• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.15 no.6
• /
• pp.2049-2068
• /
• 2021

AI combined with NLP techniques has promoted the use of Virtual Assistants and have made people rely on them for many diverse uses. Conversational Agents are the most promising technique that assists computer users through their operation. An important challenge in developing Conversational Agents globally is transferring the groundbreaking expertise obtained in English to other languages. AI is making it possible to transfer this learning. There is a dire need to develop systems that understand secular languages. One such difficult language is Hindi, which is the fourth most spoken language in the world. Semantic similarity is an important part of Natural Language Processing, which involves applications such as ontology learning and information extraction, for developing conversational agents. Most of the research is concentrated on English and other European languages. This paper presents a Corpus-based word semantic similarity measure for Hindi. An experiment involving the translation of the English benchmark dataset to Hindi is performed, investigating the incorporation of the corpus, with human and machine similarity ratings. A significant correlation to the human intuition and the algorithm ratings has been calculated for analyzing the accuracy of the proposed similarity measures. The method can be adapted in various applications of word semantic similarity or module for any other language.

• Journal of the Korea Convergence Society
• /
• v.13 no.1
• /
• pp.13-21
• /
• 2022

Recent deep learning-based research shows excellent performance in most natural language processing (NLP) fields with pre-trained language models. In particular, the auto-encoder-based language model proves its excellent performance and usefulness in various fields of Korean language understanding. However, the decoder-based Korean generative model even suffers from generating simple sentences. Also, there is few detailed research and data for the field of conversation where generative models are most commonly utilized. Therefore, this paper constructs multi-turn dialogue data for a Korean generative model. In addition, we compare and analyze the performance by improving the dialogue ability of the generative model through transfer learning. In addition, we propose a method of supplementing the insufficient dialogue generation ability of the model by extracting recommended response candidates from external knowledge information through a retrival model.

• Journal of the Korea Society of Computer and Information
• /
• v.28 no.8
• /
• pp.31-38
• /
• 2023

Style transfer is one of deep learning-based image processing techniques that has been actively researched recently. These research efforts have led to significant improvements in the quality of result images. Style transfer is a technology that takes a content image and a style image as inputs and generates a transformed result image by applying the characteristics of the style image to the content image. It is becoming increasingly important in exploiting the diversity of digital content. To improve the usability of style transfer technology, ensuring stable performance is crucial. Recently, in the field of natural language processing, the concept of Transformers has been actively utilized. Attention maps, which forms the basis of Transformers, is also being actively applied and researched in the development of style transfer techniques. In this paper, we analyze the representative techniques SANet and AdaAttN and propose a novel attention map-based structure which can generate improved style transfer results. The results demonstrate that the proposed technique effectively preserves the structure of the content image while applying the characteristics of the style image.

• Journal of IKEEE
• /
• v.3 no.1 s.4
• /
• pp.150-160
• /
• 1999

This paper presents a new register transfer level synthesis method to overcome the disadvantages of the previous register transfer level synthesis systems. The previous register transfer level synthesis systems first translate from a hardware description language to sequential circuits inadequately. Secondly, the systems separate registers and combinational circuits and then optimize only combinational circuits. This paper describes their disadvantages and then proposes a new method to overcome their shortcomings. This paper also shows the effectiveness of the proposed method by using the proposed method at designing the controller of a surveillance system and the 8-bit signed multiplier.

• Proceedings of the KSPS conference
• /
• 1996.10a
• /
• pp.430-431
• /
• 1996

Among the scholars of second language (L2) acquisition who have used prosodic considerations in syntactic analyses, pausing and intonation contours have been used to define utterances in the speech of second language learners (e.g., Sato, 1990). In recent research on conversational analysis, it has been found that lexically marked causal clause combining in the discourse of native speakers can be distinguished as "intonational subordination" and "intonational coordination(Couper-Kuhlen, Elizabeth, forthcoming.)". This study uses Pienemann's Processability Theory (1995) for an analysis of the speech of native speakers of Japanese (L1) learning English. In order to accurately assess the psycholinguistic stages of syntactic development, it is shown that pitch, loudness, and timing must all be considered together with the syntactic analysis of interlanguage speech production. Twelve Japanese subjects participated in eight fifteen minute interviews, ninety-six dyads. The speech analyzed in this report is limited to the twelve subjects interacting with two different non-native speaker interviews for a total of twenty-four dyads. Within each of the interviews, four different tasks are analyzed to determine the stage of acquisition of English for each subject. Initially the speech is segmented according to intonation contour arid pauses. It is then classified accoding to specific syntactic units and further analysed for pitch, loudness and timing. Results indicate that the speech must be first claasified prosodic ally and lexically, prior to beginning syntactic analysis. This analysis stinguishes three interlanguage lexical categories: discourse markers, coordinator $s_ordinators, and transfer from Japanese. After these lexical categories have been determined, the psycholinguistic stages of syntactic development can be more accurately assessed.d.

• Phonetics and Speech Sciences
• /
• v.11 no.2
• /
• pp.15-21
• /
• 2019

Korean and English are both known to show on-focus pitch range expansion and post-focus pitch range compression (PFC). But it is not clear if this prosodic similarity would make it easy for Korean speakers to learn English focus prosody. In the present study, we conducted a production experiment using phone number strings to examine whether Korean learners of English produce a native-like focus prosody. Korean learners of English were classified into three groups (advanced, intermediate and low) according to their English proficiency and were compared to native speakers. Results show that intermediate and low groups of speakers did not increase duration, intensity, and pitch in the focus positions, nor did they compress those cues in the post-focus positions. Advanced speakers noticeably increased the acoustic cues in the focus positions to a similar extent as native speakers. However, their performance in post-focus positions was quite far from that of native speakers in terms of pitch and excursion size. These results thus demonstrate a lack of positive transfer of focus prosody from Korean to English in L2 learning, and learners may have to relearn it from scratch, which is consistent with a previous finding. More importantly, the results provide further support for the view proposed in other works that acoustic properties of PFC were not easily transferred from one language to another.

• Phonetics and Speech Sciences
• /
• v.13 no.1
• /
• pp.25-36
• /
• 2021

This study examined the perception and production of eight English fricatives /f/, /v/, /θ/, /ð/, /s/, /z/, /ʃ/, and /ʒ/ by thirty Chinese English majors and thirty Chinese middle school students through a fricative identification test, an intelligibility test, and a goodness rating test and focused on error patterns and the perception-production relationship. The results showed that substitution errors occurred frequently in the perception and production of English fricatives by both the English majors and the middle school students. Further, the error patterns were attributed to various influencing factors such as the negative transfer from Chinese consonant inventory, hypercorrection or overcompensation mistakes, deficiency of L2 teaching, and acoustic similarities. Significant overall correlations were found between the fricative perception and production by the two subject groups but were not manifested in all the eight fricatives, indicating that Chinese learners' perceptual competence of target fricatives was not necessarily tied to their productive excellence of those sounds in all cases. Furthermore, precedences of perception over production were incompletely manifested in the eight fricatives, which suggested that perception might not always be a necessary prerequisite for production. Additionally, subject group and vowel context differences were observed. The English majors performed better than the middle school students, both perceptually and productively, and the subjects' performances in perception and production varied when vowel contexts changed.

• ETRI Journal
• /
• v.42 no.4
• /
• pp.527-533
• /
• 2020

High-level synthesis (HLS), which automatically synthesizes a register-transfer level (RTL) circuit from a behavioral description written in a high-level programming language such as C/C++, is becoming a more popular technique for improving design productivity. In general, HLS tools often generate a circuit with a larger area than those of hand-designed ones. One reason for this issue is that HLS tools often generate multiple instances of the same module from a function. To eliminate such a redundancy in circuit area in HLS, HLS tools are capable of sharing modules. Function-level module sharing at a behavioral description written in a high-level programming language may promote function reuse to increase effectiveness and reduce circuit area. In this paper, we present two HLS techniques for module sharing at the function level.

• Annual Conference on Human and Language Technology
• /
• 2003.10d
• /
• pp.293-299
• /
• 2003

인간 게놈 프로젝트가 완료됨에 다라 생체서열과 언어 사이의 대응 관계가 부각되고 있다. 본고에서는 Lewis Carroll의 언어 유희 사례를 컴퓨터생물학의 측면에서 재조명하고, Carroll이 제시한 문제 중에서 간단한 anagram 문제의 해결을 다루고자 한다. 우선 DNA 컴퓨팅의 방법론을 적용한 DNAgram의 개념을 확장하여 plasmid-DNAgram의 개념을 새롭게 도입하였다. 이 개념을 형광단백질에 대한 DNAgram의 개념을 확장하여 plasmid-DNAgram의 개념을 새롭게 도입하였다. 이 개념을 형광단백질에 대한 FRET(fluorescent resonance energy transfer)분석기법의 응용 사례인 cameleon 형광단백질에 대한 FRET 분석기법에 적용함으로써 anagram 문제의 어휘론적, 구문론적, 의미론적, 화용론적 측면에 대응하는 바이오분자 컴퓨팅 방법론을 제안하였다.

• Annual Conference on Human and Language Technology
• /
• 2018.10a
• /
• pp.567-569
• /
• 2018

한국어의 경어체는 종결어미에 따라 구분하고, 서로 다른 경어체는 각각 고유한 경어 강도가 있다. 경어체 간의 어체 변환은 규칙기반으로 진행되어 왔다. 본 논문은 어체 변환을 위한 규칙 정의의 번거로움을 줄이고 어체 변환 데이터만을 사용한 심층 학습 기반의 어체 변환 방법을 제안한다. 본 연구는 '해요체-합쇼체' 쌍의 병렬 데이터를 이용하여 Attention-based Sequence-to-Sequence 모델을 바탕으로 한 어체 변환 모델을 학습하였다. 해당 모델을 학습하고 실험하였을 때, 정확도 91%의 우수한 성과를 얻을 수 있었다.