TECH.REV: Semantic Analyser Proposal

[Nick Nicholas <nsn@MULLIAN.EE.MU.OZ.AU>]

This is the final draft of my project proposal; my thanks to all those, and
particularly Jim Carter, who mailed me with suggestions.
---

Project Proposal for 433-603: A Lojban-to-Prolog semantic analyser.

In this project, we propose developing a semantic analyser such that, given
a text in a subset of the artificial language Lojban, the analyser will
extract information from the text, store it as Prolog clauses, and be asked
simple questions on the text content (the questions and answers will both
be in Lojban, rather than explicit Prolog queries/clauses). To make the
analyser useful for non-Lojban speakers, output will also be provided in a
pidgin English, and phrase markers to the text syntactic structure may also
be displayed, time allowing.

Lojban is an artificial language intended for human use, of the type exempli-
fied by Esperanto and Interlingua. It differs from most such languages, in
that it has been explicitly based on predicate logic. Predicates serve the
role of verbs, predicates with preposed determiners serve the role of nouns,
and predications serve as sentences.

There is a number of reasons why this project is of interest. Lojban is a
simplified model of a natural language (NL), using predicate logic as its
modelling mechanism. Predicate logic also underlies the Prolog into which
Lojban text will be transformed by the analyser. Therefore the task of
transferring such information across from Lojban to Prolog will be
considerably simpler than doing so for an NL. Lojban has already been shoe-
horned into a context-free grammar using YACC (this has involved some
imaginative use of error recovery, but LALR(1) nature retained). Thus the
task of parsing Lojban text into identifiable grammatical constituencies has
already been dealt with: problems in resolving syntactic ambiguity need not
distract the analyser programmer from the more important semantic issues.

Most of the semantic issues complicating logic-based knowledge
representation of NL remain in Lojban: higher-order predicates;
metalinguistic comments and attitudinals; the ambiguous semantic
relationship between head and modifier in word compounds; the
representation of numbers, prepositional phrases, relative clauses, non-
logical connectives, negation, tense and modality; the distinction between
"the" and "a" (echoed in the language's veridical and non-veridical
determiners); the distinction between individual and collective plurals; sub-
ject-raising; and so forth.

In effect, a Lojban-to-Prolog semantic analyser would be addressing many
of the current issues in NLP knowledge representation, though biased
towards predicate logic in the way it does so. The use of a simplified model
of NL, and the way the model falls short of capturing NL nuances, will help
the analyser cover much ground quickly, and provide insights in similar
analysis of NL proper. (It is claimed that the subset of Lojban implemented
would fall short; the author believes the language itself, if it acquires a
speech community, will match NL adequately in most usages of language).
Less attention would need to be paid to syntactic issues than would be the
case with NL. Given how Lojban grammar is structured, modular subsets of
Lojban grammar can be implemented in stages in the analyser. This means
that results for simple phrases will become available a very short time into
the project.

To keep the project manageable, a subset of the language will have to be
considered; this is in line with the Lojban Canonicaliser proposed by John
Cowan (see Enclosures. The Canonicaliser will need to be implemented as a
preprocessor to what text the analyser actually sees). Lexically, the subset
of Lojban to be implemented will include roughly 500 predicates.

Grammatically, the subset is described as follows, to be implemented in
incremental, independent stages:

1. Simple predications with a known predicate, and with arguments without
internal structure (Proper names, logical variables). No quantification other
than existential. eg. mi prami da --- EXISTS X: LOVES(i, X).

2. Non-veridical arguments (cf. English "the") based on predicates, with in-
ternal arguments. eg. mi catra le prami be le pulji --- KILLS(i, x) & LOVES(x,
y) & POLICE(y): I kill the lover of the policeman. Note: strictly speaking, the
non-veridical determiner indicates that the entity the speaker has "in mind"
is described by the predicate it precedes, but not uniquely specified by it
(cf. veridical determiners). Given the absence of pragmatic content at this
early stage of the analyser, making this distinction will be problematic (it
is, after all, inherently ambiguous); it will be dealt with here exactly as
NLP deals with the "the"/"an" distinction.

3. Veridical arguments (cf. English "an") based on predicates, with internal
arguments. eg. mi catra lo prami be lo pulji --- EXISTS X EXISTS Y: KILLS(i, X)
& LOVES(X, Y) & POLICE(Y): I kill a lover of a policeman.

4. Resolution of logical connectives. eg. mi nelci do .e ko'a --> mi nelci do
.ije mi nelci ko'a --- LIKES(i, you) & LIKES(i, x1): I like you and him.

5. Anaphora and cross-indexing. eg. {le prenu}\i cu prami ri\i --- PERSON(x) &
LOVES(x, x): The person loves him/herself.

6. Restrictive and non-restrictive relative clauses. eg. mi nelci le prenu poi
do xebni ke'a --- (EXISTS x: HATES(you, x)) & LIKES(i, x) & PERSON(x): I like
the person you hate.

7. Higher order predicates. eg. lenu mi cadzu cu nandu --- DIFFICULT(event:
WALKS(i)): My walking is difficult.

8. Prepositional phrases (other than tense and location). eg. mi naumau do
nelci ko'a --> mi zmadu do leni da nelci ko'a --- EXCEEDS(i, you, quantity:
LIKES(X, x1)): I like him more than you do. eg. lo catra nesepi'o lo mrudakfu
--> lo catra poi pilno lo mrudakfu --- EXISTS X EXISTS Y: KILLS(X, _) & USES(X,
Y, event: KILLS(X, _)) & HAMMER_KNIFE(Y): an axe-murderer.

9. Attitudinals. eg. mi .ui sidju do --> mi sidju do .ije mi gleki mi va'o
lenu mi sidju do: HELP(i, you) & HAPPY(i, i) & CONTEXT((state: HAPPY(i, i),
event: HELP(i, you)): I *smile* will help you; I am happy to help you.

10. Tense (including location), and prepositions of tense (including location).
Also includes modality and event contours. eg. mi ba'o tavla --> lenu mi tavla
cu ba'o zei balvi zo'e: AFTERMATH(event: talk(i, _, _, _), _): I have spoken.

11. Masses and sets as arguments. eg. loi remna cu sipna: the mass of
humans sleep (Though it is not true at any given moment that: FORALL X:
HUMAN(X) => SLEEPS(X))

12. Non-logical connectors. eg. la gilbrt. joi la salivn. cu finti la mikadon.
--- INVENT(X, mikado) & JOINT_MASS(X, gilbert, sullivan): G & S (as a joint
unit) wrote The Mikado.

13. Quantification (including numerical, as well as subjective quantifiers
such as "enough" and "most"): eg. mu le ze mensi cu cucycau: five of the
seven sisters are barefoot.

14. Negation. Contradictory and scalar. Use of prenexes. eg. mi naku ro prenu
cu prami: NOT(FORALL X:PERSON(X), LOVES(i,X)); mi ro prenu na prami:
FORALL X:PERSON(X), NOT(LOVES(i,X))

15. Vocatives, imperatives, interrogatives, and speech protocol words: eg.
doi skami la sinderelan. mensi ma fe'o: O Computer: Cinderella is sister to
whom? (End of transmission).

Sections of Lojban Grammar not anticipated to be included in the model:

1. The mathematical subgrammar of Lojban.
2. Any analysis of word compounds.
3. Metalinguistic comments.

The detail of coverage of some sections, particularly tense, will probably
have to be curtailed due to time constraints. It is anticipated to have this
project take at most 80 hours of work.

Momenton senpretende paseman mi retenis kaj # [Victor Sadler, _Memkritiko_ 90]
   kultis kvazaux                           &  (NICK NICHOLAS. Melbourne.
      senhorlogxan elizeon                  #   Australia. IRC: nicxjo.
         (Dume:                             &   nsn@munagin.ee.mu.oz.au .)