How to Generate a Summary

How to Add the Summary generator component to AutoClassifier

  1. Navigate to the AutoClassifier Pipelines component page.

  1. Click New Component and select Summary Generator from the component list.

  1. Name your new Summary Generator component and click Add.

  1. Click Apply to save your changes.

  2. Ensure your new Summary Generator component is placed in the list of existing pipeline stages.

How to Configure Your Summary Generator

To configure your Summary Generator component, select it from the components list and complete the following fields in the Configuration section:

  1. Input Property: Specify the input metadata that the provided entities and sentences were extracted from. For example, body
  2. Sentences Property:Specify the input property containing the sentences extracted from the text metadata that are used for summary generation. These properties can be extracted by any other previous feature that you set up in the feature, if one of these features extracts sentences. For example, you might add the Natural Language Processor, which also extracts sentences, prior to this feature. If no Sentences Property is specified or the value of the provided sentences property is empty, the Summary Generator component will automatically extract sentences. 
    • Example: MicrosoftExtractedSentences
  3. Entities Property: Specify the input property containing the entities used for the summary generation. For example:
    • MSSerializedEntitiesJson
    • ComprehendSerializedEntitiesJson
    • AmazonMedicalSerializedEntitiesJson
    • SpacySerializedEntitiesJson
    • SciSpacySerializedEntitiesJson
  4. Minimum length to process summary: If the input property character length is smaller than the specified value, the component will return the entire input as summary without processing.
  5. Summery percentage: Specify the summary size as a percentage (%) of the total text size.
  6. Summary lines: Specify the total number of sentences used to build the summary. This value is used only if the Summary percentage is set to 0.
  7. Language property: Specify the input metadata containing the previously extracted data in the document. If no property is specified, or an empty value is received, this feature uses English. For example, MicrosoftExtractedLanguages
Input Property Description
Input Text
Sentences Text – Multi (Optional)

Entities

JSON - Text

Minimum Length Numeric
Summary Percentage Numeric
Summary Lines Numeric
Language Text (Optional)
Output Property Description
DocumentSummary Text

Note: Summary Generator works with the following NLP provider components: