Similarity Explorer

Compare the meaning of any two texts by overlaying their semantic fingerprints.

% overlap (based on cosine similarity)
Text 1
Text 2


The Similarity Explorer demonstrates the essence of's Retina: everything that can be described in words can be intelligently matched based on its meaning, independently of the language used. It displays topographical representations of all semantic features associated with each of the input texts and visualizes the comparison by overlaying their respective representations, making it easy to see the degree of similarity between the two.

To get started, copy-paste the texts you would like to compare into the two input fields. The demo accepts the following languages: Arabic, Chinese, Danish, English, French, German and Spanish. By pressing the “Enter” key, the system will automatically detect the language of the input text, except if it is too short. In this case, you will be prompted to select the language manually in the drop-down menu.

Next, click on the “Compare” button. The system will convert your texts into semantic fingerprints and display a combined fingerprint indicating the percentage of semantic relatedness between the two input texts. The red dots in the combined fingerprint correspond to the semantic overlap: these dots are contained in both fingerprints. Both texts share this meaning. Hovering over the fingerprints will display the terms associated with each dot. As similar meanings are represented close to each other, clusters of dots allow to visualize the main contexts associated with each text.

Use Cases

  • Semantic Search
  • Intelligent Content Filtering
  • Content Classification
  • Profile matching

How It Works

First, the Similarity Explorer demo makes a call to Retina API /text endpoint in order to detect the input languages. The /expression endpoint is then used to obtain fingerprints for each of the input texts. The resulting two fingerprints are sent to the /compare endpoint which, in turn, returns a metric object containing a variety of different similarity measures. This demo uses cosine similarity (a value in the range zero to 1, the closer the value is to 1 the more similar the items) to calculate the percentage similarity. The information about related terms when hovering over a bit is generated with the /term endpoint.

This demo uses the free trial API, which accesses general purpose Retinas trained with Wikipedia in the different languages. Please note that these Retina Databases are static, the automated content update feature being only available for commercial versions of the Retina API.

Build Your Own Application

This demo uses four different endpoints of the Retina API:

You can also use our Image API’s /image/compare endpoint to generate the comparison image (left hand side image, overlap, right hand side image).

You can start integrating's services into your own application after registering for your own free API key.