Building a Conversational Interface in 10 Steps¶
Great conversational applications require both advanced technology and solid design judgement. The most widely used conversational applications today, such as Siri, Alexa, Google Assistant, and Cortana, are all built using a similar set of techniques and technologies to ensure both high accuracy and utility. This Step-by-Step Guide outlines the methodology used to build today’s most advanced and useful conversational applications.
Taking a conversational application from conception to production typically entails completing the ten implementation steps summarized below.
|1||Select the right use case|
|2||Script your ideal dialogue interactions|
|3||Define the domain, intent, entity, and role hierarchy|
|4||Define the dialogue state handlers|
|5||Create the question answerer knowledge base|
|6||Generate representative training data|
|7||Train the natural language processing classifiers|
|8||Implement the language parser|
|9||Optimize question answering performance|
|10||Deploy trained models to production|
Taking a simple use case as an example, this Guide walks through the methodology, highlighting the key design decisions and technology components which underpin any great conversational experience. Along the way, the Guide explains how MindMeld Workbench can streamline the task of building and deploying conversational interfaces.