Model of Vietnamese linking grammar - 8

After studying and drawing out the features of this language model, the thesis will focus on solving the following problems:

• Parsing problem . This is a must-do problem when building a new syntactic representation model.
• Machine translation problem . The associative grammar model represents many distinctive features of Vietnamese that need to be transformed into another language. Therefore, the thesis chooses the Vietnamese-English translation problem to take advantage of the ability to represent the direct relationship between words of the associative grammar model.

If the elemental unit when parsing some languages ​​is a morpheme, that unit in Vietnamese is a word. According to documents of the Social Science Committee [28], each word in Vietnamese can consist of many morphemes. The word limit detection in the text is done by the automatic word separator.

Vietnamese has different characteristics from other languages, such as in semantics, there is no meaning in the morphological category (like, number, way); In sentence-making activities, grammatical relationships are not expressed in transformations but in word order [16]. The connections of the association grammar can perfectly represent these relationships.

Links appear when words are combined. According to Nguyen Tai Can [2], there are three main types of combinations: conjugation, clause, and short phrase. Conjugation and clause will be considered when performing complex parsing steps and will be covered in the next chapter. Phrases are combinations consisting of a center connected to the sub-elements by the main-sub-relation [2]. Depending on the type of center, short phrases are divided into nouns, verbs or adjectives. The association relationships will be built based on the structure of the phrases. In addition, some relationships are not expressed through word relations, for example “ my mother ”, “ Ao anh”., are two noun phrases that go side by side, the second noun will indicate the owner of the first noun. This is one of many special phenomena of Vietnamese syntax. Showing these relationships will effectively support the machine translation system with the source language being Vietnamese.

2.1.1. Link dictionary structure

The English grammar dictionary system was built by Sleator and Temperley, according to [111]. In 2003, Szolovits added a series of medical words [113]. From 2008 to 2011, the dictionary was updated by Linas Vepstas, adding clause relations, Mike Ross also added some new entries mainly related to subordinate clauses with the words “than” and words . link form “ wh ” [137].

The system is divided into 12 large sections with 7 categories for English words: nouns, determiners, pronouns, verbs, adjectives, adverbs and prepositions. Also included are the following items:

xor  selects only one of the two component links. This operator added by the thesis to the Vietnamese parser to handle the case allows to choose only one of two ways of linking, for example linking with the word "beautiful" can be "very beautiful" or " beautiful " . wonderful ” but cannot be “ very beautiful ”.

The Vietnamese linked dictionary also has the same structure as the English linked dictionary, meaning that each formula is set up for words of the same type. According to [16], Vietnamese words are divided into categories as shown in Table 2.1. down here:

Words are further divided into subcategories. In Table 2.2 below are the subcategories based on the hierarchy of [16] with the addition of the number of subcategories to meet the requirements for distinguishing links when translating according to the thesis's machine translation system.

Table 2.2 . Vietnamese word subcategories