Message-Id: <m0qHHIa-00001oC@xiron.pc.helsinki.fi>
Date:         Fri, 24 Jun 1994 13:48:39 BST
Reply-To:     C.J.Fine@BRADFORD.AC.UK
Sender:       Lojban list <LOJBAN%CUVMB.bitnet@FINHUTC.hut.fi>
From:         Colin Fine <C.J.Fine@BRADFORD.AC.UK>
Subject:      Re: sumti categories
To:           Veijo Vilva <veion@XIRON.PC.HELSINKI.FI>
Content-Length: 6404
Lines: 169

I've been away for a few days, so I haven't got back on this one,
but I think it is very important

Jorge answers me:
>
> > My suggestion is that we subcategorise on a number of features, which
> > need not always be specified.
>
> If they need not be specified, they're not useful for the purpose of
> forcing the sumti to take the feature that the tergismu requires.
> Thus:
>
>         mi fasnu
>
> forces the sumti {mi} to be +event. The sentence to me makes little sense,
> since I don't think events should be allowed to talk (to refer to
> themselves as {mi}) but
>
>         spuda mi
>
> doesn't make {mi} an event, because (according to the list) the x2 of {spuda}
> need not be one.

No, I said they need not always be specified. Sometimes they will be.
The way it works is that if the terbri is +feat for some feature, and the
sumti is -feat, then the combination fails category consistency (I agree
that this does not per se make them ungrammatical); similarly
if the terbri is -feat, and the sumti +feat. But if either happens to be
unspecified for that feature, the combination is OK as far as that feature
goes.

So to take your example,
fasnu   +abstract certainly,
mi      -abstract (plausible, but not certain)

so 'mi fasnu' fails.
But
se zanru        ~abstract (unspecified - I've just used a tilde, but I've
                        a feeling that gets sprongled by the mail server)

so both  'zanru mi' and 'zanru le fasnu' are acceptable.


> > Most terbri are -set but unspecified for mass
>
> Is any terbri unspecified for set? Probably not, in which case +/-set
> is a useful category for this purpose.
>
Yes, there are several. For example 'se badri' and 'lidne',
'se casnu', 'banro', 'se galfi'.


> +/-mass is not a category in the sense that I intend, because the
> terbri will never force its "massness" on the sumti filling it.
>
>         le plise cu gunma lei selci
>
> makes sense, even though {le plise} remains -mass, and {lei selci} +mass.

I don't agree. I would put it this way:
plise   -set +concrete (+ some others, eg +plant maybe?)
le      - (no relevant features)
gunma   -set +mass (I'm not clear whether it is +concrete or not.
                Not relevant here)
se gunma        - (no relevant features)
selci   - (no features, not even -set)
lei     +mass, but can take a -mass selgadri, in which case it massifies, but
                the mass inherits other features from the elements.

So
le plise        -set +concrete
lei selci       +mass (this is a property of the mass of selci, not the
                individual selci in the mass)

So 'le plise' is unspecified for mass, and so can match the +mass of gunma;
'lei selci' is +mass, and can match the unspecified mass of se gunma;
thus 'le plise cu gunma lei selci' is consistent; its x1 is +mass by
virtue of the 'gunma', and its x2 is +mass by virtue of the 'lei'.
I suspect there is a feature +multiple attached to se gunma, so that the
x2 is actually +multiple as well as +mass: a multiplicity of masses of
cells; I haven't worked this out in detail though.

On the other hand,
le'i plise cu gunma le'i selci
is analysed thus:

le'i    +set; if its selgadri is -set, it constructs a set of the elements,
                and the set inherits none (I think) of the elements' features.
                I rather think +set implies +abstract as well.

So
le'i plise      +set (+abstract?)
does not match
gunma   +mass -set.

Ont the other hand
le'i selci      +set (+abstract?)
does match
se gunma        (~set)

(There is a problem with my ascription of ~set to se gunma: there are two
different possible meanings to se gunma with a +set sumti: it could mean
a mass of all the sets taken jointly (as components of the mass), or a
mass of all the elements of the sets. These are different meanings, and
must be distinguished in the definition. My preference is that a 'se gunma'
cannot be a set (of the components of the mass) and so the meaning of
da gunma le'i namcu
or
da gunma le klesi
is a mass composed of individuals each of which happens to be a set).

>
> My categories, in your notation reduce to something like:
>
> dacti   -set -abstract -proposition
> fasnu   -set +abstract -proposition
> fatci   -set +abstract +proposition
> namcu   -set -abstract -proposition +something-that-dacti-isn't
> selcusku -set -abstract +proposition (?)
> selcmima +set -abstract -proposition
>
> I doubt there's a need to name all possible combinations, because
> I don't think any place is for instance +set +proposition

I agree. I think +proposition -> +abstract, and +set is incompatible with
most other features (but not all). I have been using a separate feature
+concrete, but I'm not sure whether it is needed or not.
>
> > I am certain that there are further subdivisions of +abstract (eg +/-event,
> > +/-concept) but I have not worked out in detail what they are. I doubt
> > very much though whether even they are mutually exclusive.
>
> Of course, there will always be subdivisions, but the only ones that matter
> for this purpose are the mutually exclusive ones.
>
NO. Mutually exclusive categories are a straitjacket that we do not need.

 > I think there are probably other features not in either of these categories
> > (eg +/- animate, +/-personal) but I'm not sure.
>
> +/-animate is not useful for this, because for instance, the x2 of {viska}
> can be either. I'm not sure about +/-personal. I think many gismu allow
> for the ambiguity (e.g. the x1 of gasnu)

In thinking further, I've come up with categories +personal, +volitional,
+emotional. I think I agree that +animate will not be useful.
>
> > Once all terbri, gadri, sucma'o, sumga'ima'o (eg LUhI) and a few others
> > have their features specified, it will be possible to check a sentence for
> > category consistency.
>
> Yes, but if the vocabulary requires the distinction sometimes, but not other
> times, then it will be much harder to accept that the place should force
> the category on whatever fills it.
>

I don't think of it in terms of 'forcing', but as a process of checking
consistency.

I've realised that all the BAI and tense terms will have features too. For
example I think there are features +location, +time and +punctual, that
will govern consistencies between tenses and selbri.

In the next mail, I post the beginning of my list of tergi'u with features,
for discussion and comment.

        Colin