The mobile application will be designed to be used by both Android and IOS operating systems. The conversation system will be hosted on a cloud storage provider and the user will connect via their mobile internet connection or WIFI.

The company will test the concept in South East Asia.

Vietnam has been chosen as the initial test country because of it’s young, educated population and the high proportion of English learning students. Also, it replicates in many ways the societal structure of China which is the companies key target market.

The company will undertake 500 hours of recording conversations between Vietnamese people learning English and a native speaker for the training of machine learning classifiers.

Once the recordings have been made two sets of transcription will be delivered alongside the audio. The first set a direct transcription and the second and adjusted transcription to correct errors in the English.

These recordings will have a significant value to other companies operating in the voice activated market.

The chart below shows how the user’s engagement flows and will be managed, analysed and responded to so that a natural free flowing conversation happens.

1. User speaks

2. Classifiers transcribe user’s speech

3. Conversation system identifies nature of response

4. Dialogue tree; next stage of conversation

5. Text to speech system responds with selected dialogue

As user’s talk, the conversation system evaluates the voice input and transcribes this to text.

The system uses machine learning models to identify patterns and recognise dialogue cues from the user. Once user’s input has been determined the conversation system will guide the user through a conversation based on pre-built dialogue trees and decide the next part of the conversation.

Based on the desired response from the conversation system the application will then output dialogue in a “natural” voice. It is hoped to be able to develop “emotion” into the response as the system development progresses.