PrepositionDisambiguation Google Group

Re: Is the project finished or does it continue?

2008-07-28T11:38:01Z

Thank you for your answer.
I am very interested in this topic and I am working in my thesis on
prepositions (Contrasting the use of English and Spanish
prepositions).
I am going to read the links you have suggested to learn about The
Preposition Project. I would be very pleased to contribute with the

Re: Is the project finished or does it continue?

2008-07-22T15:43:24Z

Most of the activity associated with this Google group was a discussion
of preposition disambiguation in SemEval-2007. However, the group was
named so that it could serve as a forum for discussion of preposition
disambiguation in general. Perhaps there is little interest in this
topic, although it seems to me that understanding preposition behavior

Is the project finished or does it continue?

2008-07-22T11:07:44Z

I have found no entries since 2007, and I woul like to know about the
project. Is it finished or does it continue?

The Preposition Project: Completion of Phase I

2007-08-26T22:27:26Z

The first phase of The Preposition Project (TPP) has now been completed.
Each of 673 preposition senses for 334 English prepositions (including
phrasal prepositions) has now been characterized with a variety of
information. For the 56 most common, single-word prepositions, this
characterization has made use of FrameNet instances.

Re: SemEval 2007 Task 6 (Preposition Disambiguation) Results

2007-04-07T21:18:10Z

Hi Deniz,
I think the questions (and my answers) you raise deserve full distribution.
I thought I had gotten rid of all the bad ones. I'll go back and check
these and see if there are some additional steps I can take to ensure
that any others are thrown out as well.
Traditionally, the idea is that the final results are not made known

Re: A question about the training data

2007-03-23T14:50:19Z

Hi Ken,
Thanks so much for clarifying this for me, I really appreciate it.
In my case, it would make a world of difference. If I didn't know that
the TPP versions of the data contained both the training and test
instances, my final classifiers would produce 100% accuracies on the
test sets, and that would be really bad. I suspect that anyone who

Re: A question about the training data

2007-03-23T14:13:09Z

Oh, rats. Yes, it is correct that the TPP versions are supersets, in
fact, the full sets from which the partition into training and test sets
were made (including the buggy ones that were discussed). It really
isn't an issue for the formal evaluation, but it certainly would
confound the interpretation of your own results. I.e., if you train on

A question about the training data

2007-03-23T12:50:09Z

Hi Ken,
I just realized that the actual training data from the Semeval webpage
is different from the data that we can dowload from TPP's webpage, it
seems that the TPP versions are super sets of the Semeval versions
(for the available prepositions anyway). Since I trained some of my
models based on the TPP version of the data, I'm wondering if this is

New data

2007-03-19T21:08:59Z

I've just been informed that the updated training and test sets have now
been posted. I've only downloaded the training set myself to see
whether it's ok. Both sets should have several files with a March 2
timestamp, so make sure you have the most current files.
Also, if you have any problems with the submission system at SemEval,

Errors in the test data

2007-03-19T14:35:56Z

About two weeks ago, I sent a revised set of training and test data to
the organizers to weed out the problematic instances. They have not yet
posted the revision. I sent an email asking that they do this.
However, if you are able to weed these out on your own with a simple
check (primarily, the <head> element doesn't contain the preposition of

Errors in the testing data

2007-03-19T07:23:03Z

Hi guys,
I found some errors in the xml files of the test data, I wonder what
we should do about them.
Here are the errors:
<instance id="across.p.fn.748335" docsrc="FN">
<context>
in the traffic jogged across Theobald 's Road to its north
side .<head></head>
</context>
</instance>
<instance id="across.p.fn.215850" docsrc="FN">

semeval 07 data request

2007-03-17T23:21:59Z

Ken,
Would you email us the full set of data?
Thank you
Rocio Guillen
California State University San Marcos

Re: SemEval Preposition Baselines

2007-03-15T17:04:58Z

For anyone else who wants the full set of data currently used at The
Preposition Project's Online TPP site, please send me an email. (I
haven't yet created the mechanism for downloading the source data from
that site. This is on my to-do list.)
The test data are available at the SemEval site (I think). You have to

Re: SemEval Preposition Baselines

2007-03-15T01:47:30Z

Hi Ken,
I think I must have missed something here, but where can I get the
data for the prepositions that are not linked at [link]
Also, when would you be releasing the testing data?
Thanks a lot.
Patrick

SemEval Preposition Baselines

2007-03-14T21:33:38Z

I have prepared two baselines for use in the SemEval preposition task,
one for selecting the first sense, and one for selecting the most
frequent sense as derived from the training data. They are described in
the Evaluation section of the main web page for the task
([link]), with a link to the tables