Pub Date : 2020-11-01DOI: 10.1109/SCC49832.2020.00027
Malik Khalfallah, P. Ghodous
Data provenance is a fundamental concept in scientific experimentation in general and complex event processing (CEP) in particular. For accurate determination and visualization of data provenance, efficient and user-friendly mechanisms are needed. Research in CEP optimization and visual notations can help in this process. This paper presents the extension of an optimized CEP framework to respond to data provenance requests. The extension consists in enriching the formal representation of execution plans of CEP queries to make them provenance-aware. These provenance-aware execution plans are then queried to generate a visual representation of the provenance data. We present the implementation of this framework and then its deployment and the associated evaluation in the context of an industrial use case.
{"title":"Data Provenance for Complex Event Processing Invoking Composition of Services","authors":"Malik Khalfallah, P. Ghodous","doi":"10.1109/SCC49832.2020.00027","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00027","url":null,"abstract":"Data provenance is a fundamental concept in scientific experimentation in general and complex event processing (CEP) in particular. For accurate determination and visualization of data provenance, efficient and user-friendly mechanisms are needed. Research in CEP optimization and visual notations can help in this process. This paper presents the extension of an optimized CEP framework to respond to data provenance requests. The extension consists in enriching the formal representation of execution plans of CEP queries to make them provenance-aware. These provenance-aware execution plans are then queried to generate a visual representation of the provenance data. We present the implementation of this framework and then its deployment and the associated evaluation in the context of an industrial use case.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131234986","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00055
Md. Kafil Uddin, Qiang He, Jun Han, C. Chua
App stores, such as Google Play and Apple Store, contain abundance of information, including descriptions and reviews of various apps. They provide valuable information for app developers to learn about their own apps as well as similar apps for improving their apps, e.g., adding popular features or removing unpopular features. A place to start this learning process is to identify the competitor apps of a given target app in terms of their features, popularity as well as other relevant aspects. Given the large number of apps and the large amount of information available about these apps, it is a very challenging task for app developers to effectively and efficiently identify competitor apps. In this paper, we introduce a novel approach for identifying the competitor apps of a given target app, which includes three major components. Firstly, we identify the factors that characterise the competition between apps. Then, based on the identified competition factors, we extract and process the relevant information from the app store about the target app and the apps in the same app category. Finally, we cluster the apps based on their similarity across all the competition factors and identify those apps that are in the same cluster as the target app as its competitor apps. We evaluate our approach by comparing its results with corresponding search results in (1) Google Trends and (2) Google Search. The results show that our approach is effective in identifying competitor apps.
{"title":"App Competition Matters: How to Identify Your Competitor Apps?","authors":"Md. Kafil Uddin, Qiang He, Jun Han, C. Chua","doi":"10.1109/SCC49832.2020.00055","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00055","url":null,"abstract":"App stores, such as Google Play and Apple Store, contain abundance of information, including descriptions and reviews of various apps. They provide valuable information for app developers to learn about their own apps as well as similar apps for improving their apps, e.g., adding popular features or removing unpopular features. A place to start this learning process is to identify the competitor apps of a given target app in terms of their features, popularity as well as other relevant aspects. Given the large number of apps and the large amount of information available about these apps, it is a very challenging task for app developers to effectively and efficiently identify competitor apps. In this paper, we introduce a novel approach for identifying the competitor apps of a given target app, which includes three major components. Firstly, we identify the factors that characterise the competition between apps. Then, based on the identified competition factors, we extract and process the relevant information from the app store about the target app and the apps in the same app category. Finally, we cluster the apps based on their similarity across all the competition factors and identify those apps that are in the same cluster as the target app as its competitor apps. We evaluate our approach by comparing its results with corresponding search results in (1) Google Trends and (2) Google Search. The results show that our approach is effective in identifying competitor apps.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"114 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128142236","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 addition to publishing and downloading mobile apps, Mobile App Store (MAS) has become the most important ecosystem on mobile smart devices, i.e., Mobile Software Ecosystem (MSECO). However, most of the existing work focus on the analysis of a single entity in MSECO, and rarely analyze the interaction between the entities (such as users, developers, etc.) in MSECO as well as the comprehensive effect of each entity to the entire ecosystem health. In this paper, we propose a method based on computational experiments to simulate and analyze the complex and dynamic interaction of various entities and, how these entities impact on health of MSECO. Firstly, a requirement-driven MSECO model named R-MSECO is established to break down the entire MSECO, which includes user-app interaction, developers-requirements interaction and macro-control of mobile platform three sub-models. Secondly, the health measurement method is proposed to measure the health of MSECO. Finally, we simulate the impact of the dynamic interaction of various entities on the health of MSECO based on computational experiments. The experimental results show that our method is helpful for mobile users to better understanding of the current state of MSECO as well as can provide a reference for developers and platform managers to make development and operation decisions, which is of great significance for the continued healthy development of MSECO.
{"title":"Analytics on Health of Mobile Software Ecosystem Based on the Internal Operating Mechanism","authors":"Jianmao Xiao, Shizhan Chen, Shiping Chen, Chao Gao, Hongyue Wu, Xiao Xue, Zhiyong Feng","doi":"10.1109/SCC49832.2020.00042","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00042","url":null,"abstract":"In addition to publishing and downloading mobile apps, Mobile App Store (MAS) has become the most important ecosystem on mobile smart devices, i.e., Mobile Software Ecosystem (MSECO). However, most of the existing work focus on the analysis of a single entity in MSECO, and rarely analyze the interaction between the entities (such as users, developers, etc.) in MSECO as well as the comprehensive effect of each entity to the entire ecosystem health. In this paper, we propose a method based on computational experiments to simulate and analyze the complex and dynamic interaction of various entities and, how these entities impact on health of MSECO. Firstly, a requirement-driven MSECO model named R-MSECO is established to break down the entire MSECO, which includes user-app interaction, developers-requirements interaction and macro-control of mobile platform three sub-models. Secondly, the health measurement method is proposed to measure the health of MSECO. Finally, we simulate the impact of the dynamic interaction of various entities on the health of MSECO based on computational experiments. The experimental results show that our method is helpful for mobile users to better understanding of the current state of MSECO as well as can provide a reference for developers and platform managers to make development and operation decisions, which is of great significance for the continued healthy development of MSECO.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"77 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132050067","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00018
Li Wei, Fangfang Liu
Large-scale knowledge graphs have a lot of hidden knowledge which has not been discovered, so the link prediction of the knowledge graph is an important topic. Translation models represented by TransE are the well-researched algorithms of link prediction. They project the entities and the relations in the knowledge graphs into some continuous vector spaces, and adjust the vector representations of the relations and the entities according to each piece of knowledge. However, in the case of a non-1-to-1 relationship, multiple entity vectors will compete for the same coordinate position in the space. Aiming at this problem, this paper proposes an improved method. By imposing a distance constraint on the competitive entities of a non-1-to-1 relationship, we can narrow the differences between them. Each entity will consider the other competitive entities while adapting itself to fit a triplet, so as to reach the status that each competitive entity is close to the coordinate point of the competition as a whole. Distance constraint can be applied to the existing translation models as a means of optimization. Experiments are conducted on the datasets: FB15K and WN18, and the experimental results show that the method we proposed is effective.
{"title":"The Research of Link Prediction in Knowledge Graph based on Distance Constraint","authors":"Li Wei, Fangfang Liu","doi":"10.1109/SCC49832.2020.00018","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00018","url":null,"abstract":"Large-scale knowledge graphs have a lot of hidden knowledge which has not been discovered, so the link prediction of the knowledge graph is an important topic. Translation models represented by TransE are the well-researched algorithms of link prediction. They project the entities and the relations in the knowledge graphs into some continuous vector spaces, and adjust the vector representations of the relations and the entities according to each piece of knowledge. However, in the case of a non-1-to-1 relationship, multiple entity vectors will compete for the same coordinate position in the space. Aiming at this problem, this paper proposes an improved method. By imposing a distance constraint on the competitive entities of a non-1-to-1 relationship, we can narrow the differences between them. Each entity will consider the other competitive entities while adapting itself to fit a triplet, so as to reach the status that each competitive entity is close to the coordinate point of the competition as a whole. Distance constraint can be applied to the existing translation models as a means of optimization. Experiments are conducted on the datasets: FB15K and WN18, and the experimental results show that the method we proposed is effective.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127914920","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00073
R. Anarfi, Benjamin A. Kwapong, K. K. Fletcher
Tags are critical sources of data for search, browsing and information retrieval. Manual selection of tags, over the years, have not been very effective. This paper introduces an approach to automatic mashup tag recommendation, based on reinforcement learning (RL). Our RL approach is able to carry out effective exploratory actions to automatically extract the and recommend tags for mashups. We perform experiments to evaluate our proposed method. Results from our experiments show that, the recommended mashup tags improve performance on the information retrieval task.
{"title":"desc2tag: A Reinforcement Learning Approach to Mashup Tag Recommendation","authors":"R. Anarfi, Benjamin A. Kwapong, K. K. Fletcher","doi":"10.1109/SCC49832.2020.00073","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00073","url":null,"abstract":"Tags are critical sources of data for search, browsing and information retrieval. Manual selection of tags, over the years, have not been very effective. This paper introduces an approach to automatic mashup tag recommendation, based on reinforcement learning (RL). Our RL approach is able to carry out effective exploratory actions to automatically extract the and recommend tags for mashups. We perform experiments to evaluate our proposed method. Results from our experiments show that, the recommended mashup tags improve performance on the information retrieval task.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128768808","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00067
Ruriko Kudo, Yasuharu Katsuno, Fumiko Satoh
It is crucial to ensure that server rack doors are properly closed in data centers in order to prevent serious dangers caused by the doors suddenly opening in disasters and causing accidents that hurt maintenance workers and damage facilities. In this paper, we propose multi-factor-based motion detection for server rack doors that have been left open. Our approach recognizes the status of server rack doors on the basis of maintenance workers’ motion and prevents a worker from forgetting to close the doors without using any additional devices, only smartphones, which maintenance engineers already carry for work.
{"title":"Multi-factor-based Motion Detection for Server Rack Doors Left Open","authors":"Ruriko Kudo, Yasuharu Katsuno, Fumiko Satoh","doi":"10.1109/SCC49832.2020.00067","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00067","url":null,"abstract":"It is crucial to ensure that server rack doors are properly closed in data centers in order to prevent serious dangers caused by the doors suddenly opening in disasters and causing accidents that hurt maintenance workers and damage facilities. In this paper, we propose multi-factor-based motion detection for server rack doors that have been left open. Our approach recognizes the status of server rack doors on the basis of maintenance workers’ motion and prevents a worker from forgetting to close the doors without using any additional devices, only smartphones, which maintenance engineers already carry for work.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116813922","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00044
Wictor Souza Martins, B. Kuehne, Rafael Ferreira Sobrinho, F. Preti
This paper presents a reference method for performance evaluation in Big Data architectures, called by Improvement Method for Big Data Architectures (IMBDA) aiming to increase the performance, and consequently raising the quality of service provided. The method will contribute to small businesses and startups that have limited financial re-sources (impossible to invest in market solutions). The proposed approach considers the relationship of the processes in a data processing flow to find possible bottlenecks and optimization points. To this end, IMBDA collects system logs to compose functional metrics (e.g., processing time) and non-functional metrics (e.g., CPU and memory utilization, and other cloud computing infrastructure resources). The system stores these metrics in an external data analysis tool that investigates the correlation of performance between processes. The reference method applies to the architecture of a Big Data application, which provides solutions in fleet logistics. With the use of IMBDA, it was possible to identify performance bottlenecks, allowing the reconfiguration of the architecture to increase service quality at the lowest possible cost.
本文提出了一种大数据架构性能评估的参考方法，称为IMBDA (Improvement method for Big Data architectures)，旨在提高大数据架构的性能，从而提高所提供的服务质量。该方法将有助于小型企业和创业公司，有有限的财政资源(不可能投资于市场解决方案)。该方法考虑了数据处理流中各过程之间的关系，以发现可能的瓶颈和优化点。为此，IMBDA收集系统日志，以组成功能指标(例如，处理时间)和非功能指标(例如，CPU和内存利用率，以及其他云计算基础设施资源)。系统将这些指标存储在外部数据分析工具中，该工具用于调查进程之间性能的相关性。参考方法适用于大数据应用的架构，为车队物流提供解决方案。通过使用IMBDA，可以识别性能瓶颈，从而允许对体系结构进行重新配置，以尽可能低的成本提高服务质量。
{"title":"A Reference Method for Performance Evaluation in Big Data Architectures","authors":"Wictor Souza Martins, B. Kuehne, Rafael Ferreira Sobrinho, F. Preti","doi":"10.1109/SCC49832.2020.00044","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00044","url":null,"abstract":"This paper presents a reference method for performance evaluation in Big Data architectures, called by Improvement Method for Big Data Architectures (IMBDA) aiming to increase the performance, and consequently raising the quality of service provided. The method will contribute to small businesses and startups that have limited financial re-sources (impossible to invest in market solutions). The proposed approach considers the relationship of the processes in a data processing flow to find possible bottlenecks and optimization points. To this end, IMBDA collects system logs to compose functional metrics (e.g., processing time) and non-functional metrics (e.g., CPU and memory utilization, and other cloud computing infrastructure resources). The system stores these metrics in an external data analysis tool that investigates the correlation of performance between processes. The reference method applies to the architecture of a Big Data application, which provides solutions in fleet logistics. With the use of IMBDA, it was possible to identify performance bottlenecks, allowing the reconfiguration of the architecture to increase service quality at the lowest possible cost.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117026327","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 increasing attention to ocean and the development of data-intensive sciences, a large amount of ocean data has been acquired by various observing platforms and sensors, which poses new challenges to data management and utilization. Typically, nowadays we target to move ocean data management toward the FAIR principles of being findable, accessible, interoperable, and reusable. However, the data produced and managed by different organizations with wide diversity, various structures and increasing volume make it hard to be FAIR, and one of the most critical reason is the lack of unified data representation and publication methods. In this paper, we propose novel techniques to try to solve the problem by introducing semantics with ontologies. Specifically, we first propose a unified semantic model named OEDO to represent ocean data by defining the concepts of ocean observing field, specifying the relations between the concepts, and describing the properties with ocean metadata. Then, we further optimize the state-of-the-art quick service query list (QSQL) data structure, by extending the domain concepts with WordNet to improve data discovery. Moreover, based on the OEDO model and the optimized QSQL, we propose an ocean data service publishing method called DOLP to improve data discovery and data access. Finally, we conduct extensive experiments to demonstrate the effectiveness and efficiency of our proposals.
{"title":"Bringing Semantics to Support Ocean FAIR Data Services with Ontologies","authors":"Xiaoli Ren, Xiaoyong Li, Kefeng Deng, Kaijun Ren, Aolong Zhou, Junqiang Song","doi":"10.1109/SCC49832.2020.00011","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00011","url":null,"abstract":"With the increasing attention to ocean and the development of data-intensive sciences, a large amount of ocean data has been acquired by various observing platforms and sensors, which poses new challenges to data management and utilization. Typically, nowadays we target to move ocean data management toward the FAIR principles of being findable, accessible, interoperable, and reusable. However, the data produced and managed by different organizations with wide diversity, various structures and increasing volume make it hard to be FAIR, and one of the most critical reason is the lack of unified data representation and publication methods. In this paper, we propose novel techniques to try to solve the problem by introducing semantics with ontologies. Specifically, we first propose a unified semantic model named OEDO to represent ocean data by defining the concepts of ocean observing field, specifying the relations between the concepts, and describing the properties with ocean metadata. Then, we further optimize the state-of-the-art quick service query list (QSQL) data structure, by extending the domain concepts with WordNet to improve data discovery. Moreover, based on the OEDO model and the optimized QSQL, we propose an ocean data service publishing method called DOLP to improve data discovery and data access. Finally, we conduct extensive experiments to demonstrate the effectiveness and efficiency of our proposals.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"98 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127161240","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00054
Pengwei Wang, Zhen Chen, Zhaohui Zhang
Crowdsourcing web services has received much attention in recent years, which has been widely used in many fields, and provides important human computing services for the rapid development of AI. Crowdsourcing knowledge acquisition is one of the most important applications, which includes a series of work such as image annotation and picture classification. However, due to the differences in the difficulty of these tasks and the uncertainty of workers, how to make a reasonable task assignment while ensuring the completion of tasks becomes a big challenge. To this end, we introduce WordNet external knowledge base to help determine the difficulty of picture classification tasks. We refer to the e-sports rank mechanism and use dynamic update strategy to assess the actual ability of workers. A novel criterion affinity based on the weighted Euclidean distance with penalty factor is proposed to measure the suitability between tasks and workers. On this basis, the Kuhn-Munkres (KM) algorithm is used to solve the weighted bipartite graph matching problem. Through comparative experiments, the effectiveness of our proposed method is verified.
{"title":"Suitability-based Task Assignment in Crowdsourcing Markets","authors":"Pengwei Wang, Zhen Chen, Zhaohui Zhang","doi":"10.1109/SCC49832.2020.00054","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00054","url":null,"abstract":"Crowdsourcing web services has received much attention in recent years, which has been widely used in many fields, and provides important human computing services for the rapid development of AI. Crowdsourcing knowledge acquisition is one of the most important applications, which includes a series of work such as image annotation and picture classification. However, due to the differences in the difficulty of these tasks and the uncertainty of workers, how to make a reasonable task assignment while ensuring the completion of tasks becomes a big challenge. To this end, we introduce WordNet external knowledge base to help determine the difficulty of picture classification tasks. We refer to the e-sports rank mechanism and use dynamic update strategy to assess the actual ability of workers. A novel criterion affinity based on the weighted Euclidean distance with penalty factor is proposed to measure the suitability between tasks and workers. On this basis, the Kuhn-Munkres (KM) algorithm is used to solve the weighted bipartite graph matching problem. Through comparative experiments, the effectiveness of our proposed method is verified.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"10 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124447303","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 : 2020-11-01DOI: 10.1109/SCC49832.2020.00051
Xiaohong Xu, T. He, Huazhen Wang
Existing methods for data-to-text generation have difficulty producing diverse texts with low duplication rates. In this paper, we propose a novel data-to-text generation model with Transformer planning and a Wasserstein auto-encoder, which can convert constructed data to coherent and diverse text. This model possesses the following features: Transformer is first used to generate the data planning sequence of the target text content (each sequence is a subset of the input items that can be covered by a sentence), and then the Wasserstein Auto-Encoder(WAE) and a deep neural network are employed to establish the global latent variable space of the model. Second, text generation is performed through a hierarchical structure that takes the data planning sequence, global latent variables, and context of the generated sentences as conditions. Furthermore, to achieve diversity of text expression, a decoder is developed that combines the neural network with the WAE. The experimental results show that this model can achieve higher evaluation scores than the existing baseline models in terms of the diversity metrics of text generation and the duplication rate.
{"title":"A Novel Data-to-Text Generation Model with Transformer Planning and a Wasserstein Auto-Encoder","authors":"Xiaohong Xu, T. He, Huazhen Wang","doi":"10.1109/SCC49832.2020.00051","DOIUrl":"https://doi.org/10.1109/SCC49832.2020.00051","url":null,"abstract":"Existing methods for data-to-text generation have difficulty producing diverse texts with low duplication rates. In this paper, we propose a novel data-to-text generation model with Transformer planning and a Wasserstein auto-encoder, which can convert constructed data to coherent and diverse text. This model possesses the following features: Transformer is first used to generate the data planning sequence of the target text content (each sequence is a subset of the input items that can be covered by a sentence), and then the Wasserstein Auto-Encoder(WAE) and a deep neural network are employed to establish the global latent variable space of the model. Second, text generation is performed through a hierarchical structure that takes the data planning sequence, global latent variables, and context of the generated sentences as conditions. Furthermore, to achieve diversity of text expression, a decoder is developed that combines the neural network with the WAE. The experimental results show that this model can achieve higher evaluation scores than the existing baseline models in terms of the diversity metrics of text generation and the duplication rate.","PeriodicalId":274909,"journal":{"name":"2020 IEEE International Conference on Services Computing (SCC)","volume":"333 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2020-11-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122140862","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：