New development release of AntConc with effect size measures

[Laurence Anthony]

Hi Everyone,

I haven't worked on AntConc for quite a while due to work pressures and commissions for software for other purposes. But, I finally managed to return to the software last week and updated the plot function and completely revised the word list and keyword lists tools. 

Now, the plots can be saved and offer normalized or true scaling, and the word list tool runs faster. The keyword tool has the biggest improvements. It now runs faster, offers more statistics choices, and has 10 different effect size measures with which to rank the results. The defaults have changed with this versions (I would consider these best practice in the field), but options are there to allow the results from previous versions of AntConc to be used as well. In particular, the keyword tool now calculates Log-Likelihood using 4 terms instead of 2 terms (as in the previous version), cuts off the keyword list based on p-values that use the Boneferoni correction (instead of using ad-hoc threshold values), and uses a ranking based on an effect size measure.

This version is still under the "Development" section of my website, but I think this new version is a huge improvement over earlier versions. If you are interested, please try it out:

http://www.laurenceanthony.net/software/antconc/

Once a few people have tried it out, I intend to release this version as the official AntConc 3.5.0 version in the near future.

Best wishes,

Laurence.

###############################################################
Laurence ANTHONY, Ph.D.
Professor
Center for English Language Education in Science and Engineering (CELESE)
Faculty of Science and Engineering
Waseda University
3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan
E-mail: antho...@gmail.com
WWW: http://www.laurenceanthony.net/
###############################################################

[Reed Darsey]

On 4 Jun 2016 at 22:52, Laurence Anthony wrote:

> This version is still under the "Development" section of my
> website, but I think this new version is a huge improvement
> over earlier versions. If you are interested, please try it
> out:
> http://www.laurenceanthony.net/software/antconc/

I downloaded the under listed as:
Windows 64-bit (3.5.0 Dev)

AntConc_64bit.exe; 12,995,584 bytes
It has an MD5 of: CB89E80A297BFAEB4206C764B2A8F28A

But it does not run on my Win 7 Home Premium system, where
I've run the previous versions on.


______________________________________
Reed Darsey / Grand Bay, Alabama, USA

[Laurence Anthony]

Hi Reed,

Can you confirm that your Win Home Premium is a 64 bit system? Also, did you try the 32 bit dev version?

Laurence

--
You received this message because you are subscribed to the Google Groups "AntConc-discussion" group.
To unsubscribe from this group and stop receiving emails from it, send an email to antconc+u...@googlegroups.com.
To post to this group, send email to ant...@googlegroups.com.
Visit this group at https://groups.google.com/group/antconc.
For more options, visit https://groups.google.com/d/optout.

[Reed Darsey]

On 5 Jun 2016 at 8:57, Laurence Anthony wrote:

> Can you confirm that your Win Home Premium is a 64 bit
> system?

Yes.

Also, did you try the 32 bit dev version?

I downloaded it now, and it doesn't run either.

[Reed Darsey]

On 4 Jun 2016 at 19:02, I wrote:

> On 5 Jun 2016 at 8:57, Laurence Anthony wrote:
>
> > Can you confirm that your Win Home Premium is a 64 bit
> > system?
>
> Yes.
>
> Also, did you try the 32 bit dev version?
>
> I downloaded it now, and it doesn't run either.

BUT, I moved them both to a fresh directory without an
existing 'antconc_settings.ant' and they both ran.

I guess we need to use a new directory, and manually enter
in our old settings, then?

[Laurence Anthony]

Hi,

That's right. You should be able to launch the new AntConc from any location, but as you have noticed, you'll need to be careful to not have any old settings files around. The new version has various new settings that aren't compatible with older versions. This is written in the help guide, and I've now added it to the download page.

Actually, this is the case with all updates, e.g. from 3.3 to 3.4 or 3.4 to 3.5.

Let me know how the new version performs.

Laurence.

###############################################################
Laurence ANTHONY, Ph.D.
Professor
Center for English Language Education in Science and Engineering (CELESE)
Faculty of Science and Engineering
Waseda University
3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan
E-mail: antho...@gmail.com
WWW: http://www.laurenceanthony.net/
###############################################################

[Reed Darsey]

On 5 Jun 2016 at 9:54, Laurence Anthony wrote:

> That's right. You should be able to launch the new AntConc
> from any location, but as you have noticed, you'll need to
> be careful to not have any old settings files around. The
> new version has various new settings that aren't compatible
> with older versions. This is written in the help guide, and
> I've now added it to the download page.
>
> Actually, this is the case with all updates, e.g. from 3.3
> to 3.4 or 3.4 to 3.5.

This was going from version 3.5.0 (2015) to 3.5.0 (2016),
and I don't see an associated help guide to download.

[Laurence Anthony]

Hi Reed,

Just look at the help guide in the menu. But, neither the 2005 dev version or this is the official 3.5.0. They are both development versions (3.5.0 Dev). Actually, from now on, I plan to call the dev version simply "dev" with no attached version number as it's a bit confusing.

Laurence.

###############################################################
Laurence ANTHONY, Ph.D.
Professor
Center for English Language Education in Science and Engineering (CELESE)
Faculty of Science and Engineering
Waseda University
3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan
E-mail: antho...@gmail.com
WWW: http://www.laurenceanthony.net/
###############################################################

[ED Kitis]

Hi Laurence,

I'm in the middle of writing a paper which is a (critical) discourse analysis of newspaper discourse. I had already started using version 3.4.3 of AntConc when I bumped into the newest/development version. However, I'm having some issues with version 3.5.0 and was wondering if you can help?

(a) My corpus word/token count is slightly different from what I get with version 3.4.3.

(b) I'm unable to use the concordance tool and get the following error message when I attempt to access concordance lines:

"Malformed UTF-8 character (fatal) at (followed by file address in my mac and ending in) /AntConc.app/Contents/Resources/lib/std/cavaexecscript/anconc.pl line 18476

(c) finally, can you please explain why the new default statistics/settings of the keyword tool give such radically different keywords in relation to the same corpus (and ref corpus) than version 3.4.3? Can you please recommend what statistics/settings are best for the kind of project I described above (corpus driven CDA)?

Thank you very much

Kind regards
Dimitris

[Laurence Anthony]

Hi Dimitris,

Let me answer inline:

I'm in the middle of writing a paper which is a (critical) discourse analysis of newspaper discourse. I had already started using version 3.4.3 of AntConc when I bumped into the newest/development version. However, I'm having some issues with version 3.5.0 and was wondering if you can help?

You should definitely use 3.4.3 for your paper . 3.5.0 is still in development (hence the "Dev" naming) and so should only really be used by people wanting to try out some new features.

 

(a) My corpus word/token count is slightly different from what I get with version 3.4.3.

From your error in (b) I suspect the difference is due to a problem with your encoding. The token definition in 3.5.0 dev is identical to that in 3.4.3.

 

(b) I'm unable to use the concordance tool and get the following error message when I attempt to access concordance lines:

"Malformed UTF-8 character (fatal) at (followed by file address in my mac and ending in) /AntConc.app/Contents/Resources/lib/std/cavaexecscript/anconc.pl line 18476

This seems to be a problem with the 3.5.0 development version. 3.5.0 dev attempts to automatically detect the character encoding in the files, but if it guesses wrongly the program seems to die. I suggest that you first try to set the encoding to the one used to create your corpus. If you used UTF-8, the error might mean that your UTF-8 files have not been created properly.

 

(c) finally, can you please explain why the new default statistics/settings of the keyword tool give such radically different keywords in relation to the same corpus (and ref corpus) than version 3.4.3? Can you please recommend what statistics/settings are best for the kind of project I described above (corpus driven CDA)?

3.5.0 dev has now introduced a 4-term log-likelihood keyness measure instead of the old 2-term measure. Also, it uses an effect size measure as the default method for ranking keywords. These settings have both been recommended by various people in the corpus linguistics field. If you choose the 2-term log-likelihood keyness measure, and order by keyness instead of effect size, the results should be identical to the 3.4.3 results. This is all documented in the help file.

 I hope that helps.

Laurence.

[ED Kitis]

Hi, indeed it does. Thank you and thanks for improving AntConc, looking forward to using it.

On Saturday, June 4, 2016 at 3:53:00 PM UTC+2, Laurence Anthony wrote: