{"title":"Remote Sensing Image Segmentation based on Generative Adversarial Network with Wasserstein divergence","authors":"X. Cao, Chenggang Song, Jian Zhang, Chang Liu","doi":"10.1145/3446132.3446187","DOIUrl":"https://doi.org/10.1145/3446132.3446187","url":null,"abstract":"","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"48 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122017645","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 recent years, due to the high availability and convenience of the Internet, more and more information provides corresponding services through the Internet. As people are increasingly relying on the Internet, network security issues have become increasingly prominent, and a large number of malicious web pages have also emerged. How to achieve proactive and efficient detection of malicious web pages has become a research focus in the field of network security worldwide. This paper uses the Support Vector Machine algorithm to realize autonomous learning and build the classifier; chooses the TF-IDF method to process the data, and obtains the feature matrix of the collected URL data, which is stored in the sparse matrix after normalization and standardization. To avoid the existence of relatively strong features from affecting the classification results of the classifier, the K-Means method and TruncatedSVD method are used to reduce the dimension of the data features. The linear kernel function is used for large samples, and the Gaussian kernel function is used for small samples, so that the performance of the classifier is optimal. In the training process, the grid search method is used to obtain the optimal parameters forming a complete and mature detection system. And a ten-fold cross-validation method is used to test the correct rate, recall rate, accuracy rate and F1 value of the classifier. Finally the experimental result shows the malicious web page detection model has a good reference for big data processing.
{"title":"A Malicious Web Page Detection Model based on SVM Algorithm: Research on the Enhancement of SVM Efficiency by Multiple Machine Learning Algorithms","authors":"Jingbing Chen, Jie Yuan, Yuewei Li, Yiqi Zhang, Yufan Yang, Ruiqi Feng","doi":"10.1145/3446132.3446183","DOIUrl":"https://doi.org/10.1145/3446132.3446183","url":null,"abstract":"In recent years, due to the high availability and convenience of the Internet, more and more information provides corresponding services through the Internet. As people are increasingly relying on the Internet, network security issues have become increasingly prominent, and a large number of malicious web pages have also emerged. How to achieve proactive and efficient detection of malicious web pages has become a research focus in the field of network security worldwide. This paper uses the Support Vector Machine algorithm to realize autonomous learning and build the classifier; chooses the TF-IDF method to process the data, and obtains the feature matrix of the collected URL data, which is stored in the sparse matrix after normalization and standardization. To avoid the existence of relatively strong features from affecting the classification results of the classifier, the K-Means method and TruncatedSVD method are used to reduce the dimension of the data features. The linear kernel function is used for large samples, and the Gaussian kernel function is used for small samples, so that the performance of the classifier is optimal. In the training process, the grid search method is used to obtain the optimal parameters forming a complete and mature detection system. And a ten-fold cross-validation method is used to test the correct rate, recall rate, accuracy rate and F1 value of the classifier. Finally the experimental result shows the malicious web page detection model has a good reference for big data processing.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126697082","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}
Search is one of the most common platforms used to seek information. However, users mostly get overloaded with results whenever they use such a platform to resolve their queries. Nowadays, direct answers to queries are being provided as a part of the search experience. The question-answer (QA) retrieval process plays a significant role in enriching the search experience. Most off-the-shelf Semantic Textual Similarity models work fine for well-formed search queries, but their performances degrade when applied to a domain-specific setting having incomplete or grammatically ill-formed search queries in prevalence. In this paper, we discuss a framework for calculating similarities between a given input query and a set of predefined questions to retrieve the question which matches to it the most. We have used it for the financial domain, but the framework is generalized for any domain-specific search engine and can be used in other domains as well. We use Siamese network [6] over Long Short-Term Memory (LSTM) [3] models to train a classifier which generates un-normalized and normalized similarity scores for a given pair of questions. Moreover, for each of these question pairs, we calculate three other similarity scores: cosine similarity between their average word2vec embeddings [15], cosine similarity between their sentence embeddings [7] generated using RoBERTa [17] and their customized fuzzy-match score. Finally, we develop a meta-classifier using Support Vector Machines [19] for combining these five scores to detect if a given pair of questions is similar. We benchmark our model's performance against existing State Of The Art (SOTA) models on Quora Question Pairs (QQP) dataset1 as well as a dataset specific to the financial domain. After evaluating its performance on the financial domain specific data, we conclude that it not only outperforms several existing SOTA models on F1 score but also has decent accuracy.
{"title":"Applying Transfer Learning for Improving Domain-Specific Search Experience Using Query to Question Similarity","authors":"Ankush Chopra, S. Agrawal, Sohom Ghosh","doi":"10.1145/3446132.3446403","DOIUrl":"https://doi.org/10.1145/3446132.3446403","url":null,"abstract":"Search is one of the most common platforms used to seek information. However, users mostly get overloaded with results whenever they use such a platform to resolve their queries. Nowadays, direct answers to queries are being provided as a part of the search experience. The question-answer (QA) retrieval process plays a significant role in enriching the search experience. Most off-the-shelf Semantic Textual Similarity models work fine for well-formed search queries, but their performances degrade when applied to a domain-specific setting having incomplete or grammatically ill-formed search queries in prevalence. In this paper, we discuss a framework for calculating similarities between a given input query and a set of predefined questions to retrieve the question which matches to it the most. We have used it for the financial domain, but the framework is generalized for any domain-specific search engine and can be used in other domains as well. We use Siamese network [6] over Long Short-Term Memory (LSTM) [3] models to train a classifier which generates un-normalized and normalized similarity scores for a given pair of questions. Moreover, for each of these question pairs, we calculate three other similarity scores: cosine similarity between their average word2vec embeddings [15], cosine similarity between their sentence embeddings [7] generated using RoBERTa [17] and their customized fuzzy-match score. Finally, we develop a meta-classifier using Support Vector Machines [19] for combining these five scores to detect if a given pair of questions is similar. We benchmark our model's performance against existing State Of The Art (SOTA) models on Quora Question Pairs (QQP) dataset1 as well as a dataset specific to the financial domain. After evaluating its performance on the financial domain specific data, we conclude that it not only outperforms several existing SOTA models on F1 score but also has decent accuracy.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126494749","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}
Dian Ira Putri Hutasoit, Damayanti Elisabeth, D. I. Sensuse
Globalizations makes world become smaller. It is easier to interact, communicate, connect and create business with another people in over the world. This opportunity is frequently optimized by business stakeholders to build multinational corporations, corporations that have products and branches in other countries. Culture, language and geographically differences remain and become challenges to manage knowledge over the headquarter and branches, including how to share knowledge. They influence how multinational corporations should optimize and use technology for better knowledge sharing process so they can keep benefit advantages, can compete to their competitors, or even keep their business runs well. This paper will review, compare and analyze mechanism and technology usage of knowledge sharing for multinational corporations. The sample of multinational corporations used is multinational corporations from many different countries with different area or industry. This research will result most of the mechanism and technology used to support and facilitate the knowledge share mechanism in multinational corporations.
{"title":"Knowledge Sharing for Multinational Corporations using Technology: A Systematic Literature Review","authors":"Dian Ira Putri Hutasoit, Damayanti Elisabeth, D. I. Sensuse","doi":"10.1145/3446132.3446418","DOIUrl":"https://doi.org/10.1145/3446132.3446418","url":null,"abstract":"Globalizations makes world become smaller. It is easier to interact, communicate, connect and create business with another people in over the world. This opportunity is frequently optimized by business stakeholders to build multinational corporations, corporations that have products and branches in other countries. Culture, language and geographically differences remain and become challenges to manage knowledge over the headquarter and branches, including how to share knowledge. They influence how multinational corporations should optimize and use technology for better knowledge sharing process so they can keep benefit advantages, can compete to their competitors, or even keep their business runs well. This paper will review, compare and analyze mechanism and technology usage of knowledge sharing for multinational corporations. The sample of multinational corporations used is multinational corporations from many different countries with different area or industry. This research will result most of the mechanism and technology used to support and facilitate the knowledge share mechanism in multinational corporations.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130639013","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}
Many anomaly detection methods are unsupervised e.g. they only utilize the non-anomalous data for model training. Data points that deviate from the majority of the pattern are deemed as anomalies. However, in many cases, anomaly labels are available which can help to guide the model learning for anomaly detection. We introduce an end-to-end anomaly score learning model composed of an autoencoder and a scoring neural network, which assigns an anomaly score to a given data point according to its level of abnormality. We jointly optimize the reconstruction loss and anomaly score function in an end-to-end manner. Experimental results on multiple datasets show that the proposed method appears to be superior over many state-of-the-art methods.
{"title":"DAS: Deep Autoencoder with Scoring Neural Network for Anomaly Detection","authors":"Pan Luo, Chenbo Qiu, Yuhao Wang","doi":"10.1145/3446132.3446181","DOIUrl":"https://doi.org/10.1145/3446132.3446181","url":null,"abstract":"Many anomaly detection methods are unsupervised e.g. they only utilize the non-anomalous data for model training. Data points that deviate from the majority of the pattern are deemed as anomalies. However, in many cases, anomaly labels are available which can help to guide the model learning for anomaly detection. We introduce an end-to-end anomaly score learning model composed of an autoencoder and a scoring neural network, which assigns an anomaly score to a given data point according to its level of abnormality. We jointly optimize the reconstruction loss and anomaly score function in an end-to-end manner. Experimental results on multiple datasets show that the proposed method appears to be superior over many state-of-the-art methods.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"18 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130367208","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}
Despite the remarkable success of Neural Machine Translation system, such challenges as its drawback in low-resourced conditions persist. In recent years, working mechanism of exploiting either one or both source and target side monolingual data within the Neural Machine Translation framework gained much attention in the field. Among many supervised and unsupervised proposals, back translation is increasingly seen as one of the most promising methods to improve low-resource NMT performance. Regardless of its simplicity, the effectiveness of back translation is highly dependent on performance of the backward model which is initially trained on available parallel data. To address the dilemma of back translation practices in low resource scenarios, we propose to employ target-side monolingual data to improve both backward and forward models by step-wise adoption of self-learning and back translation, which we refer to as Bidirectional Boost.Our experiments on a Tibetan-Chinese translation task attested the proposed approach with a result of producing 3.1 and 8.2 BLEU scores, respectively, both on forward and backward models over vanilla Transformers trained on genuine parallel data under supervised settings.
{"title":"Bidirectional Boost: On Improving Tibetan-Chinese Neural Machine Translation With Back-Translation and Self-Learning","authors":"Sangjie Duanzhu, Rui Zhang, Cairang Jia","doi":"10.1145/3446132.3446405","DOIUrl":"https://doi.org/10.1145/3446132.3446405","url":null,"abstract":"Despite the remarkable success of Neural Machine Translation system, such challenges as its drawback in low-resourced conditions persist. In recent years, working mechanism of exploiting either one or both source and target side monolingual data within the Neural Machine Translation framework gained much attention in the field. Among many supervised and unsupervised proposals, back translation is increasingly seen as one of the most promising methods to improve low-resource NMT performance. Regardless of its simplicity, the effectiveness of back translation is highly dependent on performance of the backward model which is initially trained on available parallel data. To address the dilemma of back translation practices in low resource scenarios, we propose to employ target-side monolingual data to improve both backward and forward models by step-wise adoption of self-learning and back translation, which we refer to as Bidirectional Boost.Our experiments on a Tibetan-Chinese translation task attested the proposed approach with a result of producing 3.1 and 8.2 BLEU scores, respectively, both on forward and backward models over vanilla Transformers trained on genuine parallel data under supervised settings.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133757549","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}
Recently, the pandemic caused by COVID-19 is severe in the entire world. The prevention and control of crimes associated with COVID-19 are critical for controlling the pandemic. Therefore, to provide efficient and convenient intelligent legal knowledge services during the pandemic, in this paper, we develop an intelligent system for answering legal questions on the WeChat platform. The data source we used for training our system is “The typical cases of national procuratorial authorities handling crimes against the prevention and control of the new coronary pneumonia pandemic following the law”, which is published online by the Supreme People’s Procuratorate of the People’s Republic of China. We base our system on BERT (a well-known pre-trained language model) and use the shared attention mechanism to capture the text information further. Then we train a model to minimise the contrastive loss. Finally, the system uses the trained model to identify the information entered by a user, and accordingly responds to the user with a reference case similar to the query case and give the reference legal gist applicable to the query case.
{"title":"Few-Shot Legal Knowledge Question Answering System for COVID-19 Epidemic","authors":"Jiaye Wu, Jie Liu, Xudong Luo","doi":"10.1145/3446132.3446409","DOIUrl":"https://doi.org/10.1145/3446132.3446409","url":null,"abstract":"Recently, the pandemic caused by COVID-19 is severe in the entire world. The prevention and control of crimes associated with COVID-19 are critical for controlling the pandemic. Therefore, to provide efficient and convenient intelligent legal knowledge services during the pandemic, in this paper, we develop an intelligent system for answering legal questions on the WeChat platform. The data source we used for training our system is “The typical cases of national procuratorial authorities handling crimes against the prevention and control of the new coronary pneumonia pandemic following the law”, which is published online by the Supreme People’s Procuratorate of the People’s Republic of China. We base our system on BERT (a well-known pre-trained language model) and use the shared attention mechanism to capture the text information further. Then we train a model to minimise the contrastive loss. Finally, the system uses the trained model to identify the information entered by a user, and accordingly responds to the user with a reference case similar to the query case and give the reference legal gist applicable to the query case.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"106 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134004614","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}
Guangwu Cui, R. Shen, Yingfeng Chen, Juan Zou, Shengxiang Yang, Changjie Fan, Jinghua Zheng
In the field of game designing, artificial intelligence is used to generate responsive, adaptive, or intelligent behaviors primarily in Non-Player-Characters (NPCs). There is a large demand for controlling game AI since a variety of players expect to be provided NPC opponents with appropriate difficulties to improve their game experience. However, to the best of our knowledge, a few works are focusing on this problem. In this paper, we firstly present a Reinforced Evolutionary Algorithm based on the Difficulty-Difference objective (REA-DD) to the DLAI problem, which combines reinforcement learning and evolutionary algorithms. REA-DD is able to generate the desired difficulty level of game AI accurately. Nonetheless, REA can only obtain a kind of game AI in each run. To improve efficiency, another algorithm based on Multi-objective Optimization is proposed, regarded as RMOEA-DD, which obtains DLAI after one run. Experiments on the game Pong from ALE and apply on a commercial game named The Ghost Story to show that our algorithms provide valid methods to the DLAI problem both in the term of controlling accuracy and efficiency.
{"title":"Reinforced Evolutionary Algorithms for Game Difficulty Control","authors":"Guangwu Cui, R. Shen, Yingfeng Chen, Juan Zou, Shengxiang Yang, Changjie Fan, Jinghua Zheng","doi":"10.1145/3446132.3446165","DOIUrl":"https://doi.org/10.1145/3446132.3446165","url":null,"abstract":"In the field of game designing, artificial intelligence is used to generate responsive, adaptive, or intelligent behaviors primarily in Non-Player-Characters (NPCs). There is a large demand for controlling game AI since a variety of players expect to be provided NPC opponents with appropriate difficulties to improve their game experience. However, to the best of our knowledge, a few works are focusing on this problem. In this paper, we firstly present a Reinforced Evolutionary Algorithm based on the Difficulty-Difference objective (REA-DD) to the DLAI problem, which combines reinforcement learning and evolutionary algorithms. REA-DD is able to generate the desired difficulty level of game AI accurately. Nonetheless, REA can only obtain a kind of game AI in each run. To improve efficiency, another algorithm based on Multi-objective Optimization is proposed, regarded as RMOEA-DD, which obtains DLAI after one run. Experiments on the game Pong from ALE and apply on a commercial game named The Ghost Story to show that our algorithms provide valid methods to the DLAI problem both in the term of controlling accuracy and efficiency.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134383597","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}
Communities are clusters of closely connected nodes in a social network. Detecting community structure can help us understand their network characteristics. Most popular algorithms are based on modularity optimization, such as the SLM algorithm. The SLM algorithm can detect non-overlapping communities. However, communities in real-world networks also overlap as nodes maybe belong to multiple clusters. Some models such as the BIGCLAM algorithm can be used to discover the overlapping community structure, but it has some problems in running community detection algorithms on large-scale networks. In this paper, we present a novel gravitation-based algorithm (GBA) which is inspired by the theory of galaxy evolution. The GBA algorithm is based on Newton's law of universal gravitation to simulate the process of community evolution. It includes the AGB algorithm and the AFMG algorithm. The AGB algorithm is used to detect community structure and find the center community. The AFMG algorithm is used to find the max gravity of the community. The experimental results show that our algorithm can detect overlapping communities in large-scale networks of tens millions of nodes, uncover good partitions of networks and are faster than compared methods by two to three orders of magnitude.
{"title":"The Gravitation-based Algorithm for Community Detecting in Large-scale Social Networks","authors":"Ming-Ray Liao, Yuanyuan Liang, Rui Wang","doi":"10.1145/3446132.3446185","DOIUrl":"https://doi.org/10.1145/3446132.3446185","url":null,"abstract":"Communities are clusters of closely connected nodes in a social network. Detecting community structure can help us understand their network characteristics. Most popular algorithms are based on modularity optimization, such as the SLM algorithm. The SLM algorithm can detect non-overlapping communities. However, communities in real-world networks also overlap as nodes maybe belong to multiple clusters. Some models such as the BIGCLAM algorithm can be used to discover the overlapping community structure, but it has some problems in running community detection algorithms on large-scale networks. In this paper, we present a novel gravitation-based algorithm (GBA) which is inspired by the theory of galaxy evolution. The GBA algorithm is based on Newton's law of universal gravitation to simulate the process of community evolution. It includes the AGB algorithm and the AFMG algorithm. The AGB algorithm is used to detect community structure and find the center community. The AFMG algorithm is used to find the max gravity of the community. The experimental results show that our algorithm can detect overlapping communities in large-scale networks of tens millions of nodes, uncover good partitions of networks and are faster than compared methods by two to three orders of magnitude.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132466401","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 view of the lack of labeled corpus and the difficulty of extracting multiple relations in the extraction of entity relation in the geographic domain, a method based on probabilistic graph is proposed for extracting the entity relation in the geographic domain. This method uses the semantic information in the knowledge base to enhance the representation of the geographic domain corpus to alleviate the problem of insufficient labeled corpus. It uses character-word hybrid vectors that can be effectively integrated into the semantic information as the feature vectors. The vectors are transmitted to Bi-LSTM and self-attention for global deep feature extraction. Finally drawing on the idea of probabilistic graph, the "semi pointer-semi annotation" method is utilized to extract the head entities, traverse the head entity, and then uses the same method to extract tail entities and relations. By comparing the experimental results on the geographic domain corpus and ACE05 corpus with other advanced methods, the probabilistic graph-based extraction method effectively improves the geographic domain entity relation extraction effect.
{"title":"Probabilistic Graph Attention for Relation Extraction for Domain of Geography","authors":"Jiaorou Yin, P. Duan, Weitao Huang, Shengwu Xiong","doi":"10.1145/3446132.3446411","DOIUrl":"https://doi.org/10.1145/3446132.3446411","url":null,"abstract":"In view of the lack of labeled corpus and the difficulty of extracting multiple relations in the extraction of entity relation in the geographic domain, a method based on probabilistic graph is proposed for extracting the entity relation in the geographic domain. This method uses the semantic information in the knowledge base to enhance the representation of the geographic domain corpus to alleviate the problem of insufficient labeled corpus. It uses character-word hybrid vectors that can be effectively integrated into the semantic information as the feature vectors. The vectors are transmitted to Bi-LSTM and self-attention for global deep feature extraction. Finally drawing on the idea of probabilistic graph, the \"semi pointer-semi annotation\" method is utilized to extract the head entities, traverse the head entity, and then uses the same method to extract tail entities and relations. By comparing the experimental results on the geographic domain corpus and ACE05 corpus with other advanced methods, the probabilistic graph-based extraction method effectively improves the geographic domain entity relation extraction effect.","PeriodicalId":125388,"journal":{"name":"Proceedings of the 2020 3rd International Conference on Algorithms, Computing and Artificial Intelligence","volume":"47 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-12-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122926059","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：