Dialogue Manager is enelai’s AI Conversational Dialogue Management engine that generates rich conversational experiences by leveraging AI-based technology to support natural, open dialogue flows. Capable of extracting both the Intent as well as Concepts (and Entity values) from a user’s request, Dialogue Manager enables the creation of a highly intelligent, contextually aware, natural conversational experience.
Task-oriented dialogue management approach with Next Best Action (NBA) has been conceived specifically for simplifying the development and for high self-service completion. In a nutshell, Dialogue Manager is persistently trying to complete the User’s Task by finding the shortest & optimum path to fulfillment, but without restricting the user -in any way- during the process.
Voiceweb’s AI Dialogue Manager with embedded rules for Next Best Conversational Action and numerous other UX and Conversational automations, reduces complexity and effort allowing for development of Q&As and multi-turn dialogues without the need to write application-specific code. Subject Matter Experts enter the functional requirements for each Intent and User Task -or modify the built in Libraries- via user friendly GUIs to create or modify the self-service dialogues & every aspect of the User Experience.
Next Best Action is a user-centric dialogue management approach that considers the alternative next actions during a customer interaction and recommends the best one. Each user has different needs & conversational preferences, therefore each dialogue flow and User Experience needs to be personalized for each user individually, while the dialogue is on-going. Enelai Dialogue Manager is adapting the flow of the dialogue with each user after each input by the user: it first evaluates the user’s phrase and the current dialogue state & context, and then generates the next best conversational step focused on the optimal dialogue path to successfully complete each user’s request (Task-oriented Dialogue Management).
Deep Learning and Rule-based (n-grams) NLP models are trained from the same corpus and operate in parallel for increased accuracy in both voice and text-based interactions resulting in highest Recognition Accuracy and unequalled User Experience: 70% to 90% decrease of the recognition errors that users would experience if only one NLP model was deployed!
During the operation of a voice/chat bot, there are cases when the users are saying entirely non-expected words/phrases, abbreviations, or lingo. In speech-based interactions, mobile signal interruptions, environmental noises and cross-talking (talking to a person while using the system) can cause entire words to be missed or invalid phrases to be understood, producing low confidence result that would cause the bot to ask the user to retry, thus degrading the User Experience. Analysis of thousands of real-world user phrases, showed that each NLU technology is robust in some cases but weak in others, and vice versa. Instead of choosing one or another, enelai platform utilizes two types of NLU models that are “listening” in parallel. With this unique approach, each language model is covering for weaknesses of the other and Dialogue Manager evaluates (with criteria) both results before deciding on the Next Best Conversational Action.