You don't have to be mad to talk to your tablet...

Voice recognition is another tool you can whip out of your toolbox in the quest to save time and money when using your Android tablet for Audits, Inspections and Assessments. Here is how you set it up, and a review of how it works ...

When it comes to entering information tablets are great in the field - unless you want to enter a paragraph of narrative. This is where voice recognition comes in. 

As you work through the fields in your Inspection, you can switch keyboards from the usual keypad (for numbers) to a full keyboard when text is required. If the text is long then it gets arduous using the tiny keyboard on the screen.

We have previously covered how clients are trialling handwriting, and also using bluetooth keyboards to complete narrative, but how about speaking those long paragraphs? How does that work out?

The Gear

For this the AssessmentNinja used their trusty Tab3, encased in its Otterbox Defender case. 

To emulate a busy environment all voice tests were done with loud music playing in the background. Today's jam was House of the Blue Light by Deep Purple, and Stunt by the Barenaked Ladies. Lots of vocals being screamed out.

Setting it up

The majority of tablets have Android KitKat, so the following guidance should suit most people. There may be some variation for hardware models other than Samsung.

Drag down from the top of the screen and tap the Settings Cog.

From Settings tap on "Language and Input" - you should see something similar to below.

Ensure 'Google voice typing' is selected

If the 'Google voice typing' is ticked then all is good. However there are some defaults you will be very interested in reviewing or changing (trust me on this). Tap the cog next to the 'Google voice typing'. You are taken to the following screen.

Now to make things interesting

If you use a bluetooth headset, then tick the box - if you have one of these, then this saves you holding the device close to your face when dictating. 

Also note that you can block offensive words - "which ones?" I hear you ask. Well I'm not going to type them all out for you. Suffice to say I tried this and the offensive words did come through into the narrative but were ****'d out.

Whilst you are on this page look at your default language. Then tap on 'Offline speech recognition'. You can then download language packs to enable offline use.

Tap on 'ALL' to select other languages to download

If the language you want to use is not shown, tap on 'ALL' and select which ones you wish to download. 

Once you have done this and gone back to the previous screen, make sure that the default language for Voice is set to one you have downloaded.

You are now ready to go whether online or offline.

Switching to Voice

Any place where you are using a keyboard you can use the Voice recognition. 

On the Samsung keyboard, just to the left of the spacebar is a key that if you press and  hold will pop up a mini-menu. 

On the menu will be a microphone icon. Tap this and you will be using Voice.

For users of the almighty Google keyboard the Voice recognition is activated by the microphone icon in the top right corner.

How does it work for shorter text entries?

For a real-world scenario I used a Property Inventory from the Clarinspect library - this has a lot of small fields to describe items in a register for a property - so lots of field switching, type switching (from numbers to text to drop downs) and the like.

For this scenario I found that the spell check and autocorrect were excellent when typing out short phrases on the keyboards - you can be quite rapid when you get into the swing of it. 

With Voice though, you finish a phrase and then wait for the processor to realise you have finished. So not as good as you'd think, especially if you want to move onto the next field.

Number fields were inherently recognised, so speaking 'three thousand' would populate the field with '3000'.

Drop down lists are not voice operated, so I still had to tap them.

What about long paragraphs?

This is where voice really took off. For this scenario I used the Pre-sale Property Inspection from the Clarinspect library. This has many more descriptive fields, as well as drop down lists and the like.

Dictating to the device was a dream - so long as I stuck to common words and did not use technical terms. As the fields were longer I was able to dictate long phrases and the processor would scribe out the words as I spoke. 

Punctuation is also supported. Saying 'comma', 'full-stop' or 'period' would result in punctuated text. 

For this Inspection type, which is quite wordy and subjective I was able to complete many more sections of the work more quickly, and also more completely than if I had been typing.

Correcting text

After dictation the text has a new highlighting you will not have seen when using the normal keyboard. 

Words that the system is unsure about are underlined in grey. You can then tap the word, and a menu appears with suggestions and also the option to delete the word.

Tap the highlighted word to get some suggestions.

Correcting text does drop you into the manual keyboard, but the Voice is just a key-press away.

Tips, tricks and summary

Overall I was pleasantly surprised by how the voice recognition worked, and the pace at which I could complete longer narratives which I would have groaned at having to do in the field. 

Being able to use Voice recognition offline is great, but there is that gotcha of making sure the default language in use is one that is downloaded (makes sense but the error that pops up doesn't steer you in this direction).

The background noise did not seem to bother the voice recognition at all, and none of the lyrics from the music ended up in any reports.

There are some things I could have wished for, for example when using the Google Keyboard the Voice recognition did not use my personal dictionary or shortcuts I had set up for technical terms.

As with any tool the use of Voice on these devices (even iPad) is not without issues - it will misunderstand from time to time, and you will have to do some edits, but as shown above this is not outrageously difficult.

I did not use a bluetooth headset, though I can see this would be very beneficial as with the Tab3 in an Otterbox you get a better level of recognition if you speak into the microphone rather than at the device as a whole. This entails holding the device flat and speaking into the end...

I have used the hands-free voice recognition for phone use, and dictating memos, and that has been a success also - but always with the caveat that this is just another tool, a better pencil, and is not suited to everything I need to do.

This is certainly a technology I will be walking my clients through.