[developers] tsdb result file

Michael Wayne Goodman goodmami at u.washington.edu
Tue Dec 2 14:01:16 CET 2008

Hi Bart,

Francis and I also noticed the following things relating to this bug:

1. Roots show up inconsistently in items parsed with cheap (we never
figured out why they do or do not show up)

2. Roots never show up with items parsed or generated with the LKB
(but at least it's consistent :P )

3. We found the following root conditions in our corpus: root_frag,
root_inffrag, root_informal, root_robust, and root_strict (Perhaps
there is an exhaustive list somewhere). So if you are going to filter
them out (like we are) until this bug gets fixed, make sure you get
all of them, or do a more generic root_* search (assuming you don't
have other types beginning with that string)

On Tue, Dec 2, 2008 at 8:10 PM, Stephan Oepen <oe at ifi.uio.no> wrote:
> hi bart,
>> I was looking at the result file in tsdb profiles, and I saw that in
>> some parses the root condition is included in the parse tree, and in
>> some parses it isn't.
>  [...]
>> Is this difference significant in some way?
> well, i would consider what you observe a bug.  which processor did you
> use in creating this profile?  incidentally, francis reported something
> along these lines last week.  we should no doubt iron this out.
> in principle, as of early 2007, all derivations should include the root
> symbol that licensed them.  [incr tsdb()] is backwards compatible with
> its treatment of derivations, i.e. it will successfully reconstruct and
> treebank both variants (the root information is not needed for that).
> but for future parse selection experiments, i would expect the root may
> prove a useful feature, hence we should aim to always have it reported.
>                                                      all best  -  oe
> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> +++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
> +++     CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
> +++       --- oe at ifi.uio.no; oe at csli.stanford.edu; stephan at oepen.net ---
> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

-Michael Wayne Goodman

More information about the developers mailing list