Natural language processing services
Configure a natural language processing service and integrate it with OpenPages® so that users have support when they classify objects and make object associations. A natural language processing service understands the intent behind text and returns suggestions, together with a confidence score.
The following natural language processing services are supported:
- IBM Watson® Discovery in IBM Cloud Pak® for Data
- IBM Watson Natural Language Understanding in IBM® Cloud
You can configure a natural language processing service to suggest either taxonomy classifications or parent and child object associations. You can use it for any objects in OpenPages but it is typically used to classify loss events, waivers, issues, and incidents or to associate them to a risk, policy, or control.
If you configure it to suggest taxonomy classifications, you can use it, for example, to support users when they classify a loss event to the correct Basel II categorization or when they classify waivers as exceptions to regulatory compliance. The text description that a user enters is used as input to a natural language processing service that has been trained with knowledge from your domain specialists.
If you configure it to suggest object associations, you can use it to support users when they, for example, create an Issue object and need to associate it to a parent Control object. Again, the text description that a user enters is used as input to a natural language processing service but in this case it returns a suggested parent Control object together with a confidence score. Object association suggestions can be configured to suggest either parent or child associations. They are synchronized with associations that users make for parent or child objects.
Using a natural language processing service is best suited to situations where users are generating a high volume of objects, hundreds or even thousands per year. Cognitive computing adds value when it is scaled to a large data set and to a large group of users. The data the classifier is trained on should be relatively small and static. For example, you would want to train a classifier to provide object association suggestions to a small number of Controls in the Controls Library, which is small and relatively static, rather than to all the Controls in your business, which are numerous and dynamic. You would not want to train a classifier to make object associations between objects that change rapidly, for example, Audit Findings and Issues, because Issues are constantly changing and there might be thousands of them.
You can link OpenPages to one or more services, either to support different purposes or multiple languages.
Terms to understand
- natural language processing service
- A natural language processing service uses machine learning algorithms to return the top-matching predefined classes for short text inputs.
- IBM Watson Discovery
- IBM Watson Discovery is a natural language processing service on IBM Cloud Pak for Data. You configure a IBM Watson Discovery service and then integrate it with OpenPages. It uses the Analyze API.
- IBM Watson Natural Language Understanding
- IBM Watson Natural Language Understanding is a language processing service in IBM Cloud. You can train and deploy text classifier models and then integrate them with OpenPages. IBM Watson Natural Language Understanding learns from your data and can predict classifications for texts that it is not trained on. IBM Watson Natural Language Understanding is a multi-label, multi-class classifier. It assumes that text likely belongs to more than one class and can better predict texts with multiple classifications.
- Classifier Configuration
- A classifier configuration in OpenPages defines connection information to an instance of a natural language processing service. For taxonomy classifications, it specifies the classifier target fields for the instance. For object associations, it specifies the object type to associate, whether it is a child or parent relationship, and other attributes.
- Classifier Field
- A classifier field is a field group in OpenPages that contains the name of a classifier configuration and a classifier input field. An IBM Watson Insights button is displayed in place of a classifier field (taxonomy classifications) or as an action (object associations).
- Classifier Input Field
- A classifier input field is a field in OpenPages that contains the short text input that a natural language processing service interprets and classifies. It is typically a Description field. The text Adding a description improves IBM Watson Suggestions is automatically displayed below a classifier input field.
- Classifier Target Fields
- For taxonomy classifications, classifier target fields are fields in OpenPages that are set when a user chooses suggestions for a classifier field.
- Watson™ Insights
- The underlying infrastructure in OpenPages is
tied to IBM Watson Insights. An IBM Watson Insights button is displayed and suggestions are
displayed in a Watson Insights panel.
An IBM Watson Insights button appears only if the classifier is able to make a suggestion based on the text entered. The button does not appear if the text does not generate suggestions.
The user interaction in the Watson Insights panel is the same regardless of whether the service is IBM Watson Discovery or IBM Watson Natural Language Understanding.