In these technical papers, authors from 3 Screen Solutions, Ruwido and TiVo describe real-word applications of machine learning to customise UI, and improve voice control to enhance personalisation for user experiences.

user interface voice

The user experience: Voice control

After years of researching digital television user experience (UX), the authors from 3 Screen Solutions explain that there is no simple and easy way for the viewer to customise the user interface (UI).

What if the software could understand and learn what a specific viewer wants and adapt accordingly? This is explored in the technical paper where the research focuses on utilising machine learning (ML) to discover and interpret behavioural patterns and to adapt the UI accordingly.

The technical paper shares the company’s solution for a truly adaptive UI, tailored to each viewer, as well as showcasing the ML engine, and examines behavioural mapping technique and the mathematical theory behind it.

Speech as an interaction mechanism for television control is perceived as fast and easy. While speech and speech to text are standard mechanisms and commonly used, the usage of voice and voice-based information like emotional state recognition of the user is underexplored.

Authors from Ruwido present the understanding of potential of voice aware and mood improving services a mixed method approach including a web-based study with 130 participants from the US and a user experience study in Austria and France.

Speech interaction is becoming increasingly available and popular in smart living rooms due to the rise of digital voice assistants such as Alexa or Google Assistant.

The main goal of this technical paper was to investigate user acceptance of emotionally aware systems and services that enhance mood, particularly in terms of a reported gulf between users’ expectations and experiences.

Authors from TiVo examine the modern consumer’s acclimatisation to using conversation services while interacting with electronic devices.

The use of voice search has seen a significant increase over the past few years with the rise of voice-enabled devices. Voice search, by construction, affords information about the user that is not available in conventional text search.

This technical paper presents a set of novel methods for inferring information about users of voice search - without explicit enrolment - and demonstrate subsequent enhancements to personalisation.

Further, it illustrates how this work helps in reducing computational cost by reducing the number of possibilities considered by our natural-language understanding (NLU) system.