Subject: Language Acquisition Learning Disorders Project
tsn19586@tsn.cc
Date: Sat 05 Feb 2000 - 03:57:39 MET
From: tsn19586@tsn.cc Subject: Language Acquisition Learning Disorders Project Date: Sat, 5 Feb 2000 13:57:39 +1100
List address to send message to everyone: ifets-discuss@LISTSERV.READADP.COM
Details of current discussion: http://ifets.ieee.org/discussions/discuss.html
---------------------------------------------------------------------------
4 February, 2000
Language Acquisition Learning Disorders Project
George Erdos (Project Director and Managing Director of Pro-Nunciation Pty. Ltd)
Dr Andrew P. Bradley
Dr Eliathamby Ambikairajah
School of Electrical Engineering and Telecommunications
The University of New South Wales
The purpose of this submission is to inform members of a project in order to gain expressions of interest in participating. Initially in devising valid testing procedures acceptable to professionals (Psychologists, Speech Pathologists, Educators, Career Counsellors). Participation would include grass roots data collection, valid testing procedures, field testing and general feedback that may assist in the success of the overall project which will span over the next 3 years.
The R&D project is being undertaken by CANCES (Centre for Advanced Numerical Computation in Engineering and Sciences) in conjunction with UNSW Faculties of Electrical Engineering Telecommunication, Education Science and Technology and Medical Schools of Optometry, Auditory Disorders, Psychology with Pro-Nunuciation under the Federal Government IRP (Industry Research Program).
Prototype Development
A research study utilising the diagnosis of learning disorders in
order to select appropriate remedial teaching is proposed between
Pro-nunciation Pty Ltd and the CANCES industry liaison program at The
University of New South Wales. As part of this collaborative research
Program, a prototype system is required to investigate the feasibility of automating the diagnosis of language acquisition disorders by measuring the performance of patients on a number of reading, writing, and speech exercises. The system will then correlate the results on these different tests to highlight any significant differences from expected, or normal, language performance. The system will provide vital, and previously unmeasured, information to an expert who will then be able to make a final diagnosis as to the specific language acquisition deficit and therefore aid the selection of the most appropriate remedial action. The final aim of the overall collaborative project is to carry out a large-scale patient trial to determine the efficacy of a computer-assisted language learning system developed by Pro-nunciation. The development of a prototype system to diagnose language acquisition disorders is an essential initial step towards this goal.
Multiple Tool Development
By making the correct decisions in the preliminary stages, simultaneous development of multi-purpose tools aimed at market segments that support each other in their interface would be most beneficial.
1. A software based (Java Script for multi platform delivery)screening tool for parents/students, educational institutions, prospective employers , medical practitioners, and government departments. Web or digital TV delivery as a method of identifying those in need of remediation. This identification process creates the demand for the self help tool and professional intervention. As in the case of many disorders and diseases that afflict humanity, many go undetected.
2. A self help tool for end users. Web or digital TV command / response delivery coupled with smart CD-ROM media. The CD-ROM would hold the media files (animation, graphics, and audio files) and accessed by remote commands based on response to multiple choice or yes/no feedback..
3. A tool for professionals in terms of diagnostics and remediation. This would include the eye scanning hardware interfaces.
4. Training Facilities for professionals to be established.
Summary
The overall aim of the project would be to create a self screening/ monitoring system that would lead to early intervention, remediation and cost saving in its social implications.
System Overview
The prototype system will consist of the following component parts:
1) Sound card with microphone and speakers (and/or headphone set) with which to play speech sequences to the patient and acquire speech samples.
2) High quality visual display unit (VDU) on which to display the test text sequences and interact with the patient and system operator.
3) Personal computer (PC), which will interface between all the component parts of the prototype system and analyse the data gathered from the sensors (both the eye scanner and the microphone), and interface with the operator and patient (via the VDU and the sound card as appropriate). In addition, the PC will host the speech processing algorithms that perform:
4) Text to speech processing, i.e., speech synthesis, so that the computer can "talk" to the patient.
5) Speech acquisition, to capture test speech samples from the patient for further analysis and/or storage in the speech database.
6) Reproduction of speech from the database of sampled test passages of speech.
7) Speech recognition, to measure both the accuracy of what the patient said and to gather timing information, e.g., the rhythm of the speech and where any significant pauses may have occurred.
8) A "point of gaze" retinal eye scanner that monitors the position on the VDU that the patient is focusing their attention. The real-time x-y co-ordinates measured by the scanner are then passed to the PC for further analysis.
9) Speech database, on which to store test speech sequences, speech data gathered from the patient(s), and speech features and templates on which to perform a comparison with the acquired patient speech data.
10) Keyboard for the patient to enter responses to tests and interact with the PC.
In order to reduce development time to a minimum we plan to use as many
third party components for the prototype system as possible. This will
include the use of software development tools to analyse the speech data and manage the speech database. However, for later models it will be necessary to re-design a number of these components in order to achieve the required level of performance, cost, size, weight, or robustness required for a more product-orientated system.
In order to diagnose a number of possible language acquisition disorders the prototype system will be capable of measuring a variety of language performance skills from each patient. This will be done by undertaking a number of predefined language tests on the patient that exercise different aspects of language comprehension and production:
1) Patients are dictated a passage of text by the computer, the patient then keys in what they perceived was said and the PC compares the original passage to what the patient typed in. Both the overall accuracy and the frequency of a number of specific types of errors are measured. The passage could then read back to the patient by the PC (using a text to speech program) so that the patient can gain feedback on their performance.
2) Patients read aloud a passage of speech so that the PC can capture a sample of their speech. The speech passage acquired from the patient is then analysed by the PC using both speech recognition software (to measure the overall accuracy of reproduction) and also temporal analysis techniques, such as template matching (to measure temporal differences between the patient's reproduction and the original passage in the speech database). It should be noted that the speech recognition would be performed with prior knowledge of the passage that the patient is trying to convey. This drastically limits the size of the vocabulary the system has to recognise which in turn simplifies the speech recognition process.
3) Finally, a passage of speech is displayed on the VDU. The student initially reads the passage silently, and then, from memory, repeats the passage out loud. Whilst the student is reading the passage on the VDU the eye scanner is capturing data on their associated eye movements which is passed to the PC for analysis. The PC also acquires a sample of the patient's speech as they read the passage out loud and then correlates this data with that measured from the eye scanner. Using the results of these tests, the system can measure features about the patient's speech, reading, and comprehension performance. The uniqueness of the system is in its ability to cross-reference and correlate the acquired patient data from the eye scanner, speech recognition, and temporal analysis. Therefore, the similarities, and differences, in all these test results can be highlighted and this used to indicate whether there is a language understanding or a language production problem. In addition, the system will highlight any sp
ecific problem phonemes, thereby indicating to an expert the most appropriate course of remediation.
Please Contact the Project Director and CC the other email addresses listed below:
George Erdos (Project Director)
Tel:- +61 2 65 854846
Fax/voice:- +61 2 65 854842
pronunce@ozemail.com.au
pronunce@bigpond.com
portcorp@tsn.cc
http://www5.zipworld.com.au/~pronunce
http://www.portcorp.tsn.cc
Andrew P. Bradley PhD CEng
Senior Lecturer
School of Electrical Engineering and Telecommunications,
The University of New South Wales, Sydney 2052, Australia.
PHONE: +61 2 9385 4039 (Office) FAX: +61 2 9385 5993
email: a.bradley@unsw.edu.au
Dr. E. Ambikairajah
Senior Lecturer
School of Electrical Engineering and Telecommunications
The University of New South Wales
Sydney 2052, AUSTRALIA
Tel: +61 2 9385 4059 (work)
Fax: +61 2 9385 5993 (work)
Email: ambi@ee.unsw.edu.au
Ana Corpuz
Industry Research Liaison Program Manager
CANCES ATP
acorpuz@cances.atp.com.au
---------------------------------------------------------
Forum website: http://ifets.ieee.org/
Forum's contact person: kinshuk@massey.ac.nz
Info on Join/Leave List: http://ifets.ieee.org/maillist.html
---------------------------------------------------------
This archive was generated by hypermail 2a24 : Sat 05 Feb 2000 - 05:12:56 MET