 |
 |
|
||||||
|
|
|
 |
|||||
 ![[Horizontal Rule]](/webart/nav_rule.gif){kind=small}            Click here to learn about |
 ![[----------------Horizontal Rule----------------]](/webart/hrule.gif){kind=small}  CADL on CPAACN Vol 13 #3, November 2000 The Communication and Assistive Device Laboratory (CADL) at the University at Buffalo, in collaboration with Enkidu Research, is conducting several interrelated experiments on the evaluation and enhancement of communication rate. According to Higginbotham: To develop AAC technologies that over-come current communication rate restrictions of AAC devices, it is necessary to specify what restrictions currently exist and what speeds augmented speakers need to achieve to sustain non-problematic social interactions. Appropriate communication and interaction speeds appear critical for competent communication performance and academic, social and employment success. However, we know little about optimal rates, or how recent AAC innovations affect interaction performance. To better understand the temporal aspects of communication, it is necessary to study various aspects of production, comprehension and interaction performance. To do so, however, required Higgin-botham and colleagues to identify and develop tools that could automate the CPA process. The research Current research underway at CADL focuses on three areas: 1. Interactive Communication Rate. They are developing transcription tech-niques to analyze interactive communication rates, including an analysis protocol that segments talk into “meaning units” (e.g., clauses, phrases). They are comparing a “meaning unit” approach to a standard “words-per-minute approach” to see whether it can accurately portray interactions between augmented and natural speakers. Included are multimodal expressions, telegraphic utterances and co-constructed events. They will investigate how communication devices affect communication speed and utterance production during conversation. 2. Narrative Comprehension. CADL researchers are determining the opti-mal speech output rates needed by listeners to comprehend narratives presented by synthetic speech (i.e., Macintalk Pro). As part of her dissertation, Kim had 50 listeners complete a battery of narrative compre-hension and subjective judgment tasks to determine the effect of communication speeds on comprehension ability. Initial results show discourse comprehension is worse at speech rates of 9-10 wpm and best at rates of 70 wpm. Comprehension of synthesized speech passages begins to decline at faster speech rates. 3. Acquisition of Scanning Skill. Researchers are evaluating the communication speeds associated with scan-ning techniques and examining the course of scanning skill acquisition. They are analyzing data from three groups of college students who have spent 15 hours learning to use a variety of scanning displays: (1) a traditional frequency optimized letter matrix; (2) a letter matrix with word or letter prediction and (3) a ten key “ambiguous keyboard.” For the study, each device automatically adjusted scanning intervals based on user performance. The analysis will focus on both objective performance measures (speed, errors, efficiency) and the students’ subjective responses to the technologies. The tools Some of the tools being used to conduct these investigations are: 1. Digital transcription tools. CADL researchers convert videotaped recordings of conversation into a digital format using Dazzle Digital Video Creator (an easy to use, con-sumer level product that is attached between the VCR and computer). Then they use Sonic Foundry’s SoundForge XP to transcribe their digital video and sound files. By dis-playing both acoustic waveform and the video image, sound and gesture can be located, marked and transcribed. They enter transcriptions as annotations that can be exported, along with timing information, into a spreadsheet and statistic program. This approach reduces transcription time by at least one-half as compared to traditional methods. Researchers also use QSR Nvivo—software to facilitate discourse analysis and provide sophisticated coding, visualization and computation facilities. 2. Logfile format and analysis. Researchers have developed an AAC device logfile format which supports most common data collection requirements and at the same time provides an extendable framework for customized logging needs. Also, the Augmentative Communication Quantitative Analysis (ACQUA) analyzes logfiles and is available on-line. It provides over 30 statistics, processes multiple files and outputs to Microsoft Excel and other statis-tical applications to analyze data at machine event, user event, word, and utterance levels. ACQUA can analyze PRC’s LAM files. CADL has organized a working group of RERC researchers and commercial manufacturers to further specify the function and structure of logfiles. 3. AAC device simulator. The Augmentative Device Simulator (ADS) emulates and records the performance of various AAC device configurations and has reconfigurable text and graphic layouts and a logfile generator. CADL will work with individuals wishing to use ADS to pursue research and other projects. 4. Usability assessment tools. Researchers have developed a set of usability tools for AAC device assessments. It includes procedures for measuring selection savings and comparing linguistic features of device output to spoken utterances, a usability questionnaire and interview protocols. They are setting up a database to catalog user-reported problems related to device use. For updates on research and tools, go to http://aac.buffalo.edu or http://www.aac-rerc.com/performance.html or contact Dr. Higginbotham cdsjeff@buffalo.edu The AAC-RERC section is partially funded by the National Institute on Disability and Rehabilitation Research under grant number H133E9 0026. The opinions are those of the grantee and do not necessarily reflect those of the U.S. Department of Education. This article appears in Augmentative Communication News, Volume 13, #3. You may order this
issue by clicking on Ordering Home Page Online Ordering 
|
