• Journal of Intelligence and Information Systems
• /
• v.28 no.2
• /
• pp.127-146
• /
• 2022

Recently, as word embedding has shown excellent performance in various tasks of deep learning-based natural language processing, researches on the advancement and application of word, sentence, and document embedding are being actively conducted. Among them, cross-language transfer, which enables semantic exchange between different languages, is growing simultaneously with the development of embedding models. Academia's interests in vector alignment are growing with the expectation that it can be applied to various embedding-based analysis. In particular, vector alignment is expected to be applied to mapping between specialized domains and generalized domains. In other words, it is expected that it will be possible to map the vocabulary of specialized fields such as R&D, medicine, and law into the space of the pre-trained language model learned with huge volume of general-purpose documents, or provide a clue for mapping vocabulary between mutually different specialized fields. However, since linear-based vector alignment which has been mainly studied in academia basically assumes statistical linearity, it tends to simplify the vector space. This essentially assumes that different types of vector spaces are geometrically similar, which yields a limitation that it causes inevitable distortion in the alignment process. To overcome this limitation, we propose a deep learning-based vector alignment methodology that effectively learns the nonlinearity of data. The proposed methodology consists of sequential learning of a skip-connected autoencoder and a regression model to align the specialized word embedding expressed in each space to the general embedding space. Finally, through the inference of the two trained models, the specialized vocabulary can be aligned in the general space. To verify the performance of the proposed methodology, an experiment was performed on a total of 77,578 documents in the field of 'health care' among national R&D tasks performed from 2011 to 2020. As a result, it was confirmed that the proposed methodology showed superior performance in terms of cosine similarity compared to the existing linear vector alignment.

• Language and Information
• /
• v.12 no.1
• /
• pp.27-65
• /
• 2008

This paper presents a set of methodologies for aligning hypernym-hyponym noun pairs between Korean and English, based on the EuroWordNet approach. Following the methods conducted in EuroWordNet, our approach makes extensive use of WordNet in four steps of the building process: 1) Monolingual dictionaries have been used to extract proper hypernym-hyponym noun pairs, 2) bilingual dictionary has converted the extracted pairs, 3) Word Net has been used as a backbone of alignment criteria, and 4) WordNet has been used to select the most similar pair among the candidates. The importance of this study lies not only on enriching semantic links between two languages, but also on integrating lexical resources based on a language specific and dependent structure. Our approaches are aimed at building an accurate and detailed lexical resource with proper measures rather than at fast development of generic one using NLP technique.

• Journal of KIISE:Software and Applications
• /
• v.41 no.7
• /
• pp.492-500
• /
• 2014

This paper presents the performance enhancement of automatic bilingual lexicon extraction by using refinement of pivot-context vectors under the standard pivot-based approach, which is very effective method for less-resource language pairs. In this paper, we gradually improve the performance through two different refinements of pivot-context vectors: One is to filter out unhelpful elements of the pivot-context vectors and to revise the values of the vectors through bidirectional translation probabilities estimated by Anymalign and another one is to remove non-noun elements from the original vectors. In this paper, experiments have been conducted on two different language pairs that are bi-directional Korean-Spanish and Korean-French, respectively. The experimental results have demonstrated that our method for high-frequency words shows at least 48.5% at the top 1 and up to 88.5% at the top 20 and for the low-frequency words at least 43.3% at the top 1 and up to 48.9% at the top 20.

• The Journal of the Korea Contents Association
• /
• v.15 no.7
• /
• pp.674-683
• /
• 2015

One of the issues with the textbooks pertinent to the current study is whether or not the Universal Design for Learning (UDL) factors have been dealt to satisfy students with different aptitudes in learning the core objectives of the lessons. This study develops a modified version of the UDL analysis criteria from the cross curricular criteria to language teaching and learning and uses it to analyze the sequence of digital English textbooks to investigate the descriptive statistics of the UDL factors in the new textbooks. The result shows that the textbook is designed most favorably to the students with the talent of linguistic aptitude and less favorably to the students with other types of aptitudes. The sequence analysis shows that sentence/word length and appearance of new words are incrementally sequenced as students advance upper grades. However, the syntactic complexity of middle school curves up steeply which is different from the elementary school textbooks. The UDL analysis will provide learning factors to consider when designing digital English textbooks to cover different aptitudinal groups.