Platform features

Feelingstream has prepared Decision Models for open use. It’s way more interesting when you get to train your model yourself, so go ahead and give it a try. No hardcore data science needed!


Evaluate text with machine learning models and automatically get text content!

Test free for 30 days!


In order to start building a Decision Model your need to do some preparations with our tools


Text Labelling

Decision Model Builder


A helpful tool for data anonymization. We clean data for protecting personal data leakage:

  • Removing numbers – replacing them with #
  • Removing names (first name + surname) – replacing them with [name]
  • Removing email addresses – replacing them with [url-or-email]
  • Removing URLS – replacing them with [url-or-email]

Text Labelling

High accuracy in machine learning models requires high quality in training data. For a customised Decision Model development internal crowdsourced taggers should be engaged in the process to increase the training data accuracy.

The Text Labelling tool helps with tagging texts into predefined categories. You need this tool when you have a lot of texts and they need to be labelled prior to training a Decision Model. 

Feelingstream labelling demo web
Feelingstream Model Builder demo web

Decision Model Builder

Train your models to detect intent or emotions from texts. This tool is needed for building Decision Models for your business case.

In the newest version the following algorithms are used to build a Model:

  • Support Vector Machine (SVM) – a supervised learning algorithm. Data labelling is required. Accuracy is 70-95%
  • Neural Network (FT) – a supervised learning algorithm. Pre-labelled texts are needed. Accuracy is 75-95%
  • Topic Model (LDA) – an unsupervised learning algorithm. No need for pre-labelled data. Accuracy is 60-90%

Unique benefits of the Feelingstream Decision Model Builder

Simple and easy to use

Suitable for regular business clients

No deep data science skills required

No previous knowledge of coding needed

Trained models are production-ready

Test Feelingstream Models

Analysing text from your customers gives you more information about them. You can find those who give high scores, yet share negative comments. And vice versa.


Do you want to conquer mountains with us?

Contact us and let’s speak!