Featured FREE Resource:

Newsletter | Advertise | CONTACT | Facebook Twitter | LinkedIn
     
Monday, April 21 2014  
Welcome to SymbianOne - smartphone and mobile industry tech news for wireless developers, device makers, and mobile industry architects
Home arrow App Reviews arrow VoiceMode: Breaking The Text Input Barrier
HomeNewsArticlesApp ReviewsDirectoryResourcesAboutLBSEVENTSDevices
Give SymbianOne a Like on facebook
GeoJobs / LBS / AnyGeo Blog / Twitter
Need A Wireless Developer?... Post Your Free Job Listing in our Career Center Today!
Mobile Tech Feature Articles from SymbianOne

Developer /Geek Reading - Free e-Book, Linux from Scratch - This 318 page eBook provides readers with the background and instruction to design and build custom Linux systems.

VoiceMode: Breaking The Text Input Barrier Print E-mail
Written by SymbianOne   
Monday, 15 October 2007
Entering text on a smartphone can be a painful experience for all but the most hardened txters. For the rest of us VoiceSignal's latest recognition product may be the answer. Richard Bloor put VoiceMode through its paces.

Smartphone makers are constantly looking for new ways to make text entry simple and effective. Generally this focuses on keyboard design and various predictive text applications. The solution to simple text entry, however, may not be a question of how best to type text into a smartphone; it might be more a case of how best to say what you want typed.

VoiceSignal has a track record of creating effective voice recognition applications for smartphones (and feature phones too). Until recently these applications were only available as built-in phone features and, at least on Symbian devices, limited to command and control applications, such as voice dialing.

That has changed with the introduction of VoiceMode. This is VoiceSignal's first recognition product available as an add-on application, in this case for S60 3rd Edition devices. VoiceMode provides continuous speech-to-text recognition capabilities. As such it promises to eliminate the need to tap out text on a numeric keypad, at least for short messages.

Once installed VoiceMode needs to be adapted before it can be used. VoiceSignal recommends this adaption training is performed in a quiet room. The process involves repeating a number of simple phrases, seventy five in all. It is possible to continue training beyond the initial seventy five phrases, but VoiceMode seems to work fine on just the basic training.

There is a second setup task that can be performed. This is to import names from your device's contacts database. This process adds up to 2500 names to VoiceMode's dictionary and assists in ensuring accurate recognition for those names.

Once the setup has been completed VoiceMode can be turned on. The first time this is done VoiceMode asks for licensing details. The choices are: a one day trial or license code entry. VoiceMode is only available officially through Nokia Software Market, which does not offer a trial download. So, you will already have a license key at this stage. Hopefully, VoiceSignal will eventually get VoiceMode into more channels and a trial download will be available.

Once running VoiceMode displays a flag whenever a field it can enter text into is active. The recognition process is activated by pressing the green "talk" key. When this is done, after a short delay, a banner displays at the bottom of the screen stating "TALK NOW".

My tests suggested that the recognition of the first word is more accurate if dictation is not started until the banner is displayed. The text required can now be dictated by talking in, as VoiceSignal recommends, a "calm, clear, and even voice". The complete message should also be broken down into small phrases, rather than dictated in one go.

During dictation it is possible to insert a range of common punctuation using key words, such as "comma", "hyphen", and "ampersand". It is also possible to insert two smilies ("happy-face" and "sad-face"). In addition, common abbreviations are substituted for the spoken version, for example "Mister" becomes "Mr". This feature also does the same with common SMS expressions, so "laughed out loud" recognizes as "LOL". The area where normal speech may not achieve the desired result is with numbers. For example "one thousand, four hundred, and ten" might be transcribed as "1 thousand 4 hundred and 10" so if the desired result was 1410 you actually need to say the numbers, "one four one zero" or "one four ten", but this is only necessary for numbers over 99. Dates do not present any problems.

Once the dictation of a phrase is complete the green key is released and the recognized text entered in the field. The last word recognized is now highlighted in red.

If a recognition error has occured the scroll keys can be used to navigate to the incorrect word. Then, by pressing the select key, a list of correction options is displayed. These options include words VoiceMode thinks could have been spoken and the option to spell the word if none of the suggestions are correct.

The correction process does more than simply correct the dictated text, VoiceMode also adapts its recognition process, so over time it becomes more accurate.

I tested VoiceSignal's VSuite back in 2006 and was impressed by the accuracy of its speaker independent voice dialing recognition in a range of environments. VoiceMode has a significantly harder job, there being a far wider vocabulary of words to recognize. However, VoiceMode takes this challenge in its stride. Recognition accuracy can be a little erratic at first, but after a few uses - and through a combination of recognition error corrections and learning the most appropriate way to talk - the recognition is impressive. It performs well in a range of environments, working well in a car traveling at motorway speeds (I was not the driver), restaurants, and city streets. This performance is even more impressive when compared to desktop voice recognition applications. These often take up hundreds of megabytes of disk space, while VoiceMode uses just a few megabytes.

While VoiceMode may be diminutive compared to similar PC applications it is relatively large by smartphone application standards. Loading VoiceMode uses approximately 7Mb of RAM, with approximately 200kb of additional memory used whenever a field VoiceMode can be used to dictate into is active. Despite this relatively large memory footprint, I found several business orientated applications could be run in parallel with VoiceMode with no memory issues on a Nokia N93. As VoiceMode worked happily on an N93 it should work acceptably on almost any S60 3rd Edition device, as the devices with the lowest available free RAM memory offer about 18Mb while the N93 offers 22Mb.Many recent devices have much more free memory available to execute applications.

However, to minimize memory used VoiceMode currently limits the number of words that can be dictated into any field to around 100. This is more than adequate for most SMS messages, but can be limiting for longer replies to emails. Considering many applications - that on the face of it probably do much less work than VoiceMode - can take up several megabytes of memory I do wonder if VoiceSignal has been a little too conservative in the current limitation. That said, VoiceSignal have a new version in development that will allow up to 500 words to be dictated with out a significant increase in memory usage. This will almost certainly make VoiceMode useful for replying to almost any email. However, this stops short of making VoiceMode a general purpose text entry system, which could for example be used to update a Word document in Quickoffice.

(With a little lateral thinking it is possible to use VoiceMode to dictate a document of more than 100 words. Using an application such as Kylom's excellent Projekt, in which the text can be entered in discrete units, VoiceMode can be used to dictate and edit a long document that is saved to text for any final formatting.)

VoiceMode is currently certified for only a small number of S60 3rd Edition devices, namely the Nokia N73, Nokia E62, Nokia E61, Nokia 5500, and Nokia 3250. I was able to test VoiceMode only on the as yet unsupported Nokia N80, Nokia E61i, and Nokia N93. On the N80 VoiceMode installed, but crashed the device each time I tried to start it, while on the E61i and N93 it worked well. I found a couple of minor issues on the E61i and N93; the Web Browser for S60 does not open with VoiceMode on, VoiceMode sometimes fails to issue an error message when it can no longer accepted text into a field, and the "pencil" button did not provide a quick link to the VoiceMode control panel, but these were relatively easy issues to work around. So, if you do not have one of the certified devices it would pay to wait until the software can be downloaded and tested before you buy. (There is a copy of the American English version of VoiceMode hosted on MOSH, although this does not look like an official posting so it may not be maintained.)

So can you throw away your keyboard? Well almost. VoiceMode is certainly accurate and much faster than tapping away on a tiny phone keypad or thumb board. Once you get accustomed to using it (which is not very long) it quickly becomes frustrating on those occasions when you can not use it to dictate a message (which mainly means when you would not want to be overheard). This is an application I will be installing on any new S60 device without any hesitation. If there is one drawback it is that you will want to use VoiceMode for all your text entry, however, given what VoiceMode can do today, I'm sure that will not be too far away.


You can find the British English version of VoiceMode here on Nokia Software Market.

 


Share

Submit Your Mobile Tech News




Social Media Strategies

Social Media Strategy Workbook: This Workbook will help you to define your goals and audiences and to decide on the channels that make the most sense for you. Ready to figure out what social media means for you and your own organization? This Social Media Planning Workbook will help you to define your goals and audiences and to brainstorm the channels that make the most sense for you.  

47 Handy Facebook Stats and Charts - Do you know the best time to engage with your customers on Facebook? Learn this and more in this free eBook. They have compiled 47 stats, charts and graphs on Facebook that are easy to share and put into presentations.  

Nokia Lumia 900 4G Windows Phone, Black (AT&T)

Contribute to the SymbianOne Symbian Search!

Top of 

Page
(c)2003 - 2008, SymbianOne - All rights reserved