Balanced edge partition has emerged as a new approach to partition an input graph data for the purpose of scaling out parallel computations, which is of interest for several modern data analytics computation platforms, including platforms for iterative computations, machine learning problems, and graph databases. This new approach stands in a stark contrast to the traditional approach of balanced vertex partition, where for given number of partitions, the problem is to minimize the number of edges cut subject to balancing the vertex cardinality of partitions. In this paper, we first characterize the expected costs of vertex and edge partitions with and without aggregation of messages, for the commonly deployed policy of placing a vertex or an edge uniformly at random to one of the partitions. We then obtain the first approximation algorithms for the balanced edge-partition problem which for the case of no aggregation matches the best known approximation ratio for the balanced vertex-partition problem, and show that this remains to hold for the case with aggregation up to factor that is equal to the maximum in-degree of a vertex. We report results of an extensive empirical evaluation on a set of real-world graphs, which quantifies the benefits of edge- vs. vertex-partition, and demonstrates efficiency of natural greedy online assignments for the balanced edge-partition problem with and with no aggregation.
{"title":"Balanced graph edge partition","authors":"F. Bourse, M. Lelarge, M. Vojnović","doi":"10.1145/2623330.2623660","DOIUrl":"https://doi.org/10.1145/2623330.2623660","url":null,"abstract":"Balanced edge partition has emerged as a new approach to partition an input graph data for the purpose of scaling out parallel computations, which is of interest for several modern data analytics computation platforms, including platforms for iterative computations, machine learning problems, and graph databases. This new approach stands in a stark contrast to the traditional approach of balanced vertex partition, where for given number of partitions, the problem is to minimize the number of edges cut subject to balancing the vertex cardinality of partitions. In this paper, we first characterize the expected costs of vertex and edge partitions with and without aggregation of messages, for the commonly deployed policy of placing a vertex or an edge uniformly at random to one of the partitions. We then obtain the first approximation algorithms for the balanced edge-partition problem which for the case of no aggregation matches the best known approximation ratio for the balanced vertex-partition problem, and show that this remains to hold for the case with aggregation up to factor that is equal to the maximum in-degree of a vertex. We report results of an extensive empirical evaluation on a set of real-world graphs, which quantifies the benefits of edge- vs. vertex-partition, and demonstrates efficiency of natural greedy online assignments for the balanced edge-partition problem with and with no aggregation.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83164796","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}
We introduce a novel graphical model, the collaborative score topic model (CSTM), for personal recommendations of textual documents. CSTM's chief novelty lies in its learned model of individual libraries, or sets of documents, associated with each user. Overall, CSTM is a joint directed probabilistic model of user-item scores (ratings), and the textual side information in the user libraries and the items. Creating a generative description of scores and the text allows CSTM to perform well in a wide variety of data regimes, smoothly combining the side information with observed ratings as the number of ratings available for a given user ranges from none to many. Experiments on real-world datasets demonstrate CSTM's performance. We further demonstrate its utility in an application for personal recommendations of posters which we deployed at the NIPS 2013 conference.
{"title":"Leveraging user libraries to bootstrap collaborative filtering","authors":"Laurent Charlin, R. Zemel, H. Larochelle","doi":"10.1145/2623330.2623663","DOIUrl":"https://doi.org/10.1145/2623330.2623663","url":null,"abstract":"We introduce a novel graphical model, the collaborative score topic model (CSTM), for personal recommendations of textual documents. CSTM's chief novelty lies in its learned model of individual libraries, or sets of documents, associated with each user. Overall, CSTM is a joint directed probabilistic model of user-item scores (ratings), and the textual side information in the user libraries and the items. Creating a generative description of scores and the text allows CSTM to perform well in a wide variety of data regimes, smoothly combining the side information with observed ratings as the number of ratings available for a given user ranges from none to many. Experiments on real-world datasets demonstrate CSTM's performance. We further demonstrate its utility in an application for personal recommendations of posters which we deployed at the NIPS 2013 conference.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"6 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80278453","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}
We study the problem of determining if an input matrix A εRm x n can be well-approximated by a low rank matrix. Specifically, we study the problem of quickly estimating the rank or stable rank of A, the latter often providing a more robust measure of the rank. Since we seek significantly sublinear time algorithms, we cast these problems in the property testing framework. In this framework, A either has low rank or stable rank, or is far from having this property. The algorithm should read only a small number of entries or rows of A and decide which case A is in with high probability. If neither case occurs, the output is allowed to be arbitrary. We consider two notions of being far: (1) A requires changing at least an ε-fraction of its entries, or (2) A requires changing at least an ε-fraction of its rows. We call the former the "entry model" and the latter the "row model". We show: For testing if a matrix has rank at most d in the entry model, we improve the previous number of entries of A that need to be read from O(d2/ε2) (Krauthgamer and Sasson, SODA 2003) to O(d2/ε). Our algorithm is the first to adaptively query the entries of A, which for constant d we show is necessary to achieve O(1/ε) queries. For the important case of d = 1 we also give a new non-adaptive algorithm, improving the previous O(1/ε2) queries to O(log2(1/ε) / ε). For testing if a matrix has rank at most d in the row model, we prove an Ω(d/ε) lower bound on the number of rows that need to be read, even for adaptive algorithms. Our lower bound matches a non-adaptive upper bound of Krauthgamer and Sasson. For testing if a matrix has stable rank at most d in the row model or requires changing an ε/d-fraction of its rows in order to have stable rank at most d, we prove that reading θ(d/ε2) rows is necessary and sufficient. We also give an empirical evaluation of our rank and stable rank algorithms on real and synthetic datasets.
研究了输入矩阵A εRm x n能否被低秩矩阵很好地逼近的问题。具体来说，我们研究了快速估计A的秩或稳定秩的问题，后者通常提供一个更稳健的秩度量。由于我们寻求重要的次线性时间算法，我们将这些问题置于性质测试框架中。在这个框架中，A要么具有低秩，要么具有稳定秩，要么远不具有这种性质。算法应该只读取a的少量条目或行，并确定a处于高概率的哪种情况。如果这两种情况都不发生，则允许输出是任意的。我们考虑两个远的概念:(1)A需要改变它的至少一个ε-分数的条目，或者(2)A需要改变它的至少一个ε-分数的行。我们称前者为“入口模型”，后者为“行模型”。为了测试一个矩阵在条目模型中是否排名最多为d，我们将a的先前需要读取的条目数从O(d2/ε2) (Krauthgamer and Sasson, SODA 2003)提高到O(d2/ε)。我们的算法是第一个自适应查询A的条目的算法，对于常数d，我们证明了实现O(1/ε)查询是必要的。对于d = 1的重要情况，我们还给出了一种新的非自适应算法，将之前的O(1/ε2)查询改进为O(log2(1/ε) /ε)。为了测试一个矩阵在行模型中是否秩最多为d，我们证明了需要读取的行数的Ω(d/ε)下界，即使对于自适应算法也是如此。我们的下界与Krauthgamer和Sasson的非自适应上界相匹配。为了检验一个矩阵在行模型中是否具有最多d的稳定秩，或者是否需要改变其行的ε/d分数才能具有最多d的稳定秩，我们证明了读取θ(d/ε2)行是必要和充分的。我们还在真实数据集和合成数据集上对我们的秩和稳定秩算法进行了实证评估。
{"title":"Improved testing of low rank matrices","authors":"Yi Li, Zhengyu Wang, David P. Woodruff","doi":"10.1145/2623330.2623736","DOIUrl":"https://doi.org/10.1145/2623330.2623736","url":null,"abstract":"We study the problem of determining if an input matrix A εRm x n can be well-approximated by a low rank matrix. Specifically, we study the problem of quickly estimating the rank or stable rank of A, the latter often providing a more robust measure of the rank. Since we seek significantly sublinear time algorithms, we cast these problems in the property testing framework. In this framework, A either has low rank or stable rank, or is far from having this property. The algorithm should read only a small number of entries or rows of A and decide which case A is in with high probability. If neither case occurs, the output is allowed to be arbitrary. We consider two notions of being far: (1) A requires changing at least an ε-fraction of its entries, or (2) A requires changing at least an ε-fraction of its rows. We call the former the \"entry model\" and the latter the \"row model\". We show: For testing if a matrix has rank at most d in the entry model, we improve the previous number of entries of A that need to be read from O(d2/ε2) (Krauthgamer and Sasson, SODA 2003) to O(d2/ε). Our algorithm is the first to adaptively query the entries of A, which for constant d we show is necessary to achieve O(1/ε) queries. For the important case of d = 1 we also give a new non-adaptive algorithm, improving the previous O(1/ε2) queries to O(log2(1/ε) / ε). For testing if a matrix has rank at most d in the row model, we prove an Ω(d/ε) lower bound on the number of rows that need to be read, even for adaptive algorithms. Our lower bound matches a non-adaptive upper bound of Krauthgamer and Sasson. For testing if a matrix has stable rank at most d in the row model or requires changing an ε/d-fraction of its rows in order to have stable rank at most d, we prove that reading θ(d/ε2) rows is necessary and sufficient. We also give an empirical evaluation of our rank and stable rank algorithms on real and synthetic datasets.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"78 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83926533","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 competitive environment of the internet, retaining and growing one's user base is of major concern to most web services. Furthermore, the economic model of many web services is allowing free access to most content, and generating revenue through advertising. This unique model requires securing user time on a site rather than the purchase of good which makes it crucially important to create new kinds of metrics and solutions for growth and retention efforts for web services. In this work, we address this problem by proposing a new retention metric for web services by concentrating on the rate of user return. We further apply predictive analysis to the proposed retention metric on a service, as a means for characterizing lost customers. Finally, we set up a simple yet effective framework to evaluate a multitude of factors that contribute to user return. Specifically, we define the problem of return time prediction for free web services. Our solution is based on the Cox's proportional hazard model from survival analysis. The hazard based approach offers several benefits including the ability to work with censored data, to model the dynamics in user return rates, and to easily incorporate different types of covariates in the model. We compare the performance of our hazard based model in predicting the user return time and in categorizing users into buckets based on their predicted return time, against several baseline regression and classification methods and find the hazard based approach to be superior.
{"title":"A hazard based approach to user return time prediction","authors":"Komal Kapoor, Mingxuan Sun, J. Srivastava, Tao Ye","doi":"10.1145/2623330.2623348","DOIUrl":"https://doi.org/10.1145/2623330.2623348","url":null,"abstract":"In the competitive environment of the internet, retaining and growing one's user base is of major concern to most web services. Furthermore, the economic model of many web services is allowing free access to most content, and generating revenue through advertising. This unique model requires securing user time on a site rather than the purchase of good which makes it crucially important to create new kinds of metrics and solutions for growth and retention efforts for web services. In this work, we address this problem by proposing a new retention metric for web services by concentrating on the rate of user return. We further apply predictive analysis to the proposed retention metric on a service, as a means for characterizing lost customers. Finally, we set up a simple yet effective framework to evaluate a multitude of factors that contribute to user return. Specifically, we define the problem of return time prediction for free web services. Our solution is based on the Cox's proportional hazard model from survival analysis. The hazard based approach offers several benefits including the ability to work with censored data, to model the dynamics in user return rates, and to easily incorporate different types of covariates in the model. We compare the performance of our hazard based model in predicting the user return time and in categorizing users into buckets based on their predicted return time, against several baseline regression and classification methods and find the hazard based approach to be superior.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"30 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82703924","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}
Qiming Diao, Minghui Qiu, Chao-Yuan Wu, Alex Smola, Jing Jiang, Chong Wang
Recommendation and review sites offer a wealth of information beyond ratings. For instance, on IMDb users leave reviews, commenting on different aspects of a movie (e.g. actors, plot, visual effects), and expressing their sentiments (positive or negative) on these aspects in their reviews. This suggests that uncovering aspects and sentiments will allow us to gain a better understanding of users, movies, and the process involved in generating ratings. The ability to answer questions such as "Does this user care more about the plot or about the special effects?" or "What is the quality of the movie in terms of acting?" helps us to understand why certain ratings are generated. This can be used to provide more meaningful recommendations. In this work we propose a probabilistic model based on collaborative filtering and topic modeling. It allows us to capture the interest distribution of users and the content distribution for movies; it provides a link between interest and relevance on a per-aspect basis and it allows us to differentiate between positive and negative sentiments on a per-aspect basis. Unlike prior work our approach is entirely unsupervised and does not require knowledge of the aspect specific ratings or genres for inference. We evaluate our model on a live copy crawled from IMDb. Our model offers superior performance by joint modeling. Moreover, we are able to address the cold start problem -- by utilizing the information inherent in reviews our model demonstrates improvement for new users and movies.
{"title":"Jointly modeling aspects, ratings and sentiments for movie recommendation (JMARS)","authors":"Qiming Diao, Minghui Qiu, Chao-Yuan Wu, Alex Smola, Jing Jiang, Chong Wang","doi":"10.1145/2623330.2623758","DOIUrl":"https://doi.org/10.1145/2623330.2623758","url":null,"abstract":"Recommendation and review sites offer a wealth of information beyond ratings. For instance, on IMDb users leave reviews, commenting on different aspects of a movie (e.g. actors, plot, visual effects), and expressing their sentiments (positive or negative) on these aspects in their reviews. This suggests that uncovering aspects and sentiments will allow us to gain a better understanding of users, movies, and the process involved in generating ratings. The ability to answer questions such as \"Does this user care more about the plot or about the special effects?\" or \"What is the quality of the movie in terms of acting?\" helps us to understand why certain ratings are generated. This can be used to provide more meaningful recommendations. In this work we propose a probabilistic model based on collaborative filtering and topic modeling. It allows us to capture the interest distribution of users and the content distribution for movies; it provides a link between interest and relevance on a per-aspect basis and it allows us to differentiate between positive and negative sentiments on a per-aspect basis. Unlike prior work our approach is entirely unsupervised and does not require knowledge of the aspect specific ratings or genres for inference. We evaluate our model on a live copy crawled from IMDb. Our model offers superior performance by joint modeling. Moreover, we are able to address the cold start problem -- by utilizing the information inherent in reviews our model demonstrates improvement for new users and movies.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"6 16","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91418298","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}
Chen Luo, Jian-Guang Lou, Qingwei Lin, Qiang Fu, Rui Ding, D. Zhang, Zhe Wang
As online services have more and more popular, incident diagnosis has emerged as a critical task in minimizing the service downtime and ensuring high quality of the services provided. For most online services, incident diagnosis is mainly conducted by analyzing a large amount of telemetry data collected from the services at runtime. Time series data and event sequence data are two major types of telemetry data. Techniques of correlation analysis are important tools that are widely used by engineers for data-driven incident diagnosis. Despite their importance, there has been little previous work addressing the correlation between two types of heterogeneous data for incident diagnosis: continuous time series data and temporal event data. In this paper, we propose an approach to evaluate the correlation between time series data and event data. Our approach is capable of discovering three important aspects of event-timeseries correlation in the context of incident diagnosis: existence of correlation, temporal order, and monotonic effect. Our experimental results on simulation data sets and two real data sets demonstrate the effectiveness of the algorithm.
{"title":"Correlating events with time series for incident diagnosis","authors":"Chen Luo, Jian-Guang Lou, Qingwei Lin, Qiang Fu, Rui Ding, D. Zhang, Zhe Wang","doi":"10.1145/2623330.2623374","DOIUrl":"https://doi.org/10.1145/2623330.2623374","url":null,"abstract":"As online services have more and more popular, incident diagnosis has emerged as a critical task in minimizing the service downtime and ensuring high quality of the services provided. For most online services, incident diagnosis is mainly conducted by analyzing a large amount of telemetry data collected from the services at runtime. Time series data and event sequence data are two major types of telemetry data. Techniques of correlation analysis are important tools that are widely used by engineers for data-driven incident diagnosis. Despite their importance, there has been little previous work addressing the correlation between two types of heterogeneous data for incident diagnosis: continuous time series data and temporal event data. In this paper, we propose an approach to evaluate the correlation between time series data and event data. Our approach is capable of discovering three important aspects of event-timeseries correlation in the context of incident diagnosis: existence of correlation, temporal order, and monotonic effect. Our experimental results on simulation data sets and two real data sets demonstrate the effectiveness of the algorithm.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"12 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82301218","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 rapidly increasing availability of electronic health records (EHRs) from multiple heterogeneous sources has spearheaded the adoption of data-driven approaches for improved clinical research, decision making, prognosis, and patient management. Unfortunately, EHR data do not always directly and reliably map to phenotypes, or medical concepts, that clinical researchers need or use. Existing phenotyping approaches typically require labor intensive supervision from medical experts. We propose Marble, a novel sparse non-negative tensor factorization method to derive phenotype candidates with virtually no human supervision. Marble decomposes the observed tensor into two terms, a bias tensor and an interaction tensor. The bias tensor represents the baseline characteristics common amongst the overall population and the interaction tensor defines the phenotypes. We demonstrate the capability of our proposed model on both simulated and patient data from a publicly available clinical database. Our results show that Marble derived phenotypes provide at least a 42.8% reduction in the number of non-zero element and also retains predictive power for classification purposes. Furthermore, the resulting phenotypes and baseline characteristics from real EHR data are consistent with known characteristics of the patient population. Thus it can potentially be used to rapidly characterize, predict, and manage a large number of diseases, thereby promising a novel, data-driven solution that can benefit very large segments of the population.
{"title":"Marble: high-throughput phenotyping from electronic health records via sparse nonnegative tensor factorization","authors":"Joyce Ho, Joydeep Ghosh, Jimeng Sun","doi":"10.1145/2623330.2623658","DOIUrl":"https://doi.org/10.1145/2623330.2623658","url":null,"abstract":"The rapidly increasing availability of electronic health records (EHRs) from multiple heterogeneous sources has spearheaded the adoption of data-driven approaches for improved clinical research, decision making, prognosis, and patient management. Unfortunately, EHR data do not always directly and reliably map to phenotypes, or medical concepts, that clinical researchers need or use. Existing phenotyping approaches typically require labor intensive supervision from medical experts. We propose Marble, a novel sparse non-negative tensor factorization method to derive phenotype candidates with virtually no human supervision. Marble decomposes the observed tensor into two terms, a bias tensor and an interaction tensor. The bias tensor represents the baseline characteristics common amongst the overall population and the interaction tensor defines the phenotypes. We demonstrate the capability of our proposed model on both simulated and patient data from a publicly available clinical database. Our results show that Marble derived phenotypes provide at least a 42.8% reduction in the number of non-zero element and also retains predictive power for classification purposes. Furthermore, the resulting phenotypes and baseline characteristics from real EHR data are consistent with known characteristics of the patient population. Thus it can potentially be used to rapidly characterize, predict, and manage a large number of diseases, thereby promising a novel, data-driven solution that can benefit very large segments of the population.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76934206","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}
We consider the problem of open-domain question answering (Open QA) over massive knowledge bases (KBs). Existing approaches use either manually curated KBs like Freebase or KBs automatically extracted from unstructured text. In this paper, we present OQA, the first approach to leverage both curated and extracted KBs. A key technical challenge is designing systems that are robust to the high variability in both natural language questions and massive KBs. OQA achieves robustness by decomposing the full Open QA problem into smaller sub-problems including question paraphrasing and query reformulation. OQA solves these sub-problems by mining millions of rules from an unlabeled question corpus and across multiple KBs. OQA then learns to integrate these rules by performing discriminative training on question-answer pairs using a latent-variable structured perceptron algorithm. We evaluate OQA on three benchmark question sets and demonstrate that it achieves up to twice the precision and recall of a state-of-the-art Open QA system.
{"title":"Open question answering over curated and extracted knowledge bases","authors":"Anthony Fader, Luke Zettlemoyer, Oren Etzioni","doi":"10.1145/2623330.2623677","DOIUrl":"https://doi.org/10.1145/2623330.2623677","url":null,"abstract":"We consider the problem of open-domain question answering (Open QA) over massive knowledge bases (KBs). Existing approaches use either manually curated KBs like Freebase or KBs automatically extracted from unstructured text. In this paper, we present OQA, the first approach to leverage both curated and extracted KBs. A key technical challenge is designing systems that are robust to the high variability in both natural language questions and massive KBs. OQA achieves robustness by decomposing the full Open QA problem into smaller sub-problems including question paraphrasing and query reformulation. OQA solves these sub-problems by mining millions of rules from an unlabeled question corpus and across multiple KBs. OQA then learns to integrate these rules by performing discriminative training on question-answer pairs using a latent-variable structured perceptron algorithm. We evaluate OQA on three benchmark question sets and demonstrate that it achieves up to twice the precision and recall of a state-of-the-art Open QA system.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"2 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76330508","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}
Bo Zong, Yinghui Wu, Jie Song, Ambuj K. Singh, H. Çam, Jiawei Han, Xifeng Yan
Performance monitor software for data centers typically generates a great number of alert sequences. These alert sequences indicate abnormal network events. Given a set of observed alert sequences, it is important to identify the most critical alerts that are potentially the causes of others. While the need for mining critical alerts over large scale alert sequences is evident, most alert analysis techniques stop at modeling and mining the causal relations among the alerts. This paper studies the critical alert mining problem: Given a set of alert sequences, we aim to find a set of k critical alerts such that the number of alerts potentially triggered by them is maximized. We show that the problem is intractable; therefore, we resort to approximation and heuristic algorithms. First, we develop an approximation algorithm that obtains a near-optimal alert set in quadratic time, and propose pruning techniques to improve its runtime performance. Moreover, we show a faster approximation exists, when the alerts follow certain causal structure. Second, we propose two fast heuristic algorithms based on tree sampling techniques. On real-life data, these algorithms identify a critical alert from up to 270,000 mined causal relations in 5 seconds; meanwhile, they preserve more than 80% of solution quality, and are up to 5,000 times faster than their approximation counterparts.
{"title":"Towards scalable critical alert mining","authors":"Bo Zong, Yinghui Wu, Jie Song, Ambuj K. Singh, H. Çam, Jiawei Han, Xifeng Yan","doi":"10.1145/2623330.2623729","DOIUrl":"https://doi.org/10.1145/2623330.2623729","url":null,"abstract":"Performance monitor software for data centers typically generates a great number of alert sequences. These alert sequences indicate abnormal network events. Given a set of observed alert sequences, it is important to identify the most critical alerts that are potentially the causes of others. While the need for mining critical alerts over large scale alert sequences is evident, most alert analysis techniques stop at modeling and mining the causal relations among the alerts. This paper studies the critical alert mining problem: Given a set of alert sequences, we aim to find a set of k critical alerts such that the number of alerts potentially triggered by them is maximized. We show that the problem is intractable; therefore, we resort to approximation and heuristic algorithms. First, we develop an approximation algorithm that obtains a near-optimal alert set in quadratic time, and propose pruning techniques to improve its runtime performance. Moreover, we show a faster approximation exists, when the alerts follow certain causal structure. Second, we propose two fast heuristic algorithms based on tree sampling techniques. On real-life data, these algorithms identify a critical alert from up to 270,000 mined causal relations in 5 seconds; meanwhile, they preserve more than 80% of solution quality, and are up to 5,000 times faster than their approximation counterparts.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"1 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79931152","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}
We show how to programmatically model processes that humans use when extracting answers to queries (e.g., "Who invented typewriter?", "List of Washington national parks") from semi-structured Web pages returned by a search engine. This modeling enables various applications including automating repetitive search tasks, and helping search engine developers design micro-segments of factoid questions. We describe the design and implementation of a domain-specific language that enables extracting data from a webpage based on its structure, visual layout, and linguistic patterns. We also describe an algorithm to rank multiple answers extracted from multiple webpages. On 100,000+ queries (across 7 micro-segments) obtained from Bing logs, our system LaSEWeb answered queries with an average recall of 71%. Also, the desired answer(s) were present in top-3 suggestions for 95%+ cases.
{"title":"LaSEWeb","authors":"Oleksandr Polozov, Sumit Gulwani","doi":"10.1145/2623330.2623761","DOIUrl":"https://doi.org/10.1145/2623330.2623761","url":null,"abstract":"We show how to programmatically model processes that humans use when extracting answers to queries (e.g., \"Who invented typewriter?\", \"List of Washington national parks\") from semi-structured Web pages returned by a search engine. This modeling enables various applications including automating repetitive search tasks, and helping search engine developers design micro-segments of factoid questions. We describe the design and implementation of a domain-specific language that enables extracting data from a webpage based on its structure, visual layout, and linguistic patterns. We also describe an algorithm to rank multiple answers extracted from multiple webpages. On 100,000+ queries (across 7 micro-segments) obtained from Bing logs, our system LaSEWeb answered queries with an average recall of 71%. Also, the desired answer(s) were present in top-3 suggestions for 95%+ cases.","PeriodicalId":20536,"journal":{"name":"Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining","volume":"60 1","pages":""},"PeriodicalIF":0.0,"publicationDate":"2014-08-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74392397","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：