We develop a highly scalable and effective contextual bandit approach towards periodical personalized recommendations. The online bootstrapping-based technique provides a principled way for UCB-type exploitation-exploration algorithms, while being able to handle arbitrary sized datasets, well suited to learn the ever evolving user preference drift from streaming data, and essentially parameter-free. We further introduce techniques to handle arbitrary sized feature spaces using feature hashing, leverage existing state-of-art machine learning via learning reduction, and increase cache hits by managing bootstrapped models in memory effectively. The resulted model trains on millions of examples and billions of features within minutes on a single personal computer. It shows persistent performance in both offline and online evaluation. We observe around 10% click through rate (CTR) and conversion lift over a collaborative filtering approach in real-world A/B testing across more than 40 million users on the major Ticketmaster email recommendation product.
{"title":"A Scalable Approach for Periodical Personalized Recommendations","authors":"Zhen Qin, I. Rishabh, John Carnahan","doi":"10.1145/2959100.2959139","DOIUrl":"https://doi.org/10.1145/2959100.2959139","url":null,"abstract":"We develop a highly scalable and effective contextual bandit approach towards periodical personalized recommendations. The online bootstrapping-based technique provides a principled way for UCB-type exploitation-exploration algorithms, while being able to handle arbitrary sized datasets, well suited to learn the ever evolving user preference drift from streaming data, and essentially parameter-free. We further introduce techniques to handle arbitrary sized feature spaces using feature hashing, leverage existing state-of-art machine learning via learning reduction, and increase cache hits by managing bootstrapped models in memory effectively. The resulted model trains on millions of examples and billions of features within minutes on a single personal computer. It shows persistent performance in both offline and online evaluation. We observe around 10% click through rate (CTR) and conversion lift over a collaborative filtering approach in real-world A/B testing across more than 40 million users on the major Ticketmaster email recommendation product.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"149 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133018768","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}
At the heart of any personalization application, such as Google Now, is a deep model for users. The understanding of users ranges from raw history to lower dimensional reductions like interest, locations, preferences, etc. We will discuss different representations of such user understanding. Going from understanding to application, we will talk about two broad applications recommendations of information and guided exploration - both in the context of Google Now. We will focus on such applications from an information retrieval perspective. Information recommendation then takes the form of biasing information retrieval, in response to a query or, in the limit, in a query-less application. Somewhere in between lies broad declaration of user intent, e.g., interest in food, and we will discuss how personalization and guided exploration play together to provide a valuable tool to the user. We will discuss valuable lessons learned along the way.At the heart of any personalization application, such as Google Now, is a deep model for users. The understanding of users ranges from raw history to lower dimensional reductions like interest, locations, preferences, etc. We will discuss different representations of such user understanding. Going from understanding to application, we will talk about two broad applications recommendations of information and guided exploration both in the context of Google Now. We will focus on such applications from an information retrieval perspective. Information recommendation then takes the form of biasing information retrieval, in response to a query or, in the limit, in a queryless application. Somewhere in between lies broad declaration of user intent, e.g., interest in food, and we will discuss how personalization and guided exploration play together to provide a valuable tool to the user. We will discuss valuable lessons learned along the way.
{"title":"Personalization for Google Now: User Understanding and Application to Information Recommendation and Exploration","authors":"Shashidhar Thakur","doi":"10.1145/2959100.2959192","DOIUrl":"https://doi.org/10.1145/2959100.2959192","url":null,"abstract":"At the heart of any personalization application, such as Google Now, is a deep model for users. The understanding of users ranges from raw history to lower dimensional reductions like interest, locations, preferences, etc. We will discuss different representations of such user understanding. Going from understanding to application, we will talk about two broad applications recommendations of information and guided exploration - both in the context of Google Now. We will focus on such applications from an information retrieval perspective. Information recommendation then takes the form of biasing information retrieval, in response to a query or, in the limit, in a query-less application. Somewhere in between lies broad declaration of user intent, e.g., interest in food, and we will discuss how personalization and guided exploration play together to provide a valuable tool to the user. We will discuss valuable lessons learned along the way.At the heart of any personalization application, such as Google Now, is a deep model for users. The understanding of users ranges from raw history to lower dimensional reductions like interest, locations, preferences, etc. We will discuss different representations of such user understanding. Going from understanding to application, we will talk about two broad applications recommendations of information and guided exploration both in the context of Google Now. We will focus on such applications from an information retrieval perspective. Information recommendation then takes the form of biasing information retrieval, in response to a query or, in the limit, in a queryless application. Somewhere in between lies broad declaration of user intent, e.g., interest in food, and we will discuss how personalization and guided exploration play together to provide a valuable tool to the user. We will discuss valuable lessons learned along the way.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"56 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132133729","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}
Automated course recommendation can help deliver personalized and effective college advising and degree planning. Nearest neighbor and matrix factorization based collaborative filtering approaches have been applied to student-course grade data to help students select suitable courses. However, the student-course enrollment patterns exhibit grouping structures that are tied to the student and course academic features, which lead to grade data that are not missing at random (NMAR). Existing approaches for dealing with NMAR data, such as Response-aware and context-aware matrix factorization, do not model NMAR data in terms of the user and item features and are not designed with the characteristics of grade data in mind. In this work we investigate how the student and course academic features influence the enrollment patterns and we use these features to define student and course groups at various levels of granularity. We show how these groups can be used to design grade prediction and top-n course ranking models for neighborhood-based user collaborative filtering, matrix factorization and popularity-based ranking approaches. These methods give lower grade prediction error and more accurate top-n course rankings than the other methods that do not take domain knowledge into account.
{"title":"Domain-Aware Grade Prediction and Top-n Course Recommendation","authors":"Asmaa Elbadrawy, G. Karypis","doi":"10.1145/2959100.2959133","DOIUrl":"https://doi.org/10.1145/2959100.2959133","url":null,"abstract":"Automated course recommendation can help deliver personalized and effective college advising and degree planning. Nearest neighbor and matrix factorization based collaborative filtering approaches have been applied to student-course grade data to help students select suitable courses. However, the student-course enrollment patterns exhibit grouping structures that are tied to the student and course academic features, which lead to grade data that are not missing at random (NMAR). Existing approaches for dealing with NMAR data, such as Response-aware and context-aware matrix factorization, do not model NMAR data in terms of the user and item features and are not designed with the characteristics of grade data in mind. In this work we investigate how the student and course academic features influence the enrollment patterns and we use these features to define student and course groups at various levels of granularity. We show how these groups can be used to design grade prediction and top-n course ranking models for neighborhood-based user collaborative filtering, matrix factorization and popularity-based ranking approaches. These methods give lower grade prediction error and more accurate top-n course rankings than the other methods that do not take domain knowledge into account.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134466341","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}
Y. Brovman, Marie Jacob, N. Srinivasan, Stephen Neola, D. Galron, Ryan Snyder, Paul Wang
This paper tackles the problem of recommendations in eBay's large semi-structured marketplace. eBay's variable inventory and lack of structured information about listings makes traditional collaborative filtering algorithms difficult to use. We discuss how to overcome these data limitations to produce high quality recommendations in real time with a combination of a customized scalable architecture as well as a widely applicable machine learned ranking model. A pointwise ranking approach is utilized to reduce the ranking problem to a binary classification problem optimized on past user purchase behavior. We present details of a sampling strategy and feature engineering that have been critical to achieve a lift in both purchase through rate (PTR) and revenue.
{"title":"Optimizing Similar Item Recommendations in a Semi-structured Marketplace to Maximize Conversion","authors":"Y. Brovman, Marie Jacob, N. Srinivasan, Stephen Neola, D. Galron, Ryan Snyder, Paul Wang","doi":"10.1145/2959100.2959166","DOIUrl":"https://doi.org/10.1145/2959100.2959166","url":null,"abstract":"This paper tackles the problem of recommendations in eBay's large semi-structured marketplace. eBay's variable inventory and lack of structured information about listings makes traditional collaborative filtering algorithms difficult to use. We discuss how to overcome these data limitations to produce high quality recommendations in real time with a combination of a customized scalable architecture as well as a widely applicable machine learned ranking model. A pointwise ranking approach is utilized to reduce the ranking problem to a binary classification problem optimized on past user purchase behavior. We present details of a sampling strategy and feature engineering that have been critical to achieve a lift in both purchase through rate (PTR) and revenue.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"76 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124937883","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}
Online social networks have become predominant in recent years and have grown to encompass massive scales of data. In addition to data scale, these networks can be heterogeneous and contain complex structures between different users, between social entities and various interactions between users and social entities. This is especially true in enterprise social networks where hierarchies explicitly exist between employees as well. In such networks, producing the best recommendations for each user is a very challenging problem for two main reasons. First, the complex structures in the social network need to be properly mined and exploited by the algorithm. Second, these networks contain millions or even billions of edges making the problem very difficult computationally. In this paper we present Guided Walk, a supervised graph based algorithm that learns the significance of different network links for each user and then produces entity recommendations based on this learning phase. We compare the algorithm with a set of baseline algorithms using offline evaluation techniques as well as a user survey. The offline results show that the algorithm outperforms the next best algorithm by a factor of 3.6. The user survey further confirms that the recommendation are not only relevant but also rank high in terms of personal relevance for each user. To deal with large scale social networks, the Guided Walk algorithm is formulated as a Pregel program which allows us to utilize the power of distributed parallel computing. This would allow horizontally scaling the algorithm for larger social networks by simply adding more compute nodes to the cluster.
{"title":"Guided Walk: A Scalable Recommendation Algorithm for Complex Heterogeneous Social Networks","authors":"R. Levin, Hassan Abassi, Uzi Cohen","doi":"10.1145/2959100.2959143","DOIUrl":"https://doi.org/10.1145/2959100.2959143","url":null,"abstract":"Online social networks have become predominant in recent years and have grown to encompass massive scales of data. In addition to data scale, these networks can be heterogeneous and contain complex structures between different users, between social entities and various interactions between users and social entities. This is especially true in enterprise social networks where hierarchies explicitly exist between employees as well. In such networks, producing the best recommendations for each user is a very challenging problem for two main reasons. First, the complex structures in the social network need to be properly mined and exploited by the algorithm. Second, these networks contain millions or even billions of edges making the problem very difficult computationally. In this paper we present Guided Walk, a supervised graph based algorithm that learns the significance of different network links for each user and then produces entity recommendations based on this learning phase. We compare the algorithm with a set of baseline algorithms using offline evaluation techniques as well as a user survey. The offline results show that the algorithm outperforms the next best algorithm by a factor of 3.6. The user survey further confirms that the recommendation are not only relevant but also rank high in terms of personal relevance for each user. To deal with large scale social networks, the Guided Walk algorithm is formulated as a Pregel program which allows us to utilize the power of distributed parallel computing. This would allow horizontally scaling the algorithm for larger social networks by simply adding more compute nodes to the cluster.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129143430","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}
An accurate and comprehensive user modeling technique is crucial for the quality of recommender systems. Traditionally, we model user preferences using only actions from the target site and may suffer from cold-start problem. As nowadays people normally engage in multiple online sites for various needs, we consider leveraging the cross-site actions to improve the user modeling accuracy. Specifically, in this paper we aim at achieving a more comprehensive and accurate user modeling by modeling user's actions in multiple aligned heterogeneous sites simultaneously. To do so, we propose a modularized probabilistic graphical model framework JUMA. We further integrate topic model and matrix factorization into JUMA for joint user modeling over text-based and item-based sites. We assemble and publish large-scale dataset for comprehensive analyzing and evaluation. Experimental results show that our framework JUMA out performs traditional within-site user modeling techniques, especially for cold-start scenarios. For cold-start users, we achieve relative improvements of 9.3% and 12.8% comparing to existing within-site approaches for recommendation in item-based and text-based sites respectively. Thus we draw the conclusion that aligning heterogeneous sites and modeling users jointly do help to improve the quality of online recommender systems.
{"title":"Joint User Modeling across Aligned Heterogeneous Sites","authors":"Xuezhi Cao, Yong Yu","doi":"10.1145/2959100.2959155","DOIUrl":"https://doi.org/10.1145/2959100.2959155","url":null,"abstract":"An accurate and comprehensive user modeling technique is crucial for the quality of recommender systems. Traditionally, we model user preferences using only actions from the target site and may suffer from cold-start problem. As nowadays people normally engage in multiple online sites for various needs, we consider leveraging the cross-site actions to improve the user modeling accuracy. Specifically, in this paper we aim at achieving a more comprehensive and accurate user modeling by modeling user's actions in multiple aligned heterogeneous sites simultaneously. To do so, we propose a modularized probabilistic graphical model framework JUMA. We further integrate topic model and matrix factorization into JUMA for joint user modeling over text-based and item-based sites. We assemble and publish large-scale dataset for comprehensive analyzing and evaluation. Experimental results show that our framework JUMA out performs traditional within-site user modeling techniques, especially for cold-start scenarios. For cold-start users, we achieve relative improvements of 9.3% and 12.8% comparing to existing within-site approaches for recommendation in item-based and text-based sites respectively. Thus we draw the conclusion that aligning heterogeneous sites and modeling users jointly do help to improve the quality of online recommender systems.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"31 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127913935","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 cold start problem arises in various recommendation applications. In this paper, we propose a tensor factorization-based algorithm that exploits content features extracted from music audio to deal with the cold start problem for the emerging application next-song recommendation. Specifically, the new algorithm learns sequential behavior to predict the next song that a user would be interested in based on the last song the user just listened to. A unique characteristic of the algorithm is that it learns and updates the mapping between the audio feature space and the item latent space each time during the iterations of the factorization process. This way, the content features can be better exploited in forming the latent features for both users and items, leading to more effective solutions for cold-start recommendation. Evaluation on a large-scale music recommendation dataset shows that the recommendation result of the proposed algorithm exhibits not only higher accuracy but also better novelty and diversity, suggesting its applicability in helping a user explore new items in next-item recommendation. Our implementation is available at https://github.com/fearofchou/ALMM.
{"title":"Addressing Cold Start for Next-song Recommendation","authors":"Szu-Yu Chou, Yi-Hsuan Yang, J. Jang, Yu-Ching Lin","doi":"10.1145/2959100.2959156","DOIUrl":"https://doi.org/10.1145/2959100.2959156","url":null,"abstract":"The cold start problem arises in various recommendation applications. In this paper, we propose a tensor factorization-based algorithm that exploits content features extracted from music audio to deal with the cold start problem for the emerging application next-song recommendation. Specifically, the new algorithm learns sequential behavior to predict the next song that a user would be interested in based on the last song the user just listened to. A unique characteristic of the algorithm is that it learns and updates the mapping between the audio feature space and the item latent space each time during the iterations of the factorization process. This way, the content features can be better exploited in forming the latent features for both users and items, leading to more effective solutions for cold-start recommendation. Evaluation on a large-scale music recommendation dataset shows that the recommendation result of the proposed algorithm exhibits not only higher accuracy but also better novelty and diversity, suggesting its applicability in helping a user explore new items in next-item recommendation. Our implementation is available at https://github.com/fearofchou/ALMM.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"412 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124405279","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}
Deployment of open source recommendation systems has been shown to be an effective way to increase sale conversions on a variety of e-commerce sites. However, there remains a large gap between deploying the core algorithm provided by these systems and delivering an application-quality recommendation system, specifically tailored to address complex and dynamically changing business needs. We will present a real-time recommendation engine built on our graph data platform that provides the following extensions to a basic recommendation model: True real-time recommendation algorithms: We provide a simple framework for customers to author and deploy real-time recommendation algorithms with no pre-computation required. Streaming updates of user behavior and product information: As quickly as data are generated, the recommendation engine applies the updates and can serve updated results. Support for offline recommendation algorithms}: Users with existing investment in a quality recommendation program can import their pre-computed results into the graph database for efficient, unified service of results. Tools for Business-centric requirements: The engine offers a range of weighting, sorting, and filtering options to tailor recommendation algorithms to business needs. For example, the engine can eliminate products that are out of stock or favor products that are known to perform well in different real-time contexts. Multiple algorithm ensemble support: There is rarely a case where one algorithm is sufficient to identify the best items to recommend to a user. Integrating points 1 through 4, the engine provides intuitive methods for specifying and combining the results of multiple recommendation algorithms to achieve the highest-performing results. Recommendation feedback tools: Pre- and Post-analysis tools, built around a business' logic, are used to generate reports to assess the value of both potential and currently deployed algorithms. Of particular interest to RecSys attendees, we will discuss the technical aspects of a graph-based implementation of the recommendation engine and how it facilitates the rapid design and deployment of an efficient real-time recommendation system under a single service. We will briefly discuss the architecture of our graph database system to show how it can efficiently serve a large user base even from a single server shared-memory architecture. Attendees will also learn about graph-based data modeling and how viewing data from this perspective can lead to new types of business insights and applications that are not easily implemented using traditional relational and/or NoSQL platforms. We will conclude with a brief demonstration of a real deployed application UI to demonstrate the ease by which recommendation systems can be implemented and deployed using the engine.
{"title":"Leveraging a Graph-Powered, Real-Time Recommendation Engine to Create Rapid Business Value","authors":"A. Anthony, Yu-Keng Shih, R. Jin, Yang Xiang","doi":"10.1145/2959100.2959126","DOIUrl":"https://doi.org/10.1145/2959100.2959126","url":null,"abstract":"Deployment of open source recommendation systems has been shown to be an effective way to increase sale conversions on a variety of e-commerce sites. However, there remains a large gap between deploying the core algorithm provided by these systems and delivering an application-quality recommendation system, specifically tailored to address complex and dynamically changing business needs. We will present a real-time recommendation engine built on our graph data platform that provides the following extensions to a basic recommendation model: True real-time recommendation algorithms: We provide a simple framework for customers to author and deploy real-time recommendation algorithms with no pre-computation required. Streaming updates of user behavior and product information: As quickly as data are generated, the recommendation engine applies the updates and can serve updated results. Support for offline recommendation algorithms}: Users with existing investment in a quality recommendation program can import their pre-computed results into the graph database for efficient, unified service of results. Tools for Business-centric requirements: The engine offers a range of weighting, sorting, and filtering options to tailor recommendation algorithms to business needs. For example, the engine can eliminate products that are out of stock or favor products that are known to perform well in different real-time contexts. Multiple algorithm ensemble support: There is rarely a case where one algorithm is sufficient to identify the best items to recommend to a user. Integrating points 1 through 4, the engine provides intuitive methods for specifying and combining the results of multiple recommendation algorithms to achieve the highest-performing results. Recommendation feedback tools: Pre- and Post-analysis tools, built around a business' logic, are used to generate reports to assess the value of both potential and currently deployed algorithms. Of particular interest to RecSys attendees, we will discuss the technical aspects of a graph-based implementation of the recommendation engine and how it facilitates the rapid design and deployment of an efficient real-time recommendation system under a single service. We will briefly discuss the architecture of our graph database system to show how it can efficiently serve a large user base even from a single server shared-memory architecture. Attendees will also learn about graph-based data modeling and how viewing data from this perspective can lead to new types of business insights and applications that are not easily implemented using traditional relational and/or NoSQL platforms. We will conclude with a brief demonstration of a real deployed application UI to demonstrate the ease by which recommendation systems can be implemented and deployed using the engine.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"82 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122554244","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}
Bart P. Knijnenburg, S. Sivakumar, Daricia Wilkinson
Every day, we are confronted with an abundance of decisions that require us to choose from a seemingly endless number of choice options. Recommender systems are supposed to help us deal with this formidable task, but some scholars claim that these systems instead put us inside a "Filter Bubble" that severely limits our perspectives. This paper presents a new direction for recommender systems research with the main goal of supporting users in developing, exploring, and understanding their unique personal preferences.
{"title":"Recommender Systems for Self-Actualization","authors":"Bart P. Knijnenburg, S. Sivakumar, Daricia Wilkinson","doi":"10.1145/2959100.2959189","DOIUrl":"https://doi.org/10.1145/2959100.2959189","url":null,"abstract":"Every day, we are confronted with an abundance of decisions that require us to choose from a seemingly endless number of choice options. Recommender systems are supposed to help us deal with this formidable task, but some scholars claim that these systems instead put us inside a \"Filter Bubble\" that severely limits our perspectives. This paper presents a new direction for recommender systems research with the main goal of supporting users in developing, exploring, and understanding their unique personal preferences.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"50 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123296330","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}
Strategic behaviour from sellers on e-commerce websites, such as faking transactions and manipulating the recommendation scores through artificial reviews, have been among the most notorious obstacles that prevent websites from maximizing the efficiency of their recommendations. Previous approaches have focused almost exclusively on machine learning-related techniques to detect and penalize such behaviour. In this paper, we tackle the problem from a different perspective, using the approach of the field of mechanism design. We put forward a game model tailored for the setting at hand and aim to construct truthful mechanisms, i.e. mechanisms that do not provide incentives for dishonest reputation-augmenting actions, that guarantee good recommendations in the worst-case. For the setting with two agents, we propose a truthful mechanism that is optimal in terms of social efficiency. For the general case of m agents, we prove both lower and upper bound results on the effciency of truthful mechanisms and propose truthful mechanisms that yield significantly better results, when compared to an existing mechanism from a leading e-commerce site on real data.
{"title":"Mechanism Design for Personalized Recommender Systems","authors":"Qingpeng Cai, Aris Filos-Ratsikas, Chang Liu, Pingzhong Tang","doi":"10.1145/2959100.2959135","DOIUrl":"https://doi.org/10.1145/2959100.2959135","url":null,"abstract":"Strategic behaviour from sellers on e-commerce websites, such as faking transactions and manipulating the recommendation scores through artificial reviews, have been among the most notorious obstacles that prevent websites from maximizing the efficiency of their recommendations. Previous approaches have focused almost exclusively on machine learning-related techniques to detect and penalize such behaviour. In this paper, we tackle the problem from a different perspective, using the approach of the field of mechanism design. We put forward a game model tailored for the setting at hand and aim to construct truthful mechanisms, i.e. mechanisms that do not provide incentives for dishonest reputation-augmenting actions, that guarantee good recommendations in the worst-case. For the setting with two agents, we propose a truthful mechanism that is optimal in terms of social efficiency. For the general case of m agents, we prove both lower and upper bound results on the effciency of truthful mechanisms and propose truthful mechanisms that yield significantly better results, when compared to an existing mechanism from a leading e-commerce site on real data.","PeriodicalId":315651,"journal":{"name":"Proceedings of the 10th ACM Conference on Recommender Systems","volume":"83 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2016-09-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115254469","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：