Search form

Are you making yourself understood?

A look at new developments in speech recognition in practice management software

Baker Heath Associates (BHA), the developers of Pearl Dental Software have been investigating the use of ‘Speech Recognition’ within their dental charting with a view to allowing practitioners to record the charting via a remote headset.  The idea being that the charted symbols would then automatically appear on the chart without having to use traditional mouse click data entry. Problems and solutions

Existing speech recognition systems use a full grammar dictionary, if you say ‘composite filling’ there is a good chance that this might be processed out to ‘compost filing’!  BHA made the first big breakthrough here, they programmed a dramatically reduced dictionary that only contained the words and phrases used within the charting process.  This dramatically improved recognition rate, instead of trying to match a spoken word or phrase against 200,000 possibilities it can look for the pattern in only 150 words and phrases. Speaking is not like clicking an item on a list, with lists your choices are limited. If you want to chart ‘Preventative Resin Restoration’ then without a list to refer to you may also say ‘Resin Restoration’, ‘Preventative Restoration’ or simply ‘PRR’.  To deal with this situation, the dictionary needs to contain the four phrases which will result in the one symbol being plotted. Has the ‘Nurse Simulator’ understood you? You assume that your dental nurse has plotted the chart item you have just said, can you achieve that level of confidence with the ‘Nurse Simulator’, I suggest the answer should be ‘no’.  Here, BHA provide feedback, the computerised voice in the headset can repeat back to you at least what it has just plotted.  If it was incorrect you can say ‘Cancel’, ‘Undo’ or ‘Rubout’ and repeat the charting instructions. There is a correct order for recording the charting. BHA force you though this sequence; say the tooth first (UR5, LR4 etc.), then the item (decayed filling, composite filling, watch, fissure sealant, lava crown etc.) and then say the surface if required (MOD, DI, MOP etc.)  The reason for this sequencing is for validation, retain canines can only be charted on UR3, UL3, LR3 or LL3 so it makes sense to identify the tooth first.  Similarly, some items such as porcelain bonded crown do not require a tooth surface in the same way as say a cavity.  BHA therefore requires the practitioner to say the item before the surface. The last problem that BHA encountered, at least in this simple trial, is the choice of headset.  Wired headsets are very good but Bluetooth headsets are not so good in that the sound quality is compromised.  There may also be other problems in practice due to hygiene masks and background noise. Have a go yourself

If you would like some light entertainment and take a look into the future then just follow these instructions: You need a PC running Windows 7 or Windows Vista Try with an inexpensive (say £10) wired headset first, one with a boom mike like this (if you can work out how to wear it you are ahead of the game!) Go into the control panel to configure ‘Speech recognition’ go as far as ‘set up microphone’, go further if you want. Note that this trial does not use the Microsoft speech interface as shown below but this will be used whilst you ‘set up microphone’ above. Close the Microsoft speech interface as shown above, it is no longer needed.  Go to http://www.pearldentalsoftware.com Click on the ‘Trial and Downloads’ tab, scroll down to ‘Other Downloads’, click on and install ‘Download: Speech recognition test’ After the software is installed, run the software from the ‘Project Speech’ icon. Have fun Your thoughts are appreciated so please let me know your comments good and bad! Paul Baker Managing Director Baker Heath Associates Ltd paul.baker@bhasoftware.com http://www.pearldentalsoftware.com Tel: 0800 027 2406  

Want to read the rest of this article?

Join now for free

Share this story

randomness