A multi-store retail chain receives hundreds of reviews, complaints and messages every day, and reading each one to work out whether the customer is angry, what they want and which branch they mean takes time. AI sentiment and intent analysis answers those questions for a single piece of text in seconds. This post walks through the five free general analysis tools BranchVoice offers publicly on its website, what each one returns, and how the same analysis runs automatically on your reviews inside the platform.
What the general analysis tools are for
The "General analysis" category on the AI tools page holds five industry-agnostic tools: Sentiment Analysis, Customer Feedback Classifier, Customer Intent Detector, Entity Extractor and News Classifier. They all work the same way: paste a review, tweet or message into the box, click analyse, and the result appears next to it.
You do not need an account. Text can be up to 1,500 characters, and the tools understand multilingual input, Turkish and English above all. The text you enter is not stored; the result is cached briefly so that the same text gets a fast answer if it is sent again. To prevent abuse the number of attempts is limited, and once you hit the limit you are asked to wait a while.
Sentiment analysis: tone and dominant emotion
Sentiment Analysis places the overall tone on a five-level scale: very negative, negative, neutral or mixed, positive, very positive. Next to it you see a breakdown across eight core emotions: joy, trust, surprise, neutral, disappointment, worry, anger and disgust.
The distinction matters because two negative reviews rarely carry the same urgency. "The product was not what I expected" is disappointment; "We waited 25 minutes at the till and nobody helped, never again" is anger and usually needs a quick reply. In mixed reviews, such as "delivery was late but the product is great", the tool captures the balance too. A customer experience team can use this to move angry messages to the front of the queue.
Feedback classifier and intent detection
The Customer Feedback Classifier flags which topics a physical store review talks about, each one separately. It ships with ten labels:
| Tool | Labels it returns |
|---|---|
| Customer Feedback Classifier | Product quality, price / value, staff and service, waiting time, cleanliness, store atmosphere, stock / size availability, payment and checkout, returns and exchanges, delivery |
| Customer Intent Detector | Complaint, praise, question, suggestion, churn risk, purchase intent, would recommend, needs a reply |
One review can cover several topics: "nice products but prices went up, and there was a long queue at the till" is flagged for product quality, price and waiting time at once. The tool computes a probability for every label.
The Customer Intent Detector looks at purpose rather than topic. "I've called three times, my refund still hasn't arrived, I'm switching to a competitor" stands out as a complaint, a churn risk and a message that needs a reply; "Do your Ankara stores have this jacket in L?" is a question with purchase intent. For a regional manager the practical value is obvious: separating customers who are waiting for an answer or about to leave from everyone else.
Entity extractor and news classifier
The Entity Extractor finds named items in a text and assigns each a type: person, organisation, brand, product, location, date and time, or money. In a review like "Mehmet at the Bursa branch was really helpful and found the phone case right away", the branch location, the staff name and the product are tagged separately, each with a confidence value. Teams use it to route praise and complaints to the right person or branch, to see how often competitor brands come up, or to add dates and amounts to complaint records as structured fields.
The News Classifier sorts news articles, columns and posts into fourteen categories: politics, economy, business, technology, sports, health, science, culture and entertainment, lifestyle, education, environment, crime and justice, world, and local. Media monitoring teams can use it to split stories mentioning the brand into economy, industry and current affairs.
How the same analysis runs automatically in BranchVoice
The free tools let you test one text at a time. BranchVoice collects every review from sources such as Google Maps, Şikayetvar (Turkey's largest consumer complaint site), Ekşi Sözlük (a popular Turkish forum) and X, and analyses each one with AI. Every review gets a sentiment on a five-level scale, plus separate scores for churn risk, likelihood to recommend, whether it contains a suggestion, risk level and response urgency. When a review names an employee, that is detected and tracked per person, as covered in employee mentions in reviews.
Topics come from analysis groups: you start with ready-made groups and add your own questions. The overall logic is explained in AI review classification, and how groups are built in analysis groups and topic taxonomy. The results do not sit in isolation; they feed dashboards, alert rules and reports by store, region and period. When negative reviews or churn-risk customers pile up, for example, critical review alerts notify the right person.
Next steps
Try a few of your own customer reviews on the AI tools page to see how the results would help your team. If you want to monitor every store's reviews with this analysis continuously, request a demo and we will walk you through the setup with your own data.