Create a Report Using Pretrained Models

5 mins read - Updated on Nov 15, 2026

You can create a Report with a pretrained model to label your customer feedback with ready-made categories built for a kind of feedback or an industry, such as mobile apps, banking or hotels. Kimola applies the model's labels to every piece of feedback, so you get structured themes without creating, training or labeling a model yourself.

A pretrained model uses labels that Kimola has already defined. The Automatic Classifier discovers themes from your data instead, and a custom model uses the labels you define. Pretrained models fit when your feedback matches an established framework, such as the main categories of the banking industry, and when you want the same labels in every Report you run.

Getting Ready

Sign in to your Kimola account and open Reports, where the report creation area is. You need your feedback as a file with a text column, as a link to a supported page or in one of the supported integrations.

Pretrained Models You Can Choose From

Kimola offers 15 pretrained models. The report wizard lists them under the Pre-built Models tab of the Classifiers step, and each one has its own page at kimola.com/pretrained-ai-models. List Pretrained Models describes them one by one.

Three models are general purpose. The Sentiment Classifier labels feedback as positive or negative. The Automatic Classifier discovers themes from the text without predefined tags. The Entity Extractor finds entities in the text and sorts them into more than 35 classes.

Ten models are built for a kind of feedback or an industry. The Consumer Feedback Classifier covers conversation about a product, service or brand in general. The E-Commerce, Mobile App, Mobile Game, Software and Online Course Feedback Classifiers each cover one kind of product or platform. The HoReCa, Banking, Automotive and Mobile Network Feedback Classifiers cover hotels, restaurants and cafes, banks, the automotive industry and mobile network operators. The models for mobile apps, mobile games, e-commerce, banking and mobile networks have 12 main categories, and those for hotels, restaurants and cafes and the automotive industry have 13.

Two models handle other content. The News Classifier sorts text into news categories, and the Hate Speech Detector identifies hate speech aimed at an ethnicity, nationality, religion, sexuality, disability or class.

Tip

Click the link icon at the right end of a model's row to open its page and read about it before you select it.

Step 1: Add Your Data

On the Reports page, add your data in one of three ways. Paste the link of a supported page into the search bar and click Start (see Create a Report by Adding Links). Upload a .xls, .xlsx, .csv or .tsv file by dragging it onto the upload area or clicking it (see Create a Report by Uploading a Custom Dataset). Or choose Intercom, Zendesk or Google Sheets under Integrations (see Create a Report Using Integrations).

Each method asks for a little setup before the classifiers. For a link, move the slider to choose the report size and confirm that you use the extracted reviews for research purposes only. For a file or an integration, the Sheets & Columns step shows a preview of your data: choose the Text Column that holds the feedback, and optionally a Url Column and a Date Column. When the setup is done, click Advanced to open the classifier steps.

Note

Clicking Continue instead of Advanced skips the classifier and structure steps and creates the Report with the Automatic Classifier and the Sentiment Classifier. To use a pretrained model, click Advanced.

Step 2: Select Pretrained Models

The Classifiers step has two tabs, Custom Models and Pre-built Models. The pretrained models are on the Pre-built Models tab, and the Sentiment Classifier and the Automatic Classifier are already selected when the step opens.

Click a model to select it. A selected model is marked with a check. You can select more than one classifier, so a Report can hold several pretrained models next to the Automatic Classifier. Type a name in the filter box to find a model quickly, then click Continue.

Example

A hotel group uploads the comments of its guest survey and selects the HoReCa Feedback Classifier and the Sentiment Classifier. Every comment then receives labels from the categories of the HoReCa model, together with its sentiment.

Note

Every model you select adds to the queries of the Report: queries = rows × models × Dimensions. See How to Calculate Queries for an example.

Step 3: Choose the Structure

The Structure step sets how many labels one piece of feedback can receive. Enable Multi-label Classification is on when the step opens, so Kimola assigns several labels when a single piece of feedback mentions different aspects, which gives a more complete view of strengths and weaknesses. Turn it off to group each piece of feedback under one main theme. See Multi-label vs. Single-label Classification for the difference in the results.

Step 4: Select Dimensions (Optional)

The Dimensions step lets you add Dimensions, the research questions Kimola answers for every piece of feedback, such as Pain Points, which asks what goes wrong or creates difficulty. Use All, Drivers, Issues, Actors, Context and Comparisons to browse them by group. For the full steps, see Add Dimensions to a Report.

Dimensions are optional. If you only need the labels of the pretrained models and their sentiment, continue without selecting any.

Tip

Check My List on the left before you continue. It shows the Dimensions that will be added to the Report.

Note

Every Dimension you select adds to the queries of the Report, in the same way as every model: queries = rows × models × Dimensions.

Step 5: Review and Create the Report

The Review step summarizes the Report before it is created. Report Title names the Report so you can find it later. Report Language sets the language of the results: English, Spanish, French, German, Portuguese, Dutch or Turkish (see Language Options in Reports). Source names where the data comes from, and Required Query shows how many queries the Report will use.

When everything looks right, click Create Report!. Kimola shows the progress while it prepares the Report, and the processing time depends on the amount of data. When it is done, the wizard shows "Here we go!" and "We have completed creating your report." Click View Report to open it.

After the Report Is Created

You can open the Report from the wizard or find it on the Reports page at any time. The labels of the pretrained models you selected appear as themes, next to the sentiment results and the Dimensions you chose. Start with Overview in Reports, read the records under each label with Filter Records in Reports and compare how themes perform in Performance in Reports.

Was this article helpful?

Tell us about your thoughts and experiences regarding the article.