KJC Korea
Journal Central

This study focused on the need to consider the feature of documents for automatic language processing such as QA, and focuses on the need for detailed guidelines for semantic roles annotation. The purpose of this paper is to improve the performance of the QA system of descriptive legal documents. In the legal document, the analysis was focused on <conditions> among high-frequency semantic roles of adjuncts, and the necessity of a new semantic roles of adjuncts was discussed. Chapter 2, explained the importance of automatic recognition of semantic roles of adjuncts in the Q & A system. Chapter 3 introduced the semantic roles of Prop Bank and ETRI, and explained the necessity of classifyingsemantic roles of adjuncts for perception of boundary clauses. In order to annotate semantic roles of adjuncts for boundary clauses, it is necessary to examine high-frequency connective endings and analyze vocabularies which is distributed before and after the connective endings. In addition, it was analyzed whether a specific semantic roles of adjuncts can be recognized according to the distribution of the connective endings. In Chapter 4, a method of sequentially applying semantic roles of adjuncts when automatically finding answers to questions about the law was proposed. In a question about the law, <time> or <extent> can be interpreted as <condition> in an answer.

This study aims to examine the grammatical basis of ‘NP+Adv’ constructions and to type them according to the syntactic relation between their internal constituents(that is a noun phrase and an adverb) and a predicate. ‘NP+Adv’ constructions are based on the semantic relation between a noun phrase and an adverb. An adverb in ‘NP+Adv’ construction is incomplete in meaning, and a noun phrase supplements the incomplete meaning of an adverb. ‘NP+Adv’ constructions are syntactically classified into two types, that is an adverb expansion AdvP and an adverbial consolidation AdvP. The adverb expansion AdvP is a construction that an adverb is syntactically related to a predicate and it is expanded by a noun phrase. The adverbial consolidation AdvP is a construction that two adverbial constituents(that is an adverbial noun phrase and an adverb) are consolidated into one. But not all adverbial constituents can be consolidated into one. They have to be semantically related to each other.

This study investigates whether three syllabic features, which are the classes of words, the types of words, and the positions in words from which orthographic syllables are drawn, have an effect on the probability for a pair of the sublexical items to be a legitimate Korean word when they are combined at random. As a result of three computational experiments, three main findings were obtained as follows: (1) the combination of a pair of syllables generated more real words when their positions in words were preserved than not; (2) the combination of a pair of syllables from a homogeneous word type produced more real words than from heterogeneous word types and the combination of a pair of sino-Korean syllables showed the higher productivity than of native Korean ones as well; and (3) the combination of a pair of syllables from a homogeneous word class created more real words than from heterogeneous word classes and the combination of a pair of noun syllables was more productive than of verb ones as well. These findings suggest that the word classes, the word types, and the positions in words have a significant impact on the probability of inter-syllabic combinations in Korean word formation, and that the syllables as sublexical units play an important role in organization of lexical representations in the mental lexicon of native speakers of Korean.

The purpose of this study is to find the best methods to interpret the meaning of the Korean collocation ‘mas(-i) issta’. Blunly, it attempts to clarify why the combination of the words ‘taste(mas)’ and ‘be, exist(issta)’, which have the meaning of ‘have taste’, is interpreted to have the meaning of ‘have good taste’. We first examined three kinds of interpretation methods.: (1) The predicate ‘issta(be, exist)’ is interpreted to have the meaning of ‘good’. However, this is difficult to select as it can also be interpreted as having meanings such as ‘bad’, ‘long’, ‘a lot (of money)’ (2) The default value of the noun ‘mas(taste)’ is assumed to be ‘good taste’. However, this is rejected as ‘taste’ cannot always be interpreted as ‘good taste’. (3) There is a method that abandons uni-direction and views ‘taste’ and ‘be, exist’ to be subjects of co-selection. However, co-selection is a hypothesis that cannot be selected as it falls into a circularity of explanations. An alternative method is to hypothesize that ‘have taste’ forms a single semantic unit. This views that monosemy of meaning is established through configuration stages of phrases and not words. Therefore, units of meaning are formed in phrasal units.

I explained some issues about Korean adnominals(listed-words of ‘wulimalsaym’) on the perspective of word-formation. By cross-classifying Korean adnominals on two criteria(e.g. formal criterion, semantic/functional criterion), I could get conclusions as follows. First, based on the semantic/functional criterion, Korean adnominals can be classified into grammatical adnominals(e.g. deictic adnominals, interrogative/infinitive adnominals, modal adnominals) and lexical adnominals(e.g. state/property adnominals, numeral adnominals, modal adnominals). Second, based on the formal criterion, Korean adnominals can be classified into simple adnominals(e.g. root adnominals, base-modificated adnominals, partial-reduplicated adnominals) and complex adnominals(e.g. compound adnominals, derivative adnominals, full-reduplicated adnominals, ending-combined adnominals, cosa-combined adnominals). Third, most of grammatical adnominals are simple deictic adnominals, and most of lexical adnominals are simple state/property adnominals. Fourth, some adnominals can be classified into two or more categories(e.g. deictic/modal adnominals, deictic/state/property adnominals, infinite/modal adnominals, state/property/numeral adnominals).

I argue that the inflectional element ‘-teo-’ in Korean is an element essentially representing a modal meaning, rather than a marker for tense or a marker solely for evidentiality, and show that its semantic description contains an intensional function in terms of Kratzer’s(1981,1991) formal semantic theory of modality, and a definedness condition on the existence of the direct evidence about the preceding proposition. After discussing previous three different approaches, i.e., tense based, solely modal based, and solely evidential based ones, I show that my definition combining the modality and the presupposition of direct evidence enables capturing the various semantic constraints of ‘-teo-’ sentences. In particular, I show that there is a general semantic requirement of an epistemic modal element on the judge dependent elements and this requirement triggers to resolve the so-called 1st person restriction of ‘-teo-’ sentences.

The purpose of this study is to define the types classified as speech error as morphological neology and to examine their features and factors of use. The morphological neology is formed by the speaker’s intention. The types of morphological neology are categorized according to the construction and movement units. First, there are modified NPs and idiomatic constructions in the construction of morphological neology. Also, movement occurs due to morphological and phonetic similarities. Gradually expanding the unit, syllable unit movement and word unit movement appear. Next, the factors of use of morphological neology were divided into two reasons. The first reason is that the speaker seeks the novelty of expression. The second reason is to maintain the understanding of meaning to communicate well. We confirm that speech errors can now be included in neologism. Because it was recognized and used as a morphological neology by many speakers.

This paper shows that there are three different uses of the Korean ending “-tani”; as an exclamative ending, as an echo question ending, and as a rhetorical question ending. These can be distinguished from one another in three ways. First, they are grammaticalized from different reported speech constructions. Second, different kinds of endings and particles can be used with each usage. Finally, unlike the exclamative or rhetorical question usage, as an echo question ending, it does not guarantee the factuality of the proposition—alternatively, the speaker of –tani echo questions requires more information about the prior utterance before they can judge whether the proposition is true or not. Furthermore, while -tani has been suggested as a mirative marker, it expresses different kinds of mirative meanings according to its type. The exclamative and rhetorical question uses express a range of mirative meanings, such as counterexpectation or new information. The echo question ending use, however, only expresses the speaker’s unprepared mind towards the prior utterance.

This study builds up a methodological pipeline to compare the lexical difference between the South Korean data and the North Korean data using the recent techniques of natural language processing. Assuming that the Chosun-ilbo and Rodong-shinmun are the counterpart of each other, we created the word embedding models to compare the distributional properties. We collected the data published in the newspapers from 2015 to 2017, preprocessed the texts, and then ran the Word2Vec library with the manipulated data. Classifying the patterns of lexical difference into four subtypes, we delved into how the two Korean languages have been lexically divergent.

This study was conducted with the aim of revealing the semantic function of the dependent noun ‘Deung’ used in newspaper articles. We have set up two hypothesis on why the ambiguous dependent noun ‘Deung’ is frequently used in newspaper articles that call for facuality and objectivity. First, on a semantic level, there will be a context in which the ambiguity of the dependent noun “Deung” is rather used as a useful language device for the writer of the newspaper article. Second, at the formal level, the grammatical characteristics of the quasi-grammatical element, the dependent noun ‘Deung’, will contribute to structurally describing newspaper text. As a result of the study, we were able to derive four semantic functions of the dependent noun ‘Deung’ used in newspaper articles as follows: First, a function that increases the acceptability as a fact by increasing the ambiguity of information in relation to the nature of newspaper articles that must defend ‘factuality’. Second, the function of indirectly conveying the intent(subject) by allowing users to choose the amount and type of information to be labeled in relation to the characteristics of newspaper articles that must protect fairness and objectivity. Third, when writing newspaper articles that require a lot of information in a limited space, it naturally connects these information. Fourth, in relation to the characteristics of newspaper articles in which substitute expressions are often used, the it functions to increase the cohesiveness of text as an substitute expression.