLM Top 10 (https://owasp.org/www-project-top-10-for-large-language-model-applications) and the NIST AI Risk Management Framework (https://www.nist.gov/itl/ai-risk-management-framework). ReviewMiner can surface potential concerns, but it does not perform a formal risk assessment.
Very short reviews, emojis only, or reviews in languages not supported by the underlying model may be omitted from the aspect counts. Check the “skipped rows” report after each upload to see if any data was lost.
ReviewMiner turns a pile of customer reviews into sentiment and theme summaries — aspect coverage on one screen for product and CX teams who need signal without reading every line.
Decision-support only — confirm before you publish or act.
Want to try the product? ReviewMiner
Generated 2026-09-17 · primary sources only · no fabricated traffic metrics