meeting_080619

Meeting notes, Pedagogical programs, 080619

Participants

Biret Ánne, Lene, Saara, Trond.

Identical lemmas

Some words have identical lemma forms, but different inflection. How should they be marked?

bassi-prof vs bassi-time bassit basit

  <entry id="bassi_actor">
    <lemma>bassi</lemma>
	<pos class="N"/>
    <translations>
      <tr xml:lang="nob">reingjører</tr>
      <tr xml:lang="nob">rengjøringsassistent</tr>
      <tr xml:lang="nob">vasker</tr>
    </translations>
    <semantics>
      <sem class="PROFESSION"/>
    </semantics>
    <stem class="bisyllabic" diphthong="no" gradation="no" rime="i"/>
    <dialect class=""/>
    <sources>
    </sources>
    <forms>
    <form tag="N+Sg+Gen">bassi</form>
    </forms>
  </entry>

  <entry id="bassi_time">
    <lemma>bassi</lemma>
	<pos class="N"/>
    <translations>
      <tr xml:lang="nob">helligdag</tr>
    </translations>
    <semantics>
      <sem class="TIME"/>
    </semantics>
    <stem class="bisyllabic" diphthong="no" gradation="yes" rime="i"/>
    <dialect class=""/>
    <sources>
    </sources>
    <forms>
    <form tag="N+Sg+Gen">basi</form>
    </forms>
  </entry>

In the questions, we use lemma as id, cf:

<q> <qtype>N-ILL</qtype> <question> <text>Masa SUBJ MAINV</text> <element type="SUBJ"> <grammar pos="Pron"/> </element> <element type="MAINV"> <grammar tag="V+Ind+Tense+Person-Number"/> <id>liikot</id> </element> </question> <answer> <text>(CLADV) (SUBJ) MAINV (CLADV) N-ILL (CLADV)</text> <element type="N-ILL"> <sem class="OBJECT"/> <grammar pos="N"/> </element> </answer> </q>

So, instead of referring to <lemma>, we refer to <id>.

The questions_nouns.xml etc. documents will be changed from <lemma> to <id>, and conversely for the relevant dtds (Saara).

        <grammar tag="V+Ind+Tense+Person-Number"/>
        <lemma>liikot</lemma>

Solution:

        <grammar tag="V+Ind+Tense+Person-Number"/>
        <id>liikot</id>

Saara looks into this.

But this solution works only as long as we are within the ped universe. When we take words from the big lexicon there is no way to refer to identical entries. Here both are bassi and not bassi_actor and bassi_time . When we analyse the answers, and when we go to isme.fst, we do not have a way to distinguish between the two bassis in isme.fst:

bassi BASSI ;        ! II-I cg Acc: bassi+N+Sg+Acc:basi
bassi:bassi¤ ACTOR ; ! no cg Acc: bassi+N+Sg+Acc:bassi

The answer is that we store our generated forms, but use sme.fst only for analysis. So, generation is not a problem.

Reference to grammar in questions

<q>
<qtype>N-ILL</qtype>
  <question>
      <text>Masa SUBJ MAINV</text>
      <element type="SUBJ">
        <grammar pos="Pron"/>
      </element>
      <element type="MAINV">
        <grammar tag="V+Ind+Tense+Person-Number"/>
        <lemma>liikot</lemma>
      </element>
    </question>
    <answer>
      <text>(CLADV) (SUBJ) MAINV (CLADV) N-ILL (CLADV)</text>
      <element type="N-ILL">
	<sem class="OBJECT"/>
        <grammar pos="N"/>
      </element>
    </answer>
  </q>
  
  <q>
    <qtype>V-POT</qtype>
  <question>
    <text>MAINV go SUBJ</text>
    <element type="MAINV">
      <grammar tag="V+Ind+Prs+Person-Number"/>
      <sem class="ACTIVITY"/>
    </element>
    <element type="SUBJ">
      <sem class="HUMAN"/>
      <grammar pos="N"/>
    </element>
  </question>
  <answer>
    <text>In dieđe MAINV go </text>
      <element type="MAINV">
        <grammar tag="V+Pot+1sg"/>
      </element>
    </answer>
  </q>

<q>
  <qtype>V-GO</qtype>
  <question>
    <text>MAINV go SUBJ luomus</text>
    <element type="MAINV">
      <grammar tag="V+Cond+Prs+Person-Number"/>
      <sem class="ACTIVITY"/>
    </element>
    <element type="SUBJ">
      <sem class="HUMAN"/>
      <grammar pos="N"/>
    </element>
  </question>
  <answer>
    <text>(CLADV) SUBJ MAINV (luomus) (CLADV)</text>
  </answer>
    <answer>
      <text>(CLADV) SUBJ MAINV (luomus) (CLADV)</text>
      <element type="SUBJ">
        <lemma>son</lemma>
        <grammar tag="Pron+Pers+Person-Number"/>
      </element>
    </answer>
    <answer>
	  <type>NEG</type>
	  <type>CONNEG</type>
      <text>(luomus) (SUBJ) NEG CONNEG (luomus) (CLADV)</text>
      <element type="NEG">
        <grammar tag="Neg+Ind+Person-Number"/>
      </element>
      <element type="CONNEG">
        <grammar tag="V+Cond+Prs+ConNeg"/>
        <sem class="ACTIVITY"/>
      </element>
    </answer>
  </q>

The <type> element means that when this answer in the contextual Morfa game is presented to the user, the NEG and CONNEG should be left empty.

If <type> is left unspecified, then the left-out element is the default element for each question-type.

PN for the verb in the answer inherited from SUBJ in question. The PN is given to the most prominent verb form of this ordered list:

NEG > AUX > MAINV

For imperative we need to give contextual clues to the use, to trigger an imperative form:

  <q>
    <qtype>V-IMPRT</qtype>
   <question>
      <text>SUBJ MAINV MAINV2</text>
      <element type="SUBJ">
	<grammar tag="Pron+Pers+Sg1"/>
      </element>
      <element type="MAINV">
	<lemma>áigut</lemma> 
	<grammar tag="V+Cond+Prs+Sg1"/>
      </element>
       <element type="MAINV2">
	<sem class="ACTIVITY"/>
	<grammar tag="V+Inf"/>
      </element>
   </question>
    <answer> 
      <text>MAINV2</text>
      <element type="MAINV2">
	<sem class="ACTIVITY"/>
	<grammar tag="V+Imprt+Prs+Sg2"/>
      </element>
    </answer>
  </q>

We need to distinguish between inclusive and exclusive we for the imperative:

Here are the sets:

Sihke don ja mun MAINV MAINV2 - includes the user
Sihke don ja mii MAINV MAINV2 - includes the user
Sihke son ja mun MAINV MAINV2 - excludes the user
Sihke son ja mii MAINV MAINV2 - excludes the user
Sihke don ja son MAINV MAINV2 - excludes the machine
Sihke don ja sii MAINV MAINV2 - excludes the machine

For imperative, we need to force the student to use 1Du and 1Pl imperatives:

M: We should eat (Better: you and I, etc.)
Ua: Eat-Imp-1Du
Ub: eat-Imp-2Du

Since we is ambiguous in non-Polynesian languages, we should consider acepting both first and second person answers, at least in Sahka, perhaps also in Vasta. In contextual Morfa we can always specify.

Saara: Let us not put this in the xml file. Perhaps we will have to take this into account in Sahka.

We will need marking of which elements to leave blank, Saara will look at it.

Noun questions

Essive

Hello

In the question of whether the SUBJ should be marked as optional or obligatory in the answers, we have encountered a problem. Grammatically, the third person subject is always obligatory, whereas the first and second person are optional. We cannot differentiate between first and second person optional subjects on the one hand, and third person obligatory subjects on the other hand, because number and person are merged in the pronouns and we cannot access just 1st and 2nd person. This is not a problem for Morfa, since the first answer option can show the subject as obligatory regardless of person.

If we have the next answer with an optional subject, this will apply to all persons, not just first and second person. For the QA program, called Vasta from now on, this is a problem because users can generate subject-less sentences that are ungrammatical, without being told so.

We want to be able to allow optional subjects for 1st and 2nd person, but not for 3rd. How could this be done most efficiently?

Q: Maid SUBJ MAINV [oaidnit]
A: (CLADV) (SUBJ) MAINV N-ACC (CLADV)

Q: Gosa SUBJ MAINV [mannat]
A: (CLADV) (SUBJ) MAINV N-ILL

Could the pattern be the start of the question?

ANSWERSTART: (CLADV) (SUBJ) vs. (CLADV) SUBJ

Or should it concern just the subject?

ANSWERSUBJECT: SUBJ vs. (SUBJ)

In the second case, we could have two choices for the subject in questions_defaults.xml (I have added dual, although I don't know if it patterns this way):

  <part name="ANSWERSUBJECT">
     <answer>
     <text>SUBJ</text>
     <element type="SUBJ">
        <grammar tag="N+Sg+Nom"/>
        <grammar tag="N+Pl+Nom"/>
        <grammar tag="Pron+Pers+Sg3+Nom"/>
        <grammar tag="Pron+Pers+Pl3+Nom"/>
        <grammar tag="Pron+Pers+Du3+Nom"/>
      </element>
    </answer>
    <answer>
     <text>(SUBJ)</text>
     <element type="SUBJ">
        <grammar tag="Pron+Pers+Sg1+Nom"/>
        <grammar tag="Pron+Pers+Sg2+Nom"/>
        <grammar tag="Pron+Pers+Pl1+Nom"/>
        <grammar tag="Pron+Pers+Pl2+Nom"/>
        <grammar tag="Pron+Pers+Du1+Nom"/>
        <grammar tag="Pron+Pers+Du2+Nom"/>
      </element>
    </answer>
  </part>

Now we could replace the start in the answers in questions_nouns.xml like this:

(CLADV) (SUBJ) MAINV N-ACC (CLADV) -->
(CLADV) ANSWERSUBJECT MAINV N-ACC (CLADV)

  <q>
    <qtype>N-ESS</qtype>
    <agreement>SUBJ/N-ESS</agreement> 
    <question>
      <text>Manin SUBJ MAINV PLACE</text>
      <element type="SUBJ">
	<grammar tag="N+Sg+Nom"/>       <!--example Number inherited from here-->
	<sem class="FAMILY"/> 
      </element>
      <element type="PLACE">
	<grammar tag="N+Sg+Loc"/>
	<sem class="PLACE-NAME"/>
      </element>
       <element type="MAINV">
	<grammar tag="V+Ind+Tense+Person-Number"/>
	<id>bargat</id>
      </element>
   </question>
    <answer> 
      <text>(CLADV) SUBJ MAINV (PLACE) (CLADV) N-ESS (PLACE) (CLADV)</text>
      <element type="N-ESS">
	<grammar tag="N+Ess"/>
	<sem class="PROFESSION"/>  <!--The example word should agree with the SUBJ in number-->
      </element>
    </answer>
     <answer> 
      <text>(CLADV) SUBJ MAINV (PLACE) (CLADV) N-ESS (PLACE) (CLADV)</text>
    <element type="SUBJ">
	<grammar tag="Pron+Pers+Person-Number+Nom"/>
      </element>
    <element type="N-ESS">
     <grammar tag="N+Ess"/>
      <sem class="PROFESSION"/>
    </element>
     </answer>
  </q>

<q>
    <qtype>N-COM</qtype>
    <agreement>INTERR/N-COM</agreement>
    <question>
      <text>INTERR SUBJ MAINV</text>
      <element type="INTERR">
        <grammar tag="Pron+Interr+Number+Com"/>
        <lemma>mii</lemma>
      </element>
      <element type="SUBJ">
        <grammar pos="Pron"/>
      </element>
       <element type="MAINV">
        <grammar tag="V+Ind+Tense+Person-Number"/>
        <lemma>čierastit</lemma>
      </element>
   </question>
    <answer>
      <text>(CLADV) (SUBJ) MAINV (CLADV) N-COM (CLADV)</text>
      <element type="N-COM">
        <sem class="SLIDE-TOOL"/>
      </element>
    </answer>
  </q>

So, the conclusion is to specify the agreement in a separate xml entity, above as following the <qtype>.

Place names excluded from comitative and plural forms and essive

The placenames should not be put in comitative, essive or in plural, in the exercises.

TODO: Put it in a Bugzilla error, to remind Saara about it.

Vasta

Saara has questions, but will have to code more before they can be formulated. We take them in the next meeting.

Game names

The numeral games should be called Numra . We have published Logut , so we would like to keep a working link telling it is Numra now.

We will also need one more name, for the clock (but not today).

Work until next meeting

Saara: Coding Vasta, jotting dowm open issues.
Lene/BA:
- Put ANSWERSUBJECT and agreementinformation into the question-files
- Lemma-id into the nouns.xml
Trond: Documentation

Next meeting

Tuesday afternoon, 13.00 Norw time.