Excel reads numeric cells in column E as exponents rather than cell value


Rick
 

Hello NVDA List:

 

The other day I was reading an Excel  spreadsheet sent to me by a colleague. It had numeric values in column E. When reading these cells with Microsoft One Core voices (I am using Microsoft David) the cell name was read as an exponent to the power of 10. For instance, if the value 1.24 was present in cell E4, the speech produced, “1.24 time 10 to the power of 4” rather than 1.24 E4. This does not happen when using Espeak voices or SAPI 5 voices. Note, if the number in the cell is the result of a formula and not a direct value, the phrase, “has formula” is inserted between the number and the cell name, so it works as expected (i.e. 1.24 has formula E4).

 

Note: This does not happen using Narrator, since Narrator reads the cell name first.

 

I am running the latest released copy of NVDA, Office 365 desktop apps and Windows 10 Professional.

 

Rick

 


Quentin Christensen
 

Rick,

Well spotted, and interesting pickup.  This is exactly why NVDA itself doesn't try to interpret information, but simply present it as it is displayed.  Many synthesizers do try to interpret data, which is where you end up with St being read as "street" when it is actually used in the context of "saint" or dr for Doctor instead of drive.

In this case, it is processing the whole string (value and cell coordinates) as one piece of information: 1.24E4 is read as the scientific formatted number "1.24 times 10 to the power of 4".  If the value was in the cell above it is read as "1.24 times 10 cubed" or in E2 as "1.24 times 10 squared".  It is interesting to note that this affects the OneCore voices, and even then only some of them.  On my system, I have ten OneCore voices.  Of those:
- David, Zira and Mark interpret the coordinates as you encountered
- George, Catherine, Hazel, Susan, Linda, Richard or James, simply read it as it should be read: 1.24 E4

I just checked and of the David, Hazel and Zira Microsoft Speech API version 5 voices I have, David reads it as a scientific number, but Hazel and Zira both read 1.24 E4.

I will raise this with Microsoft and look to see if we can do anything with how we format the string, maybe we can put a comma in-between the pieces of information.

On Mon, Nov 19, 2018 at 9:47 PM Rick <softwarethatworks@...> wrote:

Hello NVDA List:

 

The other day I was reading an Excel  spreadsheet sent to me by a colleague. It had numeric values in column E. When reading these cells with Microsoft One Core voices (I am using Microsoft David) the cell name was read as an exponent to the power of 10. For instance, if the value 1.24 was present in cell E4, the speech produced, “1.24 time 10 to the power of 4” rather than 1.24 E4. This does not happen when using Espeak voices or SAPI 5 voices. Note, if the number in the cell is the result of a formula and not a direct value, the phrase, “has formula” is inserted between the number and the cell name, so it works as expected (i.e. 1.24 has formula E4).

 

Note: This does not happen using Narrator, since Narrator reads the cell name first.

 

I am running the latest released copy of NVDA, Office 365 desktop apps and Windows 10 Professional.

 

Rick

 



--
Quentin Christensen
Training and Support Manager

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess 


Rick
 

Hello Quentin:

 

Thanks for taking my observation a bit further by trying additional voices.

 

On a somewhat related topic, if I remember correctly from my sighted days, Excel showed the selected cell reference in a small window in the upper left corner, just under the menu. If a cell was named, the name was shown instead of the cell reference. Is there an easy way to have NVDA speak a cell’s name instead of its reference value?

 

Thanks, Rick

 

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Quentin Christensen
Sent: Monday, November 19, 2018 6:14 PM
To: nvda@nvda.groups.io
Subject: Re: [nvda] Excel reads numeric cells in column E as exponents rather than cell value

 

Rick,

 

Well spotted, and interesting pickup.  This is exactly why NVDA itself doesn't try to interpret information, but simply present it as it is displayed.  Many synthesizers do try to interpret data, which is where you end up with St being read as "street" when it is actually used in the context of "saint" or dr for Doctor instead of drive.

 

In this case, it is processing the whole string (value and cell coordinates) as one piece of information: 1.24E4 is read as the scientific formatted number "1.24 times 10 to the power of 4".  If the value was in the cell above it is read as "1.24 times 10 cubed" or in E2 as "1.24 times 10 squared".  It is interesting to note that this affects the OneCore voices, and even then only some of them.  On my system, I have ten OneCore voices.  Of those:

- David, Zira and Mark interpret the coordinates as you encountered

- George, Catherine, Hazel, Susan, Linda, Richard or James, simply read it as it should be read: 1.24 E4

 

I just checked and of the David, Hazel and Zira Microsoft Speech API version 5 voices I have, David reads it as a scientific number, but Hazel and Zira both read 1.24 E4.

 

I will raise this with Microsoft and look to see if we can do anything with how we format the string, maybe we can put a comma in-between the pieces of information.

 

On Mon, Nov 19, 2018 at 9:47 PM Rick <softwarethatworks@...> wrote:

Hello NVDA List:

 

The other day I was reading an Excel  spreadsheet sent to me by a colleague. It had numeric values in column E. When reading these cells with Microsoft One Core voices (I am using Microsoft David) the cell name was read as an exponent to the power of 10. For instance, if the value 1.24 was present in cell E4, the speech produced, “1.24 time 10 to the power of 4” rather than 1.24 E4. This does not happen when using Espeak voices or SAPI 5 voices. Note, if the number in the cell is the result of a formula and not a direct value, the phrase, “has formula” is inserted between the number and the cell name, so it works as expected (i.e. 1.24 has formula E4).

 

Note: This does not happen using Narrator, since Narrator reads the cell name first.

 

I am running the latest released copy of NVDA, Office 365 desktop apps and Windows 10 Professional.

 

Rick

 


 

--

Quentin Christensen
Training and Support Manager

 

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

 

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess 


Matthew Horspool
 

Hi Quentin,

Doing something with how the string is formatted would do wonders for me! I'm a long time JAWS user and technically minded, but I still cannot come to terms with how NVDA "gabbles" out all of its information in one go. Perhaps I've got too used to my creature comforts, but having differentiation between things like this is such a big deal for me that I simply cannot switch to NVDA full-time. If it was improved I would consider a switch much more viable.

Inserting a comma between the cell data and its coordinates would be a good step in the right direction. Could the same be done, please, for table cell coordinates generally, table header information, access keys in menus, the name of a menu and the first control in that menu, etc?

It is only a quick fix though. Longer term, you could do worse than to implement something akin to the JAWS model, e.g. by introducing a message voice which could read coordinates, access keys and suchlike. If this is set differently, a pause will occur naturally as the voice changes. If it's set the same, pauses could be eliminated for those who like the way NVDA does things currently.

Even longer term, something like the Speech and Sounds Manager would be simply fantastic! I have perfect pitch and so, in JAWS, I currently have heading level 1 play a note C, heading level 2 play a D, level 3 an E and so on. Switching to NVDA which must report to heading and its level in speech, because what else is it supposed to do other than not report it at all, feels very sluggish. I also have sounds for buttons, radio buttons, combo and list boxes, tree views and so on and some of this can be replicated in an add-on, but not as neatly. I heard somewhere about something nicknamed "Speech Refactor" which might make some of this stuff possible? Where are we with that?

Thanks,

Matthew

 

From: nvda@nvda.groups.io [mailto:nvda@nvda.groups.io] On Behalf Of Quentin Christensen
Sent: 19 November 2018 23:14
To: nvda@nvda.groups.io
Subject: Re: [nvda] Excel reads numeric cells in column E as exponents rather than cell value

 

Rick,

 

Well spotted, and interesting pickup.  This is exactly why NVDA itself doesn't try to interpret information, but simply present it as it is displayed.  Many synthesizers do try to interpret data, which is where you end up with St being read as "street" when it is actually used in the context of "saint" or dr for Doctor instead of drive.

 

In this case, it is processing the whole string (value and cell coordinates) as one piece of information: 1.24E4 is read as the scientific formatted number "1.24 times 10 to the power of 4".  If the value was in the cell above it is read as "1.24 times 10 cubed" or in E2 as "1.24 times 10 squared".  It is interesting to note that this affects the OneCore voices, and even then only some of them.  On my system, I have ten OneCore voices.  Of those:

- David, Zira and Mark interpret the coordinates as you encountered

- George, Catherine, Hazel, Susan, Linda, Richard or James, simply read it as it should be read: 1.24 E4

 

I just checked and of the David, Hazel and Zira Microsoft Speech API version 5 voices I have, David reads it as a scientific number, but Hazel and Zira both read 1.24 E4.

 

I will raise this with Microsoft and look to see if we can do anything with how we format the string, maybe we can put a comma in-between the pieces of information.

 

On Mon, Nov 19, 2018 at 9:47 PM Rick <softwarethatworks@...> wrote:

Hello NVDA List:

 

The other day I was reading an Excel  spreadsheet sent to me by a colleague. It had numeric values in column E. When reading these cells with Microsoft One Core voices (I am using Microsoft David) the cell name was read as an exponent to the power of 10. For instance, if the value 1.24 was present in cell E4, the speech produced, “1.24 time 10 to the power of 4” rather than 1.24 E4. This does not happen when using Espeak voices or SAPI 5 voices. Note, if the number in the cell is the result of a formula and not a direct value, the phrase, “has formula” is inserted between the number and the cell name, so it works as expected (i.e. 1.24 has formula E4).

 

Note: This does not happen using Narrator, since Narrator reads the cell name first.

 

I am running the latest released copy of NVDA, Office 365 desktop apps and Windows 10 Professional.

 

Rick

 


 

--

Quentin Christensen
Training and Support Manager

 

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

 

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess 


Quentin Christensen
 

Hi Rick,

Good question.  There actually isn't currently a way to have cell names read automatically, though it is a good idea.  I created an issue for it: https://github.com/nvaccess/nvda/issues/8963  Feel free to subscribe for updates, or add extra details.

Currently, you could press alt+f3 to jump to the name box, read the name and then press escape, but it would be good to have it read automatically.

Regards

Quentin.


On Tue, Nov 20, 2018 at 12:23 PM Rick <softwarethatworks@...> wrote:

Hello Quentin:

 

Thanks for taking my observation a bit further by trying additional voices.

 

On a somewhat related topic, if I remember correctly from my sighted days, Excel showed the selected cell reference in a small window in the upper left corner, just under the menu. If a cell was named, the name was shown instead of the cell reference. Is there an easy way to have NVDA speak a cell’s name instead of its reference value?

 

Thanks, Rick

 

From: nvda@nvda.groups.io <nvda@nvda.groups.io> On Behalf Of Quentin Christensen
Sent: Monday, November 19, 2018 6:14 PM
To: nvda@nvda.groups.io
Subject: Re: [nvda] Excel reads numeric cells in column E as exponents rather than cell value

 

Rick,

 

Well spotted, and interesting pickup.  This is exactly why NVDA itself doesn't try to interpret information, but simply present it as it is displayed.  Many synthesizers do try to interpret data, which is where you end up with St being read as "street" when it is actually used in the context of "saint" or dr for Doctor instead of drive.

 

In this case, it is processing the whole string (value and cell coordinates) as one piece of information: 1.24E4 is read as the scientific formatted number "1.24 times 10 to the power of 4".  If the value was in the cell above it is read as "1.24 times 10 cubed" or in E2 as "1.24 times 10 squared".  It is interesting to note that this affects the OneCore voices, and even then only some of them.  On my system, I have ten OneCore voices.  Of those:

- David, Zira and Mark interpret the coordinates as you encountered

- George, Catherine, Hazel, Susan, Linda, Richard or James, simply read it as it should be read: 1.24 E4

 

I just checked and of the David, Hazel and Zira Microsoft Speech API version 5 voices I have, David reads it as a scientific number, but Hazel and Zira both read 1.24 E4.

 

I will raise this with Microsoft and look to see if we can do anything with how we format the string, maybe we can put a comma in-between the pieces of information.

 

On Mon, Nov 19, 2018 at 9:47 PM Rick <softwarethatworks@...> wrote:

Hello NVDA List:

 

The other day I was reading an Excel  spreadsheet sent to me by a colleague. It had numeric values in column E. When reading these cells with Microsoft One Core voices (I am using Microsoft David) the cell name was read as an exponent to the power of 10. For instance, if the value 1.24 was present in cell E4, the speech produced, “1.24 time 10 to the power of 4” rather than 1.24 E4. This does not happen when using Espeak voices or SAPI 5 voices. Note, if the number in the cell is the result of a formula and not a direct value, the phrase, “has formula” is inserted between the number and the cell name, so it works as expected (i.e. 1.24 has formula E4).

 

Note: This does not happen using Narrator, since Narrator reads the cell name first.

 

I am running the latest released copy of NVDA, Office 365 desktop apps and Windows 10 Professional.

 

Rick

 


 

--

Quentin Christensen
Training and Support Manager

 

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

 

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess 



--
Quentin Christensen
Training and Support Manager

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess 


Quentin Christensen
 

Hi Matthew,

There's a couple of suggestions there, and although I'm sure I've heard them before, I can't immediately locate them in our issue tracker.  If you have a look and find any related issues, feel free to either subscribe, or comment with more info: https://github.com/nvaccess/nvda/issues

Or, feel free to add new issues for them.  Ideally, if you could make each a different issue, so I think there's three there:
- Add option to pause between information (perhaps a combo box with "comma", "full stop", and "no pause")
- Add a "secondary" information profile.  EG the cell contents would be read in the regular profile for that app, and coordinates, name, etc in the secondary information profile.  The same could then be used in other places - eg text read in normal voice on the web, but the fact that it is a link or a heading would be given in the secondary information voice.
- Add "earcon" options for information such as headings and other elements.  The trick with this one will be not making too complex for the average user to understand, and having default sounds which make a bit of sense (I guess for headings, pick a sound and play it at a different pitch for different heading levels)

Speech refactor was our internal name for updating a heap of code that was done this year, which doesn't have an obvious difference at first, but yes, allows things like these to be implemented a lot easier in future.

Regards

Quentin.

On Wed, Nov 21, 2018 at 12:47 AM Matthew Horspool <mhorspool@...> wrote:

Hi Quentin,

Doing something with how the string is formatted would do wonders for me! I'm a long time JAWS user and technically minded, but I still cannot come to terms with how NVDA "gabbles" out all of its information in one go. Perhaps I've got too used to my creature comforts, but having differentiation between things like this is such a big deal for me that I simply cannot switch to NVDA full-time. If it was improved I would consider a switch much more viable.

Inserting a comma between the cell data and its coordinates would be a good step in the right direction. Could the same be done, please, for table cell coordinates generally, table header information, access keys in menus, the name of a menu and the first control in that menu, etc?

It is only a quick fix though. Longer term, you could do worse than to implement something akin to the JAWS model, e.g. by introducing a message voice which could read coordinates, access keys and suchlike. If this is set differently, a pause will occur naturally as the voice changes. If it's set the same, pauses could be eliminated for those who like the way NVDA does things currently.

Even longer term, something like the Speech and Sounds Manager would be simply fantastic! I have perfect pitch and so, in JAWS, I currently have heading level 1 play a note C, heading level 2 play a D, level 3 an E and so on. Switching to NVDA which must report to heading and its level in speech, because what else is it supposed to do other than not report it at all, feels very sluggish. I also have sounds for buttons, radio buttons, combo and list boxes, tree views and so on and some of this can be replicated in an add-on, but not as neatly. I heard somewhere about something nicknamed "Speech Refactor" which might make some of this stuff possible? Where are we with that?

Thanks,

Matthew

 

From: nvda@nvda.groups.io [mailto:nvda@nvda.groups.io] On Behalf Of Quentin Christensen
Sent: 19 November 2018 23:14
To: nvda@nvda.groups.io
Subject: Re: [nvda] Excel reads numeric cells in column E as exponents rather than cell value

 

Rick,

 

Well spotted, and interesting pickup.  This is exactly why NVDA itself doesn't try to interpret information, but simply present it as it is displayed.  Many synthesizers do try to interpret data, which is where you end up with St being read as "street" when it is actually used in the context of "saint" or dr for Doctor instead of drive.

 

In this case, it is processing the whole string (value and cell coordinates) as one piece of information: 1.24E4 is read as the scientific formatted number "1.24 times 10 to the power of 4".  If the value was in the cell above it is read as "1.24 times 10 cubed" or in E2 as "1.24 times 10 squared".  It is interesting to note that this affects the OneCore voices, and even then only some of them.  On my system, I have ten OneCore voices.  Of those:

- David, Zira and Mark interpret the coordinates as you encountered

- George, Catherine, Hazel, Susan, Linda, Richard or James, simply read it as it should be read: 1.24 E4

 

I just checked and of the David, Hazel and Zira Microsoft Speech API version 5 voices I have, David reads it as a scientific number, but Hazel and Zira both read 1.24 E4.

 

I will raise this with Microsoft and look to see if we can do anything with how we format the string, maybe we can put a comma in-between the pieces of information.

 

On Mon, Nov 19, 2018 at 9:47 PM Rick <softwarethatworks@...> wrote:

Hello NVDA List:

 

The other day I was reading an Excel  spreadsheet sent to me by a colleague. It had numeric values in column E. When reading these cells with Microsoft One Core voices (I am using Microsoft David) the cell name was read as an exponent to the power of 10. For instance, if the value 1.24 was present in cell E4, the speech produced, “1.24 time 10 to the power of 4” rather than 1.24 E4. This does not happen when using Espeak voices or SAPI 5 voices. Note, if the number in the cell is the result of a formula and not a direct value, the phrase, “has formula” is inserted between the number and the cell name, so it works as expected (i.e. 1.24 has formula E4).

 

Note: This does not happen using Narrator, since Narrator reads the cell name first.

 

I am running the latest released copy of NVDA, Office 365 desktop apps and Windows 10 Professional.

 

Rick

 


 

--

Quentin Christensen
Training and Support Manager

 

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

 

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess 



--
Quentin Christensen
Training and Support Manager

Official NVDA Training modules and expert certification now available: http://www.nvaccess.org/shop/

Facebook: http://www.facebook.com/NVAccess 
Twitter: @NVAccess