Tags (Top 20)
Blogs have been "tagged" with keywords:
- dyslexia (82)
- visual impairment (69)
- Books for All (68)
- accessible formats (48)
- accessibility (38)
- SQA (30)
- literacy (29)
- text_to_speech (28)
- digital exams (27)
- print disability (25)
- physical disability (24)
- PDF (20)
- text-to-speech (19)
- curriculum for excellence (18)
- eBooks (17)
- AAC (17)
- iPad (16)
- digital question papers (15)
- assessment (12)
- additional_support_needs (12)
Search results for the Tag keyword: speech recognition
By Paul Nisbet on Friday 17th April, 2015 at 1:11pm
Speech recognition has been around for many years, and many people have tried it without much success. It could be made to work, but often involved a lot of training, time and effort. Today though, computers are much more powerful, speech recognition software is much more accurate and reliable, and we believe it is now a viable option for many more learners.
There has been a lot of interest in speech recognition recently in Scotland, partly because the technology is now more common and better, and partly because of the introduction of the National Literacy assessment, where scribes cannot be used for assessment of writing, but technology, including speech recognition, can.
On 15th January 2015 we held a seminar, supported by SQA, where we discussed the use of speech recognition software in assessments and examinations. You can view a recording of the seminar on CALL's web site: scroll down to 'Speech Recognition in Practice'.
We heard very positive reports about speech recognition from practitioners in East Lothian, Scottish Borders and Stirling, and the participants on the day were keen to continue the conversation and try out speech recognition. So, we thought - how can CALL help?
The Talking in Exams Project is our response, and this is the plan:
Create guidance materials for getting started with speech recognition.
We are creating web pages on the CALL site, with general information covering the SR software and links to tutorials, videos and research. The web pages initially cover Dragon Naturally Speaking, Windows Speech Recognition, WordQ+SpeakQ and Siri on the iPad, but later we will add more for MacOS Dictation, Android and Google Chrome tools.
Build a community of practice where we can share what works and what doesn’t.
We will organise some more free sessions where we can get together and share experiences. We will set up online collaboration via CALL's web site, and/or via a Glow blog / wiki / Learning Space for project partners to talk and share. We anticipate running these sessions during this term so that work with students can start before the end of term.
Provide (a limited number of) Dragon and SpeakQ+WordQ licences to schools.
Schools who take part in the project can use the free speech recognition tools built into Windows, MacOS and on tablets, but we also want to include Dragon NaturallySpeaking and WordQ+SpeakQ in the project, so we have a small number of licences for both programs that we can provide free to schools. We anticipate having more schools involved than we have licences and so we will probably choose who gets the software by drawing lots.
Schools can use one or more of the above, e.g. Dragon NaturallySpeaking on one machine, Windows SR on another, and/or Siri.
Support schools to trial speech recognition software
As well as the web pages, we will organise (free) sessions to introduce the speech recognition tools. We’ll have these on a few dates across the country.
We will suggest a procedure for staff to follow to teach students and record results, possibly based on Speech Recognition as AT for Writing, by Daniel Cochrane and Kelly Key, or the Speech Recognition Trial Protocol, by Cindy Cavanagh.
Gather and publish case studies / reports.
We hope that participating schools will share case studies or reports on their experiences and we intend to provide an outline format for schools to use to collect information about learners as they learn to use SR. The main question is whether SR is viable for implementation at the end of the trial.
If you are interested in taking part, register an interest by emailing Paul.Nisbet@ed.ac.uk by Thursday April 30th. We will get back to you after this date to discuss next steps.
Share or bookmark this post:
By Paul Nisbet on Wednesday 2nd April, 2014 at 12:13pm
Recently there has been renewed interest in the potential of speech recognition for learners with writing and literacy difficulties (partly as a result of the restriction on use of a scribe for assessing writing at National 3/4 Literacy). Dragon NaturallySpeaking is we think the best speech recognition software for Windows PC, and I was interested whether it would run on the relatively low powered Acer TravelNote laptop that is available from XMA through the Scottish Tablet and Notebook Procurement Scheme. (There are of course scores of lightweight laptops around but it's often easier and cheaper for schools and local authorities to buy machines through this national procurement scheme.)
So we did an experiment - I installed Dragon NaturallySpeaking Premium 12 on the Acer and on my own Dell laptop, dictated into both at the same time, and tried to see if there was any time lag or lack of response between the two machines. The Acer has a Celeron 1.5Gz processor with 2 GB of RAM, while the Dell has an i5 2.5 GHz processor and 4 GB RAM, so the Dell should be noticeably faster. Both machines were running Windows 7. I didn't bother to train Dragon to my voice, and the accuracy was pretty good 'out of the box'. I looked like an even bigger prat than usual by wearing two identical headsets (Andrea NC181VM USB)...
In fact, for basic dictation, we couldn't see much difference between the two. The Acer seemed slightly slower to load programs and Dragon said that the natural language processing facility wouldn't work because of the lack of RAM and processor speed, but apart from that it was fine. (The natural language commands let you give commands to Word in simpler English (e.g.. 'Turn on bold') but not having them is not a huge disadvantage because you can still usually use the more formal commands (e.g.. 'Set Font Bold') for most tasks.)
The Acer costs £216; Dragon Naturally Speaking Premium Education is £99; so for £315 schools can get a lightweight laptop running a good speech recognition system. (You would probably need MS Office which your local authority would install, and we strongly recommend a USB headset like the Andrea at around £30 but the total cost still seems pretty good value.)
Alternatively, for about the same price you can get an iPad Air and try the free built-in Siri speech recognition - you do need an internet connection but we think it's just as good as Dragon and it seems more forgiving of strong accents and also very simple to use.
Share or bookmark this post:
By Paul Nisbet on Wednesday 28th September, 2011 at 5:45pm
The new Scottish male computer voice is now available for download from CALL's web site. 'Stuart' works with most text-to-speech programs including for example ClaroRead, Co:Writer, PDFaloud, Penfriend, Read and Write Gold and WordTalk.
However, if you install Stuart on your computer, you won't see it in the list of voices offered by PDFaloud. This is because PDFaloud offers you voices from a list of 'safe voices' that have been tested with PDFaloud. This doesn't necessarily mean another voice won't work - it may just mean that Texthelp haven't tested it. Since Stuart is brand new, he isn't in the safe voices list and so you won't see him.
You can add Stuart to the safe voices list by opening the 'safevoices.ini' file that is usually to be found in C:\Program Files\Adobe\Reader 9.0\Reader\plug_ins\Texthelp, adding the voice, and then saving the file again.
Step 1: Go to C:\Program Files\Adobe\Reader 9.0\Reader\plug_ins\Texthelp and double click on "safevoice.ini" so that it opens in Notepad.
Step 2: Scroll down to the bottom of the list and type in "CereVoice Stuart - English (Scotland)". Make sure the name of the voice that you type in is exactly as it appears in the Speech tab of the Windows Control Panel.
Step 3: Save the file.
Step 4: Restart Adobe Reader and Stuart should now be in the PDFaloud list of voices.
Step 5: Enjoy reading your digital papers with a bloke's voice ;).
Share or bookmark this post:
By Paul Nisbet on Friday 23rd September, 2011 at 11:58am
A common question we get from staff, parents and students is "Can I use speech recognition software to dictate my answers into the computer in an examination?" and so SQA funded us to spend some time trying to answer this. We've written a report with the results of the tests we've carried out on Dragon NaturallySpeaking, Windows 7 speech recognition, and WordQ+SpeakQ and you can download it from here.
We found that:
The accuracy and reliability of speech recognition software has improved considerably in recent years and all the programs tested were functional and seemed effective when dictating into a word processor. So if you want to use speech recognition to dictate extended answers into Microsoft Word for the Standard Grade English Writing paper, or Higher History, for example, then all of the programs can be used.
However, Windows speech recognition is not functional for dictating into SQA digital question papers, and so we do not recommend it for use in examinations unless the candidate is only intending to dictate into a word processor.
Dragon NaturallySpeaking is the most well known speech recognition program and can be used to dictate into both digital question papers and to a word processor. It is probably the most accurate, is relatively easy to train and use and gives voice control over formatting and over the computer in general. Dragon has text-to-speech for reading back the dictated text, and the Premium version can also play back a recording of the dictation to help with finding and correcting errors. For single user copies, Dragon NaturallySpeaking Premium is available with an educational discount (£68) and the 100-user Professional school license at £895 would seem to be relatively good value for schools who wish to make the software available to a large number of pupils. The educational discounts are availabel through Pugh or Dyslexic.com.
WordQ + SpeakQ is speech recognition software specifically designed for users who have difficulties with literacy. It uses the Windows speech recognition system, but accessed using a different, simpler interface. It has text-to-speech to help get through the training process; it can read back each phrase as it is dictated; it has text-to-speech for proof-reading; and it provides word prediction. SpeakQ can be used to dictate into SQA digital papers and also to word processors. WordQ + SpeakQ is arguably simpler to use than Dragon and the integrated text-to-speech and word prediction does make it a more attractive option for writers with reading and writing difficulties. WordQ + SpeakQ requires use of the keyboard and so it is not suitable for users who wish to control the computer completely by voice. A single user license for WordQ + SpeakQ is £199 and a site licence is £1995 from Assistive Solutions.
Speech recognition software may have considerable potential to enable some candidates to work independently and to rely less on scribes, and we are thinking it would be useful to organise some trials in schools to investigate this potential and to look at the practicalities of using speech recognition in exams. If you are interested please contact us.