Nxaʔamxcín (Moses) Dictionary Blog

Wrote, tested and deployed some XSLT that converts or comments out some now-obsolete types of feature structures.

Made a start on three different views of entries. I have pipelines set up, and need to tweak the XSLT for each case; following that, I have to decide how to save the setting (session variable, cookie or URL parameter, or combination).

Met for most of the day with EB, SC and the rest of our team, and made some sound decisions about different views of the database. These are some notes:

Dictionary views

Elders' view (perhaps not needed, because EB and SC interface between them and the dictionary):

• Orthographical form
• Meaning
• Hyphenated form

Learner's view:

• Orthography
• Sound (EB has recordings of contributors K and A, as well as Nellie Moses Friedlander. Or perhaps PC could record the entries.)
• Phonetic (an entry could have more than one pron:seg type="n")
• Meaning
• Hyphenated form
• Usage
• Paradigm (for verbs)
• Related words: all with the same root or stem
• Cross-references: <xr>s

Print dictionaries

• Both directions
• Moses-English organized by root, but with prefixed versions included with "see root".
• Additional front matter? Grammar background information per Nancy Mattina's work in latest community-published dictionary? Charts of subject and object suffixes (where possible, using the same root with different transitive markers)?

Informants

We should gather and include information on informants (genealogy, dates, etc.) so that we can generate relationship info and trees, and provide a good credits page if we need to, but keep that info confidential for the moment. One source for info on deceased participants would be here, especially the St. Mary's Mission, Nespelem City Cemetery, Little Nespelem Cemetery, and Nespelem Catholic Cemetery (Sacred Heart?) records. Other cemetery records which might be helpful are Cashmere Cemetery and ____ Creek Cemetery in Monse.

Extra features

Portal page - with photo of Moses Mountain from EB? The web database should have the same title as the community's print dictionary: Nxaʔamxčín Nwwáwəlxtənt. EB and SC will confirm the spelling with PC.

A good map of the area, on which we could also put the placenames and link to their entries, would be good.

Names need to be divided into personal, place, story characters, orgs, flora and fauna. For flora, we could try to get a copy of Ethnobotany of the Southern Okanagan. Personal names will eventually be hidden in entries.

Generating orthography

• Start from the hyph.
• c goes to c-wedge.
• x-dot goes to x-wedge.
• any other dotted character loses the dot and is doubled.
• Some schwas will be inserted, rules to be determined.

DONE 2013-01-29.

1. Change all name feature structures from

   <f name="category">
   <symbol value="proper-noun"/>
   </f>
   

   to:

   <f name="name">
   <binary value="true"/>
   </f>
   

2. Change all loanword feature structures from

   <f name="baseType">
   <symbol value="loanword"/>
   </f>
   

   to:

   <f name="loanword">
   <binary value="true"/>
   </f>
   

3. Comment out existing compound feature structures:

   <fs>
   <f name="baseType">
   <symbol value="compound"/>
   </f>
   </fs>
   

4. Comment out existing compound lexical suffix feature structures:

   <fs>
   <f name="baseType">
   <symbol value="suffix"/>
   </f>
   <f name="derivational">
   <symbol value="lexical-suffix-compound"/>
   </f>
   </fs>
   

Met with ECH and SMK to rework the whole feature structure thing, and we went back to a flat model. We'll work with a flat model until we're sure everything is functional and accounted for, and then we'll start adding structure and dependencies as appropriate. This seems the cleanest approach, and will let us get productive again. I've encoded a clean new version of the feature system with no @dcr: attributes or comments, and we'll go from there. I'll start adding back the @dcr attributes soon.

Following this morning's discussion of feature structures, we are no longer treating names and loanwords as baseTypes (since names, and potentially loanwords too, can be multi-morphemic).

The feature structure for names is being changed from

<fs>
<f name="category">
<symbol value="proper-noun"/>
</f>
</fs>

to:

<fs>
<f name="name">
<binary value="true"/>
</f>
</fs>

Monomorphemic names will also have the <fs> of a root:

<fs>
<f name="baseType">
<symbol value="root"/>
</f>
<f name="name">
<binary value="true"/>
</f>
</fs>

The feature structure for loanwords is being changed from:

<fs>
<f name="baseType">
<symbol value="loanword"/>
</f>
</fs>

to:

<fs>
<f name="loanword">
<binary value="true"/>
</f>
</fs>

A loanword stem entry would have this feature structure:

<fs>
<f name="baseType">
<symbol value="stem"/>
</f>
<f name="loanword">
<binary value="true"/>
</f>
</fs>

Spent the day tidying up the last of the things on my list. I've implemented authentication through eXist rather than Tomcat, which makes things cleaner in the deployment department, but slightly more messy when it comes to users, groups and permissions in the db itself. I've also got the Snowball analyzer working with English stemming, and it works out of the box, with syntax highlighting and everything. Things I'll need to remember for the next one:

• In controller.xql, when doing a dispatch/forward to an XQuery library, url parameters are passed on automatically. If you erroneously pass them explicitly as well, you get an array in the XQuery.
• CSS background-size sometimes fails on Chrome with an SVG image if the root <svg> element doesn't have preserveAspectRatio="none".
• To use the Snowball analyzer, in collection.xconf:

  <lucene>
              <analyzer class="org.apache.lucene.analysis.snowball.SnowballAnalyzer">
                  <param name="name" type="java.lang.String" value="English"/>
                  <param name="stopWords" type="java.util.Set">
      				<!-- using set from StopAnalyzer.ENGLISH_STOP_WORDS_SET -->
                      <value>a</value>
                      <value>an</value>
              [...]
              </analyzer>
  </lucene>
  

Today I ported the old Cocoon app to a fresh build of eXist, and got everything working. There are two or three little niggles to sort out, but basically everything is up and running. In the process, I've cleaned out a whole pile of obsolete code, cleaned up a lot of things, and simplified a bit. Another few hours and we'll be ready to deploy this as a replacement for the current app, which is pretty flaky sometimes, and which I suspect is responsible for the death of Tomcat once in a while.

Notes for future ports:

• Don't forget to put any jars you created into WEB-INF/lib.
• Edit controller-config.xml to set up your "/site" location for the app code.
• There's still a bug with the passing of duplicate parameters into XQuery functions. I suspect the add-parameter thing is unnecessary, and params get passed through automatically -- have to test this.

"Questions for elders" now come with an email link that pre-populates an email message with the id and the question.

Meeting to discuss what we'll do during the visit of the CT folks next week. Agreed on a dinner on Sunday, presentation of the db on Monday morning, and discussions re orthography, dictionary components, a user view of the db, etc.

Aspect

Need to make some decisions about how to indicate transitive/intransitive distinctions here. The categories in green are the affixes that need to be added to the feature system.

Imperfective-aspect ʔac (allomorph c)

This aspect prefix is used with transitive stems

Imperfective-aspect sac (allomorph sc)

This aspect prefix is used with intransitive stems and cooccurs with -mix. However, there are a few examples where it shows up on transitive stems (see MLW 2003: 309).

Imperfective-aspect s-IMPF

This aspect prefix is used with intransitive stems and usually cooccurs with -mix; it is not clear whether its meaning/usage is different from that of sac-. There are a few examples where it shows up on transitive stems (see MLW 2003: 309).

Imperfective-aspect ʔas This aspect prefix may be composed of a combination of ʔac- and s-IMPF. It occurs on transitive predicates and is rare (see MLW 2003: 310).

Nonperfective-aspect mix-IMPF

This aspect suffix is used with two non-perfective constructions: imperfectives (prefixed by sac- or s- according to MDK 1982) and irrealis/unrealized forms (prefixed by kas- (i.e. kaɬ-s)).