Event extraction is a sub-task of information extraction and is an important part of natural language processing. Depending on the range of features used, event extraction methods are classified as sentence-level or document-level. However, document-level event extraction is more practical for practical tasks. Document-level event extraction is a difficult task, as it requires features to be extracted from a larger amount of text to determine which span of text is the desired event element. However, most methods do not utilize both sentence-level and document-level features. In order to utilize hybrid feature information and fuse it, this paper proposes a document-level event extraction method that integrating hybrid features. The event extraction method is based on Dynamic Multi-Pooling Convolutional Neural Network (DMCNN) and Bi-directional Long Short-Term Memory (BiLSTM), combined with self-attention mechanisms and Conditional Random Field (CRF). We evaluate the model proposed in this paper on the MUC-4 dataset and the experimental results show that our proposed model outperforms previous work.
{"title":"Integrating Hybrid Features for Document-Level Event Role Extraction Method","authors":"Jingyao Zhang, Tao Xu","doi":"10.1145/3573942.3574051","DOIUrl":"https://doi.org/10.1145/3573942.3574051","url":null,"abstract":"Event extraction is a sub-task of information extraction and is an important part of natural language processing. Depending on the range of features used, event extraction methods are classified as sentence-level or document-level. However, document-level event extraction is more practical for practical tasks. Document-level event extraction is a difficult task, as it requires features to be extracted from a larger amount of text to determine which span of text is the desired event element. However, most methods do not utilize both sentence-level and document-level features. In order to utilize hybrid feature information and fuse it, this paper proposes a document-level event extraction method that integrating hybrid features. The event extraction method is based on Dynamic Multi-Pooling Convolutional Neural Network (DMCNN) and Bi-directional Long Short-Term Memory (BiLSTM), combined with self-attention mechanisms and Conditional Random Field (CRF). We evaluate the model proposed in this paper on the MUC-4 dataset and the experimental results show that our proposed model outperforms previous work.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115928438","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}
In order to ensure the correctness of module interaction between distributed systems in the analysis and design stages, this paper proposes a visual model checking method for distributed systems. The component diagram and sequence diagram are used to visually model the system and describe the interaction between subsystems. The object property specification language was used to annotate the properties of the model, and the properties were extracted and converted into projection temporal logic formulas, and then converted into property non-automata. The sequence diagram model is transformed into a system automaton. Finally, the model checking tool is used to verify whether the model satisfies the system properties. The experimental results show that this method can realize the verification of distributed system and the modeling is more intuitive and convenient.
{"title":"Visual Model Checking Distributed System","authors":"Yiyang Jia, Xinfeng Shu","doi":"10.1145/3573942.3574023","DOIUrl":"https://doi.org/10.1145/3573942.3574023","url":null,"abstract":"In order to ensure the correctness of module interaction between distributed systems in the analysis and design stages, this paper proposes a visual model checking method for distributed systems. The component diagram and sequence diagram are used to visually model the system and describe the interaction between subsystems. The object property specification language was used to annotate the properties of the model, and the properties were extracted and converted into projection temporal logic formulas, and then converted into property non-automata. The sequence diagram model is transformed into a system automaton. Finally, the model checking tool is used to verify whether the model satisfies the system properties. The experimental results show that this method can realize the verification of distributed system and the modeling is more intuitive and convenient.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131351309","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}
The traditional industrial Internet platform adopts a centralized data storage model, and the security sharing of data generated by terminal devices is a major bottleneck that hinders the development of industrial Internet. Along with the geometric growth of terminal data, protecting the security and integrity of data has become the core research area of the Industrial Internet. Blockchain is distributed, open, transparent and tamper-evident, and can provide a reliable underlying service to realize a distributed data security sharing system. Therefore, this paper proposes a blockchain-based data security sharing model for industrial Internet platforms, with a distributed blockchain network as the core to build up a decentralized data security sharing service. Meanwhile, to address the problems of high consensus latency, low throughput and performance, and no support for node dynamic management of the practical Byzantine fault-tolerant (PBFT) algorithm used in blockchain, a simplified consistency protocol and a new node management mechanism are introduced to achieve dynamic management of nodes while reducing the complexity of algorithm communication.
{"title":"Research on Blockchain Consensus Algorithm for Secure Data Sharing on Industrial Internet Platform","authors":"Chao Jia, Lili Gao, Min Li","doi":"10.1145/3573942.3573973","DOIUrl":"https://doi.org/10.1145/3573942.3573973","url":null,"abstract":"The traditional industrial Internet platform adopts a centralized data storage model, and the security sharing of data generated by terminal devices is a major bottleneck that hinders the development of industrial Internet. Along with the geometric growth of terminal data, protecting the security and integrity of data has become the core research area of the Industrial Internet. Blockchain is distributed, open, transparent and tamper-evident, and can provide a reliable underlying service to realize a distributed data security sharing system. Therefore, this paper proposes a blockchain-based data security sharing model for industrial Internet platforms, with a distributed blockchain network as the core to build up a decentralized data security sharing service. Meanwhile, to address the problems of high consensus latency, low throughput and performance, and no support for node dynamic management of the practical Byzantine fault-tolerant (PBFT) algorithm used in blockchain, a simplified consistency protocol and a new node management mechanism are introduced to achieve dynamic management of nodes while reducing the complexity of algorithm communication.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"44 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133474221","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}
Accurate identification of MFR working mode recognition is an essential prerequisite for target threat assessment. To solve the problem of lower recognition rate of radar pulse signals with overlapping parameters, a hybrid recognition model based on CNN-BILSTM-SoftAttention is proposed. Firstly, We utilize the combined CPI parameters to describe pluse stream and capture local characteristics with CNN. Then, the BILSTM Network is used to analyze the timing regularity of radar pulse sequences, and to discover the inter-class rule between different working modes and the intra-class rule of the same working mode. Finally, combined with the attention mechanism model, we can distinguish different working mode by assigning higher weights to parameters with overlapping. Through simulation analysis, the proposed algorithm is compared with SVM, CNN, CNN_LSTM method, the accuracy of model can reach 92.48% in the strong noise environment, increasing by 20%. The results show that the proposed method has better classification ability and higher performance than existing work pattern classification methods.
{"title":"MFR Working Mode Recognition Based on CNN-BILSTM-SoftAttention Model","authors":"Jie Yang, Jinghua Tian","doi":"10.1145/3573942.3574115","DOIUrl":"https://doi.org/10.1145/3573942.3574115","url":null,"abstract":"Accurate identification of MFR working mode recognition is an essential prerequisite for target threat assessment. To solve the problem of lower recognition rate of radar pulse signals with overlapping parameters, a hybrid recognition model based on CNN-BILSTM-SoftAttention is proposed. Firstly, We utilize the combined CPI parameters to describe pluse stream and capture local characteristics with CNN. Then, the BILSTM Network is used to analyze the timing regularity of radar pulse sequences, and to discover the inter-class rule between different working modes and the intra-class rule of the same working mode. Finally, combined with the attention mechanism model, we can distinguish different working mode by assigning higher weights to parameters with overlapping. Through simulation analysis, the proposed algorithm is compared with SVM, CNN, CNN_LSTM method, the accuracy of model can reach 92.48% in the strong noise environment, increasing by 20%. The results show that the proposed method has better classification ability and higher performance than existing work pattern classification methods.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"66 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133512762","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}
With the rapid development of microelectronics technology, the amount of data information is becoming larger and larger, and the speed of data processing is becoming higher and higher. In order to meet the needs of today's data cache and solve a series of problems such as unstable data transmission and data loss caused by the common data cache technology due to its small capacity and slow data processing speed, a synchronous dynamic random access memory (DDR3 SDRAM) based data cache design method with high speed and large capacity and multi-channel is proposed to achieve fast and efficient real-time storage of eight-channel video data. Based on Vivado MIG IP core and Kintex-7 FPGA as the control core, asynchronous FIFO with read/write bit width ratio of 8:1 is realized, and the read/write cache control module is designed, and the real-time data is finally cached to the corresponding address of DDR3 SDRAM. Improved DDR3 SDRAM bandwidth utilization. The experimental results show that the system can access 8-channel high speed video data, and the data transmission is stable and reliable. The design is mainly composed of multi-channel data acquisition module, cross-clock domain data processing module, read and write priority arbitration and other modules, with a working frequency of up to 400M Hz. It has been verified that the design can be used for real-time acquisition system of space-borne video storage.
{"title":"High Speed Multi-channel Data Cache Design Based on DDR3 SDRAM","authors":"Xiaofeng Yang, Ancheng Liu, Jinjin Wang","doi":"10.1145/3573942.3573972","DOIUrl":"https://doi.org/10.1145/3573942.3573972","url":null,"abstract":"With the rapid development of microelectronics technology, the amount of data information is becoming larger and larger, and the speed of data processing is becoming higher and higher. In order to meet the needs of today's data cache and solve a series of problems such as unstable data transmission and data loss caused by the common data cache technology due to its small capacity and slow data processing speed, a synchronous dynamic random access memory (DDR3 SDRAM) based data cache design method with high speed and large capacity and multi-channel is proposed to achieve fast and efficient real-time storage of eight-channel video data. Based on Vivado MIG IP core and Kintex-7 FPGA as the control core, asynchronous FIFO with read/write bit width ratio of 8:1 is realized, and the read/write cache control module is designed, and the real-time data is finally cached to the corresponding address of DDR3 SDRAM. Improved DDR3 SDRAM bandwidth utilization. The experimental results show that the system can access 8-channel high speed video data, and the data transmission is stable and reliable. The design is mainly composed of multi-channel data acquisition module, cross-clock domain data processing module, read and write priority arbitration and other modules, with a working frequency of up to 400M Hz. It has been verified that the design can be used for real-time acquisition system of space-borne video storage.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"91 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133593182","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}
The selection of wireless sensor networks (WSNs) communication topology and the application of unmanned aerial vehicles (UAV) in data collection are studied in the scenario of being far away from the base station. The topology consists of a set of cluster head nodes that communicate with the UAV. After considering the network energy consumption factors, the optimal number of cluster heads of the network is derived, the network is clustered and the cluster head node is selected. Finally, the cruising path of the drone to collect data from the cluster head is optimized. Simulation results show that the proposed algorithm can effectively save the energy of nodes, prolong network life, and reduce the total distance of the UAV collects data.
{"title":"Wireless Sensor Networks Based on UAV Auxiliary Energy-Saving Data Collection Algorithms","authors":"Aijing Sun, Yijia Li, Shichang Li","doi":"10.1145/3573942.3573982","DOIUrl":"https://doi.org/10.1145/3573942.3573982","url":null,"abstract":"The selection of wireless sensor networks (WSNs) communication topology and the application of unmanned aerial vehicles (UAV) in data collection are studied in the scenario of being far away from the base station. The topology consists of a set of cluster head nodes that communicate with the UAV. After considering the network energy consumption factors, the optimal number of cluster heads of the network is derived, the network is clustered and the cluster head node is selected. Finally, the cruising path of the drone to collect data from the cluster head is optimized. Simulation results show that the proposed algorithm can effectively save the energy of nodes, prolong network life, and reduce the total distance of the UAV collects data.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115349427","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}
Affected by the change in daytime illumination sequence and by the shooting angle in the complex field environment, the kiwifruit images possess the unfriendly features of uneven illumination, such as local darkness and local brightness. The ill-posed image with uneven illumination will seriously constraint the subsequent image analysis processing. Current deep learning methods have achieved satisfactory results, and a large number of paired images (one is the input image, one is the ground truth image) is required to train the better network performance. However, it is difficult to capture ground truth images of the kiwifruit in the field. Based on this, the paper proposed Convolutional Block Attention Module Deep Curve Estimation (CBAM-DCE) to accomplish a non-reference illumination unevenness correction for field kiwifruit images. A deep learning network model is used to estimate the image-specific curve for image enhancement, and a non-reference loss function is applied to evaluate the image enhancement effect. Compared with seven related enhancement algorithms, the presented algorithm shakes off uneven illumination or normal-light image pairs for training. Five different public datasets and our Kiwifruit dataset were used in the experiments. Experiments demonstrate that our proposed CBAM-DCE is superior to other state-of-the-art algorithms for enhancing natural images under different lighting conditions.
{"title":"CBAM-DCE: A Non-Reference Image Correction Algorithm for Uneven Illumination","authors":"Mengyu Fan, Jinjun Lu, Xianguang Kong, Wei Sun, Wei Sun, Yijun Sun","doi":"10.1145/3573942.3574098","DOIUrl":"https://doi.org/10.1145/3573942.3574098","url":null,"abstract":"Affected by the change in daytime illumination sequence and by the shooting angle in the complex field environment, the kiwifruit images possess the unfriendly features of uneven illumination, such as local darkness and local brightness. The ill-posed image with uneven illumination will seriously constraint the subsequent image analysis processing. Current deep learning methods have achieved satisfactory results, and a large number of paired images (one is the input image, one is the ground truth image) is required to train the better network performance. However, it is difficult to capture ground truth images of the kiwifruit in the field. Based on this, the paper proposed Convolutional Block Attention Module Deep Curve Estimation (CBAM-DCE) to accomplish a non-reference illumination unevenness correction for field kiwifruit images. A deep learning network model is used to estimate the image-specific curve for image enhancement, and a non-reference loss function is applied to evaluate the image enhancement effect. Compared with seven related enhancement algorithms, the presented algorithm shakes off uneven illumination or normal-light image pairs for training. Five different public datasets and our Kiwifruit dataset were used in the experiments. Experiments demonstrate that our proposed CBAM-DCE is superior to other state-of-the-art algorithms for enhancing natural images under different lighting conditions.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124313876","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}
An image encryption algorithm based on hyperchaotic system is proposed to satisfy the security of image information transmission, the encryption algorithm consists of two times of scrambling and diffusion. The relation between plaintext and key is constructed to generate plaintext related chaotic sequence. Two rounds of scrambling operation are performed to make the pixel positions of the image are sufficiently scrambled, and two-way diffusion is carried out after scrambling, in this way, the value of any one pixel can affect other pixel values as much as possible. The information entropy, histogram distribution, adjacent pixel correlation, plaintext sensitivity, key sensitivity and key space are analyzed. As show by simulation results, our algorithm has good encryption effect and high security.
{"title":"An Efficient Image Encryption Algorithm Based on Hyperchaotic and Two Times of Scrambling Diffusion","authors":"Wei Bu, Shu-cui Xie","doi":"10.1145/3573942.3574063","DOIUrl":"https://doi.org/10.1145/3573942.3574063","url":null,"abstract":"An image encryption algorithm based on hyperchaotic system is proposed to satisfy the security of image information transmission, the encryption algorithm consists of two times of scrambling and diffusion. The relation between plaintext and key is constructed to generate plaintext related chaotic sequence. Two rounds of scrambling operation are performed to make the pixel positions of the image are sufficiently scrambled, and two-way diffusion is carried out after scrambling, in this way, the value of any one pixel can affect other pixel values as much as possible. The information entropy, histogram distribution, adjacent pixel correlation, plaintext sensitivity, key sensitivity and key space are analyzed. As show by simulation results, our algorithm has good encryption effect and high security.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"37 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115988147","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}
Aiming at the problem of semantic lack of sentence similarity calculation in the process of metamorphosis test of neural machine translation system, an NMT sentence granularity similarity calculation method based on improved Cosine Distance is proposed. Text vectors are constructed through the improved TF-IDF weights, and the combination of Edit Distance and Jaccard similarity coefficient is used as a suppressor for cosine similarity. Experiments on neural machine translation systems such as Alibaba Translation and Baidu Translation on the UM-Corpus dataset show that, compared with the method based on Edit Distance, this method improves the Pearson correlation coefficient and Spearman correlation coefficient of the reference translation method by 20.5% and 12%, respectively. And this method is closer to the BLEU and METEOR evaluation results based on the reference translation, the evaluation accuracy is higher.
{"title":"NMT Sentence Granularity Similarity Calculation Method Based on Improved Cosine Distance","authors":"Shuyan Wang, Jingjing Ma","doi":"10.1145/3573942.3574021","DOIUrl":"https://doi.org/10.1145/3573942.3574021","url":null,"abstract":"Aiming at the problem of semantic lack of sentence similarity calculation in the process of metamorphosis test of neural machine translation system, an NMT sentence granularity similarity calculation method based on improved Cosine Distance is proposed. Text vectors are constructed through the improved TF-IDF weights, and the combination of Edit Distance and Jaccard similarity coefficient is used as a suppressor for cosine similarity. Experiments on neural machine translation systems such as Alibaba Translation and Baidu Translation on the UM-Corpus dataset show that, compared with the method based on Edit Distance, this method improves the Pearson correlation coefficient and Spearman correlation coefficient of the reference translation method by 20.5% and 12%, respectively. And this method is closer to the BLEU and METEOR evaluation results based on the reference translation, the evaluation accuracy is higher.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116016347","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}
Fei Wang, Long Chen, Xiaohua Huang, Cai Xu, Wei Zhao, Ziyu Guan, Guangyue Lu
Sentiment classification is a hot topic in the field of natural language processing. Currently, state-of-the-art classification models follow two steps: pre-training a large language model on upstream tasks, and then using human-labeled data to fine-tune a task-related model. However, there is a large gap between the upstream tasks of the pre-trained model and the downstream tasks being performed, resulting in the need for more labeled data to achieve excellent performance. Manually annotating data is expensive. In this paper, we propose a few-shot sentiment classification method based on Prompt and Contrastive Learning (PCL), which can significantly improve the performance of large-scale pre-trained language models in low-data and high-data regimes. Prompt learning aims to alleviate the gap between upstream and downstream tasks, and the contrastive learning is designed to capture the inter-class and intra-class distribution patterns of labeled data. Thanks to the integration of the two strategies, PCL markedly exceeds baselines with low resources. Extensive experiments on three datasets show that our method has outstanding performance in the few-shot settings.
{"title":"Prompt and Contrastive Learning for Few-shot Sentiment Classification","authors":"Fei Wang, Long Chen, Xiaohua Huang, Cai Xu, Wei Zhao, Ziyu Guan, Guangyue Lu","doi":"10.1145/3573942.3573969","DOIUrl":"https://doi.org/10.1145/3573942.3573969","url":null,"abstract":"Sentiment classification is a hot topic in the field of natural language processing. Currently, state-of-the-art classification models follow two steps: pre-training a large language model on upstream tasks, and then using human-labeled data to fine-tune a task-related model. However, there is a large gap between the upstream tasks of the pre-trained model and the downstream tasks being performed, resulting in the need for more labeled data to achieve excellent performance. Manually annotating data is expensive. In this paper, we propose a few-shot sentiment classification method based on Prompt and Contrastive Learning (PCL), which can significantly improve the performance of large-scale pre-trained language models in low-data and high-data regimes. Prompt learning aims to alleviate the gap between upstream and downstream tasks, and the contrastive learning is designed to capture the inter-class and intra-class distribution patterns of labeled data. Thanks to the integration of the two strategies, PCL markedly exceeds baselines with low resources. Extensive experiments on three datasets show that our method has outstanding performance in the few-shot settings.","PeriodicalId":103293,"journal":{"name":"Proceedings of the 2022 5th International Conference on Artificial Intelligence and Pattern Recognition","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117112159","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：