NORM/Vowels.R Change Log

As of April 2009, we use this page to announce and document changes made to the NORM website and Vowels R package. Most posts are from Tyler - email him if you have any questions about these things (email addresses at bottom of page). Starting in November 2011, we now post changes and other announcements to the new blog.

April 11, 2010

Numerous updates have been posted to vowels.R and to NORM this weekend with the release of versions 1.1. (It may take a few days for the new vowels.R version to post to CRAN.) Many changes are small, but users should notice several improvements. The main changes are listed here:

New & fixed in vowels.R version 1.1:

• New compute.medians() function to generate (and plot) vowel medians instead of only vowel means
• Faster standard deviation ellipsis drawing (and smaller resulting file sizes). This should no longer be sooo slow
• Added a 'separate' parameter to the add.spread.vowelplot() function so that it can be used to plot group-based standard deviations without doing cumbersome manipulations of the data (set separate=FALSE, the default is TRUE to mirror the current implementation even though this is the opposite of the default behavior for the compute.means(), compute.sds(), and compute.medians() functions)
• Fixed some bugs in the generation of legends for certain combinations of vowel coloring and shaping

New & fixed in NORM version 1.1:

• NORM now offers the modified version of the Watt & Fabricius method - cf. Fabricius et al. 2009 (this has been available in vowels.R for some time, but is now an option via NORM)
• NORM now generates inline PNG format images instead of JPEG format images. This makes nicer looking (i.e. less fuzzy) graphics. (Look for further improves in image quality in the future.)
• NORM also can make larger inline plots than before. The plots should be crisper and larger
• The new NORM11_stage script generates better fitting axes for when standard deviation ellipses and bars are plotted with grouped values. Ellipses and bars should no longer go off-plot

July 24, 2009

There was a small bug in the ellipsis plotting function that would make ellipses fail to draw in black & white. This has been fixed on NORM and vowels version 1.0-3, which fixes the bug (and has no other changes), will be posted to CRAN shortly. It should be up and available in the next few days.

I also fixed a bug in the NORM10_stage script that was making standard deviation error bars and ellipses fail to draw when plotting vowel means with the Watt & Fabricius function. This was a problem in the staging script only and not the vowels package. (Means were being generated and stored for the normalized values before determining the standard deviations.)

July 15, 2009

I modified the NORM10_stage script (the interface between NORM's web-interface and the vowels package backend) to ignore any blank lines that may be in a user's input file. These blank lines caused various NORM functions (such as Lobanov normalization) to fail and could easily 'sneak' into the end of an input file via MS Excel.

Tyler can't think of a reason why a user would want to have a blank field for a speaker or vowel, so now uses blank entries in those two fields to determine a blank 'line'. That is, even if there is formant data in an input file line, that line will be ignored by NORM if there is no speaker and vowel specified. (Likely, NORM would have failed anyway if this was the case, but it seems better to be explicit about this...)

Blank lines are removed in the NORM10_stage script by the following code. You can use this manually, in R with the vowels package, as well.

> vowels <- vowels[vowels[,1]!="" & vowels[,2]!="",]

April 22, 2009

Thanks to Santiago Barreda (UAlberta), vowels and NORM now plot standard deviation ellipses in addition to bars. Note that the information the ellipses provide is no different than that from the standard deviation bars, but I admit this is a slightly nicer view of the data and have added the feature due to popular demand.

Vowels version 1.0-2 is posted to CRAN (though takes a day or two to show up) and is available from the About NORM page.

New options in vowels_1.0-2:

• Standard deviation ellipses
• Better support for overriding (and removing) plot shapes
• Added the ability to override the x- and y-axis limits (i.e., you can now plot, for example, only the high front vowels, but still generate axes that would fit the entire vowel space; this is handy if you want to build up a vowel plot piece-by-piece)

New option in NORM: Plot standard deviation ellipses instead of bars. Note that this is slightly computationally heavy and seems to slow the server a little. It's actually the first thing I've noticed that seems to slow the server's response. I'll see about speeding this up in future versions (if possible).

March 19, 2009

There was a minor bug in the vowels R package (and therefore NORM) that was causing Nearey1 plots to be labeled as Nearey2 and vice-versa. The processing was correct - the bug was only in the labeling of the plot (down at the bottom, where it says "Variant..."). This has now been fixed on the website and a new vowels package, version 1.0-1, is available. It will be a few days until this update posts on CRAN, but it is available now on the About NORM page.

February, 2009

New NORM version 1.0! Since its first release at NWAV 2007 NORM has been in beta development (version 0.9). We are happy to announce NORM version 1.0! The web-based version (where you're at) is a bit different than it used to be. In particular, you can now set custom colors, run only one normalization procedure (or alternatively more than two), download all result files with one click, generate group means, generate standard deviation bars, and more. The actual NORM processing engine is a redesigned and complete R package. If you use R, even just a little, you are urged to download the vowels R package - see the About NORM page for more information and for links to the files. If you prefer the old version of NORM (v. 0.9), or need normalization options no longer available here, you can still use the old version.