Читать книгу Cognitive Engineering for Next Generation Computing - Группа авторов - Страница 52

This can be started by collecting the sorts of inquiries that will be posted by a delegate gathering of clients. On collecting this information an information base can be constructed to respond to the inquiries and train the framework successfully. Although you might be enticed to start by looking into information resources, as a result, you can fabricate your insight base or corpus for your framework, best practices demonstrate that you have to make a stride back and characterize your general application technique. The problem to start with corpus is it is likely to aim to the inquiries to sources that have been already assembled. If you start with the corpus, you may discover you can’t address the issues of your end clients when you move to an operational state.

These underlying inquiries need to speak to the different kinds of clients that always question the application. What would clients like to ask and by what means will they ask inquiries?

While building the application we need to consider whether it is a consumer-based application utilized by an all-inclusive community of clients, or are you building up a framework that is destined to be utilized by technicians? The future performance of the application depends on gathering the right questions. A large number of these questions and answers pairs should be collected and used in the system as machine learning algorithms are used to train it. We need at least 1,000 to 2,000 question– answer pairs to kick start the procedure. The subject expert’s help should be taken and the questions are posed by the clients using their voice to the system.