• Journal of KIISE:Software and Applications
• /
• v.34 no.2
• /
• pp.112-122
• /
• 2007

Training a statistical model for semantic role labeling requires a large amount of manually tagged corpus. However. such corpus does not exist for Korean and constructing one from scratch is a very long and tedious job. This paper suggests a modified algorithm of self-training, an unsupervised algorithm, which trains a semantic role labeling model from any raw corpora. For initial training, a small tagged corpus is automatically constructed iron case frames in Sejong Electronic Dictionary. Using the corpus, a probabilistic model is trained incrementally, which achieves 83.00% of accuracy in 4 selected adverbial cases.

• Annual Conference on Human and Language Technology
• /
• 2001.10d
• /
• pp.341-347
• /
• 2001

세종전자사전이 궁극적으로 범용전자사전을 지향한다는 점에 비추어 볼 때, 텍스트 자동 분석과 생성, 정보 검색 및 자동 번역 등에 활용될 데이터베이스로서의 전자사전은 자연 언어 어휘의 내적 구조와 기능방식에 관한 정보들, 가령 음운 통사 의미 화용적 가치와 실현 조건 등에 관한 정보들을 체계적이고도 정교하게 담고 있어야만 한다. 의존명사, 대명사, 수사 범주에 속하는 언어 단위들은 단일 명사와 구별되는 어휘 통사적 속성들을 지니며, 사전의 기술 구조에는 그 정보 값들을 체계적으로 명시화할 수 있는 정보 항목과 표상 구조가 설정되어야 한다. 가령 의존명사처럼 통사 의미적 자율성을 지니지 않는 언어 요소의 경우, 어휘 관계 정보보다는 인접하는 여타 언어 단위들과의 호응관계나 결합제약 조건들이 더 중요한 정보일 수 있다. 본 사전이 체언사전의 하위사전으로 별도로 구축되는 것은 단일어 사전에서 그러한 정보들을 효과적으로 표상하기 어렵기 때문이다. 그러나 본 사전은 실제적으로는 체언사전에 통합되어 운영된다는 점에서 이중적 지위를 누린다고 하겠다.

• Annual Conference on Human and Language Technology
• /
• 2001.10d
• /
• pp.348-355
• /
• 2001

세종전자사전은 한국어 어휘의 정보를 총체적으로 표상하는 전산어휘자료체이다. 세종전자사전은 궁극적으로 다양한 유형과 기능의 한국어 자동처리에의 활용을 목표로 한다. 세종체언전자사전은 최종적으로 5만여 항목을 대상으로 구축될 것이다. 세종명사전자사전은 전산적 효율성을 고려하여 명사 어휘의 정보를 8개의 하위정보구획과 50여개의 세부정보항목으로 구분하여 표상한다. 특히, 명사의 어휘 통사적인 결합에 관한 정보는 한국어 명사구와 문장의 자동생성에 직접 활용될 수 있는 정보이다. 명사는 수식어 요소 또는 조사와 결합하여 명사구를 형성하고, 동사 및 형용사와 결합하여 문장의 생성에 참여한다. 개별 명사들은 이 과정에서 다양한 제약적인 양상들을 보여주고 있으며, 세종명사전자사전은 이 정보들을 명시적으로 표상한다. 또한 명사는 기능동사와 결합하는 술어명사와 그렇지 않은 비술어명사로 구분이 된다. 술어 명사가 기능동사와 결합하여 문장을 형성할 때, 명사와 그 논항들은 다양한 통사적 기능을 담당한다. 또한 술어명사는 논항과 결합하여 명사구를 형성한다. 그러나, 술어명사의 명사구 형성과 기능동사의 결합은 불규칙적이고, 명사와 기능동사의 의미적 특성에 따라 다양한 제약이 발견된다. 이 정보들의 정밀한 표상은 개별 술어명사로부터 생성될 수 있는 가능한 명사구와 문장의 형태를 구체적으로 밝혀주게 된다. 세종명사전자사전의 어휘 통사적인 결합에 관한 정보들은 명사구 또는 문장의 자동생성의 정확성과 효율성을 높여줄 것이다.

• Korean Journal of Cognitive Science
• /
• v.19 no.2
• /
• pp.143-161
• /
• 2008

The goal of this paper is to extract the hyponymy relation between words in the corpus. Adopting the basic algorithm of Hearst (1992), I propose a method of pattern-based extraction of semantic relations from the corpus. To this end, I set up a list of hypernym-hyponym pairs from Sejong Electronic Dictionary. This list is supplemented with the superordinate-subordinate terms of CoroNet. Then, I extracted all the sentences from the corpus that include hypemym-hyponym pairs of the list. From these extracted sentences, I collected all the sentences that contain meaningful constructions that occur systematically in the corpus. As a result, we could obtain 21 generalized patterns. Using the PERL program, we collected sentences of each of the 21 patterns. 57% of the sentences are turned out to have hyponymy relation. The proposed method in this paper is simpler and more advanced than that in Cederberg and Widdows (2003), in that using a word net or an electronic dictionary is generally considered to be efficient for information retrieval. The patterns extracted by this method are helpful when we look fer appropriate documents during information retrieval, and they are used to expand the concept networks like ontologies or thesauruses. However, the word order of Korean is relatively free and it is difficult to capture various expressions of a fired pattern. In the future, we should investigate more semantic relations than hyponymy, so that we can extract various patterns from the corpus.

• The KIPS Transactions:PartB
• /
• v.11B no.4
• /
• pp.491-500
• /
• 2004

This paper presents an ontology construction method using various computational language resources, and an ontology-based word sense disambiguation method. In order to acquire a reasonably practical ontology the Kadokawa thesaurus is extended by inserting additional semantic relations into its hierarchy, which are classified as case relations and other semantic relations. To apply the ontology to disambiguate word senses, we apply the previously-secured dictionary information to select the correct senses of some ambiguous words with high precision, and then use the ontology to disambiguate the remaining ambiguous words. The mutual information between concepts in the ontology was calculated before using the ontology as knowledge for disambiguating word senses. If mutual information is regarded as a weight between ontology concepts, the ontology can be treated as a graph with weighted edges, and then we locate the weighted path from one concept to the other concept. In our practical machine translation system, our word sense disambiguation method achieved a 9% improvement over methods which do not use ontology for Korean translation.

• Korean Journal of Cognitive Science
• /
• v.21 no.1
• /
• pp.95-126
• /
• 2010

This study proposes a bottom-up and inductive manual mapping methodology for integrating two heterogenous fine-grained ontologies which were built by a top-down and deductive methodology, namely the Sejong semantic classes (SJSC) and the upper nodes in KorLexNoun 1.5 (KLN), for HLP applications. It also discusses various problematics in the mapping processes of two language resources caused by their heterogeneity and proposes the solutions. The mapping methodology of heterogeneous fine-grained ontologies uses terminal nodes of SJSC and Least Upper Bounds (LUB) of KLN as basic mapping units. Mapping procedures are as follows: first, the mapping candidate groups are decided by the lexfollocorrelation between the synsets of KLN and the noun senses of Sejong Noun Dfotionaeci(SJND) which are classified according to SJSC. Secondly, the meanings of the candidate groups are precisely disambiguated by linguistic information provided by the two ontologies, i.e. the hierarchicllostructures, the definitions, and the exae les. Thirdly, the level of LUB is determined by applying the appropriate predicates and definitions of SJSC to the upper-lower and sister nodes of the candidate LUB. Fourthly, the mapping possibility ic inthe terminal node of SJSC is judged by che aring hierarchicllorelations of the two ontologies. Finally, the ituorrect synsets of KLN and terminologiollocandidate groups are excluded in the mapping. This study positively uses various language information described in each ontology for establishing the mapping criteria, and it is indeed the advantage of the fine-grained manual mapping. The result using the proposed methodology shows that 6,487 LUBs are mapped with 474 terminal and non-terminal nodes of SJSC, excluding the multiple mapped nodes, and that 88,255 nodes of KLN are mapped including all lower-level nodes of the mapped LUBs. The total mapping coverage is 97.91% of KLN synsets. This result can be applied in many elaborate syntactic and semantic analyses for Korean language processing.