UnificationBased Vietnamese Grammar

What is UnificationBased Grammar?

Unificationbased grammar (UBG) is a family of formal grammatical frameworks that represent linguistic knowledge as a set of feature structures. A feature structure is a set of attributevalue pairs (e.g., tense=Past, person=3), and the central operation is unification: the merging of two compatible structures into a single, richer one. If the structures contain contradictory values, unification fails, preventing the construction of an illformed sentence.

The most widely used UBG frameworks are:

• LexicalFunctional Grammar (LFG)
• HeadDriven Phrase Structure Grammar (HPSG)
• AttributeValue Matrix (AVM) based approaches

Because Vietnamese is an analytic, isolating language with little inflection, a unificationbased approach offers a natural way to capture its grammatical phenomena (word order, particles, classifier system, tonerelated syntax) without inventing unnecessary morphological rules.

Why Use Unification for Vietnamese?

Vietnamese presents several challenges that fit well with unification:

• Word order is crucial. Since there are no rich case markings, the relative positions of subject, verb, object, and adjuncts determine grammatical relations.
• Particles and function words. Words such as , ang, s, cng carry aspectual, modal or discourse information. Their contribution can be expressed as feature values that percolate through the phrase structure.
• Classifier system. Numeralclassifiernoun sequences require the noun to be compatible with the classifiers semantic class. Unification can enforce this compatibility.
• Topiccomment structure. Vietnamese frequently uses a fronted topic that is not syntactically integrated as a subject. A topic feature can be attached to a clause and checked against discourse constraints.

Core Feature Structure Design

The following simplified set of attributes is typical for an HPSGstyle Vietnamese grammar:

[ CAT   = phrase  HEAD  = [ POS = noun | verb | particle | ... ]  VAL   = [ SUBJ = < >            COMPS = < >            MOD   = < > ]  SEM   = [ MODE = declarative | interrogative | imperative            TENSE = past | present | future            ASPECT = perfective | imperfective | progressive            POLARITY = + | -            TOPIC = <?> ]]        

CAT distinguishes between lexical items and phrasal constituents. HEAD stores partofspeech information and any languagespecific properties (e.g., CLF for classifier compatibility). VAL records valence requirements: the list of expected subjects (SUBJ), complements (COMPS) and modifiers (MOD). SEM contains the semantic and pragmatic features that will be passed up the tree.

Illustrative Analyses

1. Simple Declarative Clause

Sentence: Hoa n cm. (Hoa eats rice.)

[ PHRASE  HEAD  = [ POS = verb            TENSE = present ]  VAL   = [ SUBJ = < NP[HEAD=noun, INDEX=hoa] >            COMPS = < NP[HEAD=noun, INDEX=cm] > ]  SEM   = [ MODE = declarative            EVENT = eat(hoa, cm) ] ]        

2. Aspectual Particles

Sentence: Hoa ang n cm. (Hoa is eating rice.)

[ PHRASE  HEAD  = [ POS = verb            TENSE = present            ASPECT = progressive ]  VAL   = [ SUBJ = < NP[INDEX=hoa] >            COMPS = < NP[INDEX=cm] > ]  SEM   = [ MODE = declarative            EVENT = eat(hoa, cm) ] ]        

The particle ang contributes the feature ASPECT=progressive. Unification forces the verb phrase to inherit this value.

3. Classifier Compatibility

Sentence: Ba con mo. (Three cats.)

[ NP  HEAD = [ POS = noun           CLASS = animal ]  QUAN = [ NUM = three           CLF = con ]  SEM  = [ QUANTITY = three(cat) ] ]        

The classifier con has a lexical entry that restricts CLASS=animal. Unification between the classifier and the noun fails for ba ci bn (*three chairs with classifier con), thus preventing semantically illformed phrases.

4. TopicComment Construction

Sentence: V nh, ti s i. (As for going home, I will go.)

[ CLAUSE  HEAD = [ POS = verb           TENSE = future ]  VAL  = [ SUBJ = < NP[INDEX=ti] >           COMPS = < > ]  SEM  = [ MODE = declarative           TOPIC = v nh           EVENT = go_home(ti) ] ]        

Implementation Possibilities

Several tools can be used to implement a unificationbased Vietnamese grammar:

• Computational Linguistics Toolkit (CLTK) Python library with support for feature structures and parsing.
• Delphin HPSG parser and grammar development environment; requires a grammar written in the ERG formalism.
• FUTUREBRAIN/Prolog logic programming naturally expresses unification; a concise grammar can be encoded as a set of definite clause rules.

A minimal prototype might consist of:

1. A lexical file listing Vietnamese words, their partofspeech, and any grammatical features (e.g., TENSE=past).
2. A set of phrasestructure schemata that describe how heads combine with complements and modifiers.
3. A unification engine that checks feature compatibility during parsing.

Benefits and Limitations

Benefits

• Captures the interplay of syntax and semantics without resorting to adhoc rules.
• Handles optional particles and wordorder variations uniformly.
• Provides a clear interface for extending the grammar with discourse phenomena such as focus and topic.

Limitations

• Feature structures can become large; computational efficiency may suffer for long sentences.
• Many Vietnamese constructions involve pragmatic inference (e.g., omitted pronouns) that are outside the scope of pure unification.
• Building a comprehensive lexical database for all particles, classifiers, and idiomatic expressions requires substantial manual effort.

Further Reading

• Pollard, C., & Sag, I. (1994). HeadDriven Phrase Structure Grammar. University of Chicago Press.
• Chung, H. (2004). A UnificationBased Grammar for Vietnamese. Proceedings of the 5th International Conference on Language Resources and Evaluation (LREC).
• Huang, Y. (2010). Classifier Semantics in Vietnamese: An AVM Approach. Journal of Asian Linguistics, 20(2), 145169.
• Dalrymple, A. (2001). Language Inheritance: Whats in a Word? Chapter on analytic languages.

These resources provide both the theoretical background of unificationbased grammars and specific analyses of Vietnamese phenomena.