Recently, the demand for remote counseling has been on the rise using conversational agents with Large Language Models (LLM) in many areas. This LLM trend is expanding beyond generating simple words to constructing complex sentences, showing progress in various fields. The significant development of LLM is attributed to the progress in Natural Language Processing (NLP) technology, built upon extensive language data. This study introduces a client technology of Open API System based conversational agent using LLM. The target agent is a Chatbot which can generate health counseling messages including empathetic conversations for users with caring his chronic conditions requiring ongoing health management. The overall system can be linked with the community care system through user APPs and APIs for digital healthcare context management. The advantage with the client with REST API is that APIs make it easy to integrate new applications with existing software systems, allowing target systems to meet requirements across a variety of platforms.

In the realm of Generative AI, where various models are introduced, prompt engineering emerges as a significant technique within natural language processing-based Generative AI. Its primary function lies in effectively enhancing the results of sentence generation by large language models (LLMs). Notably, prompt engineering has gained attention as a method capable of improving LLM performance by modifying the structure of input prompts alone. In this study, we apply prompt engineering to Korean-based LLMs, presenting an efficient approach for generating specific conversational responses with less data. We achieve this through the utilization of the query transformation module (QTM). Our proposed QTM transforms input prompt sentences into three distinct query methods, breaking them down into objectives and key points, making them more comprehensible for LLMs. For performance validation, we employ Korean versions of LLMs, specifically SKT GPT-2 and Kakaobrain KoGPT-3. We compare four different query methods, including the original unmodified query, using Google SSA to assess the naturalness and specificity of generated sentences. The results demonstrate an average improvement of 11.46% when compared to the unmodified query, underscoring the efficacy of the proposed QTM in achieving enhanced performance.

This study examined ethical ways to use welfare technology in a situation where the demand for non-face-to-face welfare services using Cloud based healthcare systems had increased rapidly in caring for elderly people who live alone. Through focus group interviews with social workers related to the care of elderly people who live alone, in-depth interviews were conducted on the current situation, problems, ethical issues, and development directions arising in the implementation of welfare technology. The main areas of interest were focused on improving safety in caring them using IoT technology and enhancing emotional support in preventing lonely deaths using companion robot and AI speaker. Issues such as the need for individualization, client-centeredness, privacy, self-determination, competence, informed consent, right to know, convenience, and advocacy were identified as important ethical considerations related to use of welfare technology. The research results suggested that various stakeholders should participate in the development of ethical indicators and welfare technology for the ethical use of welfare technology.

The development of wireless communication technology and mobile devices has brought about the advent of an era of sharing text data that overflows on social media and the web. In particular, social media has become a major source of storing people’s sentiments in the form of opinions and views on specific issues in the form of unstructured information. Therefore, the importance of emotion analysis is increasing, especially with machine learning for both personal life and companies’ management environments. At this time, data reliability is an essential component for data classification. The accuracy of sentiment classification can be heavily determined according to the reliability of data, in which case noise data may also influence this classification. Although there is stopword that does not have meaning in such noise data, data that does not fit the purpose of analysis can also be referred to as noise data. This paper aims to provide an analysis of the impact of profanity data on deep learning-based sentiment classification. For this purpose, we used movie review data on the Web and simulated the changes in performance before and after the removal of the profanity data. The accuracy of the model trained with the data and the model trained with the data before removal were compared to determine whether the profanity is noise data that lowers the accuracy in sentiment analysis. The simulation results show that the accuracy dropped by about 2% when judging profanity as noise data in the sentiment classification for review data.

In this letter, we study a scenario based on degenerate unmixing estimation technique (DUET) that separates original signals from mixture of FHSS signals with two antennas. We have shown that the assumptions for separating mixed signals in DUET can be applied to drone based digital signage recognition signals and proposed the DUET-based separation scheme (DBSS) to classify the mixed recognition drone signals by extracting the delay and attenuation components of the mixture signal through the likelihood function and the short-term Fourier transform (STFT). In addition, we propose an iterative algorithm for signal separation with the conventional DUET scheme. Numerical results showed that the proposed algorithm is more separation-efficient compared to baseline schemes. DBSS can separate all signals within about 0.56 seconds when there are fewer than nine signage signals.

Human activity recognition (HAR) is a key technology in many applications, such as smart signage, smart healthcare, smart home, etc. In HAR, deep learning-based methods have been proposed to recognize activity data effectively from video streams. In this paper, the end-to-end model based on convolutional long short-term memory (LSTM) is proposed to recognize human activities. Convolutional LSTM can learn features of spatial and temporal simultaneously from video stream data. Also, the number of learning weights can be diminished by employing convolutional LSTM with an end-to-end model. The proposed HAR model was optimized with various simulation environments using activities data from the AI hub. From simulation results, it can be confirmed that the proposed model can be outperformed compared with the conventional model.

In a computing environment, higher resolutions generally require more memory bandwidth, which inevitably leads to the consumption more power. This may become critical for the overall performance of mobile devices and graphic processor units with increased amounts of memory access and memory bandwidth. This paper proposes a lossless compression algorithm with a multiple differential pulse-code modulation variable sign code Golomb-Rice to reduce the memory bandwidth requirement. The efficiency of the proposed multiple differential pulse-code modulation is enhanced by selecting the optimal differential pulse code modulation mode. The experimental results show compression ratio of 1.99 for high-efficiency video coding image sequences and that the proposed lossless compression hardware can reduce the bus bandwidth requirement.

Deep reinforcement learning technique combines reinforcement learning and neural network for various applications. This paper is to propose an effective lazy training method for deep reinforcement learning, especially for deep Q-network combining neural network with Q-learning to be used for the obstacle avoidance and path planning applications. The proposed method can reduce the overall training time by designing a lazy learning method and a method removing unnecessary repetitions in the training step. These two methods can reduce a significant portion of total execution time without losing any required accuracy. The proposed method is evaluated for the obstacle avoidance and path planning tasks, where an agent trapped in an unknown environment is trying to find out the shortest path to the destination without any collision, through its self-study. And the experiment results show that the proposed method reduces 53.38% of training time on average, compared to the traditional method with no performance loss and make the training procedure more stable.