• ETRI Journal
• /
• v.42 no.1
• /
• pp.90-100
• /
• 2020

Named entity recognition (NER) continues to be an important task in natural language processing because it is featured as a subtask and/or subproblem in information extraction and machine translation. In Urdu language processing, it is a very difficult task. This paper proposes various deep recurrent neural network (DRNN) learning models with word embedding. Experimental results demonstrate that they improve upon current state-of-the-art NER approaches for Urdu. The DRRN models evaluated include forward and bidirectional extensions of the long short-term memory and back propagation through time approaches. The proposed models consider both language-dependent features, such as part-of-speech tags, and language-independent features, such as the "context windows" of words. The effectiveness of the DRNN models with word embedding for NER in Urdu is demonstrated using three datasets. The results reveal that the proposed approach significantly outperforms previous conditional random field and artificial neural network approaches. The best f-measure values achieved on the three benchmark datasets using the proposed deep learning approaches are 81.1%, 79.94%, and 63.21%, respectively.

• Annual Conference on Human and Language Technology
• /
• 1997.10a
• /
• pp.94-101
• /
• 1997

한국어의 언어분석을 위한 가공코퍼스의 하나인 품사부착 코퍼스는 형태소 언어분석의 기초가 되는 자료로서 각종 언어분석 모델의 학습자료와 관측자료 또는 검증자료로서 중요한 역할을 한다. 품사부착 코퍼스의 구축은 많은 노력과 시간이 요구되는 어려운 작업이다. 기존의 구축방법은 자동 태거의 결과를 일일이 사람이 확인해 가면 오류를 발견하고 수정하는 단순 작업이었다. 이러한 단순 작업은 한번 수정된 자동태거의 반복적 오류, 미등록어에 의한 오류 들을 계속적으로 수정해야하는 비효율성을 내포하고 있었다. 본 논문에서는 HMM기반의 자동 태거를 사용하여 1차적으로 한국어 문서를 자동 태깅한다. 자동 태깅 결과로부터 규칙기반의 오류 수정을 추가적으로 행한다. 이렇게 구축된 결과를 사용자에게 제시하여 최종 오류를 수정하고 이를 앞으로의 태깅작업에 반영하는 품사부착 워크벤치에 대해 기술한다.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.15 no.1
• /
• pp.1-11
• /
• 2015

A concept of Deep Level Situation Understanding is proposed to realize human-like natural communication (called casual communication) among multi-agent (e.g., humans and robots/machines), where the deep level situation understanding consists of surface level understanding (such as gesture/posture understanding, facial expression understanding, speech/voice understanding), emotion understanding, intention understanding, and atmosphere understanding by applying customized knowledge of each agent and by taking considerations of thoughtfulness. The proposal aims to reduce burden of humans in humans-robots interaction, so as to realize harmonious communication by excluding unnecessary troubles or misunderstandings among agents, and finally helps to create a peaceful, happy, and prosperous humans-robots society. A simulated experiment is carried out to validate the deep level situation understanding system on a scenario where meeting-room reservation is done between a human employee and a secretary-robot. The proposed deep level situation understanding system aims to be applied in service robot systems for smoothing the communication and avoiding misunderstanding among agents.

• Journal of the Korean Society of Laryngology, Phoniatrics and Logopedics
• /
• v.10 no.1
• /
• pp.50-57
• /
• 1999

A study to get the data of normal stop consonants production as preliminary study for cleft palate patients was performed. Normal Korean native 21 speakers were pronounced $VCV(V=/a/,C=/P/,/{P^1}/,/P^h/,/t/,/{t^1}/,/{t^h}/,/k/,/{k^1}/,/{k^h}/)$ syllables with natural speech rate and intensity. With intrapharyngeal pressure waveforms by pressure sensor in oropharyngeal cavity, amplitude and time duration of intrapharyngeal pressure were analyzed, and with acoustic waveforms and spectrograms, closure duration and VOT were analyzed. As results, the highest amplitude of intrapharyngeal pressure showed in alveolars and heavily aspirated consonants. Velars and unaspirated consonants were higher than bilabials and slightly aspirated ones each, in intrapharyngeal pressure. Bilabilas, alveolars and velars showed similar rise time of intrapharyngeal pressure build-up, but in decay time and total duration time, bilabials were slightly shorter than alveolars and velars, with no statistic significance. In the aspects of tensity, unaspirated consonants showed the longest rise time, heavily aspirated, the second, and slightly aspirated consonants, the shortest, which were also seen in decay time and total duration time. In closure duration, slightly aspirated consonants had the shortest closure duration, and the heavily aspirated ones, the second, and unaspirated consonants showed the longest. In VOT, heavily aspirated consonants showed the longest, slightly aspirated ones, the second and unaspirated ones showed the shortest.

• Proceedings of the KSPS conference
• /
• 1996.10a
• /
• pp.245-245
• /
• 1996

When language A borrows words, it borrows them according to its own phonetic rules. In other words, language B, where borrowed words coming from, has to comply with the phonetic requirements of language A. It may be added that language A only borrows the elements, the types of syllables and accentuation that already exist in its own phonetic struture and rejects all the rest that are not compatible. It operates exactly like a sieve. That is why borrowed words offer an excellent observation post to notice how react in phonetic contexts. The Japanese language has borrowed and is borrowing extensively from other languages and cultures, mainly from the English ones in the fields of sports, medicine, industry, commerce, and natural sciences. Relatively very few new words are created using the ancient Chinese or native backgrounds. This presentation will look for the rules of borrowing and try to show that this way of borrowing represents an organized system of its own. Three levels would be particularly studied : - the phonemic level - the syllable level and - the accentual level. This last point would be specially targeted with the question of syllable tension-relaxation. Such a study of languages in phonetics contacts could shed some new light on the phonetic charaCteristics of Japanese language and will confirm or weaken some conclusion already demonstated otherwise. We will be aming specially at the endings of the borrowed words where, it seems, Japanese language manifests itself very strongly.

• The Journal of the Korea Contents Association
• /
• v.10 no.4
• /
• pp.1-9
• /
• 2010

Authoring tools for sketching the motion of characters to be animated have been studied. However the natural interface for sound editing has not been sufficiently studied. In this paper, I present a novel method that sound sample is selected by speaking sound-imitation words(onomatopoeia). Experiment with the method based on statistical models, which is generally used for pattern recognition, showed up to 97% in the accuracy of recognition. In addition, to address the difficulty of data collection for newly enrolled sound samples, the GLR Test based on only one sample of each sound-imitation word showed almost the same accuracy as the previous method.

• Proceedings of the KSPS conference
• /
• 2006.11a
• /
• pp.14-18
• /
• 2006

The role of dialogue manager is to select proper actions based on observed environment and inferred user intention. This paper presents stochastic model for dialogue manager based on Markov decision process. To build a mixed initiative dialogue manager, we used accumulated user utterance, previous act of dialogue manager, and domain dependent knowledge as the input to the MDP. We also used dialogue corpus to train the automatically optimized policy of MDP with reinforcement learning algorithm. The states which have unique and intuitive actions were removed from the design of MDP by using the domain knowledge. The design of dialogue manager included the usage of natural language understanding and response generator to build short message based remote control of home networked appliances.

• Proceedings of the KSPS conference
• /
• 2006.11a
• /
• pp.109-112
• /
• 2006

This paper presents a dialogue interface using the dialogue management system as a method for controlling home appliances in Home Network Services. In order to realize this type of dialogue interface, we first investigated the user requirements for Home Network Services by analyzing the dialogues entered by users. Based on the analysis, we were able to extract 15 user intentions and 22 semantic components. In our study, example dialogues were collected from WOZ (Wizard-of-OZ) environment to implement a reasoning model for generating meaningful responses for example-based dialogue modeling technique. An overview of the Home Network Control System using proposed dialogue interface will be presented. Lastly, we will show that the Dialogue Management System trained with our collected dialogues behaves properly to achieve its task of controlling Home Network appliances by going through the steps of natural language understanding, response reasoning, response generation.

• MALSORI
• /
• no.21_24
• /
• pp.90-110
• /
• 1992

The purpose of this thesis is to investigate the intonation features of the last two syllables of rhythmic units, with the exception of the sentence final unit, in the Seoul dialect of Korean. The Model 5500 Sona-graph was used to measure the pitch and duration of the target syllables. There are two classes of materials. One class was designed to determine the intonation of rhythmic units in a natural situation and the other to investigate the intonation of rhythmic units in an artificial situation, in which speakers were asked to read the materials pausing only at the marked boundaries, with a view to identifying the intonation of Seoul dialect more clearly. The findings of this investigation are as follows: (1) Korean averages an 11% rising intonation between the two syllables at the end of a rhythmic unit. (2) The rising rate between the final two syllables' pitch values at the subject rhythmic unit is generally higher than those at other units in a sentence and it seems to be meaningful syntactically. (3) Before a boundary the rhythmic units undergo 'pre-lowering', in which the pitch gradually lowers from the first syllable to the penultimate. (4) Every syllable in each rhythmic unit tends to lengthen when speakers read the materials with a pause between units and the tendency is most salient at the final syllable before a boundary.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2015.10a
• /
• pp.175-178
• /
• 2015

This paper presents a VR-simulated sailor training platform for emergency in order to prevent a human error that causes 60~80% of domestic/ abroad marine accidents. Through virtual reality technology, the proposed platform provides an interaction method for proficiency of procedures in emergency, and a crowd control method for controlling crowd agents in a virtual ship environment. The interaction method uses speech recognition and gesture recognition to enhance the immersiveness and efficiency of the training. The crowd control method provides natural simulations of crowd agents by applying a behavior model that reflects the social behavior model of human. To examine the efficiency of the proposed platform, a prototype whose virtual training scenario describes the outbreak of fire in a ship was implemented as a standalone system.