Earticle

The deficiency of competent native English speaker raters and the inherent problem with intra-rater and inter-rater reliability of the oral proficiency interview (OPI) has precluded the full-fledged implementation of English performance testing, inevitably ushering in the computer- based oral proficiency interview (COPI) as its viable alternative with the help of automatic speech recognition (ASR). The plausibility and feasibility of implementing ASR-based COPI has recently been investigated with favorable results, which warrants more sophisticated research focusing on development of desirable test methods that will meet the rigorous criteria required by high-stakes language tests. In this respect, employing varied statistical methods as correlational, regression analyses, and ANOVA, the present study attempts to explore strengths and limitations of test method facets and to identify valid test methods to maximize the validity and reliability of ASR-based COPⅠ. Within the theoretical framework of communicative language components to be measured, the statistical findings reveal that some test methods prove to be more effective than others in producing COPI test results with better discriminability and reliability. The survey of students and teachers also suggest their favorable attitudes toward utilizing the COPI for in-class evaluation. Both findings strongly corroborates potential of the COPI in question as a valid performance testing tool to measure overall communicative competence. The current research is expected not only to shed light on advancement of performance testing, but also to serve the purpose of enhancing communicative English teaching.

Serious inherent problems with practicality, intra-rater and inter-rater reliability overshadow the known positive washback effects of performance assessment in language education. In particular, it has been welldocumented that inter-rater reliability poses a serious threat to overall test validity, since individual raters necessarily measure performance according to their own subjective severity criteria in language proficiency. However, language testing has witnessed a remarkable series of breakthroughs in performance assessment during the recent advent of the information era. One such breakthrough utilizes state-of-the-art automatic speech recognition (ASR) technology for oral proficiency interviews(OPI). Granting that current forms of ASR technologies may not produce results with the reliability needed to accommodate highstakes standardized test administration, they do offer aid in approaching the thorny issues of practicality and inherent human inter-rater subjectivity. Accordingly, this paper is intended to investigate the degree to which ASR-based OPI ratings match similar human-conducted OPI ratings by employing correlational analyses on the basis of degrees of rater severity. Furthermore, this paper attempts to explore a method of enhancing the robustness of ASR-based OPI ratings which capitalizes on suprasegmental information by measuring fluency based principally on the test-takers’ response time length.

The purpose of this study is to investigate the alkali-silica reaction through the addition of reactive aggregate in the alkali-activated geopolymer mortar mixed with copper slag, as well as the mechanical properties and ASR reactions. As a binder for geopolymer mortar, a unary and binary combination of fly ash, ground granulated blast furnace slag, and copper slag was carried out. Table flow, bending strength, and compressive strength of geopolymer mortar were tested according to the mixing characteristics. In addition ASR expansion experiments and AEW measurements were performed. As a result of the experiment, the table flow increased as the mixing rate of copper slag increased. The binary mortar of fly ash and copper slag was generally lower in strength than ordinary portland cement, but all specimens of ground granulated blast furnace slag and copper slag were higher in strength than ordinary portland cement. As a result of the ASR expansion experiment, all binary mortar specimens showed an expansion rate of less than 0.1% as specified in ASTM C 1260. As a result of AEW measurement, copper slag was the highest when using a unary binder, and the binary mortar mixed with ground granulated blast furnace slag was lower than the binary mortar mixed with fly ash. Based on the previous results, when manufacturing copper slag geopolymer mortar, a binary system using a mixture of copper slag and ground granulated furnace slag is considered advantageous.

The purpose of this paper is to review the evolution of computer-assisted interpreting(CAI) and to examine the state of the art in CAI-related research and development. Pointing to the lacunae in CAI research in Korea, the current paper presents the argument that calls for a new perspective supporting human-machine collaboration in interpreting. It looks into the trends and developments related to CAI, mainly driven by research efforts conducted in Europe. In particular, CAI tools are described according to their use in interpreting phases and generation. The main part of the paper is focused on presenting research aimed at developing the latest generation of CAI tools integrated with automatic speech recognition (ASR) and/or machine translation (MT) systems. Latest research on integrated CAI models is presented in four categories: proof-of-concept research with manually built systems; the use of general-purpose, commercial ASR and/or MT programs in the interpreting process; the advancement of existing CAI tools with the integration of ASR/MT engines; and finally, the development of a dedicated CAI system equipped with an ASR engine that can be adapted to the specific domains of interpreting sessions.

Onset detection is a primary task in audio processing for any higher-level audio processing such as music information retrieval or automatic speech recognition. Onset detection using data driven approach is hard due to labeled data scarcity. In this work we use some in build dataset for training our convolutional neural network (CNN) work and make some test data for Nepalese traditional music. The CNN with raw waveform of input audio signal performs well in this study for onset detection. This network performs well in diversified audio type where 50 millisecond windows are set in each audio file to identify the presence or absence of onset.

폐자동차 ASR의 소각재를 대상으로 물리화학적 물성측정 및 리싸이클링을 위한 경량재료 제조실험을 수행하였다. 대상시료는 국내 ASR 소각장에서 채취한 바닥재 2종류와 비산재 4종류이었으며, 이들의 주요 성분 및 입도분석을 실시하고 공정시험법에 의한 중금속 용출량을 조사하였다. 또한, 비산재인 boiler ash를 원료로 하여 경량물질과 무기바인더를 첨가하여 성형 및 소성하는 방법으로 경량재료를 제조하였다. 바닥재에는 Cu 함량이 3wt% 내외로 상당히 높은 것으로 나타나 Cu의 사전 분리가 필수적인 것으로 나타났다. 수용성물질을 많이 함유한 SDR(semi-dry reactor) ash와 Bag filter ash의 주성분은 각각 CaCl2·Ca(OH)2·H2O 및 CaCl2·4H2O 인 것으로 나타났다. boiler ash를 원료로 사용하여 제조한 경량재료 시편의 경우 중금속 용출이 크게 감소하였으며, 그 이유는 중금속 성분이 불용성 화합물로 안정화 또는 encapsulation 되었기 때문으로 판단되었다.

폐자동차 ASR으로부터 염소성분을 제거하기 위하여 풍력선별 및 비중선별 실험을 수행하였다. 또한, ASR중의 폐플라스틱만을 분리하여 물을 매체로 한 비중선별을 실시하였다. ASR 시료는 모두 8 mm이하로 재분쇄하여 선별실험에 사용하였으며, 플라스틱은 3가지 입도로 나누어 비중선별 실험을 하였다. 풍력선별은 공기유량 9~20 M3/hr 범위의 1단계와 공기유량 25~34 M3/hr 범위의 2단계로 나누어 실시하고, 각 산물의 비율과 재질분포를 조사하였다. 1단계 풍력선별후 underflow 산물의 비율은 62~66%인 것으로 나타났으며, 공기유량이 큰 2단계 풍력선별에서는 overflow 산물의 비율이 크게 증가하였다. 폐플라스틱만을 대상으로 한 비중선별 실험결과 부유물질이 침강물질에 비해 다소 많게 나타났으며, 염소함량에 있어서는 최대 수백배의 염소함량 차이를 보여 순수 플라스틱의 경우 매우 우수한 염소함유 재질의 분리효과를 얻을 수 있었다.

A study on the air and gravity separation has been performed for the removal of chlorine containing materials from ASR ofpreviously shredded to pass through 8 m sieve prior to separation tests and the gravity separation of waste plastics was con-ducted for three different particle sizes. The two-stage air classification was conducted with the range of air flow rate of 9~20 M3/hr at first stage and 25~34 M3/hr at second stage, respectively. The fraction of overflow product was remarkably increased inthe 2nd stage air classification because of high air flow rate while that of underflow product obtained from 1st stage air clas-sification was found to be 62~66%. From the results of gravity separation on waste plastics, it was also found that the amountof the float product was much greater than sink product. It is believed that the gravity separation may be used very efficiently for the removal of chlorine bearing materials from waste plastics.

폐자동차파쇄잔재물(Automobile Shredder Residue, ASR)은 폐차 재활용시 발생되는 최종 폐기물로써 파분쇄, 공기분급, 자력선별 및 정전선별법과 같은 자원처리 공정을 이용하여 선별할 수 있다. 본 연구에서는 유도형 정전선별기를 이용하여 ASR의 선별효율 향상 및 예측을 위한 전도체(구리) 및 비전도체(유리)의 궤적분석이 수행되었다. 전도체의 궤적분석 결과, 0.5와 0.25 mm 조립자 구리선의 모사궤적은 실제궤적과 거의 일치하였다. 반면 0.06 mm 구리선의 관찰궤적은 () 전극으로 편향되었다. 이는 입자특성 및 상대습도에 의한 전하량의 영향 때문으로 판단된다. 비전도체의 경우 절연체 유리의 관찰궤적이 전도성 입자들의 궤적과 유사한 특징을 보이면서 () 전극으로 편향되었다. 현미경, SEM & EDS 분석결과 유리표면에서 미립의 철과 전도성 유기물과 같은이물질 발견되었다. 이와 같이 이물질이 부착된 유리는 ASR 재활용을 위한 정전선별시 비철금속의 선별효율을 저하시키는 것으로판단된다. 향후 연구에서 유리에 부착된 이물질 제거를 위한 전처리 기술개발 및 개선된 궤적모사 연구가 요구된다.

Automobile shredder residue (ASR) is the final waste produced when end-of-life vehicles (ELVs) are shredded. ASR can be separated using mineral-processing operations such as comminution, air classification, magnetic separation, and/or electrostatic separation. In this work, trajectory analyses of conductors (copper) and non-conductors (glass) in the ASR have been carried out using induction electrostatic separator for predicting or improving the ASR-separation efficiency. From results of trajectory analysis for conductors, the trajectories of copper wire by observation versus simulation for coarse particles of 0.5 and 0.25 mm showed consistent congruity. The observed 0.06 mm fine-particles trajectory was deflected toward the (?) attractive electrode owing to the charge-density effects due to the particle characteristics and relative humidity. In the case of non?conductors, the actual trajectory of dielectric glass deflected toward the (?) electrode, showing characteristics similar to those of conductive particles. The analyses of stereoscopic microscope and SEM & EDS found heterologous materials (fine ferrous particles and conductive organics) on the glass surface. This demonstrates the glass decreasing separation efficiency for non?ferrous metals during electrostatic separation for the recycling of ASR. Future work will require a pretreatment process for eliminating impurities from the glass and advanced trajectory-simulation processes.

2010년 기준 전 세계 자동차 등록대수는 약 10억 대에 이르며, 약 4천만 대의 폐 자동차가 발생하였다. 이에 EU를 비롯한 선진국들은 2015년까지 폐 자동차의 재활용률을 95%까지 높일 것을 요구하고 있다. 우리나라도 ‘전기·전자제품 및 자동차의 자원순환에 관한 법률’을 제정하여 폐 자동차의 95% 재활용을 목표로 하고 있다. 이러한 요구조건을 충족시키기 위해 현재 충분히 재활용되지 못하고 파쇄잔재물의 형태로 매립되고 있는 플라스틱, 비철금속 등의 처리 문제가 중요한 과제로 부상하고 있다. 본 연구에서는 폐 자동차 처리 선진국의 재활용 현황 조사의 일환으로, EU 폐차처리규정 설정 방향과 EU의 ELVs & SLF/ASR 처리 현황에대한 조사를 수행하여, ELVs & SLF/ASR 처리의 세계적인 흐름을 확인하였다.

The statistics showed that about 1 billion automobiles were registered and about 40million ELVs occurred on the world in2010. So all advanced countries including EU had plan to increase the ELVs recycling rate up to 95% of total by 2015. TheKorean government also established a target for raising up to 95% of ELVs recycling rate according to ‘Act on the ResourceCirculation of Electrical and Electronic Equipment and Vehicles’. Before being satisfied with the requirement of recycling ofELVs however, the problem is issued on the scraps of plastic and non-ferrous metals which are now being abandoned andreclaimed with no adequate reuse. Therefore, as a part of preceding investigation on the present state of ELVs recycling in theworld, this preliminary investigation study was carried out focusing on the state of EU's disposal and management regulationsof ELVs and SLF/ASR including the world trend of disposal and management regulations of ELVs and SLF/ASR.

본 연구에서는 미국에서 개발된 성인 행동평가척도 자기보고용(ASR)을 한국어로 번안하고 한국 성인을 대상으로 했을 때 원검사의 요인구조 및 문항구성을 그대로 유지할 수 있는지를 확인하고 신뢰도 및 타당도를 평가해보았다. 2005년 대한민국 인구 및 주택 총조사 자료를 참조하여 총 1,003명(남자 507명, 여자 496명)의 18~60세 성인에 대한 ASR 자료를 수집하였다. 확인적 요인분석을 통해 미국판 원검사의 요인구조를 한국판에도 동일하게 적용할 수 있음을 확인했고, 각 하위척도들의 내적 합치도 수준(Cronbach’s α=.53~.96)이 양호했다. 내재화 및 외현화 척도에 속하는 하위척도들은 각각의 상위척도와 높은 상관을 보였고, 간이정신진단검사(SCL-90-R)와 ASR 관련 척도 간에 .50 이상의 높은 상관관계를 보이는 것으로 나타나 수렴 및 공존 타당도가 지지되었다. 또한 모든 하위척도에서 잠재적 임상집단이 비임상집단에 비해 중간 이상의 효과크기를 보이며 유의미하게 높은 점수를 보였다. 문항반응이론 분석 결과, ASR의 문제행동 문항들이 높은 문제행동을 변별하는데 기여하고 하고 있으며 모든 문항의 변별지수가 유의미한 것으로 나타나 문항의 타당도를 지지했다. 한국판 ASR의 임상 및 연구 장면에서의 활용에 대한 논의가 이루어졌다.

The purpose of this study was to develop and investigate the reliability and validity of the Korean version of the Adult Self Report (ASR). Selection of a sample for the ASR (507 men and 496 women) was based on data from the 2005 Korean Population and Housing Census. Results of the confirmatory factor analysis showed that the factor structure of the original ASR can be plausibly applied to the Korean version. The Korean ASR demonstrated good internal consistency (.53~.96). The subscales of the ASR showed high correlations with its higher-order-factors, such as internalizing and externalizing. High correlation was also observed between ASR and SCL-90-R related scales. The mean differences and Cohen's effect sizes of the subscales between latent clinical and normative sample supported the scale's discriminant validity. Item and test validity were also supported by Item Response Theory. Implications of using the ASR as a clinical and research instrument are discussed.

The current study explored the usefulness of mobile-based automatic speech recognition (ASR) pronunciation practice by investigating a) its effects on the production of four English vowels, and b) learners’ perception of ASR as a learning tool. A total of 19 Korean university students produced 28 minimal pair sentences containing the English vowel contrasts /i/-/ɪ/ and /ɛ/-/æ/ (e.g., I said beat, I said bit) at pretest and posttest, and completed six sessions of ASR practice outside of class that involved voice-typing a short text, minimal pairs in sentences, and decontextualized minimal pairs. Results of acoustic analysis of F1 and F2 formant frequencies showed a meaningful improvement in frontness for the vowel /i/, but no changes for the other vowels. Overall, the majority of the participants perceived ASR as useful for pronunciation practice, but some showed skepticism and frustration regarding the current state of the technology. Further discussed are the problems and limitations that EFL learners experienced during the ASR training.

The purpose of this talk is to focus on the implications of COVID-induced/ accelerated changes on the teaching and training of interpreters. Specifically, this talk will begin with an experiment that investigated how automatic speech recognition (ASR) technologies could help improve simultaneous interpreting (SI) quality. With the increase in the use of English as lingua franca (ELF), interpreters will face speakers with different speech features. Quality of SI may suffer because of source speeches with unfamiliar accents. Understanding accented speech could be cognitively challenging as interpreters need to spend more effort. The increase in cognitive load resulted from trying to comprehend accented speech may lead to omissions, errors and disfluencies in the target language renditions. Experiment results showed that ASR might have the potential to improve the accuracy of SI rendition as the accuracy scores of the captioned segments were higher than those of the non-captioned segments. Because of the speaker’s unfamiliar accent, the participants might have had to exert extra effort to understand what the speaker was saying, which could have led to increased errors and omissions. However, the amount of effort exerted by the interpreter to understand what the speaker said may also be reduced by the use of captions. As a result, the rendition of captioned segments tended to have higher accuracy scores than the rendition of non-captioned segments. Machine-aided interpreting could be realized on platforms that provide RSI also provide caption generated by ASR. Remote simultaneous interpreting (RSI) is likely to stay post COVID. Online conferences may democratize conference attendance of speakers and attendees who might have otherwise not been able to attend some conferences because of various constraints. With wider participation of individuals from different backgrounds, speakers’ accents may be even more diverse. T&I curriculums should introduce machine-aided interpreting subjects to leverage new technologies.

This study evaluates whether automatic speech recognition (ASR) systems can support scalable spoken learner corpus development without compromising the analytical rigor required in second language acquisition research. Using recordings from 60 Taiwanese learners of English at A2–B1 proficiency levels, this study compares three ASR platforms—Microsoft 365 Word, OpenAI Whisper, and Vocol.ai—across 1,200 elicited sentences and 60 weather-broadcast tasks. Transcriptions are assessed for lexical accuracy, sentence segmentation, and semantic coherence using multiple AI-based evaluators. Results show substantial variation across systems: Microsoft 365 Word frequently produced distorted or implausible lexical substitutions; Whisper maintained overall meaning but introduced systematic tense changes; and Vocol.ai provided the highest fidelity, preserving both lexical and structural details. No system generated error-free output, particularly with L2-specific pronunciation patterns. A threephase validation framework is proposed to ensure transcription reliability. The findings demonstrate that ASR can enhance corpus creation only when embedded within rigorous, multi-step validation procedures.

One of the most recent advances in the field of computer technology is the area of Automatic Speech Recognition(ASR). There have been many attempts to harness ASR to help in English speaking testing. In this paper I attempt to design such a program which will facilitate English speaking Testing. This program concerns elementary school level because there are more demands for it here. For this purpose I developed a prototype testing tool using Talking Max as an authoring program. First I analyzed the function of Talking Max. Then I constructed new algorithm for the discourse model of ASR. I also drew the communicative functions in the contents of testing units to design testing scripts. I submitted this prototype to a pilot test which includes computer simulations. Finally, I verified the goodness of this prototype as a testing tool by estimating testing validity, reliability and practicability. The result of the study indicated that the testing tool with powerful ASR functions enables us to measure student’s speaking ability effectively, especially for the formative testing. It is also shown that the prototype also has considerable accuracy in measuring student’s speaking ability.

Deep learning is a technique of machine learning in artificial intelligence area. Deep learning is a refined "machine learning" algorithm that surpasses a considerable lot of its forerunners in its capacity to perceive syllables and pictures. As of now Deep learning is a greatly dynamic examination territory in machine learning and example acknowledgment society. It has increased colossal triumphs in an expansive zone of utilizations, for example, speech recognition, computer vision and natural language processing and numerous industry items. Neural networks are used to implement the machine learning or to design intelligent machines. In this paper thorough survey to all machine learning paradigms and application areas of deep machine learning and different types of neural networks with applications are discussed.

외국어 학습에서 발음학습은 가장 중요한 부분 중 하나이다. 발음학습 과정은 학습자의 발음에 대해 정확한 평가와 잘못된 발음이 있을 경우 적절한 피드백을 주어 이를 개선시키는 작업을 포함한다. 숙련된 평가자의 평가는 비용에서, 비숙련 원어민들의 평가는 일관성에서 문제가 있기 때문에 이를 보완할 수 있는 자동발음평가 시스템에 대한 연구가 진행되고 있으며 자동음성인식 기술의 활용이 각광받고 있다. 본 연구에서는 자동음성인식 기술과 비원어민 교수자의 음성 모델을 기반으로 단어 수준에서 학습자의 발음 정확성과 유창성을 평가하는 시스템을 구축하였고, 이를 통해 학습자들이 자신의 발음을 정확히 평가받고 평가결과에 따라 적절한 피드백을 받을 수 있도록 하였다. 또한 시스템의 성능평가를 통해 발음 정확성과 유창성에 대한 자동평가결과가 전반적으로 학습자의 실제 영어실력을 정확히 구분한다는 것을 확인하였다.

An appropriate evaluation on learner's pronunciation has been an important part of foreign language education. The learners should be evaluated and receive proper feedback for pronunciation improvement. Due to the cost and consistency problem of human evaluation, automatic pronunciation evaluation system has been studied. The most of the current automatic evaluation systems utilizes underlying Automatic Speech Recognition (ASR) technology. We suggest in this work to evaluate learner’s pronunciation accuracy and fluency in word-level using the ASR and non-native teacher's speech model. Through the performance evaluation on our system, we confirm the overall evaluation result of pronunciation accuracy and fluency actually represents the learner's English skill level quite accurately.

The goal of automatic speech recognition (ASR) system is to accurately and efficiently convert a speech signal into a text message independent of the device, speaker or the environment. In general the speech signal is captured and pre-processed at front-end for feature extraction and evaluated at back-end using the Gaussian mixture hidden Markov model. In this statistical approach since the evaluation of Gaussian likelihoods dominate the total computational load, the appropriate selection of Gaussian mixtures is very important depending upon the amount of training data. As the small databases are available to train the Indian languages ASR system, the higher range of Gaussian mixtures (i.e. 64 and above), normally used for European languages, cannot be applied for them. This paper reviews the statistical framework and presents an iterative procedure to select an optimum number of Gaussian mixtures that exhibits maximum accuracy in the context of Hindi speech recognition system.

Arabic, the world’s second most spoken language in terms of number of speakers, has not received much attention from the traditional speech processing research community. This study is specifically concerned with the analysis of vowels in modern standard Arabic dialect. The first and second formant values in these vowels are investigated and the differences and similarities between the vowels explored using consonant-vowels-consonant (CVC) utterances. For this purpose, a Hidden Markov Model (HMM) based recognizer is built to classify the vowels and the performance of the recognizer analyzed to help understand the similarities and dissimilarities between the phonetic features of vowels. The vowels are also analyzed in both time and frequency domains, and the consistent findings of the analysis are expected to enable future Arabic speech processing tasks such as vowel and speech recognition and classification.