Research Output
1 - 6 of 6
Publication
Emotional response language education: a first ‘in-the-wild’ evaluation
2019-07-10, Sloan, John, Luo, Beate, Carson-Berndsen, Julie
This paper reports on the development and testing of Version 3 of the Emotional Response Language Education (ERLE) e-learning platform. An ‘in-the-wild’, heuristic user evaluation with five English as a Foreign Language students from Feng Chia University in Taiwan and one native English speaker in Ireland was performed over three months, with feedback from students informing changes and improvements. The primary goal of the study was to assess the robustness and reliability of a newly integrated speech recognition system to the ERLE platform. The feedback garnered led to the introduction of a tutorial prior to the initial class, a redesign of the buttons and presentation of the ASR output, and an animated response to loud input which causes difficulty for the ASR system. The improved system has since been implemented as a complimentary aid to a first-year English speaking and listening course at the same university in a larger, longitudinal study.
Publication
BoTest: a Framework to Test the Quality of Conversational Agents Using Divergent Input Examples
2018-03-11, Ruane, Elayne, Faure, Théo, Smith, Ross, Bean, Dan, Carson-Berndsen, Julie, Ventresque, Anthony
Quality of conversational agents is important as users have high expectations. Consequently, poor interactions may lead to the user abandoning the system. In this paper, we propose a framework to test the quality of conversational agents. Our solution transforms working input that the conversational agent accurately recognises to generate divergent input examples that introduce complexity and stress the agent. As the divergent inputs are based on known utterances for which we have the 'normal' outputs, we can assess how robust the conversational agent is to variations in the input. To demonstrate our framework we built ChitChatBot, a simple conversational agent capable of making casual conversation.
Publication
An Agent-based Framework for Speech Investigation
2005-09-04, Walsh, Michael, O'Hare, G. M. P. (Greg M. P.), Carson-Berndsen, Julie
This paper presents a novel agent-based framework for investigating
speech recognition which combines statistical data
and explicit phonological knowledge in order to explore strategies
aimed at augmenting the performance of automatic speech
recognition (ASR) systems. This line of research is motivated
by a desire to provide solutions to some of the more notable
problems encountered, including in particular the problematic
phenomena of coarticulation, underspecified input, and out-ofvocabulary
items. This research also seeks to promote the use
of deliberative reasoning agents in the speech and natural language
processing arenas.
Publication
The effect of phoneme distribution on perceptual similarity in English
2019-09-19, O'Neill, E. (Emma), Carson-Berndsen, Julie
This paper investigates the extent to which native speaker perceptions regarding the similarity between phonemes of English are influenced by their distributional properties. A similarity hierarchy model based on the distribution of consonantal phonemes in the English language was generated by creating phoneme-embeddings from contextual information. We compare this to similarity models based on phonological feature theory and on native speaker perception. Characteristics of the perception-based model are shown to appear in the distribution-based model whilst not being captured by the feature-based model. This not only provides evidence of similarity perceptions being influenced by distributional properties but is an argument for incorporating distributional information alongside phonological features when modelling perceptual similarity.
Publication
The effect of soft, modal and loud voice levels on entrainment in noisy conditions
2015-09-10, Szekely, Eva, Keane, Mark T., Carson-Berndsen, Julie
Conversation partners have a tendency to adapt their vocal intensity to each other and to other social and environmental factors. A socially adequate vocal intensity level by a speech synthesiser that goes beyond mere volume adjustment is highly desirable for a rewarding and successful human-machine or machine mediated human-human interaction. This paper examines the interaction of the Lombard effect and speaker entrainment in a controlled experiment conducted with a confederate interlocutor. The interlocutor was asked to maintain either a soft, a modal or a loud voice level during the dialogues. Through half of the trials, subjects were exposed to a cocktail party noise through headphones. The analytical results suggest that both the background noise and the interlocutors voice level affect the dynamics of speaker entrainment. Speakers appear to still entrain to the voice level of their interlocutor in noisy conditions, though to a lesser extent, as strategies of ensuring intelligibility affect voice levels as well. These findings could be leveraged in spoken dialogue systems and speech generating devices to help choose a vocal effort level for the synthetic voice that is both intelligible and socially suited to a specific interaction.
Publication
A Multi-Agent Computational Linguistic Approach to Speech Recognition
2003-08-09, Walsh, Michael, Kelly, Robert, O'Hare, G. M. P. (Greg M. P.), Carson-Berndsen, Julie, Abu-Amer, Tarek
This paper illustrates how a multi-agent system implements and governs a computational linguistic model of phonology for syllable recognition. We describe how the Time Map model can be recast as a multi-agent architecture and discuss how constraint relaxation, output extrapolation, parse-tree pruning, clever task allocation, and distributed processing are all achieved in this new architcture.