Jill Hewitt

Speech-Based Real-Time Subtitling Services

Recent advances in technology have led to the availability of powerful speech recognizers at low cost and to the possibility of using speech interaction in a variety of new and exciting practical applications. The purpose of this research was to investigate and develop the use of speech recognition in live television subtitling. This paper describes how the “SpeakTitle” project met the challenges of real time speech recognition and live subtitling through the development of a customisable speaker interface and use of ‘Topics’ for specific subject domains. In the prototype system (described in Hewitt et al., 2000; Bateman et al., 2001) output from the speech recognition system (the IBM ViaVoice® engine) is passed in to a custom-built editor from where it can be corrected and passed on to an existing subtitling system. The system was developed to the extent that it was acceptable for the production of subtitles for live television broadcasts and it has been adopted by three subtitle production facilities in the UK.The evolution of the product and the experiences of users in developing the system in a live subtitling environment are considered, and the system is analysed against industry standards. Ease-of-use and accuracy are also discussed and further research areas are identified.

Structure for user-oriented dialogues in computer-aided telephony

Abstract Members of The Speech Project, at Oxford Brookes University, have been experimenting with alternative structures for ‘goal-seeking’ dialogues. A dialogue prototyping system has been developed that can instantiate different dialogue structures kept on file. These become active after being installed using dynamic data structures. Input is achieved using speech recognition for a small range of words, and output is in the form of pre-recorded speech messages. The dialogues thus created employ a new approach, and replace the usual menustructure with a simpler question-answer process. The nature of dialogues using computer aided telephony is analysed, and then the work done by members of The Speech Project on an alternative approach using a simpler dialogue structure is described.

Change analysis and management: a process model and its application within a commercial setting

This paper discusses the need for systematic and methodical approaches to Change Analysis and Management (CHAM) on software projects. Our main contribution is a framework for CHAM, called the Change-Cycle, which is concerned with identifying and formalising reusable change knowledge. We review current methods and techniques for handling change and discuss the desirable features of a change process. We describe the Change-Cycle in detail and our experience of using if within a commercial setting.

The quest for the last 5%: interfaces for correcting real-time speech-generated subtitles

This paper relates to ongoing work in relation to the creation of live television subtitles by speaking them. It describes an editing interface which has been developed to rapidly correct errors produced by the speech recogniser.

Critical success factors for automatic speech recognition in the classroom

This study looked at continuous automated speech recognition (ASR) to an audience in a university lecture theatre and ran an evaluation based on a previous experiment by Ryba, McIvor, Shakir and Paez, which found that nonnative speakers of English were much more favorable towards the use of ASR in class than native speakers. Our evaluation was done on a class of 29 students composed entirely of non-native speakers/ A strong indicator of the level of engagement with the technology was the linguistic ability of the user -- the weaker the students English, the more he or she tended to look at the textual output, the greater distraction experienced through poor recognition and also the greater impatience felt with slow recognition. There also seemed to be cultural differences -- the Chinese students appeared to look at the textual output much more than Indian students. We conclude that the 2 axes around which successful classroom speech recognition occurs are those of accuracy and unobtrusiveness. The more accurate and unobtrusive the technology, the more successful will be the automatically transcribed lecture.

Design of an intelligent interface to standard PC applications which maximises the ability of the disabled user

The paper reports on work on the design of speech and switch-driven interfaces. It presents a design for a generic user-interface management system (UIMS) which allows the user access to any PC application via both speech and switched input. This generic UIMS allows an instantiation of an interface which is tailored to the requirements of a particular individual, and which can, if necessary, be changed over time as his/her condition changes. Interaction is aided by a knowledge of the users requirements at the task level, drawing on interaction history and the knowledge of the application gained from a generic task model. A user-oriented task model is derived during the interaction, thus providing enhanced performance over time.

How to Build a Pan-European Multilingual Information Distribution System

Researchers who are involved in EC programs need to exchange and disseminate information. A system that supports such access to multimedia information from across Europe has many requirements, including the need to present the information in several natural languages, to have local variants of some parts of the data and to maintain a consistent style despite multiple sources of information.

Providing public access to information in complex and weakly structured domains: A 3-layered model for hypermedia information systems

Abstract Advice systems are defined as information systems which present users with both information and more subjective expert advice about complex and weakly structured domains. This paper presents a Generic Advice System Architecture (GASA) to assist in developing such systems. It describes how the architecture was used in developing the SPIRE system, whose aim is to provide advice and information which will assist the integration of students with disabilities into higher education. The paper discusses the way in which the GASA addresses key issues in the development of hypermedia advice systems including the need to make such systems available to users with little training and limited knowledge of the domain; the requirement to support diverse information exploration strategies; the provision of purely ‘point and select access; ’ and the minimisation of user disorientation and cognitive overhead.

Experiences with Web content creation from a database

Web content can be created from data stored in a database for the specific purpose of making Web pages.The data is appropriately structured, is stored in a relational database, and is transformed into HTML when required. A single Web page will contain data from many database tables. lt is relatively simple to change the database structure and to reflect the database structure in the Web site. Each datum is held just once in the database, so if updated is correct on all pages. Data is updated by downloading from other authoritative data sources, for example a personnel database or by a data owner using an on-line form to update the database.

The value of minimal prosodic information in spoken language corpora

This paper reports on an investigation into representing tone unit boundaries (pauses) as well as words in a corpus of spoken English. An analysis of data from MARSEC (Machine Readable Spoken English Corpus) shows that, for professional speakers, the inclusion of this minimal prosodic information will lower the perplexity of a language model. The analysis is based on information theoretic techniques, and an objective method of evaluation is provided by entropy indicators, which are explained. This result is of general interest, and supports the development of improved language models for many applications. The automated capture of pauses seems to be technically feasible, and warrants further investigation. The specific issue which prompted this investigation is a task in broadcasting technology: the semi-automated production of online subtitles for live television programmes. This task is described, and an approach to it using speech recognition technology is explained.