Deep Neural Networks (DNN) are fundamental part of current ASR. State-of-the-art are hybrid models in which acoustic models (AM) are designed using neural networks. However, there is an increasing interest in developing end-to-end Deep Learning solutions where a neural network is trained to predict character/grapheme or sub-word sequences which can be converted directly to words. Though several promising results have been reported for end-to-end ASR systems, it is still not clear if they are capable to unseat hybrid systems. In this contribution, we evaluate open-source state-ofthe-art hybrid and end-to-end Deep Learning ASR under the IberSpeech-RTVE Speech to Text Transcription Challenge. The hybrid ASR is based on Kaldi and Wav2Letter will be the endto-end framework. Experiments were carried out using 6 hours of dev1 and dev2 partitions. The lowest WER on the reference TV show (LM-20171107) was 22.23% for the hybrid system (lowercase format without punctuation). Major limitation for Wav2Letter has been a high training computational demand (between 6 hours and 1 day/epoch, depending on the training set). This forced us to stop the training process to meet the Challenge deadline. But we believe that with more training time it will provide competitive results with the hybrid system.
Delivering TV services via internet protocols over high-speed connections is commonly referred to as IPTV (internet protocol television). Similar to the app-stores of smartphones, IPTV platforms enable the emergence of IPTV service ecologies in which 3 rd party developers provide services to consumer that add value to the IPTV experience. A key issue in the IPTV ecosystem is the resilience of its services. This paper focuses on the issues related resilience in digital service ecologies. Starting with a review of resilience research in the area of biological ecologies we focus on differences and solutions for digital ecologies. Within the context of service ecologies we compare SOA and REST as design options and their impact with respect to resilience. Resilience, Ecology, IPTV, REST, SOA I. INTERACTIVE IPTVDelivering TV services via internet protocols over high-speed connections is commonly referred to as IPTV (internet protocol television). IPTV platforms differ from internet-based multimedia platforms (e.g. Netflix, YouTube, iTunes, etc.) in terms of content, delivery and costs. IPTV offers its subscribers live TV content in addition to the stored content of multi-media platforms. The IPTV video-ondemand streaming and/or downloading services are therefore a superset of the services provided by the internet-based multimedia platforms.To ensure that content-providers grant access to premium content, IPTV platforms offer very dependable (secure, safe, reliable and available) service delivery. Using bandwidth provisioning and locked-down set-top boxes, it becomes possible to allow subscribers instant access to highly soughtafter digital assets (e.g. new tv-shows, new movie releases in HD) without compromising the DRM constraints of the content owners. However all this comes at nearly twice the costs. In Canada (Saskatchewan) a Netflix service can costs a user $38 a month (30$ internet + 8$ Netflix), a medium IPTV service package is however, $72 (internet included).To combat the migration of customers from IPTV to basic internet services, providers have begun the move towards interactive IPTV platforms that allow for apps on the TV.Interactive IPTV platforms allow providers to blur the lines between classical TV and computers. Platforms like Microsoft Mediaroom, allow IPTV providers to embed applications into the video-stream and thus increase the interactivity of TV (see figure 1). Microsoft Mediaroom [1] is one of the leading platforms in this market segment. Especially telcos across the globe (e.g. AT&T, Bell, Deutsche Telekom, BTVision, etc.) favor Mediaroom since it offers backend support that fits their specific needs. The Microsoft Mediaroom platform is a servercentric IPTV solution that is based on XML documents delivered over HTTP. This server-centric design is partially due to resource limitations and the need for a secure solution. Figure 1: A simple MRML code snippetSet-top boxes are optimized to transform in a fast, secure and reliable manner IPTV data into TV displayable content. To avoid costs, set-t...
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.
customersupport@researchsolutions.com
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.