Search

In real-life situations where human care robots are deployed, there are custom sound events whose acoustic characteristics change depending on the user’s choice unlike general sound events so that the human care robots cannot recognize custom sound events correctly in a conventional way. To solve this critical problem, a real-time sound event recognition system with customization process is proposed. The human care robot collects custom sound samples of a specific user and customizes a sound event recognition model. The overfitting-based customized model shows the best recognition performance by improving F-scores by 66.4% on average compared to the conventional recognition model. After the customization process, the human care robot performs a real-time sound recognition by consistently streaming robot’s real-time microphone signals into the overfitting-based customized SER model. In this process, an optimized overlap is applied on subsequent audio inputs on SER to achieve sufficiently fast response and robust performance. As a pilot test of the human care robot implemented in actual environment, the real-time sound recognition system shows the best average F-score of 0.982 with 75% overlap for sound events including custom sounds. This pilot test result confirms that the real-time sound recognition system with customization process can be successfully applied to human care robots to respond to the custom sounds.

Sequential recommendation systems aim to predict users’ subsequent interactions by analyzing their historical behavior. Significant advancements have been achieved in this field through the utilization of models such as RNN, GRU, and models incorporating the self-attention mechanism. Most state-of-the-art methods enhance recommendation effectiveness by incorporating side information, such as category or brand, within the attention layer to enrich item representations. However, in our pilot experiments, we found that increasing the amount of side information could negatively impact the recommendation performance when integrated directly into the attention layer due to the learning of interdependency among them. To alleviate this problem, we propose Independent Representation of side Information on each item for Sequential recommendation (IRIS). IRIS learns pairs of both item and side information independently, without any fusion among side information in attention layer. In contrast to existing approaches, IRIS avoids direct fusion among side information in the attention layer, opting instead for a novel strategy of learning independent pairs of both item and side information. Our proposed method constructs an attention layer specifically designed to learn these pairs, thus avoiding potential distortions to item representations caused by intermixing side information. Experiments on four public datasets show that IRIS outperforms state-of-the-art sequential recommendation models. Ablation study also indicates robust scalability of side information within IRIS.

Deep learning models have significantly addressed the challenges of multivariate time series forecasting. Recently, Transformer-based models which have primarily focused on either temporal or inter-variate (spatial) dependencies have demonstrated exceptional performance. These models decide whether to embed multivariate time series data temporally or spatially. Hence, we propose the dual embedded transformer (DEformer) which simultaneously considers both temporal and spatial dependencies. Our model enables capturing these dependencies independently through two distinct encoders. The temporal encoder, which processes the entire time series of each variate, is designed to learn temporal dependency. Conversely, the spatial encoder, which processes the multivariate data at each time step, is optimized to capture spatial dependency. Both encoders share an identical architecture, comprising an attention mechanism to model correlations and a feed-forward network to enhance feature representation. Through empirical studies on challenging real-world datasets, we not only demonstrate that our method can outperforms state-of-the-art approaches, but also prove the performance of independently embedding both dependencies through ablation study.

Recently, recommender systems have been used in various fields. However, they are still plagued by many issues, including cold-start and sparsity problems. The cold-start problem occurs when users are unable to make recommendations to other users owing to a complete lack of information about certain items. This problem can exist both at the user side and the item side. User-side cold-start problems occur when new users access the systems; item-side cold-start problems occur when new items are added to databases. In this study, we addressed the item-side cold-start problem using the concept of weak supervision. First, a new process for identifying feature based representative reviewers in a rater group was designed. Then, we developed a method to predict the expected preferences for new items by combining content-based filtering and the preferences of representative users. Through extensive experiments, we first confirmed that in comparison to existing methods, the proposed approach provided enhanced accuracy, which was evaluated by determining a mean absolute error for the average ratings. Then, we compared the proposed scheme with the collaborative filtering (CF) and neural CF approaches (NCF). The estimation by the proposed approach was 21% and 38% more accurate than CF and NCF in terms of mean absolute error (MAE), respectively. In future, the proposed framework can be applied in various recommender systems as a core function.