Lab 6 (due 2/12 11:59 pm)

Overview

The goals for this lab are to reduce the MT output, by working with the VPM (variable property mapping) and doing further morphological clean up and/or adding constraints to semantically empty auxiliaries, and to get one more phenomenon working (preferably adnominal possession or clausal complements). We will primarily be working by editing tdl this week, but might consult the customization system for inspiration on how to implement things. You will also use [incr tsdb()] to compare the initial and final state of your grammar for the week over the testsuites.

For tdl editing, please practice incremental development: Test as frequently as you possibly can, both by compiling the grammar and by testing specific sentences.

This lab entails the following general steps, which are not (fully) ordered with respect to each other.

Refine your semi.vpm file to constrain the possible values of e.g. aspect in generated sentences.
Clean up morphology OR semantically empty words OR other sources of spurpious generator output.
Continue testing MT to see the effects of those improvements.
Process the test corpus and your testsuite using [incr tsdb()], the LKB, and the grammar resulting from your updated choices file.
Examine the results of the second test run for coverage, accuracy and ambiguity, including as a diff to the final lab 5 test run.
Write it all up :)

Variable property mapping

You may have noticed that you get many variants on generation if you start with a form that is underspecified for e.g., aspect or evidentiality. We can get a handle on this by using variable property mapping to supply default values in the unmarked case (either in monolingual generation or in the MT scenario). The basic strategy is to take any underspecified values in variable properties and translate them, via vpm, to something that conflicts with any more specific values your grammar can produce.

The file semi.vpm provides a mapping between grammar-external features of indices (referential indices and events) and their values, and grammar-internal ones. For background on VPM, see the DELPH-IN wiki. As soon as you start using a VPM file, then only variable properties (features on indices) that are handled in the file are actually preserved.

You should already have a semi.vpm file provided by the customization system. Open it up and see which variable properties are there, and then look in your grammar to see what is missing. In general, we'd expect to see all of the features of the types event and ref-ind represented in a mature semi.vpm file.
Your script file tells the lkb to load the semi.vpm file with the following line:
```
(mt:read-vpm (lkb-pathname (parent-directory) "semi.vpm") :semi)
```
The semi.vpm files currently generated by the customization system are not fully ready for MT. In particular, the feature paths should include PNG. and E. only on the internal (left-hand) side. Thus you'll want to edit blocks like this:
```
E.TENSE : E.TENSE
  present <> present
  past <> past
  future <> future
  * <> *
```
To look like this:
```
E.TENSE : TENSE
  present <> present
  past <> past
  future <> future
  * <> *
```
If your grammar uses a PERNUM feature, you'll need to map separate PER and NUM features from the external (right-hand side) of the VPM to a single PRENUM feature on the internal (left-hand side). See the example under "Properties: An Example" on the DELPH-IN wiki page.
If you have any other features you have added on indices, you will need to provide VPM entries for them as well. (If you added them through the customization system, they may be there already.)
If your language has aspect marked in some sentences but other forms that are just underspecified for aspect, you'll want to have the default aspect be "no-aspect". Define this as a subtype of aspect in your grammar, but don't have anything other than the semi.vpm mention it otherwise. In semi.vpm, replace * <> * under aspect with the following:
```
* >> no-aspect
no-aspect << [e]
```
If all forms in your language are marked for aspect, you may still want some default value in cases where the input MRS has no aspect, so that you don't get all the forms. If you wanted perfective as your defaul aspect, replace * <> * with:
```
perfective << [e]
```

You can do a similar trick for other kinds of generation ambiguity relating to variable properties.

Test your semi.vpm file by doing self-translation (i.e. translating from your language to your language). You should see fewer strings coming out.

Morphology or semantically empty word clean up

Here you'll be doing tdl editing to either remove morphology that is insufficiently constrained (and so showing up all over the place....) or adding constraints to that morphology. If the problem is with semantically empty words, we'll be adding constraints to the words or possibly refining trigger rules.

Please post to Canvas Tuesday with:

A description of the kinds of extra realizations you're seeing
A description of what the actual purpose of the forms (affixes, words) appears to be
Anything you already have planned in terms of clean-up.

I'll reply with suggestions about how to proceed.

For the write-up for this section, I'd like to see both a description of what changes you made and some quantiative observations about how the number of generator outputs is affected.

Additional phenomena

Here, we'd like to get adnominal possession or clausal complements working, i.e. the equivalents of these sentences from the eng.txt:

The dog s car sleeps
My dogs sleep
I think that you know that dogs chase cars
I ask whether you know that dogs chase cars

Please post to Canvas Tuesday with:

An indication of which of these two phenomena you are working on
A description of the phenomenon in the language
A description of what is already working (if anything) in your grammar wrt the phenomenon.

I will advise on Canvas how to proceed in each case!

Run both the test corpus and the testsuite

Following the same procedure as usual, do test runs over both the testsuite and the test corpus.

Collect the following information to provide in your write up:

How many items parsed?
What is the average number of parses per parsed item?
How many parses did the most ambiguous item receive?
What NEW sources of ambiguity can you identify?

Write up

Your write up should be a plain text file (not .doc, .rtf or .pdf) which includes the following:

A description of the chnages you made to your semi.vpm file, why you made those specifc changes, and how that affected generator output.
A description of the phenomena you improved in the grammar (including both the morphology etc clean up and adnominal possession or clausal complements), including:
- Prose description of the phenomenon
- Prose description of your analysis
- The specific changes you made to the tdl (paste in the actual tdl)
- Specific IGT I can use to test the analysis / investigate if something isn't working and you need help.
A description of the performance of your final grammar for this week on the test suite and test corpus, as compared to your starting grammar (see details above).

Submit your assignment

Be sure your write up and the text-file version of your test suite are included in your grammar directory.
Likewise, make sure that tsdb/home includes four profiles:
1. Final testsuite with initial grammar for the week
2. Final testsuite with final grammar for the week
3. Test corpus with initial grammar for the week
4. Test corpus with final grammar for the week
If you're using svn, export the grammar so I don't get all your .svn files:
```
svn export yourgrammar iso-lab6

For git, please do the equivalent.
```

Create a tarball:

      tar czf iso-lab6.tgz iso-lab6

Upload the tarball to Canvas.

Back to course page

Last modified: