• ETRI Journal
• /
• v.46 no.1
• /
• pp.118-126
• /
• 2024

In human conversations, listeners often utilize brief backchannels such as "uh-huh" or "yeah." Timely backchannels are crucial to understanding and increasing trust among conversational partners. In human-machine conversation systems, users can engage in natural conversations when a conversational agent generates backchannels like a human listener. We propose a method that simultaneously predicts backchannels and recognizes speech in real time. We use a streaming transformer and adopt multitask learning for concurrent backchannel prediction and speech recognition. The experimental results demonstrate the superior performance of our method compared with previous works while maintaining a similar single-task speech recognition performance. Owing to the extremely imbalanced training data distribution, the single-task backchannel prediction model fails to predict any of the backchannel categories, and the proposed multitask approach substantially enhances the backchannel prediction performance. Notably, in the streaming prediction scenario, the performance of backchannel prediction improves by up to 18.7% compared with existing methods.

• The Journal of the Korea Contents Association
• /
• v.17 no.1
• /
• pp.25-38
• /
• 2017

This study investigates non-native speakers(NNS) of English use of backchannels with beginner-intermediate learners' use of 'uh', 'um(uhm)', 'and' and 'hm' suggesting a view as a possible pedagogical implication. The initial aim of this study was to learn this phenomenon and observe their conversation patterns to compare with previous studies. Based on the previous findings, the analyzed data using conventional Conversation Analysis (CA) methods indicate the possible presence of L1 topic markers, '-un' and '-nun' in the form of L2 backchannel tokens when uttered by beginning and intermediate level speakers of English and the presences of L2 backchannel tokens appear only in front of noun phrases. Additionally, these same words with these tokens and when translated back to Korean also require topic markers of '-un' and '-nun.' Finally, This study discusses possible pedagogical implications with the initial analysis of backchannel tokens for Korean EFL learners. In addition, the ultimate goal of this study is to refine this analysis with follow up experiments to validate this investigation into a working hypothesis generating discussions of this backchannel phenomenon from being viewed as a hindrance to as an positive influence that needs to be understood.

• Journal of KIISE:Databases
• /
• v.31 no.4
• /
• pp.422-431
• /
• 2004

In recent years, different models in data delivery have been explored in mobile computing systems. Particularly, there were a lot of research efforts in the periodic push model where the server repetitively disseminates information without explicit request. However, average waiting time per data operation highly depends on the length of a broadcast cycle and different access pattern among clients may deteriorate the response time considerably. In this case, clients are preferably willing to send a data request to the server explicitly through backchannel in order to obtain optimal response time. We call the broadcast model supporting backchannel as hybrid broadcast. In this paper, we devise a new transaction processing algorithm(O-PreH) in hybrid broadcast environments. The data objects which the server maintains are divided into Push_Data for periodic broadcasting and Pull_Data for on-demand processing. Clients tune in broadcast channel or demand the data of interests according to the data type. Periodic invalidation reports from the server support maintaining transactional consistency. If one or more conflicts are found, conflict orders are determined not to violate the consistency(pre-reordering) and then the remaining operations have to be executed pessimistically. Through extensive simulations, we demonstrate the improved throughput of the proposed algorithm.

• Proceedings of the Korean Society for Language and Information Conference
• /
• 1996.02a
• /
• pp.297-306
• /
• 1996

A spontaneously spoken, natural Japanese discourse contains many instances of the so-called redundant interjections and of backchannel utterances. These expressions have not hitherto received much attention and few systematic analyses have been made. We show that these utterances are characterizable as discourse markers, and that they comprise a well-defined category, characterizable in a regular manner by their phonologico-prosodic properties. Our report is based on an experiment involving spontaneously spoken conversations, recorded in a laboratory environment and analyzed using digital devices. Prosodic patterns of discourse markers occurring in the recorded conversations have been analyzed. Several pitch patterns have been found that characterize the most frequently used Japanese discourse markers

• Annual Conference on Human and Language Technology
• /
• 2023.10a
• /
• pp.263-268
• /
• 2023

백채널은 화자의 말에 언어 및 비언어적으로 반응하는 것으로 상대의 대화 참여를 유도하는 역할을 한다. 백채널은 보편형 대화 참여와 반응형 대화 참여로 나뉠 수 있다. 보편형 대화 참여는 화자에게 대화를 장려하도록 하는 단순한 반응이다. 반면에 반응형 대화 참여는 화자의 발화 의도를 파악하고 그에 맞게 반응하는 것이다. 이때 발화의 의미를 파악하기 위해서는 표면적인 의미뿐만 아니라 대화의 맥락을 이해해야 한다. 본 논문에서는 대화 맥락을 반영한 백채널 예측 모델을 제안하고 예측 성능을 개선하고자 한다. 대화 맥락을 요약하기 위한 방법으로 전체 요약과 선택 요약을 제안한다. 한국어 상담 데이터를 대상으로 실험한 결과는 현재 발화만 사용했을 때보다 제안한 방식으로 대화 맥락을 반영했을 때 성능이 향상되었다.

• Asian Journal for Public Opinion Research
• /
• v.1 no.4
• /
• pp.249-267
• /
• 2014

Social media is being increasingly utilized in disaster relief work to identify safety issues, locate displaced-victims, and seek or provide support for those who need help. The presence of social media in disaster management has changed significantly in recent years, as it was not prevalent in the 2008 Wenchuan earthquake, but had become a powerful force in the 2013 Ya'an earthquake. This paper discusses the development of social media in disaster management via making a comparison between how it functioned in the two earthquakes. It examines the following aspects: who are the stakeholders that use social media in the earthquake management; how do they adopt this means in response to the earthquakes; and what are the outcomes of adopting social media with regards to public engagement and collaboration in an emergency event. As Sina Weibo acts as the equivalent of China's Twitter, the methodology relies on an analysis of posts in Weibo. The outcomes primarily show that: (1) authorities, celebrities and the public actively adopted social media for the purpose of information dissemination and resource mobilization; and (2) social media users are both content consumers and content generators in the times of earthquakes. The study concludes that social media as a backchannel communication tool is helpful for government institutions, corporations, and nonprofit organizations to build relationships with their stakeholders in the disaster management cycle. The result will interest academics and emergency management practitioners who are concerned with improving disaster communication.

• Journal of KIISE:Databases
• /
• v.31 no.3
• /
• pp.297-306
• /
• 2004

Push-based broadcasting in wireless information services is a very effective technique to disseminate information to a massive number of clients when the number of data items is small. When the database is large, however, it nay be beneficial to integrate a pull-based (client-to-server) backchannel with the push-based broadcast approach, resulting in a hybrid data delivery. In this paper, we analyze the performance behavior of a predeclaration-based transaction processing, which was originally devised for a push-based data broadcast, in the hybrid data delivery through an extensive simulation. Our results show that the use of predeclaration-based transaction processing can provide significant performance improvement not only in a pure push data delivery, but also in a hybrid data delivery.