Semantic Similarity

Out of the Box Packages > UiPath Language Analysis > Semantic Similarity

Note: This ML package will soon be deprecated. For more information, check the Deprecation timeline page from the Overview guide.

Note:

The Semantic similarity model is currently in public preview.

UiPath® is committed to stability and quality of our products, but preview features are always subject to change based on feedback that we receive from our customers. Using preview features is not recommended for production deployments.

This model allows you to compare a single reference sentence with bunch of other candidate sentences and ranks these candidate sentences in order of similarity.

Model details

Input type

JSON

Input description

JSON with one string, called "reference" and one list of strings, called, "candidates". This means that multiple reference "sentences" are not accepted. "Candidates" is a list of candidate sentences.

 {"reference": "I like trains because they are fast", "candidates": ["I like trains because they are quick", "I like trains because they are comfortable", "I do not like buses because they are slow", "I do not like trains because they are uncomfortable"]}{"reference": "I like trains because they are fast", "candidates": ["I like trains because they are quick", "I like trains because they are comfortable", "I do not like buses because they are slow", "I do not like trains because they are uncomfortable"]}

Output description

JSON with reference and most similar candidates, and the associated score on that similarity (between 0-1) ordered in descending order of score

Example:

 {
  "response": [
    {
      "candidate": "I like trains because they are quick",
      "score": 0.96463942527771
    },
    {
      "candidate": "I like trains because they are comfortable",
      "score": 0.81790685653686523
    },
    {
      "candidate": "I do not like trains because they are uncomfortable",
      "score": 0.53707438707351685
    },
    {
      "candidate": "I do not like buses because they are slow",
      "score": 0.48663735389709473
    }
  ]
}{
  "response": [
    {
      "candidate": "I like trains because they are quick",
      "score": 0.96463942527771
    },
    {
      "candidate": "I like trains because they are comfortable",
      "score": 0.81790685653686523
    },
    {
      "candidate": "I do not like trains because they are uncomfortable",
      "score": 0.53707438707351685
    },
    {
      "candidate": "I do not like buses because they are slow",
      "score": 0.48663735389709473
    }
  ]
}

Recommend GPU

A GPU is recommended when number of candidate sentences per reference exceeds 100.

Training enabled

Training is not enabled and you do not need to train or retrain this model. You can just deploy the model as needed.

AI Center

Was this page helpful?