Agustin Zuniga, Huber Flores, Eemil Lagerspetz, P. Nurmi, S. Tarkoma, P. Hui, J. Manner
We contribute by quantifying the effect of network latency and battery consumption on mobile app performance and retention, i.e., user's decisions to continue or stop using apps. We perform our analysis by fusing two large-scale crowdsensed datasets collected by piggybacking on information captured by mobile apps. We find that app performance has an impact in its retention rate. Our results demonstrate that high energy consumption and high latency decrease the likelihood of retaining an app. Conversely, we show that reducing latency or energy consumption does not guarantee higher likelihood of retention as long as they are within reasonable standards of performance. However, we also demonstrate that what is considered reasonable depends on what users have been accustomed to, with device and network characteristics, and app category playing a role. As our second contribution, we develop a model for predicting retention based on performance metrics. We demonstrate the benefits of our model through empirical benchmarks which show that our model not only predicts retention accurately, but generalizes well across application categories, locations and other factors moderating the effect of performance.
{"title":"Tortoise or Hare? Quantifying the Effects of Performance on Mobile App Retention","authors":"Agustin Zuniga, Huber Flores, Eemil Lagerspetz, P. Nurmi, S. Tarkoma, P. Hui, J. Manner","doi":"10.1145/3308558.3313428","DOIUrl":"https://doi.org/10.1145/3308558.3313428","url":null,"abstract":"We contribute by quantifying the effect of network latency and battery consumption on mobile app performance and retention, i.e., user's decisions to continue or stop using apps. We perform our analysis by fusing two large-scale crowdsensed datasets collected by piggybacking on information captured by mobile apps. We find that app performance has an impact in its retention rate. Our results demonstrate that high energy consumption and high latency decrease the likelihood of retaining an app. Conversely, we show that reducing latency or energy consumption does not guarantee higher likelihood of retention as long as they are within reasonable standards of performance. However, we also demonstrate that what is considered reasonable depends on what users have been accustomed to, with device and network characteristics, and app category playing a role. As our second contribution, we develop a model for predicting retention based on performance metrics. We demonstrate the benefits of our model through empirical benchmarks which show that our model not only predicts retention accurately, but generalizes well across application categories, locations and other factors moderating the effect of performance.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"57 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74505905","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}
''User reviews” are becoming an essential component of e-commerce. When buyers write a negative or doubting review, ideally, the sellers need to quickly give a response to minimize the potential impact. When the number of reviews is growing at a frightening speed, there is an urgent need to build a response writing assistant for customer service providers. In order to generate high-quality responses, the algorithm needs to consume and understand the information from both the original review and the target product. The classical sequence-to-sequence (Seq2Seq) methods can hardly satisfy this requirement. In this study, we propose a novel deep neural network model based on the Seq2Seq framework for the review response generation task in e-commerce platforms, which can incorporate product information by a gated multi-source attention mechanism and a copy mechanism. Moreover, we employ a reinforcement learning technique to reduce the exposure bias problem. To evaluate the proposed model, we constructed a large-scale dataset from a popular e-commerce website, which contains product information. Empirical studies on both automatic evaluation metrics and human annotations show that the proposed model can generate informative and diverse responses, significantly outperforming state-of-the-art text generation models.
{"title":"Review Response Generation in E-Commerce Platforms with External Product Information","authors":"Lujun Zhao, Kaisong Song, Changlong Sun, Qi Zhang, Xuanjing Huang, Xiaozhong Liu","doi":"10.1145/3308558.3313581","DOIUrl":"https://doi.org/10.1145/3308558.3313581","url":null,"abstract":"''User reviews” are becoming an essential component of e-commerce. When buyers write a negative or doubting review, ideally, the sellers need to quickly give a response to minimize the potential impact. When the number of reviews is growing at a frightening speed, there is an urgent need to build a response writing assistant for customer service providers. In order to generate high-quality responses, the algorithm needs to consume and understand the information from both the original review and the target product. The classical sequence-to-sequence (Seq2Seq) methods can hardly satisfy this requirement. In this study, we propose a novel deep neural network model based on the Seq2Seq framework for the review response generation task in e-commerce platforms, which can incorporate product information by a gated multi-source attention mechanism and a copy mechanism. Moreover, we employ a reinforcement learning technique to reduce the exposure bias problem. To evaluate the proposed model, we constructed a large-scale dataset from a popular e-commerce website, which contains product information. Empirical studies on both automatic evaluation metrics and human annotations show that the proposed model can generate informative and diverse responses, significantly outperforming state-of-the-art text generation models.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"3 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78812517","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}
Liheng Chen, Yanru Qu, Zhenghui Wang, Lin Qiu, Weinan Zhang, Ken Chen, Shaodian Zhang, Yong Yu
In graphs with rich texts, incorporating textual information with structural information would benefit constructing expressive graph embeddings. Among various graph embedding models, random walk (RW)-based is one of the most popular and successful groups. However, it is challenged by two issues when applied on graphs with rich texts: (i) sampling efficiency: deriving from the training objective of RW-based models (e.g., DeepWalk and node2vec), we show that RW-based models are likely to generate large amounts of redundant training samples due to three main drawbacks. (ii) text utilization: these models have difficulty in dealing with zero-shot scenarios where graph embedding models have to infer graph structures directly from texts. To solve these problems, we propose a novel framework, namely Text-driven Graph Embedding with Pairs Sampling (TGE-PS). TGE-PS uses Pairs Sampling (PS) to improve the sampling strategy of RW, being able to reduce ~ 99% training samples while preserving competitive performance. TGE-PS uses Text-driven Graph Embedding (TGE), an inductive graph embedding approach, to generate node embeddings from texts. Since each node contains rich texts, TGE is able to generate high-quality embeddings and provide reasonable predictions on existence of links to unseen nodes. We evaluate TGE-PS on several real-world datasets, and experiment results demonstrate that TGE-PS produces state-of-the-art results on both traditional and zero-shot link prediction tasks.
{"title":"Sampled in Pairs and Driven by Text: A New Graph Embedding Framework","authors":"Liheng Chen, Yanru Qu, Zhenghui Wang, Lin Qiu, Weinan Zhang, Ken Chen, Shaodian Zhang, Yong Yu","doi":"10.1145/3308558.3313520","DOIUrl":"https://doi.org/10.1145/3308558.3313520","url":null,"abstract":"In graphs with rich texts, incorporating textual information with structural information would benefit constructing expressive graph embeddings. Among various graph embedding models, random walk (RW)-based is one of the most popular and successful groups. However, it is challenged by two issues when applied on graphs with rich texts: (i) sampling efficiency: deriving from the training objective of RW-based models (e.g., DeepWalk and node2vec), we show that RW-based models are likely to generate large amounts of redundant training samples due to three main drawbacks. (ii) text utilization: these models have difficulty in dealing with zero-shot scenarios where graph embedding models have to infer graph structures directly from texts. To solve these problems, we propose a novel framework, namely Text-driven Graph Embedding with Pairs Sampling (TGE-PS). TGE-PS uses Pairs Sampling (PS) to improve the sampling strategy of RW, being able to reduce ~ 99% training samples while preserving competitive performance. TGE-PS uses Text-driven Graph Embedding (TGE), an inductive graph embedding approach, to generate node embeddings from texts. Since each node contains rich texts, TGE is able to generate high-quality embeddings and provide reasonable predictions on existence of links to unseen nodes. We evaluate TGE-PS on several real-world datasets, and experiment results demonstrate that TGE-PS produces state-of-the-art results on both traditional and zero-shot link prediction tasks.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"239 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79302356","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}
Revealing important vertices is a fundamental task in network analysis. As such, many indicators have been proposed for doing so, which are collectively called centralities. However, the abundance of studies on centralities blurs their differences. In this work, we compare centralities based on their sensivitity to modifications in the graph. Specifically, we introduce a quantitative measure called (average-case) edge sensitivity, which measures how much the centrality value of a uniformly chosen vertex (or an edge) changes when we remove a uniformly chosen edge. Edge sensitivity is applicable to unweighted graphs, regarding which, to our knowledge, there has been no theoretical analysis of the centralities. We conducted a theoretical analysis of the edge sensitivities of six major centralities: the closeness centrality, harmonic centrality, betweenness centrality, endpoint betweenness centrality, PageRank, and spanning tree centrality. Our experimental results on synthetic and real graphs confirm the tendency predicted by the theoretical analysis. We also discuss an extension of edge sensitivity to the setting that we remove a uniformly chosen set of edges of size k for an integer k = 1.
{"title":"Sensitivity Analysis of Centralities on Unweighted Networks","authors":"Shogo Murai, Yuichi Yoshida","doi":"10.1145/3308558.3313422","DOIUrl":"https://doi.org/10.1145/3308558.3313422","url":null,"abstract":"Revealing important vertices is a fundamental task in network analysis. As such, many indicators have been proposed for doing so, which are collectively called centralities. However, the abundance of studies on centralities blurs their differences. In this work, we compare centralities based on their sensivitity to modifications in the graph. Specifically, we introduce a quantitative measure called (average-case) edge sensitivity, which measures how much the centrality value of a uniformly chosen vertex (or an edge) changes when we remove a uniformly chosen edge. Edge sensitivity is applicable to unweighted graphs, regarding which, to our knowledge, there has been no theoretical analysis of the centralities. We conducted a theoretical analysis of the edge sensitivities of six major centralities: the closeness centrality, harmonic centrality, betweenness centrality, endpoint betweenness centrality, PageRank, and spanning tree centrality. Our experimental results on synthetic and real graphs confirm the tendency predicted by the theoretical analysis. We also discuss an extension of edge sensitivity to the setting that we remove a uniformly chosen set of edges of size k for an integer k = 1.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"11 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"78573560","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The understanding of talent flow is critical for sharpening company talent strategy to keep competitiveness in the current fast-evolving environment. Existing studies on talent flow analysis generally rely on subjective surveys. However, without large-scale quantitative studies, there are limits to deliver fine-grained predictive business insights for better talent management. To this end, in this paper, we aim to introduce a big data-driven approach for predictive talent flow analysis. Specifically, we first construct a time-aware job transition tensor by mining the large-scale job transition records of digital resumes from online professional networks (OPNs), where each entry refers to a fine-grained talent flow rate of a specific job position between two companies. Then, we design a dynamic latent factor based Evolving Tensor Factorization (ETF) model for predicting the future talent flows. In particular, a novel evolving feature by jointly considering the influence of previous talent flows and global market is introduced for modeling the evolving nature of each company. Furthermore, to improve the predictive performance, we also integrate several representative attributes of companies as side information for regulating the model inference. Finally, we conduct extensive experiments on large-scale real-world data for evaluating the model performances. The experimental results clearly validate the effectiveness of our approach compared with state-of-the-art baselines in terms of talent flow forecast. Meanwhile, the results also reveal some interesting findings on the regularity of talent flows, e.g. Facebook becomes more and more attractive for the engineers from Google in 2016.
{"title":"Large-Scale Talent Flow Forecast with Dynamic Latent Factor Model?","authors":"Le Zhang, Hengshu Zhu, Tong Xu, Chen Zhu, Chuan Qin, Hui Xiong, Enhong Chen","doi":"10.1145/3308558.3313525","DOIUrl":"https://doi.org/10.1145/3308558.3313525","url":null,"abstract":"The understanding of talent flow is critical for sharpening company talent strategy to keep competitiveness in the current fast-evolving environment. Existing studies on talent flow analysis generally rely on subjective surveys. However, without large-scale quantitative studies, there are limits to deliver fine-grained predictive business insights for better talent management. To this end, in this paper, we aim to introduce a big data-driven approach for predictive talent flow analysis. Specifically, we first construct a time-aware job transition tensor by mining the large-scale job transition records of digital resumes from online professional networks (OPNs), where each entry refers to a fine-grained talent flow rate of a specific job position between two companies. Then, we design a dynamic latent factor based Evolving Tensor Factorization (ETF) model for predicting the future talent flows. In particular, a novel evolving feature by jointly considering the influence of previous talent flows and global market is introduced for modeling the evolving nature of each company. Furthermore, to improve the predictive performance, we also integrate several representative attributes of companies as side information for regulating the model inference. Finally, we conduct extensive experiments on large-scale real-world data for evaluating the model performances. The experimental results clearly validate the effectiveness of our approach compared with state-of-the-art baselines in terms of talent flow forecast. Meanwhile, the results also reveal some interesting findings on the regularity of talent flows, e.g. Facebook becomes more and more attractive for the engineers from Google in 2016.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"103 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77014634","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The spread of content produced by fake news publishers was one of the most discussed characteristics of the 2016 U.S. Presidential Election. Yet, little is known about the prevalence and focus of such content, how its prevalence changed over time, and how this prevalence related to important election dynamics. In this paper, we address these questions using tweets that mention the two presidential candidates sampled at the daily level, the news content mentioned in such tweets, and open-ended responses from nationally representative telephone interviews. The results of our analysis highlight various important lessons for news consumers and journalists. We find that (i.) traditional news producers outperformed fake news producers in aggregate, (ii.) the prevalence of content produced by fake news publishers increased over the course of the campaign-particularly among tweets that mentioned Clinton, and (iii.) changes in such prevalence were closely following changes in net Clinton favorability. Turning to content, we (iv.) identify similarities and differences in agenda setting by fake and traditional news media and show that (v.) information individuals most commonly reported to having read, seen or heard about the candidates was more closely aligned with content produced by fake news outlets than traditional news outlets, in particular for information Republican voters retained about Clinton. We also model fake-ness of retained information as a function of demographics characteristics. Implications for platform owners, news consumers, and journalists are discussed.
{"title":"What happened? The Spread of Fake News Publisher Content During the 2016 U.S. Presidential Election","authors":"Ceren Budak","doi":"10.1145/3308558.3313721","DOIUrl":"https://doi.org/10.1145/3308558.3313721","url":null,"abstract":"The spread of content produced by fake news publishers was one of the most discussed characteristics of the 2016 U.S. Presidential Election. Yet, little is known about the prevalence and focus of such content, how its prevalence changed over time, and how this prevalence related to important election dynamics. In this paper, we address these questions using tweets that mention the two presidential candidates sampled at the daily level, the news content mentioned in such tweets, and open-ended responses from nationally representative telephone interviews. The results of our analysis highlight various important lessons for news consumers and journalists. We find that (i.) traditional news producers outperformed fake news producers in aggregate, (ii.) the prevalence of content produced by fake news publishers increased over the course of the campaign-particularly among tweets that mentioned Clinton, and (iii.) changes in such prevalence were closely following changes in net Clinton favorability. Turning to content, we (iv.) identify similarities and differences in agenda setting by fake and traditional news media and show that (v.) information individuals most commonly reported to having read, seen or heard about the candidates was more closely aligned with content produced by fake news outlets than traditional news outlets, in particular for information Republican voters retained about Clinton. We also model fake-ness of retained information as a function of demographics characteristics. Implications for platform owners, news consumers, and journalists are discussed.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"223 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73207980","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}
Crowdsourcing has become a popular tool for large-scale data collection where it is often assumed that crowd workers complete the work independently. In this paper, we relax such independence property and explore the usage of peer communication-a kind of direct interactions between workers-in crowdsourcing. In particular, in the crowdsourcing setting with peer communication, a pair of workers are asked to complete the same task together by first generating their initial answers to the task independently and then freely discussing the task with each other and updating their answers after the discussion. We first experimentally examine the effects of peer communication on individual microtasks. Our results conducted on three types of tasks consistently suggest that work quality is significantly improved in tasks with peer communication compared to tasks where workers complete the work independently. We next explore how to utilize peer communication to optimize the requester's total utility while taking into account higher data correlation and higher cost introduced by peer communication. In particular, we model the requester's online decision problem of whether and when to use peer communication in crowdsourcing as a constrained Markov decision process which maximizes the requester's total utility under budget constraints. Our proposed approach is empirically shown to bring higher total utility compared to baseline approaches.
{"title":"Leveraging Peer Communication to Enhance Crowdsourcing","authors":"Wei Tang, Ming Yin, Chien-Ju Ho","doi":"10.1145/3308558.3313554","DOIUrl":"https://doi.org/10.1145/3308558.3313554","url":null,"abstract":"Crowdsourcing has become a popular tool for large-scale data collection where it is often assumed that crowd workers complete the work independently. In this paper, we relax such independence property and explore the usage of peer communication-a kind of direct interactions between workers-in crowdsourcing. In particular, in the crowdsourcing setting with peer communication, a pair of workers are asked to complete the same task together by first generating their initial answers to the task independently and then freely discussing the task with each other and updating their answers after the discussion. We first experimentally examine the effects of peer communication on individual microtasks. Our results conducted on three types of tasks consistently suggest that work quality is significantly improved in tasks with peer communication compared to tasks where workers complete the work independently. We next explore how to utilize peer communication to optimize the requester's total utility while taking into account higher data correlation and higher cost introduced by peer communication. In particular, we model the requester's online decision problem of whether and when to use peer communication in crowdsourcing as a constrained Markov decision process which maximizes the requester's total utility under budget constraints. Our proposed approach is empirically shown to bring higher total utility compared to baseline approaches.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"13 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73588197","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}
Jai Gupta, Zhen Qin, Michael Bendersky, Donald Metzler
Spell correction is a must-have feature for any modern search engine in applications such as web or e-commerce search. Typical spell correction solutions used in production systems consist of large indexed lookup tables based on a global model trained across many users over a large scale web corpus or a query log. For search over personal corpora, such as email, this global solution is not sufficient, as it ignores the user's personal lexicon. Without personalization, global spelling fails to correct tail queries drawn from a user's own, often idiosyncratic, lexicon. Personalization using existing algorithms is difficult due to resource constraints and unavailability of sufficient data to build per-user models. In this work, we propose a simple and effective personalized spell correction solution that augments existing global solutions for search over private corpora. Our event driven spell correction candidate generation method is specifically designed with personalization as the key construct. Our novel spell correction and query completion algorithms do not require complex model training and is highly efficient. The proposed solution has shown over 30% click-through rate gain on affected queries when evaluated against a range of strong commercial personal search baselines - Google's Gmail, Drive, and Calendar search production systems.
{"title":"Personalized Online Spell Correction for Personal Search","authors":"Jai Gupta, Zhen Qin, Michael Bendersky, Donald Metzler","doi":"10.1145/3308558.3313706","DOIUrl":"https://doi.org/10.1145/3308558.3313706","url":null,"abstract":"Spell correction is a must-have feature for any modern search engine in applications such as web or e-commerce search. Typical spell correction solutions used in production systems consist of large indexed lookup tables based on a global model trained across many users over a large scale web corpus or a query log. For search over personal corpora, such as email, this global solution is not sufficient, as it ignores the user's personal lexicon. Without personalization, global spelling fails to correct tail queries drawn from a user's own, often idiosyncratic, lexicon. Personalization using existing algorithms is difficult due to resource constraints and unavailability of sufficient data to build per-user models. In this work, we propose a simple and effective personalized spell correction solution that augments existing global solutions for search over private corpora. Our event driven spell correction candidate generation method is specifically designed with personalization as the key construct. Our novel spell correction and query completion algorithms do not require complex model training and is highly efficient. The proposed solution has shown over 30% click-through rate gain on affected queries when evaluated against a range of strong commercial personal search baselines - Google's Gmail, Drive, and Calendar search production systems.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"60 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79867354","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}
Finding diagrams that contain a specific part or a similar part is important in many engineering tasks. In this search task, the query part is expected to match only a small region in a complex image. This paper investigates several local matching networks that explicitly model local region-to-region similarities. Deep convolutional neural networks extract local features and model local matching patterns. Spatial convolution is employed to cross-match local regions at different scale levels, addressing cases where the target part appears at a different scale, position, and/or angle. A gating network automatically learns region importance, removing noise from sparse areas and visual metadata in engineering diagrams. Experimental results show that local matching approaches are more effective for engineering diagram search than global matching approaches. Suppressing unimportant regions via the gating network enhances accuracy. Matching across different scales via spatial convolution substantially improves robustness to scale and rotation changes. A pipelined architecture efficiently searches a large collection of diagrams by using a simple local matching network to identify a small set of candidate images and a more sophisticated network with convolutional cross-scale matching to re-rank candidates.
{"title":"Local Matching Networks for Engineering Diagram Search","authors":"Zhuyun Dai, Zhen Fan, Hafeezul Rahman, Jamie Callan","doi":"10.1145/3308558.3313500","DOIUrl":"https://doi.org/10.1145/3308558.3313500","url":null,"abstract":"Finding diagrams that contain a specific part or a similar part is important in many engineering tasks. In this search task, the query part is expected to match only a small region in a complex image. This paper investigates several local matching networks that explicitly model local region-to-region similarities. Deep convolutional neural networks extract local features and model local matching patterns. Spatial convolution is employed to cross-match local regions at different scale levels, addressing cases where the target part appears at a different scale, position, and/or angle. A gating network automatically learns region importance, removing noise from sparse areas and visual metadata in engineering diagrams. Experimental results show that local matching approaches are more effective for engineering diagram search than global matching approaches. Suppressing unimportant regions via the gating network enhances accuracy. Matching across different scales via spatial convolution substantially improves robustness to scale and rotation changes. A pipelined architecture efficiently searches a large collection of diagrams by using a simple local matching network to identify a small set of candidate images and a more sophisticated network with convolutional cross-scale matching to re-rank candidates.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"57 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80160979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The proliferation of publicly accessible urban data provide new insights on various urban tasks. A frequently used approach is to treat each region as a data sample and build a model over all the regions to observe the correlations between urban features (e.g., demographics) and the target variable (e.g., crime count). To define regions, most existing studies use fixed grids or pre-defined administrative boundaries (e.g., census tracts or community areas). In reality, however, definitions of regions should be different depending on tasks (e.g., regional crime count prediction vs. real estate prices estimation). In this paper, we propose a new problem of task-specific city region partitioning, aiming to find the best partition in a city w.r.t. a given task. We prove this is an NP-hard search problem with no trivial solution. To learn the partition, we first study two variants of Markov Chain Monte Carlo (MCMC). We further propose a reinforcement learning scheme for effective sampling the search space. We conduct experiments on two real datasets in Chicago (i.e., crime count and real estate price) to demonstrate the effectiveness of our proposed method.
{"title":"Learning Task-Specific City Region Partition","authors":"Hongjian Wang, P. Jenkins, Hua Wei, Fei Wu, Z. Li","doi":"10.1145/3308558.3313704","DOIUrl":"https://doi.org/10.1145/3308558.3313704","url":null,"abstract":"The proliferation of publicly accessible urban data provide new insights on various urban tasks. A frequently used approach is to treat each region as a data sample and build a model over all the regions to observe the correlations between urban features (e.g., demographics) and the target variable (e.g., crime count). To define regions, most existing studies use fixed grids or pre-defined administrative boundaries (e.g., census tracts or community areas). In reality, however, definitions of regions should be different depending on tasks (e.g., regional crime count prediction vs. real estate prices estimation). In this paper, we propose a new problem of task-specific city region partitioning, aiming to find the best partition in a city w.r.t. a given task. We prove this is an NP-hard search problem with no trivial solution. To learn the partition, we first study two variants of Markov Chain Monte Carlo (MCMC). We further propose a reinforcement learning scheme for effective sampling the search space. We conduct experiments on two real datasets in Chicago (i.e., crime count and real estate price) to demonstrate the effectiveness of our proposed method.","PeriodicalId":23013,"journal":{"name":"The World Wide Web Conference","volume":"35 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2019-05-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90349693","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：