[Corpora-List] Frequency Class / H äufigkeitsklasse

[Reinhard Rapp]

> we have recently proposed the Zipf scale of word frequency,
> which is easy to understand and leads to a correct understanding
> of the word frequency effect in psychological research.

Let me confirm this and provide some more support.

>>From psychophysics the Weber-Fechner law is known. According to Wikipedia, 
it describes "the relationship between the physical magnitudes of stimuli 
and the perceived intensity of the stimuli. Fechner's law states that 
subjective sensation is proportional to the logarithm of the stimulus 
intensity."

It (at least roughly) applies, for example, to brightness and loudness. The 
logarithmic decibel scale reflects this finding.

The interesting thing is now that the Weber-Fechner law also seems to apply 
to subjective word frequencies (= word familiarities). That is, if test 
persons are asked to estimate word familiarities, a word which occurs twice 
as often than another one is (on average) estimated to occur only log(2) as 
often.

I verified this as follows: I computed Pearson's correlation coefficient 
between several thousand word familiarity estimates as obtained from test 
persons (taken from the MRC psycholinguistic database) and the same words' 
frequencies in the British National Corpus. The correlation was only 0.12 if 
the words' corpus frequencies were used directly, but went up to 0.75 if the 
logarithms of the word frequencies were taken. This is not much lower than 
the correlation between different groups of test persons which was between 
0.77 and 0.88.

If you are interested in details, please have a look at my following paper 
which also cites some of the previous work:

http://www.fb06.uni-mainz.de/user/rapp/papers/rapp-gldv2005.pdf

Kind regards,

Reinhard


-----Ursprüngliche Nachricht----- 
From: Marc Brysbaert
Sent: Wednesday, November 6, 2013 2:21 PM
To: 'Cedric Krummes' ; Corpora at uib.no
Subject: Re: [Corpora-List]Frequency Class / Häufigkeitsklasse

As far as I know, these are rarely used in psycholinguistics. The only
example I can think of are the word family lists compiled by Paul Nation for
second language acquisition
(http://www.victoria.ac.nz/lals/about/staff/publications/paul-nation/Informa
tion-on-the-BNC_COCA-word-family-lists.pdf).

Dividing a continuum in classes is likely to decrease the power of the
variable. As an alternative we have recently proposed the Zipf scale of word
frequency, which is easy to understand and leads to a correct understanding
of the word frequency effect in psychological research. The scale simply is
log10(frequency per billion words). It gives values from 1 to 7 , with the
values 1-3 representing low-frequency words and 4-7 high frequency words.

More information can be found here: http://crr.ugent.be/archives/1352

I hope this is helpful?

Kind regards,

Marc Brysbaert

-----Original Message-----
From: corpora-bounces at uib.no [mailto:corpora-bounces at uib.no] On Behalf Of
Cedric Krummes
Sent: woensdag 6 november 2013 13:55
To: Corpora at uib.no
Subject: [Corpora-List] Frequency Class / Häufigkeitsklasse

Dear colleagues,

Does anybody know a reason and/or application for frequency classes
(German: Häufigkeitsklassen)? I've read the Wikipedia entry and I know how
to calculate them but I'm unsure what to do with them. Could I do something
comparativey?

Best wishes,

Cédric Krummes
--
Dr. Cédric Krummes
"SMS Communication in Switzerland"

Universität Leipzig · +49-341-97-37404
http://www.cedrickrummes.org/contact/

_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora


_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora 


_______________________________________________
UNSUBSCRIBE from this page: http://mailman.uib.no/options/corpora
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora