High accuracy speech recognition requires a large amount of transcribed data for supervised training. In the absence of such data, domain adaptation of a well-trained acoustic model can be performed, but even here, high accuracy usually requires significant labeled data from the target domain. In this work, we propose an approach to domain adaptation that does not require transcriptions but instead uses a corpus of unlabeled parallel data, consisting of pairs of samples from the source domain of the well-trained model and the desired target domain. To perform adaptation, we employ teacher/student (T/S) learning, in which the posterior probabilities generated by the source-domain model can be used in lieu of labels to train the target-domain model. We evaluate the proposed approach in two scenarios, adapting a clean acoustic model to noisy speech and adapting an adults' speech acoustic model to children's speech. Significant improvements in accuracy are obtained, with reductions in word error rate of up to 44% over the original source model without the need for transcribed data in the target domain. Moreover, we show that increasing the amount of unlabeled data results in additional model robustness, which is particularly beneficial when using simulated training data in the target-domain.
This paper presents the results of a study that evaluates audio description (AD) and visitor experience with a group of blind and partially sighted (BPS) visitors to a real-world visitor attraction—Titanic Belfast. We apply the 10-facet model of visitor experience of Packer and Ballantyne (2016) for the first time in the context of accessibility, and through this we highlight accessibility issues which arose during the study. We identify two categories in our qualitative analysis that the model (Packer & Ballantyne, 2016) cannot cover. We also model the factors that influence visitor experience and apply them to the later approach of Packer, Ballantyne, & Bond’s (2018) Dimensions of Visitor Experience (DoVE) Adjective Checklist. The checklist is based on their previous 10-facet model, and translated and refined into 15 dimensions. Although the DoVE checklist is not specifically designed for the context of accessibility, we found that it is sufficiently comprehensive to model accessibility aspects of the museum AD and visitor experience for BPS visitors.
This paper proposes a new approach to universal access based on the premise that humans have the universal capacity to engage emotionally with a story, whatever their ability. Our approach is to present the “story” of museum resources and knowledge as a journey, and then represent this journey physically as a smart map. The key research question is to assess the extent to which our “story” to journey to smart map’ (SJSM) approach provides emotional engagement as part of the museum experience. This approach is applied through the creation of a smart map for blind and partially sighted (BPS) visitors. Made in partnership with Titanic Belfast, a world-leading tourist attraction, the interactive map tells the story of Titanic’s maiden voyage. The smart map uses low-cost technologies such as laser-cut map features and software-controlled multi-function buttons for the audio description (AD). The AD is enhanced with background effects, dramatized personal stories and the ship’s last messages. The results of a reception study show that the approach enabled BPS participants to experience significant emotional engagement with museum resources. The smart model also gave BPS users a level of control over the AD which gave them a greater sense of empowerment and independence, which is particularly important for BPS visitors with varying sight conditions. We conclude that our SJSM approach has considerable potential as an approach to universal access, and to increase emotional engagement with museum collections. We also propose several developments which could further extend the approach and its implementation.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.