Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00012
Yuxun Lu, R. Ichise
Knowledge graph completion (KGC) models aim to provide a feasible way of manipulating facts in knowledge graphs. Most KGC models do not consider type constraint in relations due to the scarcity of type information in training data. We proposed an unsupervised method for inferring type constraint based on existing bilinear KGC models. Our method induces two type indicators into every relation and adjusts the location of entity embeddings in feature space to match the type indicators. Our approach eliminates the external feature space for entity types and type constraints in relations and has a consistent feature space; therefore, it has fewer parameters than other methods. Experiments show that our methods can improve the performance of the base models and outperform other methods on datasets about general knowledge.
{"title":"Unsupervised Type Constraint Inference in Bilinear Knowledge Graph Completion Models","authors":"Yuxun Lu, R. Ichise","doi":"10.1109/ICKG52313.2021.00012","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00012","url":null,"abstract":"Knowledge graph completion (KGC) models aim to provide a feasible way of manipulating facts in knowledge graphs. Most KGC models do not consider type constraint in relations due to the scarcity of type information in training data. We proposed an unsupervised method for inferring type constraint based on existing bilinear KGC models. Our method induces two type indicators into every relation and adjusts the location of entity embeddings in feature space to match the type indicators. Our approach eliminates the external feature space for entity types and type constraints in relations and has a consistent feature space; therefore, it has fewer parameters than other methods. Experiments show that our methods can improve the performance of the base models and outperform other methods on datasets about general knowledge.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"24 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123059353","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00062
Wenjun Wu, Lingling Zhang, Yiwei Chen, Xuan Luo, Bifan Wei, Jun Liu
The clustering technique plays an important role in data mining and machine learning fields. Clustering for high-dimensional data, such as texts, images, and videos, remains a challenging task due to the existence of many noise features. The widely used methods for this issue focus on mining a effective pattern in high-dimensional data using some dimensionality reduction techniques before clustering. This strategy slightly mitigates the effects of irrelevant and redundant features, but cannot significantly improve the clustering performance because the captured pattern by dimensionality reduction is not directly related to the clustering task. In this paper, we propose a unified framework to achieve discriminative dimensionality reduction and fuzzy clustering for high-dimensional data simultaneously. The proposed framework not only utilizes the clustering results to directly guide or supervise the process of discriminative dimensionality reduction, but also controls the clustering fuzziness more easily by a $F$ -norm regularization term. An efficient optimization algorithm is exploited to address the objective function of our method, which is proved to converge to the local optimal solution in theory. We evaluate the proposed method on three large-scale fine-grained image datasets, including Birds, Flowers, and Cars, for clustering and retrieval two tasks. The experimental results on metrics ACC, NMI, ARI and Recall@K indicate that our method achieves the comparable performance over the state-of-the-art methods.
{"title":"Fuzzy c-Means Clustering with Discriminative Projection","authors":"Wenjun Wu, Lingling Zhang, Yiwei Chen, Xuan Luo, Bifan Wei, Jun Liu","doi":"10.1109/ICKG52313.2021.00062","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00062","url":null,"abstract":"The clustering technique plays an important role in data mining and machine learning fields. Clustering for high-dimensional data, such as texts, images, and videos, remains a challenging task due to the existence of many noise features. The widely used methods for this issue focus on mining a effective pattern in high-dimensional data using some dimensionality reduction techniques before clustering. This strategy slightly mitigates the effects of irrelevant and redundant features, but cannot significantly improve the clustering performance because the captured pattern by dimensionality reduction is not directly related to the clustering task. In this paper, we propose a unified framework to achieve discriminative dimensionality reduction and fuzzy clustering for high-dimensional data simultaneously. The proposed framework not only utilizes the clustering results to directly guide or supervise the process of discriminative dimensionality reduction, but also controls the clustering fuzziness more easily by a $F$ -norm regularization term. An efficient optimization algorithm is exploited to address the objective function of our method, which is proved to converge to the local optimal solution in theory. We evaluate the proposed method on three large-scale fine-grained image datasets, including Birds, Flowers, and Cars, for clustering and retrieval two tasks. The experimental results on metrics ACC, NMI, ARI and Recall@K indicate that our method achieves the comparable performance over the state-of-the-art methods.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"31 4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122445511","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00045
Yin Wang, Nan Xia, Xiangfeng Luo, Jinhui Li
Chinese financial event detection has a great significance in the application of financial risk analysis, en-terprise management and decision-making. The existing tasks of Chinese event detection are mainly regarded as character-based or word-based classification, which suffers from the ambiguity of trigger words. These tasks only concentrate on local information (e.g character and word), which loses sight of global information like sentence semantics. Furthermore, in the finance field, there exists the problem of fuzzy boundary between different event types. In this paper, we propose a global semantics with boundary constraint knowledge graph (BCKG) for Chinese financial event detection, which considers both sentence semantics and boundary knowledge. At first, Chinese financial dataset (CFD) is constructed by considering the complexity in financial area. And then, the sentence seman-tics embedding is obtained by pre-training BERT fine-tuning mechanism to address the problem of ambiguity of trigger words, which considers both syntactic information and context sentence semantics comprehensively. Finally, we construct the BCKG for financial event, which can add additional prior knowledge to solve fuzzy boundary problem. The proposed method for event detection achieves outstanding performance on standard ACE 2005 Chinese dataset and constructed CFD. The experimental results demonstrate the effectiveness of the proposed method.
{"title":"Global Semantics with Boundary Constraint Knowledge Graph for Chinese Financial Event Detection","authors":"Yin Wang, Nan Xia, Xiangfeng Luo, Jinhui Li","doi":"10.1109/ICKG52313.2021.00045","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00045","url":null,"abstract":"Chinese financial event detection has a great significance in the application of financial risk analysis, en-terprise management and decision-making. The existing tasks of Chinese event detection are mainly regarded as character-based or word-based classification, which suffers from the ambiguity of trigger words. These tasks only concentrate on local information (e.g character and word), which loses sight of global information like sentence semantics. Furthermore, in the finance field, there exists the problem of fuzzy boundary between different event types. In this paper, we propose a global semantics with boundary constraint knowledge graph (BCKG) for Chinese financial event detection, which considers both sentence semantics and boundary knowledge. At first, Chinese financial dataset (CFD) is constructed by considering the complexity in financial area. And then, the sentence seman-tics embedding is obtained by pre-training BERT fine-tuning mechanism to address the problem of ambiguity of trigger words, which considers both syntactic information and context sentence semantics comprehensively. Finally, we construct the BCKG for financial event, which can add additional prior knowledge to solve fuzzy boundary problem. The proposed method for event detection achieves outstanding performance on standard ACE 2005 Chinese dataset and constructed CFD. The experimental results demonstrate the effectiveness of the proposed method.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"3 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115069927","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00067
Panpan Lei, Xiao Sun
Conversational Question Generation (CQG) is a new concern in Question Generation (QG) study. Recently Seq2Seq neural network model has been widely used in the QG area. CQG model is also based on the Seq2Seq neural network model. We note a problem: the CQG model's input is not a single sentence, but a long text and conversation history. Seq2Seq model can't effectively process long input, the model will generate questions not related to the answer. To solve this problem, we propose an answer-centric local and global information fusion model. We extract the evidence sentence containing the answer in the passage and encode the evidence sentence and the passage information separately. On the one hand, we add answer-centered position tags in the passage to reinforce the attention of information related to the answer. On the other hand, we put the key sentence into the question type prediction model. By combining the answer position embedding to predict the question type, and then put the predicted question types in the key sentence to guide the generation of the question. Finally, we use a gate mechanism to merge key sentence information and passage information. The experimental results show that we have achieved better results.
{"title":"Answer-Centric Local and Global Information Fusion for Conversational Question Generation","authors":"Panpan Lei, Xiao Sun","doi":"10.1109/ICKG52313.2021.00067","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00067","url":null,"abstract":"Conversational Question Generation (CQG) is a new concern in Question Generation (QG) study. Recently Seq2Seq neural network model has been widely used in the QG area. CQG model is also based on the Seq2Seq neural network model. We note a problem: the CQG model's input is not a single sentence, but a long text and conversation history. Seq2Seq model can't effectively process long input, the model will generate questions not related to the answer. To solve this problem, we propose an answer-centric local and global information fusion model. We extract the evidence sentence containing the answer in the passage and encode the evidence sentence and the passage information separately. On the one hand, we add answer-centered position tags in the passage to reinforce the attention of information related to the answer. On the other hand, we put the key sentence into the question type prediction model. By combining the answer position embedding to predict the question type, and then put the predicted question types in the key sentence to guide the generation of the question. Finally, we use a gate mechanism to merge key sentence information and passage information. The experimental results show that we have achieved better results.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124959532","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00053
Xiao Zhang, Heqi Zheng, Heyan Huang, Zewen Chi, Xian-Ling Mao
Chinese GaoKao Reading Comprehension is a chal-lenging NLP task. It requires strong logical reasoning ability to capture deep semantic relations between the questions and answers. However, most traditional models cannot learn sufficient inference ability, because of the scarcity of Chinese GaoKao reading comprehension data. Intuitively, there are two methods to improve the reading comprehension ability for Chinese GaoKao reading comprehension task. 1). Increase the scale of data. 2). Introduce additional related knowledge. In this paper, we propose a novel method based on adversarial training and knowledge distillation, which can be trained in other knowledge-rich datasets and transferred to the Chinese GaoKao reading comprehension task. Extensive experiments show that our proposed model performs better than the state-of-the-art baselines. The code and the relevant dataset will be publicly avaible.
{"title":"A Knowledge Enhanced Chinese GaoKao Reading Comprehension Method","authors":"Xiao Zhang, Heqi Zheng, Heyan Huang, Zewen Chi, Xian-Ling Mao","doi":"10.1109/ICKG52313.2021.00053","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00053","url":null,"abstract":"Chinese GaoKao Reading Comprehension is a chal-lenging NLP task. It requires strong logical reasoning ability to capture deep semantic relations between the questions and answers. However, most traditional models cannot learn sufficient inference ability, because of the scarcity of Chinese GaoKao reading comprehension data. Intuitively, there are two methods to improve the reading comprehension ability for Chinese GaoKao reading comprehension task. 1). Increase the scale of data. 2). Introduce additional related knowledge. In this paper, we propose a novel method based on adversarial training and knowledge distillation, which can be trained in other knowledge-rich datasets and transferred to the Chinese GaoKao reading comprehension task. Extensive experiments show that our proposed model performs better than the state-of-the-art baselines. The code and the relevant dataset will be publicly avaible.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"6 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125878522","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00042
Riddho Ridwanul Haque, Chowdhury Farhan Ahmed, M. Samiullah, C. Leung
Large graph data repositories are becoming in-creasingly common. Identifying frequently appearing subgraph patterns in such databases can reveal useful information, and such patterns have been used for a variety of applications. Im-perfections and stochasticity are often unavoidable in real-world graph data, and the existence of edges in the graphs within such databases is often uncertain. Taking this uncertainty into account while mining frequent patterns poses considerable computational challenges. However, doing so is crucial for accurately mining relevant patterns. Existing frequent subgraph mining approaches that consider uncertainty rely on approximation schemes, and are both inefficient and inaccurate. In this paper, we present UFreS, an exact algorithm for mining frequent subgraph patterns from uncertain graph databases. We also introduce Edge-Embedding graphs, the first data structure designed to efficiently and exactly infer the expected support of a subgraph pattern in an uncer-tain graph. Experimental evaluations conducted on real-world datasets show that UFreS is efficient, scalable, and outperforms the existing approaches in terms of runtime, memory usage and accuracy.
{"title":"UFreS: A New Technique for Discovering Frequent Subgraph Patterns in Uncertain Graph Databases","authors":"Riddho Ridwanul Haque, Chowdhury Farhan Ahmed, M. Samiullah, C. Leung","doi":"10.1109/ICKG52313.2021.00042","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00042","url":null,"abstract":"Large graph data repositories are becoming in-creasingly common. Identifying frequently appearing subgraph patterns in such databases can reveal useful information, and such patterns have been used for a variety of applications. Im-perfections and stochasticity are often unavoidable in real-world graph data, and the existence of edges in the graphs within such databases is often uncertain. Taking this uncertainty into account while mining frequent patterns poses considerable computational challenges. However, doing so is crucial for accurately mining relevant patterns. Existing frequent subgraph mining approaches that consider uncertainty rely on approximation schemes, and are both inefficient and inaccurate. In this paper, we present UFreS, an exact algorithm for mining frequent subgraph patterns from uncertain graph databases. We also introduce Edge-Embedding graphs, the first data structure designed to efficiently and exactly infer the expected support of a subgraph pattern in an uncer-tain graph. Experimental evaluations conducted on real-world datasets show that UFreS is efficient, scalable, and outperforms the existing approaches in terms of runtime, memory usage and accuracy.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"85 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126022676","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00060
Hongxuan Li, Bifan Wei, Jun Liu, Zhaotong Guo, Jingchao Qi, Bei Wu, Yong Liu, Yuanyuan Shi
Mining the facets of topics is an essential task for information retrieval, information extraction and knowledge base construction. For the topics in courses, there are three challenges: different topics have different facet, the labels of facets rarely appear in the topic description text and not all topics have enough textural information to mine facets. In this paper we propose a weakly-supervised algorithm for topic-specific facet mining (ToFM for short) based on our finding that similar topics in a cluster have similar facet sets. For example, topics Binary Search Tree, Suffix Tree and AVL tree in Tree cluster have example, insertion, deletion, traversal and other similar facets. ToFM first splits topics in a domain into several topic clusters based on the topic description text. Then ToFM extracts initial facet sets for all topics from the corresponding Wikipedia article pages. Finally, ToFM performs a normalized facet propagation within each topic cluster to acquire final facet sets of every topic. We evaluate the performance of ToFM on six real-world datasets and experimental results show that ToFM achieves better performance than the existing facet mining algorithms.
{"title":"ToFM: Topic-specific Facet Mining by Facet Propagation within Clusters","authors":"Hongxuan Li, Bifan Wei, Jun Liu, Zhaotong Guo, Jingchao Qi, Bei Wu, Yong Liu, Yuanyuan Shi","doi":"10.1109/ICKG52313.2021.00060","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00060","url":null,"abstract":"Mining the facets of topics is an essential task for information retrieval, information extraction and knowledge base construction. For the topics in courses, there are three challenges: different topics have different facet, the labels of facets rarely appear in the topic description text and not all topics have enough textural information to mine facets. In this paper we propose a weakly-supervised algorithm for topic-specific facet mining (ToFM for short) based on our finding that similar topics in a cluster have similar facet sets. For example, topics Binary Search Tree, Suffix Tree and AVL tree in Tree cluster have example, insertion, deletion, traversal and other similar facets. ToFM first splits topics in a domain into several topic clusters based on the topic description text. Then ToFM extracts initial facet sets for all topics from the corresponding Wikipedia article pages. Finally, ToFM performs a normalized facet propagation within each topic cluster to acquire final facet sets of every topic. We evaluate the performance of ToFM on six real-world datasets and experimental results show that ToFM achieves better performance than the existing facet mining algorithms.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"70 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132840288","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00071
Y. He, Xian-Ling Mao, Wei Wei, Heyan Huang
Traditional Query Suggestion (TQS) aims to retrieve or generate completed queries given input keywords and query logs, which plays a vital role in information retrieval. Nearly all existing TQS methods obtain suggested queries, which are usually in the form of keywords or phrases. However, queries like keywords or phrases suffer from incomplete or ambiguous se-mantics. Ideally, question-formed queries are more intuitive and closer to the information needs of users, which can improve their satisfaction during a search. Motivated by this idea, thus, this paper defines a novel question-formed query suggestion task that generates question-formed queries given input keywords and web page texts. Moreover, we also propose a novel pipeline method for this novel task. Specifically, a query generation module is first employed to generate related question-formed queries given keywords and web page texts. Then, a selection module selects the most representative tops among all generated queries as the final suggestion. Extensive experiments demonstrate that our method outperforms the state-of-the-art baselines in human evaluation.
{"title":"Question-formed Query Suggestion","authors":"Y. He, Xian-Ling Mao, Wei Wei, Heyan Huang","doi":"10.1109/ICKG52313.2021.00071","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00071","url":null,"abstract":"Traditional Query Suggestion (TQS) aims to retrieve or generate completed queries given input keywords and query logs, which plays a vital role in information retrieval. Nearly all existing TQS methods obtain suggested queries, which are usually in the form of keywords or phrases. However, queries like keywords or phrases suffer from incomplete or ambiguous se-mantics. Ideally, question-formed queries are more intuitive and closer to the information needs of users, which can improve their satisfaction during a search. Motivated by this idea, thus, this paper defines a novel question-formed query suggestion task that generates question-formed queries given input keywords and web page texts. Moreover, we also propose a novel pipeline method for this novel task. Specifically, a query generation module is first employed to generate related question-formed queries given keywords and web page texts. Then, a selection module selects the most representative tops among all generated queries as the final suggestion. Extensive experiments demonstrate that our method outperforms the state-of-the-art baselines in human evaluation.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133573391","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}
Pub Date : 2021-12-01DOI: 10.1109/ICKG52313.2021.00019
Motaher Hossain, Youxi Wu, Philippe Fournier-Viger, Zhao Li, Lei Guo, Yan Li
Sequential pattern mining (SPM) under the nonoverlapping condition (or nonoverlapping SPM) is a type of data mining used to extract frequent gapped subsequences (known as patterns) from sequences, which is more valuable and versatile than other related methods. In nonoverlapping SPM, two occurrences cannot reuse the same sequence letter in the exact location as the occurrences. This method evaluates the frequency of the patterns in the sequence, and ignores the impact of external utility (item price or profit). Therefore, some low-frequency and essential patterns are overlooked. To address this issue, this paper introduces High Utility Self-adaptive Nonoverlapping Pattern (HSNP) mining and proposes HSNP-Miner, which includes two steps: support calculation and candi-date pattern generation. To calculate the support, we propose the NoSup algorithm, which can effectively calculate support while avoiding the creation of redundant nodes. An advanced upper bound method is employed to generate the candidate patterns more efficiently. Compared to other competitive methods, the experimental results demonstrate the efficiency of the proposed algorithm and the uniqueness of nonoverlapping sequence pat-tarns.
非重叠条件下的顺序模式挖掘(SPM)是一种用于从序列中提取频繁间隙子序列(称为模式)的数据挖掘方法，它比其他相关方法更有价值和通用性。在非重叠SPM中，两个序列不能在相同的位置重复使用相同的序列字母。该方法评估序列中模式的频率，并忽略外部效用(项目价格或利润)的影响。因此，忽略了一些低频和基本模式。为了解决这一问题，本文引入了HSNP (High Utility Self-adaptive non - overlap Pattern)挖掘方法，并提出了HSNP- miner算法，该算法包括支持度计算和候选数据模式生成两个步骤。为了计算支持度，我们提出了NoSup算法，该算法可以有效地计算支持度，同时避免冗余节点的产生。采用一种先进的上界方法，更有效地生成候选模式。与其他竞争方法相比，实验结果证明了该算法的有效性和非重叠序列模式的唯一性。
{"title":"HSNP-Miner: High Utility Self-Adaptive Nonoverlapping Pattern Mining","authors":"Motaher Hossain, Youxi Wu, Philippe Fournier-Viger, Zhao Li, Lei Guo, Yan Li","doi":"10.1109/ICKG52313.2021.00019","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00019","url":null,"abstract":"Sequential pattern mining (SPM) under the nonoverlapping condition (or nonoverlapping SPM) is a type of data mining used to extract frequent gapped subsequences (known as patterns) from sequences, which is more valuable and versatile than other related methods. In nonoverlapping SPM, two occurrences cannot reuse the same sequence letter in the exact location as the occurrences. This method evaluates the frequency of the patterns in the sequence, and ignores the impact of external utility (item price or profit). Therefore, some low-frequency and essential patterns are overlooked. To address this issue, this paper introduces High Utility Self-adaptive Nonoverlapping Pattern (HSNP) mining and proposes HSNP-Miner, which includes two steps: support calculation and candi-date pattern generation. To calculate the support, we propose the NoSup algorithm, which can effectively calculate support while avoiding the creation of redundant nodes. An advanced upper bound method is employed to generate the candidate patterns more efficiently. Compared to other competitive methods, the experimental results demonstrate the efficiency of the proposed algorithm and the uniqueness of nonoverlapping sequence pat-tarns.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132428957","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 the online education system, predicting the knowledge of exercises is a fundamental task of many applications, such as cognitive diagnosis. Usually, experts consider this problem as Hierarchical Multi-label Classification (HMC), since the knowledge concepts exhibit a multi-level structure. However, existing methods either sacrificed knowledge consistency for classification accuracy or sacrificed classification accuracy for knowledge consistency. Maintaining the balance is difficult. To forgo this dilemma, in this paper, we develop a novel frame-work called Consistency-Aware Multi-modal Network (Cam-Net). Specifically, we develop a multi-modal embedding module to learn the representation of the multi-modal exercise. Then, we adopt a hybrid prediction method consisting of the flat prediction module and the local prediction module. The local prediction module deals with the relation between the knowledge hierarchy and the input exercise. The flat prediction module focuses on maintaining knowledge consistency. Finally, to balance classification accuracy and knowledge consistency, we combine the outputs of two modules to make a final prediction. Extensive experimental results on two real-world datasets demonstrate the high performance and the ability to reduce knowledge inconsistency of CamNet.
{"title":"Consistency-aware Multi-modal Network for Hierarchical Multi-label Classification in Online Education System","authors":"Siqi Lei, Wei Huang, Shiwei Tong, Qi Liu, Zhenya Huang, Enhong Chen, Yu Su","doi":"10.1109/ICKG52313.2021.00063","DOIUrl":"https://doi.org/10.1109/ICKG52313.2021.00063","url":null,"abstract":"In the online education system, predicting the knowledge of exercises is a fundamental task of many applications, such as cognitive diagnosis. Usually, experts consider this problem as Hierarchical Multi-label Classification (HMC), since the knowledge concepts exhibit a multi-level structure. However, existing methods either sacrificed knowledge consistency for classification accuracy or sacrificed classification accuracy for knowledge consistency. Maintaining the balance is difficult. To forgo this dilemma, in this paper, we develop a novel frame-work called Consistency-Aware Multi-modal Network (Cam-Net). Specifically, we develop a multi-modal embedding module to learn the representation of the multi-modal exercise. Then, we adopt a hybrid prediction method consisting of the flat prediction module and the local prediction module. The local prediction module deals with the relation between the knowledge hierarchy and the input exercise. The flat prediction module focuses on maintaining knowledge consistency. Finally, to balance classification accuracy and knowledge consistency, we combine the outputs of two modules to make a final prediction. Extensive experimental results on two real-world datasets demonstrate the high performance and the ability to reduce knowledge inconsistency of CamNet.","PeriodicalId":174126,"journal":{"name":"2021 IEEE International Conference on Big Knowledge (ICBK)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-12-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122395402","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：