Suggestion for Graph application

47 views
Skip to first unread message

Louis de Viron

unread,
Apr 1, 2014, 11:09:03 AM4/1/14
to unitex-...@googlegroups.com

Dear all,

I have a suggestion for future Unitex/Gramlab developments. Currently,
it is not possible to choose the application mode for a subgraph. If the
user chooses to apply "locate" with the merge mode, this mode is applied
to all the subgraphs of the main graph.

That implies a problem: if I have a subgraph in which a certain part is
tagged, the only ways to ignore the tags is either applying the main
graph without grammar outputs (which is not ok if I want the outputs of
the main graph), either making an untagged copy of the graph (which is
not easy for the maintenance).

An interesting idea would be to allow the user to choose the application
mode of each subgraph.

Do you think it is possible?

Thanks for your answers

Louis

Denis Maurel

unread,
Apr 1, 2014, 1:26:20 PM4/1/14
to Louis de Viron, unitex-...@googlegroups.com


Hi Louis,

If you use a cascade of graphs, you can choose for each one if it is in merge or in replace mode...

Best regards,

Denis Maurel


____________________________________
Professor Denis Maurel
Université François Rabelais Tours
LI (Computer Science Research Laboratory)
EPU-DI
64 avenue Jean-Portalis
37200 Tours
France
Phone: 33-2.47.36.14.35
Fax: 33-2.47.36.14.22
mailto:denis....@univ-tours.fr

http://www.univ-tours.fr/maurel

http://www.li.univ-tours.fr
http://tln.li.univ-tours.fr/


----- Mail original -----
| --
| You received this message because you are subscribed to the Google Groups
| "Unitex-GramLab" group.
| To unsubscribe from this group and stop receiving emails from it, send an
| email to unitex-gramla...@googlegroups.com.
| To post to this group, send email to unitex-...@googlegroups.com.
| Visit this group at http://groups.google.com/group/unitex-gramlab.
| To view this discussion on the web visit
| https://groups.google.com/d/msgid/unitex-gramlab/533AD68F.6090701%40earlytracks.com.
| For more options, visit https://groups.google.com/d/optout.
|

Louis de Viron

unread,
Apr 2, 2014, 4:13:50 AM4/2/14
to denis....@univ-tours.fr, unitex-...@googlegroups.com
Hi Denis,

Thanks for your answer. Unfortunately, if I understand correctly, the
casacades don't help if I don't want to use the output of the graphs
(neither in merge nor in replace mode). Here is an example with the
subgraph "det" and the main graph "main" (in attachment). Both have
outputs. I would like to use the subgraph sometimes with and sometimes
without outputs but with the main graph always in merge mode.

For example, I would like to tag the DET only if it is followed by an
adjective and obtain the results :
- <Sentence>This is a graph.</Sentence>
- <Sentence>This is <DET>a</DET> nice graph.</Sentence>

The best solution I consider is the subgraph duplication, but it's heavy
in term of maintenance.

Do you understand my problem?

Thanks for your help,

Louis
det.png
main.png

Daniel Stein

unread,
Apr 2, 2014, 4:20:50 AM4/2/14
to unitex-...@googlegroups.com, denis....@univ-tours.fr
Hi Louis,

if I understand you correct you suggest something like

:det

in order to call a subgraph in merge mode and (e.g.)

:!det

in order to ignore the output of this subgraph in this instance. Indeed, this way you could use the same subgraph with and without producing an output in one graph - I like this idea.

Kind regards
Daniel

Louis de Viron

unread,
Apr 2, 2014, 4:25:19 AM4/2/14
to unitex-...@googlegroups.com
Exactly!
Regards,
Louis
--
You received this message because you are subscribed to the Google Groups "Unitex-GramLab" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unitex-gramla...@googlegroups.com.
To post to this group, send email to unitex-...@googlegroups.com.
Visit this group at http://groups.google.com/group/unitex-gramlab.

Denis Maurel

unread,
Apr 2, 2014, 6:57:26 AM4/2/14
to Louis de Viron, unitex-...@googlegroups.com


Hi Louis,

Why you did not use <A> in your png file?

<DET> [<DET>] </DET> [<A>]

Another possibility with cascade is to insert output with graph 1, use it with graph 2 and remove it with graph 3...

Best regards,

Denis Maurel


____________________________________
Professor Denis Maurel
Université François Rabelais Tours
LI (Computer Science Research Laboratory)
EPU-DI
64 avenue Jean-Portalis
37200 Tours
France
Phone: 33-2.47.36.14.35
Fax: 33-2.47.36.14.22
mailto:denis....@univ-tours.fr

http://www.univ-tours.fr/maurel

http://www.li.univ-tours.fr
http://tln.li.univ-tours.fr/


----- Mail original -----

Louis de Viron

unread,
Apr 2, 2014, 7:16:18 AM4/2/14
to denis....@univ-tours.fr, unitex-...@googlegroups.com
Sorry it was'nt the right attachment. Here you'll find the good one.

I understand the possibility with cascade, but it seems to be a bit
heavy in a complex architecture. Daniel's proposal would be a great
solution, I think.

Regards,

Louis
det.png
main.png

Denis Maurel

unread,
Apr 2, 2014, 7:48:45 AM4/2/14
to Louis de Viron, unitex-...@googlegroups.com


Hi Louis,

To have only output when there is an adjective, you can add <A> in the subgraph DET after <E>/,.DET}

Best regards,

Denis Maurel


____________________________________
Professor Denis Maurel
Université François Rabelais Tours
LI (Computer Science Research Laboratory)
EPU-DI
64 avenue Jean-Portalis
37200 Tours
France
Phone: 33-2.47.36.14.35
Fax: 33-2.47.36.14.22
mailto:denis....@univ-tours.fr

http://www.univ-tours.fr/maurel

http://www.li.univ-tours.fr
http://tln.li.univ-tours.fr/


----- Mail original -----

Louis de Viron

unread,
Apr 2, 2014, 8:29:03 AM4/2/14
to denis....@univ-tours.fr, unitex-...@googlegroups.com
Hi Denis,

This example was created to illustrate my problem with a simple case. In fact, this problem is more common and occurs in several more complex cases for which a graph duplication is too heavy, mainly for maintenance reasons (every change needs to be replicated in two places).

Another Example:

I have a graph matching all possible NP's and tagging their heads : N-PREP-N, A-N, N-A ...
- When the NP is subject, I want to tag the NP and its head
- In the other cases, I want to use NP as an insertion, without any annotation.

At this stage, I can build two graphs (NP-tag and NP-notag) and call the first one in the first case, and the second one in the other cases. It implies that every change in the NP-tag graph must be done in NP-notag, which is very costly...

I am aware of the alternative solutions. My only goal is to suggest an improvement for the future unitex developments.

Kind regards,

Louis

Nebojsa Vasiljevic

unread,
Apr 2, 2014, 11:26:22 AM4/2/14
to unitex-...@googlegroups.com, denis....@univ-tours.fr
Hello Louis,

Have you tried to catch the undesired output into an output variable, or in your particular case to surround the subgraph call by an output variable declaration?

From the section 6.8 of Manual: "Note that when an output variable is being declared, the outputs
are not emitted in the output occurrence; they are just stored into the pending output variable(s)".

Regards,
Nebojša Vasiljević

Louis de Viron

unread,
Apr 2, 2014, 11:33:16 AM4/2/14
to unitex-...@googlegroups.com
Hi Nebojsa,
It works perfectly! Thanks a lot for your help!
Regards,
Louis
--
You received this message because you are subscribed to the Google Groups "Unitex-GramLab" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unitex-gramla...@googlegroups.com.
To post to this group, send email to unitex-...@googlegroups.com.
Visit this group at http://groups.google.com/group/unitex-gramlab.
Reply all
Reply to author
Forward
0 new messages