MRS in Wikiwoods Data

I just downloaded the Wikiwoods 1212 data. When I process the itsdb with delphin.itsdb.TestSuite, I could not find any MRS even if there is a derivation analysis. Does the data actually provide MRS? Thanks in advance for the help!


The server hosting the Wikiwoods data is not responding for me right now so I cannot confirm, but judging by your screenshot it appears that the mrs field is indeed empty.

In theory, with the derivation and the same version of the grammar used to produce it, you could trivially reconstruct the analysis to get the MRS, but in practice I have no idea how to do that. It may be that the MRS and some other output formats are included in the “export” files of the Wikiwoods data, but I don’t know if those files are readable by PyDelphin.

1 Like

First, it is true that the itsdb profiles in Wikiwoods 1212 do not have the MRSs stored (that is, the mrs' field in each profile is empty). You have two options to obtain the MRSs: (1) Use the stored parse results in the export* directories instead of using the itsdb* profiles. You'll find for each item the MRS recorded in several different formats for convenience. (2) Load the 1212 ERG and the itsdb* profiles into LKB+itsdb, and "reconstruct" any particular item's derivation so you can use normal LKB viewing tools to inspect that item's MRS. More specifically, (a) download a copy of the 1214 ERG (compatible with WW1212 data): svn checkout (b) start up a Lisp image with LKB-FOS, including [incr tsdb()]: see (c) load the ERG via the LKB Top menu: Load--Complete grammar and navigate to your copy of the 1214 grammar's ./lkb/script file (d) point the [incr tsdb()] to one of Wikiwoods 1212's itsdb* directories, say itsdb0’, by selecting from the podium window the menu item Options–Database Root and havigating to your copy
(e) browse one of the profiles, say 00110' , by clicking on it and then selecting the podium's Browse--Results (answer 'yes' to its questions about the size of the tables, which used to be a possible concern) (f) choose one of the items (say the fourth one down) and scroll the view window to the right so you see the column labeled "derivation". Double-click on the "1" in the cell for the fourth row, which will pop up a window containing the recipe’ stored for the treebanked parse for that item. Double-click on the single line in that pop-up window, which will cause the LKB to recreate that parse and show you its parse tree. You can then use normal LKB graphical tools to inspect the MRS for that parse.

1 Like

Thanks for the replies!

@Dan: For the first approach, may I know if there are any parsers for reading the files in export#.tar into something that I can work with in Python? It seems that they are plain texts: