question about spanish MOR

[Janet Bang]

Hi Brian,

Thank you for your help! Yes we have noted that for Spanish that the
taggers are quite good in terms of the break down. In our own review of a
couple of transcripts we've also noted about 95% accuracy for part of
speech and lemma each and even a higher percentage if you consider accuracy
of the lemma alone, so we are pretty happy with how we can automate such a
complicated step.

Our version of the SPA MOR was downloaded in 2018, so we will download the
newer version.

For the prepost rules, I am not familiar with how to incorporate these. How
would I include this in the SPA MOR?

Janet

On Thu, Jan 16, 2020 at 6:16 PM Brian MacWhinney <macw at andrew.cmu.edu>
wrote:

> Dear Janet,
>     Typically, automatic taggers get about 95% of the words right.  The
> best for adult written English are up to 97%.  For CHILDES English, we can
> hit about 97%.  For Spanish, we are probably down around 95%, although
> sometimes it seems better.
>
> In the case of ratito, you must be using some much earlier version of SPA
> MOR.  The current version doesn't have this problem.  The case of bonita is
> different. It created a few test sentences and it seems that this problem
> arises when bonita is the last word in the sentences, as in "es una niña
> bonita".  The problem is that the association of "co" with final position
> is so strong that it overrides the association of an adjective with the
> preceding noun.  This can be corrected with a couple of prepost rules that
> take words that can be either co or adj and force them to be adj when they
> follow a noun.  Here are the rules:
>
> # es una gata bonita
> n|* co:voc|*^adj|* => n|* adj|*
> # es una niña bonita
> co:voc|*^n|* co:voc|*^adj|* => n|* adj|*
>
> These work for these cases, but (1) you may have other cases I don't know
> about, and (2) prepost rules can also mess up other things.  So, please
> give this a try.
>
> --Brian
>
> On Jan 16, 2020, at 8:47 PM, Janet Y Bang <jbang at stanford.edu> wrote:
>
> Hello,
>
> We are working with the Spanish MOR and noticed a few errors in the MOR
> line:
>
>
>    1. bonita oftentimes gets broken down into co:voc|bonita=pretty in
>    cases where the word "bonita" is being used to modify a noun, e.g.,
>    utterance is "niña bonita.
>    2. In cases were "rato" or "ratito" is used to indicate a period of
>    time, oftentimes the breakdown is n|rato&m-DIM=rat OR n|rato-m=rat.
>
> What would be the most efficient way to fix these errors? We have a
> lab-internal .cut file where we are adding new words to our internal mor
> dictionary that were not in the downloaded dictionary (so that we can keep
> track of differences between our lab lexical items and the downloaded
> dictionary), but we weren't sure how to override words that were already in
> the dictionary? Would it be best to fix these in the relevant cut files in
> the lex folder (i.e., adj.cut, n.cut, respectively)?
>
> Thank you,
> Janet
>
> --
> Janet Y. Bang, Ph.D.
> Postdoctoral Fellow
> Department of Psychology
> Stanford University
>
> jbang at stanford.edu
>
>
>
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "chibolts" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to chibolts+unsubscribe at googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/chibolts/DM5PR02MB3275154B3A345DAA4BD9B5A2D7310%40DM5PR02MB3275.namprd02.prod.outlook.com
> <https://groups.google.com/d/msgid/chibolts/DM5PR02MB3275154B3A345DAA4BD9B5A2D7310%40DM5PR02MB3275.namprd02.prod.outlook.com?utm_medium=email&utm_source=footer>
> .
>
>
> --
> You received this message because you are subscribed to the Google Groups
> "chibolts" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to chibolts+unsubscribe at googlegroups.com.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/chibolts/ECE843B9-F0F4-4D19-920D-BE7D4DD5F425%40andrew.cmu.edu
> <https://groups.google.com/d/msgid/chibolts/ECE843B9-F0F4-4D19-920D-BE7D4DD5F425%40andrew.cmu.edu?utm_medium=email&utm_source=footer>
> .
>

-- 
You received this message because you are subscribed to the Google Groups "chibolts" group.
To unsubscribe from this group and stop receiving emails from it, send an email to chibolts+unsubscribe at googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/chibolts/CAC5V4hjo%3DopXfGrbh3r1mPBRhJyubDVkD2nCP7f9a_7wBwmEfw%40mail.gmail.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/chibolts/attachments/20200117/72cc4c57/attachment.htm>