As a new concept, the metaverse has been widely concerned by the industry, academia, media and the public. Many domestic and foreign companies have also set up in the field of the metaverse. The traditional 2D and 3D virtual fitting has not achieved breakthrough and development because of the technical problems of authenticity and timeliness. With this in mind, we developed a virtual fitting system based on the metaverse community, which includes two modules: parameterized virtual digital human modeling and multi-scene and multi-action fitting. The system realizes the construction of individualized virtual digital person. As the medium between the metaverse clothing community and the real world, it is used to achieve multi-category dynamic fitting display actions in the meta-universe clothing community platform for multi-category scenes. The system integrates the high simulation and synchronization of the metaverse with the virtual fitting system, to break the barriers of traditional virtual fitting technology and realize the combination of the garment industry and the metaverse. The experimental results show that the system can realize the construction of virtual digital human in 2.1ms at the fastest and realize the dynamic display of multi-action and multi-scene fitting.
{"title":"Garment Metaverse: Parametric Digital Human and Dynamic Scene Try-on","authors":"Hua Wang, Xiaoxiao Liu, Minghua Jiang, Changlong Zhou","doi":"10.1145/3590003.3590014","DOIUrl":"https://doi.org/10.1145/3590003.3590014","url":null,"abstract":"As a new concept, the metaverse has been widely concerned by the industry, academia, media and the public. Many domestic and foreign companies have also set up in the field of the metaverse. The traditional 2D and 3D virtual fitting has not achieved breakthrough and development because of the technical problems of authenticity and timeliness. With this in mind, we developed a virtual fitting system based on the metaverse community, which includes two modules: parameterized virtual digital human modeling and multi-scene and multi-action fitting. The system realizes the construction of individualized virtual digital person. As the medium between the metaverse clothing community and the real world, it is used to achieve multi-category dynamic fitting display actions in the meta-universe clothing community platform for multi-category scenes. The system integrates the high simulation and synchronization of the metaverse with the virtual fitting system, to break the barriers of traditional virtual fitting technology and realize the combination of the garment industry and the metaverse. The experimental results show that the system can realize the construction of virtual digital human in 2.1ms at the fastest and realize the dynamic display of multi-action and multi-scene fitting.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128921393","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Although face recognition technology is extensively used, it is vulnerable to various face spoofing attacks, such as photo and video attacks. Face anti-spoofing is a crucial step in the face recognition process and is particularly important for the security of identity verification. However, most of today's face anti-spoofing algorithms regard this task as an image binary classification problem, which is easy to over-fit. Therefore, this paper builds the basic deep supervised network as the baseline model and designs the central gradient convolution to extract the pixel difference information within the local region. To reduce the redundancy of gradient features, the central gradient convolution is decoupled to replace the vanilla convolution in the baseline model to form two cross-central gradient networks. A cross-feature interaction module is then built to effectively fuse the networks. And a depth uncertainty module is built for the problem that most face datasets are noisy and it is difficult for the model to extract fuzzy region features. Compared with existing methods, the proposed method performs well on the OULU-NPU, CASIA-FASD, and Replay-Attack datasets.
{"title":"Face Anti-spoofing Method Based on Deep Supervision","authors":"Hongxia Wang, Li Liu, Ailing Jia","doi":"10.1145/3590003.3590023","DOIUrl":"https://doi.org/10.1145/3590003.3590023","url":null,"abstract":"Although face recognition technology is extensively used, it is vulnerable to various face spoofing attacks, such as photo and video attacks. Face anti-spoofing is a crucial step in the face recognition process and is particularly important for the security of identity verification. However, most of today's face anti-spoofing algorithms regard this task as an image binary classification problem, which is easy to over-fit. Therefore, this paper builds the basic deep supervised network as the baseline model and designs the central gradient convolution to extract the pixel difference information within the local region. To reduce the redundancy of gradient features, the central gradient convolution is decoupled to replace the vanilla convolution in the baseline model to form two cross-central gradient networks. A cross-feature interaction module is then built to effectively fuse the networks. And a depth uncertainty module is built for the problem that most face datasets are noisy and it is difficult for the model to extract fuzzy region features. Compared with existing methods, the proposed method performs well on the OULU-NPU, CASIA-FASD, and Replay-Attack datasets.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"87 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130540008","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper proposed an electronic nose system that utilized a SnO2 semiconductor sensor array to detect volatile ammonia gas in farmland. All sensors were controlled by the Arduino development board. The system could collect data during both the steady-state and transient phases of sensor operation. The collected data was analyzed using PCA (principal component analysis) and MLP (Multi-layer perceptron) neural networks. The experiment was divided into two parts: The first part analyzed four concentrations of ammonia (100ppm, 200ppm, 400ppm, and Air) using PCA and MLP, which successfully distinguished the concentrations with an identification rate of over 95%. In the second part, four gases (air mixed with ammonia, pure ammonia gas, air mixed with ethanol, and pure ethanol) were analyzed using PCA and MLP, with the electronic nose system successfully distinguishing between the four types of gases. The system could read and process data during the transient phase of the sensor, and the constructed sensor array electronic nose system and acquisition method has significant potential for ammonia detection in agricultural environments.
{"title":"ENOSE Performance in Transient Time and Steady State Area of Gas Sensor Response for Ammonia Gas: Comparison and Study","authors":"Kuan Geng, Jahangir Moshayedi Ata, Jing-hao Chen, Jiandong Hu, Hao Zhang","doi":"10.1145/3590003.3590046","DOIUrl":"https://doi.org/10.1145/3590003.3590046","url":null,"abstract":"This paper proposed an electronic nose system that utilized a SnO2 semiconductor sensor array to detect volatile ammonia gas in farmland. All sensors were controlled by the Arduino development board. The system could collect data during both the steady-state and transient phases of sensor operation. The collected data was analyzed using PCA (principal component analysis) and MLP (Multi-layer perceptron) neural networks. The experiment was divided into two parts: The first part analyzed four concentrations of ammonia (100ppm, 200ppm, 400ppm, and Air) using PCA and MLP, which successfully distinguished the concentrations with an identification rate of over 95%. In the second part, four gases (air mixed with ammonia, pure ammonia gas, air mixed with ethanol, and pure ethanol) were analyzed using PCA and MLP, with the electronic nose system successfully distinguishing between the four types of gases. The system could read and process data during the transient phase of the sensor, and the constructed sensor array electronic nose system and acquisition method has significant potential for ammonia detection in agricultural environments.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127875618","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Haoran Wang, Tianyun Xue, Zhaoran Wang, Xiangyu Bai
As an integral part of the ecosystem, grassland plays an important role in protecting water and soil, preventing wind and fixing sand and protecting biodiversity. However, some grasslands are degraded at this stage, so a grassland monitoring method is urgently needed to prevent desertification from spreading. With the rapid rise of deep learning, it is more and more popular to apply artificial intelligence methods to grassland degradation monitoring. This paper systematically and comprehensively analyzes that almost all semantic segmentation methods have been applied to relevant research on grassland degradation areas since semantic segmentation methods were applied to grassland monitoring. Then, according to the different algorithm structures of grassland extraction methods, the principles of representative algorithms are introduced in turn. Then we made a statistical analysis of the publication status, research space distribution and the number of citations of papers in this field. Finally, the analysis results are discussed, and the possible research hotspots in the future are discussed.
{"title":"Comparison of regional monitoring methods for grassland degradation based on remote sensing images","authors":"Haoran Wang, Tianyun Xue, Zhaoran Wang, Xiangyu Bai","doi":"10.1145/3590003.3590083","DOIUrl":"https://doi.org/10.1145/3590003.3590083","url":null,"abstract":"As an integral part of the ecosystem, grassland plays an important role in protecting water and soil, preventing wind and fixing sand and protecting biodiversity. However, some grasslands are degraded at this stage, so a grassland monitoring method is urgently needed to prevent desertification from spreading. With the rapid rise of deep learning, it is more and more popular to apply artificial intelligence methods to grassland degradation monitoring. This paper systematically and comprehensively analyzes that almost all semantic segmentation methods have been applied to relevant research on grassland degradation areas since semantic segmentation methods were applied to grassland monitoring. Then, according to the different algorithm structures of grassland extraction methods, the principles of representative algorithms are introduced in turn. Then we made a statistical analysis of the publication status, research space distribution and the number of citations of papers in this field. Finally, the analysis results are discussed, and the possible research hotspots in the future are discussed.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125506916","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Query-based object detection, including DETR and Sparse R-CNN, has gained considerable attention in recent years. However, in dense scenes, end-to-end object detection methods are prone to false positives. To address this issue, we propose a graph convolution-based post-processing component to refine the output results from Sparse R-CNN. Specifically, we initially select high-scoring queries to generate true positive predictions. Subsequently, the query updater refines noisy query features using GCN. Lastly, the label assignment rule matches accepted predictions to ground truth objects, eliminates matched targets, and associates noisy predictions with the remaining ground truth objects. Our method significantly enhances performance in crowded scenes. Our method achieves 92.3% AP and 41.6% on CrowdHuman dataset, which is a challenging objection detection dataset.
{"title":"A Component for Query-based Object Detection in Crowded Scenes","authors":"Shuo Mao","doi":"10.1145/3590003.3590039","DOIUrl":"https://doi.org/10.1145/3590003.3590039","url":null,"abstract":"Query-based object detection, including DETR and Sparse R-CNN, has gained considerable attention in recent years. However, in dense scenes, end-to-end object detection methods are prone to false positives. To address this issue, we propose a graph convolution-based post-processing component to refine the output results from Sparse R-CNN. Specifically, we initially select high-scoring queries to generate true positive predictions. Subsequently, the query updater refines noisy query features using GCN. Lastly, the label assignment rule matches accepted predictions to ground truth objects, eliminates matched targets, and associates noisy predictions with the remaining ground truth objects. Our method significantly enhances performance in crowded scenes. Our method achieves 92.3% AP and 41.6% on CrowdHuman dataset, which is a challenging objection detection dataset.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126241978","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper proposes a data mining algorithm based on multimodal decision fusion, which is mainly used to solve the correlation relationship of multi-level and multi-level multimodal data, the algorithm combines the methods of statistics, queueing study, machine learning and Bayesian decision fusion, compared with the results obtained by single modality, single data and single method, the algorithm proposed in this paper retains the information contained in the data to the maximum extent, and the algorithm is applied to the analysis of both numerical and text-based data. The proposed algorithm can be further extended by modifying the data types and methods to form new methods.
{"title":"Speech image data mining algorithm based on multimodal decision fusion","authors":"Cong Lu, Danxing Wang, Daquan Zhang, Aiqun Yu","doi":"10.1145/3590003.3590007","DOIUrl":"https://doi.org/10.1145/3590003.3590007","url":null,"abstract":"This paper proposes a data mining algorithm based on multimodal decision fusion, which is mainly used to solve the correlation relationship of multi-level and multi-level multimodal data, the algorithm combines the methods of statistics, queueing study, machine learning and Bayesian decision fusion, compared with the results obtained by single modality, single data and single method, the algorithm proposed in this paper retains the information contained in the data to the maximum extent, and the algorithm is applied to the analysis of both numerical and text-based data. The proposed algorithm can be further extended by modifying the data types and methods to form new methods.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121532552","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Fuhu Song, Jifeng Hu, Che Wang, Jiao Huang, Haowen Zhang, Yi Wang
The goal of cross-modal audio-text retrieval is to retrieve the target audio clips (textual descriptions), which should be relevant to a given textual (audial) query. It is a challenging task because it necessitates learning comprehensive feature representations for two different modalities and unifying them into a common embedding space. However, most existing cross-modal audio-text retrieval approaches do not explicitly learn the sequential representation in audio features. Moreover, their method of directly employing a fully connected neural network to transform the different modalities into a common space is detrimental to sequential features. In this paper, we introduce a sequential feature augmentation framework based on reinforcement learning and feature fusion to enhance the sequential feature for cross-modal features. First, we adopt reinforcement learning to explore effective sequential features in audial and textual features. Then, a recurrent fusion module is applied as a feature enhancement component to project heterogeneous features into a common space. Extensive experiments are conducted on two prevalent datasets: the AudioCaps and the Clotho. The results demonstrate that our method gains a significant improvement over previous state-of-the-art methods.
{"title":"Cross-Modal Audio-Text Retrieval via Sequential Feature Augmentation","authors":"Fuhu Song, Jifeng Hu, Che Wang, Jiao Huang, Haowen Zhang, Yi Wang","doi":"10.1145/3590003.3590056","DOIUrl":"https://doi.org/10.1145/3590003.3590056","url":null,"abstract":"The goal of cross-modal audio-text retrieval is to retrieve the target audio clips (textual descriptions), which should be relevant to a given textual (audial) query. It is a challenging task because it necessitates learning comprehensive feature representations for two different modalities and unifying them into a common embedding space. However, most existing cross-modal audio-text retrieval approaches do not explicitly learn the sequential representation in audio features. Moreover, their method of directly employing a fully connected neural network to transform the different modalities into a common space is detrimental to sequential features. In this paper, we introduce a sequential feature augmentation framework based on reinforcement learning and feature fusion to enhance the sequential feature for cross-modal features. First, we adopt reinforcement learning to explore effective sequential features in audial and textual features. Then, a recurrent fusion module is applied as a feature enhancement component to project heterogeneous features into a common space. Extensive experiments are conducted on two prevalent datasets: the AudioCaps and the Clotho. The results demonstrate that our method gains a significant improvement over previous state-of-the-art methods.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"173 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121533331","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
This paper presents a deep context-aware model with a copy mechanism based on reinforcement learning for abstractive text summarization. Our model is optimized using weighted ROUGEs as global prediction-based rewards and the self-critical policy gradient training algorithm, which can reduce the inconsistency between training and testing by directly optimizing the evaluation metrics. To alleviate the lexical diversity and component diversity problems caused by global prediction rewards, we improve the richness of the multi-head self-attention mechanism to capture context through global deep context representation with copy mechanism. We conduct experiments and demonstrate that our model outperforms many existing benchmarks over the Gigaword, LCSTS, and CNN/DM datasets. The experimental results demonstrate that our model has a significant effect on improving the quality of summarization.
{"title":"Deep Reinforcement Learning with Copy-oriented Context Awareness and Weighted Rewards for Abstractive Summarization","authors":"Caidong Tan","doi":"10.1145/3590003.3590019","DOIUrl":"https://doi.org/10.1145/3590003.3590019","url":null,"abstract":"This paper presents a deep context-aware model with a copy mechanism based on reinforcement learning for abstractive text summarization. Our model is optimized using weighted ROUGEs as global prediction-based rewards and the self-critical policy gradient training algorithm, which can reduce the inconsistency between training and testing by directly optimizing the evaluation metrics. To alleviate the lexical diversity and component diversity problems caused by global prediction rewards, we improve the richness of the multi-head self-attention mechanism to capture context through global deep context representation with copy mechanism. We conduct experiments and demonstrate that our model outperforms many existing benchmarks over the Gigaword, LCSTS, and CNN/DM datasets. The experimental results demonstrate that our model has a significant effect on improving the quality of summarization.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121193004","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Donat Scharnitzky, Zsolt Kramer, S. Molnár, A. Mihály
Tools for real-time emulation of mobile networks are valuable for researchers due to the high amount of time and resources it allows to save compared to carrying out measurements in live networks. In this paper we present the rationale, design and prototype implementation of a novel net device in the ns-3 open source network simulator that allows for end-to-end real-time emulation of LTE networks with real endpoints. We then show the performance evaluation of a QUIC proxy built on MASQUE using our emulated LTE setup. Our results confirm the intended behavior of the implementation, however, we also show the limitations of the real-time capabilities of ns-3.
{"title":"Real-time Emulation of MASQUE-based QUIC Proxying in LTE Networks using ns-3","authors":"Donat Scharnitzky, Zsolt Kramer, S. Molnár, A. Mihály","doi":"10.1145/3590003.3590995","DOIUrl":"https://doi.org/10.1145/3590003.3590995","url":null,"abstract":"Tools for real-time emulation of mobile networks are valuable for researchers due to the high amount of time and resources it allows to save compared to carrying out measurements in live networks. In this paper we present the rationale, design and prototype implementation of a novel net device in the ns-3 open source network simulator that allows for end-to-end real-time emulation of LTE networks with real endpoints. We then show the performance evaluation of a QUIC proxy built on MASQUE using our emulated LTE setup. Our results confirm the intended behavior of the implementation, however, we also show the limitations of the real-time capabilities of ns-3.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121708101","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Frequent customer service conversations focus on hot topics of communication users, and automatic hot topic discovery is critical to improving user experience. Traditionally, Customer service relies on operator to write traffic summaries. It leads to the source of the conversation difficult to analyze, which makes difficult to spot aggregated hotspot events. In this paper, we propose a Customer Service hot event Discovery based on dynamic dialogue embedding (CShe-D). This model includes dynamic semantic representation of customer service dialogue, clustering-based customer service hot event discovery and new hot event prediction. In the dialogue semantic embedding module, we obtain the dynamic embedding of each dialogue with combining word importance and word length based on the pre-trained language model to capture richer semantic information in different contexts. We further apply a clustering iterative algorithm with dynamic dialogue embedding to discover customer service hotspots. It can monitor the change trend of events in real time, optimize the accuracy of hot event discovery in operator customer service. Finally, the effectiveness of our CShe-D model is verified by experiments on real dialogue data in the field of customer service.
{"title":"Customer Service Hot event Discovery Based on Dynamic Dialogue Embedding","authors":"Fei Li, Yanyan Wang, Ying Feng, Qiangzhong Feng, Yuan Zhou, Dexuan Wang","doi":"10.1145/3590003.3590011","DOIUrl":"https://doi.org/10.1145/3590003.3590011","url":null,"abstract":"Frequent customer service conversations focus on hot topics of communication users, and automatic hot topic discovery is critical to improving user experience. Traditionally, Customer service relies on operator to write traffic summaries. It leads to the source of the conversation difficult to analyze, which makes difficult to spot aggregated hotspot events. In this paper, we propose a Customer Service hot event Discovery based on dynamic dialogue embedding (CShe-D). This model includes dynamic semantic representation of customer service dialogue, clustering-based customer service hot event discovery and new hot event prediction. In the dialogue semantic embedding module, we obtain the dynamic embedding of each dialogue with combining word importance and word length based on the pre-trained language model to capture richer semantic information in different contexts. We further apply a clustering iterative algorithm with dynamic dialogue embedding to discover customer service hotspots. It can monitor the change trend of events in real time, optimize the accuracy of hot event discovery in operator customer service. Finally, the effectiveness of our CShe-D model is verified by experiments on real dialogue data in the field of customer service.","PeriodicalId":340225,"journal":{"name":"Proceedings of the 2023 2nd Asia Conference on Algorithms, Computing and Machine Learning","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2023-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125175187","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：