Skip Navigation | ANU Home | Search ANU | Students | Staff

Using Voice Recognition Software for Examinations

(Based on using Dragon NaturallySpeaking, Version 8)

By Trevor Allan, DLO, The Australian National University

Voice recognition software is able to be used successfully by a number of students with disabilities for sitting examinations. At ANU we have been successfully using voice recognition in examinations for a number of years now. Recent experiences indicate that voice recognition is a realistic option under some circumstances for some students. There are also a number of considerations and procedures to ensure fairness and equity in the use of this examination method both for the student with a disability and for other students.

Issues to Consider

1. Who should use the software to sit exams?

Voice recognition is not the type of technology to be used without adequate experience and preparation. The use of this technology is not recommended for students who have not been using it for some time beforehand, to develop a mature voice file with an appropriate subject-specific vocabulary. A student should have at least a couple of months experience using voice recognition successfully before attempting to sit an exam with this technique.

It is also important that the student has adapted to the techniques of using dictation to create formal written English. This is perhaps one of the major adjustments that new users need to make. It is a very different process to typing or writing, and requires reasonably extensive practice before the student develops sufficient fluency and confidence to be able to use this technique in a stressful situation like an exam. Once the student has had sufficient experience to feel confident in the use of the technology, sitting exams using voice technology is often the most efficient, effective and appropriate means of dealing with the examinations. The student is able to operate independently, fluently and under relatively similar circumstances to other students.

Students with manipulative problems such as RSI, arthritis, arm injuries, paraplegia and quadriplegia, some students with Learning Disabilities, brain injuries and vision impairments may benefit from the use of this technology. Again, it is important for the student to have had sufficient experience in the use of voice recognition to be confident and fluent in its use.

2. Security measures

To ensure that students do not breach security and have access to other files on the computer, and thereby have an advantage over other students, several security measures may be instituted.

• Clear the documents menu before the exam begins and check that it only contains the exam paper after the examination has been completed

This can be done by clicking on "Start" then "Settings" then "Taskbar and Start Menu "then "Start Menu Programs" then click the "Clear" button in the "Documents menu" section.

Figure 1: Clearing the "Documents" Menu

• The exam invigilator should monitor the computer screen to ensure no other programs or screens are loaded.

With the closer supervision provided by the invigilator and the technical check on whether or not other documents have been loaded, the level of security with this form of examination is equal to, or better than regular examination procedures. It is possible for an invigilator to supervise several students, by a regular visual check of each student's computer. Usually we find that, given the potential for a person using Voice Recognition to disturb other students doing exams, VR-based exams are scheduled to be in a separate area, with direct supervision by an Invigilator, so the level of supervision and security is generally much higher than in other exams. At The ANU, we have recently networked Dragon, and so we can schedule VR exams in any student computer lab on campus.

• After the exam is completed, the invigilator should check the "Documents" section of the Start menu to see if any other documents were loaded.

This procedure should confirm that no other documents were loaded during the period of the exam. If the Document menu was cleared before the exam, the only document on the menu should be the examination paper.

Figure 2: Checking the "Documents" section

• Regularly save the file to disk as the exam progresses.

Given the nature of computers, it is possible that crashes or other problems may occur during the course of the exam. The student should create a file name and save the document at the beginning of the exam, then regularly save the document as the exam progresses. Then, if problems do occur, the whole document is not lost, only the material entered since the last save.

• Save the exam to floppy disk, USB Drive or CD-ROM and print out the completed examination paper.

Again, this provides a double insurance for the examination paper. It provides the person marking the paper with the opportunity to mark the printed paper, or to enter marks and comments on the electronic version.

3. Procedural Adjustments

• Provide rest breaks as needed. (At least five minutes per hour)

If the student is to avoid stressing the voice, with the potential for permanent problems, or, at the very least, the deterioration in recognition accuracy due to changes in voice quality, then the student needs to take a brief rest, have a drink of water, and allow the voice to recover.

• Allow 10 % extra writing time for proofreading purposes.

Even though the latest versions of continuous speech recognition programs are extremely accurate, there is still the likelihood of a number of misrecognitions in the exam paper, which will require close proofreading and correction by the student to avoid possible inaccuracies and misleading information in their answers. 10 % extra writing time should be adequate for such purposes.

• Provide exam papers in electronic and paper formats, where possible.

By providing the exam paper in both electronic and paper formats, the student's needs in accessing and inputting information are met. The student is able to read the written paper, and, where appropriate, is able to input information in the relevant sections of the electronic paper. The system can work very well for short answer, essay and multiple-choice questions. It does not work particularly well for exams that require mathematical calculations, graphs or drawings. The major advantage of NaturallySpeaking is in the input and transcription of text, control of programs, including pulldown menus and the desktop. It is possible to do these other tasks using voice, but that requires different programs and extensive experience. Other techniques will need to be used, depending on the examination requirements and the impact of the student's disability. These could include the use of a scribe, perhaps the student may be able to do some writing, or use a computer.

These changes should be in addition to any other relevant examination adjustments for particular students.

One further issue to consider is that people's voices may also be affected by other factors, which can affect their capacity to sit exams using VR. For example, a student whose voice is altered by a cold could have difficulty using the software. In circumstances such as this, the exam may need to be deferred.

With these provisions in place, students who have difficulty in writing or using a keyboard are able to compete equally with other students. The provisions cover most of the needs of the student and the concerns of the university to ensure that academic integrity is protected. The independent access of the student is enhanced, and the student is able to meet in the requirements of the course by their own means.

It is important to remember that using voice recognition for examinations is not a short-term solution, even though current training times are very rapid. It is best used by experienced users with mature voice files, appropriate subject-specific vocabularies and sufficient experience to feel fluent and confident in the use of this technique.

4. Student Comments:

Emma:

"I have used Dragon NaturallySpeaking, a voice recognition program, to do three exams and have been using Dragon NaturallySpeaking for six months. I found the idea of doing my exam on the computer using voice recognition a little daunting at first having only ever handwritten an exam, and it takes a while to get used to dictating to the computer. To do an exam using VR you need to be using it often so that the voice file is up-to-date and knows all words you will need to use in your exam. I find that when the computer is having difficulties understanding all my big words it does take up a bit of time having to correct it due to recognition errors

Recognition errors that occur when dictating to the computer take time to correct more than the usual time spent re-reading, and then there is the fact that editing using voice recognition is far more time-consuming than would ordinarily be on a computer, and it is doing corrections that you would not make when writing.

I find with the longer exams it is harder to judge how much longer it takes to use voice recognition in comparison to handwriting, although I do find with the long ones that more time is needed to check for recognition errors and for corrections. The last exam I did was a one hour exam which took me one hour and 15 minutes, in comparison to the rest of the class where most had finished after half an hour or so. It was after doing this exam that I feel I'm able to make the best comparison on the amount of time it takes to use voice recognition versus handwriting as if I had handwritten the exam. I would have taken 45 minutes to complete the exam and check it."

Alison:

"I'm very glad to have the option to take exams using VR.

It takes some of the pressure off my arms at exam time. It also means that I don't have to worry about getting severe pain during an exam and not being able to continue writing. RSI-type injuries are unpredictable and a student may be able to write an exam by hand one week, but not the next.

Recently I have been using Dragon Natspeak for a lot of my written work and I've become fairly comfortable and efficient. However, I often find it frustrating, and feel that jotting things down would be faster. It's hard to put a figure on how much faster, because it depends on what sort of work I'm doing. I try to train all the necessary vocab into the memory before each exam. If I didn't have to correct and edit, it would probably be faster than writing. But I think using Dragon is, overall, slower than the intense writing I used to do in exams. I think it's essential that the University continue with this practice. It's far preferable to having to use scribes, as the student has more control and will feel less frustration. Most importantly, it means that students with writing difficulties can avoid putting strain on their arms, which is crucial to recovery." Using Voice Recognition for Exams