Alain Désilets

Are wikis usable

Wikis are simple to use, asynchronous, Web-based collaborative hypertext authoring systems which are quickly gaining in popularity. In spite of much anecdotal evidence to the effect that wikis are usable by non technical experts, this has never been studied formally. In this paper, we studied the usability of a wiki through observation and problem-solving interaction with several children who used the tool to collaboratively author hypertext stories over several sessions. The children received a minimal amount of instruction, but were able to ask for help during their work sessions. Despite minimal instruction, 5 out of 6 teams were able to complete their story. Our data indicate that the major usability problems were related to hyperlink management. We report on this and other usability issues, and provide suggestions for improving the usability of wikis. Our analysis and conclusions also apply to hypertext authoring with non wiki-based tools.

Semantic Similarity for Detecting Recognition Errors in Automatic Speech Transcripts

Browsing through large volumes of spoken audio is known to be a challenging task for end users. One way to alleviate this problem is to allow users to gist a spoken audio document by glancing over a transcript generated through Automatic Speech Recognition. Unfortunately, such transcripts typically contain many recognition errors which are highly distracting and make gisting more difficult. In this paper we present an approach that detects recognition errors by identifying words which are semantic outliers with respect to other words in the transcript. We describe several variants of this approach. We investigate a wide range of evaluation measures and we show that we can significantly reduce the number of errors in content words, with the trade-off of losing some good content words.

The Cross-Lingual Wiki Engine: enabling collaboration across language barriers

In this paper, we present the Cross-Lingual Wiki Engine (CLWE), a system designed to support concurrent, collaborative authoring and translation of content in multiple languages. We start by showing how collaborative translation differs from conventional translation environments. In particular, we show how conventional industrial translation processes and tools are based on assumptions that often do not hold in collaborative environments. We then provide a detailed storyboard which shows how the CLWE can be used by groups of users, to collaboratively author and translate content without having to make those assumptions. We then discuss the implementation of the CLWEs change tracking infrastructure, which turns out to be the critical component in enabling this sort of open-ended translation workflow. We show how the problem of tracking changes in multiple languages at once can be greatly simplified using abstract change tokens which are independent of language and textual content. The system has been deployed in several communities, including SUMO (the Firefox documentation site), and preliminary feedback is encouraging.

VoiceCode: an innovative speech interface for programming-by-voice

In this paper we describe VoiceCode, a system for programming-by-voice. With VoiceCode, programmers can dictate code in an easy to pronounce syntax, which the system translates to native syntax in the current programming language. We illustrate how this approach addresses most of the usability issues for programming-by-voice.

VoiceGrip: A Tool for Programming-by-Voice

In recent years, there has been an increase in the number of computer programmers suffering from Repetitive Strain Injury (RSI)—an umbrella term covering a series of musculoskeletal disorders caused by repetitive motion of the hands and arms. For those individuals, or any programmer with a handicap that precludes keyboard and/or mouse input, Speech Recognition (SR) is an attractive alternative because it could allow them to do their work without using such devices. Unfortunately, programming-by-voice with current SR systems is awkward because programming languages are not meant to be spoken. In this paper we describe various usability problems with programming-by-voice and show that none of the existing programming-by-voice tools address all of those barriers. We then present VoiceGrip, a programming-by-voice tool that adresses the widest range of programming-by-voice problems to date. VoiceGrip uses a unique approach where programmers first dictate code using an easy to utter pseudo-syntax, and then translate that automatically to native code in the appropriate programming language. The system has been downloaded by 343 individuals, and postings on a neutral programming-by-voice mailing list indicate that it is being used by at least some of them. We also present an experiment evaluating the performance of the systems symbol translation algorithm. In this experiment, the system exhibited low error rates in the range of 2.7% when confusion between homophonic symbols (i.e. symbols that have the same spoken pseudo code form) was ignored and 6.6% when confusion between homophonic symbols was taken into account. Finally, even though VoiceGrip is the tool that currently addresses the widest range of programming-by-voice problems, we conclude that a better tool can be developed by combining features of VoiceGrip with features of other existing programming-by-voice tools.

Tell Me a Story

The next time you find yourself struggling with the details of your softwares functionality, back up and tell a story. Talking through simple user scenarios can help keep you on the right track.

Babel wiki workshop: cross-language collaboration

The Internet has by now become a truly global and multilingual community: non-English speakers presently account for two-thirds of Internet users. Yet, while dozens of languages are being used simultaneously on the same network, the language barriers make it difficult for connection and collaboration to happen across languages.