Recommenders in a personalized,collaborative digital library environment

We envisage an information source not only as an information resource where users may submit queries to satisfy their daily information need, but also as a collaborative working and meeting space of people sharing common interests. Indeed, we will present a highly personalized environment where not only users may organize (and search into) the information space according to their individual taste and use, but which provides advanced features of collaborative work among the users. It is up to the system to discover interesting properties about the users’ interests, relationships between users and user communities and to make recommendations based on preference patterns of the users, which is the main topic of this paper.     

LunchTime: a slow-casual game for long-term dietary behavior change

Eating out has recently become part of our lifestyle. However, when eating out in restaurants, many people find it difficult to make meal choices consistent with their health goals. Bad eating choices and habits are in part responsible for the alarming increase in the prevalence of chronic diseases such as obesity, diabetes, and high blood pressure, which burden the health care system. Therefore, there is a need for an intervention that educates the public on how to make healthy choices while eating away from home. In this paper, we propose a goal-based slow-casual game approach that addresses this need. This approach acknowledges different groups of users with varying health goals and adopts slow technology to promote learning and reflection. We model two recognized determinants of well-being into dietary interventions and provide feedback accordingly. To demonstrate the suitability of our approach for long-term sustained learning, reflection, and attitude and/or behavior change, we develop and evaluate LunchTime—a goal-based slow-casual game that educates players on how to make healthier meal choices. The result from the evaluation shows that LunchTime facilitates learning and reflection and promotes positive dietary attitude change.     

A novel method for personalized music recommendation

With the development of digital music technologies, it is an interesting and useful issue to recommend the ‘favored music’ from large amounts of digital music. Some Web-based music stores can recommend popular music which has been rated by many people. However, three problems that need to be resolved in the current methods are: (a) how to recommend the ‘favored music’ which has not been rated by anyone, (b) how to avoid repeatedly recommending the ‘disfavored music’ for users, and (c) how to recommend more interesting music for users besides the ones users have been used to listen. To achieve these goals, we proposed a novel method called personalized hybrid music recommendation, which combines the content-based, collaboration-based and emotion-based methods by computing the weights of the methods according to users’ interests. Furthermore, to evaluate the recommendation accuracy, we constructed a system that can recommend the music to users after mining users’ logs on music listening records. By the feedback of the user’s options, the proposed methods accommodate the variations of the users’ musical interests and then promptly recommend the favored and more interesting music via consecutive recommendations. Experimental results show that the recommendation accuracy achieved by our method is as good as 90%. Hence, it is helpful for recommending the ‘favored music’ to users, provided that each music object is annotated with the related music emotions. The framework in this paper could serve as a useful basis for studies on music recommendation.     

A privacy-preserving recommendation method with clustering and locality-sensitive hashing

Nowadays, there is a significant increase in information, resulting in information overload. Recommendation systems have been widely adopted, and they can help users find information relevant to their interests. However, a malicious attacker can infer users' private information via recommendations. To solve problems of data sparseness, enormous high-dimensional data, the cold start problem and privacy protection in an intelligent recommender system, this study proposes a privacy-preserving collaborative filtering recommendation method with clustering and locality-sensitive hashing. First, we cluster users according to their characteristic information to obtain sub-rating matrices. We use the latent factor model to predict and fill in the missing ratings in those matrices. Second, we combine the sub-rating matrices into a complete rating matrix, subsequently, we obtained the neighbors of the target user by analyzing the similarity of the users. We use a locality-sensitive hashing algorithm to reduce the dimensionality of the user rating data and build an index that could quickly obtain the neighbors of the target user. Finally, we predict the target user's ratings and provide recommendations to the target user. Through experiments, our study shows that our method can deal with the problems of data sparseness and cold start problems well and the accuracy of the intelligent recommendation system has been improved. In addition, we use hash techniques to search for the neighbors, which effectively protects the privacy of the user.     

Implicit: a multi-agent recommendation system for web search

For people with non-ordinary interests, it is hard to search for information on the Internet because search engines are impersonalized and are more focused on “average” individuals with “standard” preferences. In order to improve web search for a community of people with similar but specific interests, we propose to use the implicit knowledge contained in the search behavior of groups of users. We developed a multi-agent recommendation system called Implicit, which supports web search for groups or communities of people. In Implicit, agents observe behavior of their users to learn about the “culture” of the community with specific interests. They facilitate sharing of knowledge about relevant links within the community by means of recommendations. The agents also recommend contacts, i.e., who in the community is the right person to ask for a specific topic. Experimental evaluation shows that Implicit improves the quality of the web search in terms of precision and recall.     

Followee recommendation based on text analysis of micro-blogging activity

Nowadays, more and more users keep up with news through information streams coming from real-time micro-blogging activity offered by services such as Twitter. In these sites, information is shared via a followers/followees social network structure in which a follower receives all the micro-blogs from his/her followees. Recent research efforts on understanding micro-blogging as a novel form of communication and news spreading medium have identified three different categories of users in these systems: information sources, information seekers and friends. As social networks grow in the number of registered users, finding relevant and reliable users to receive interesting information becomes essential. In this paper we propose a followee recommender system based on both the analysis of the content of micro-blogs to detect users' interests and in the exploration of the topology of the network to find candidate users for recommendation. Experimental evaluation was conducted in order to determine the impact of different profiling strategies based on the text analysis of micro-blogs as well as several factors that allows the identification of users acting as good information sources. We found that user-generated content available in the network is a rich source of information for profiling users and finding like-minded people.     

基于贝叶斯网络模型的用户兴趣联合推送

The association push of interesting information is implemented according to the information that other users with similar interests have read, the refore user can easily find high quality interesting information from the immense Web information. The association push of interesting information based on Bayesian network model ispresented in this paper, it stores the conditional probability and semantic meaning between the terms, so the similarity between the users' interests and the similarity between the user's interest and document can be computed according to the semantic meaning of the terms, therefore the association push of interesting information can beimplemented. The experience result indicates this method is effective.     

TWILITE: A recommendation system for Twitter using a probabilistic model based on latent Dirichlet allocation

Twitter provides search services to help people find users to follow by recommending popular users or the friends of their friends. However, these services neither offer the most relevant users to follow nor provide a way to find the most interesting tweet messages for each user. Recently, collaborative filtering techniques for recommendations based on friend relationships in social networks have been widely investigated. However, since such techniques do not work well when friend relationships are not sufficient, we need to take advantage of as much other information as possible to improve the performance of recommendations.In this paper, we propose TWILITE, a recommendation system for Twitter using probabilistic modeling based on latent Dirichlet allocation which recommends top-K users to follow and top-K tweets to read for a user. Our model can capture the realistic process of posting tweet messages by generalizing an LDA model as well as the process of connecting to friends by utilizing matrix factorization. We next develop an inference algorithm based on the variational EM algorithm for learning model parameters. Based on the estimated model parameters, we also present effective personalized recommendation algorithms to find the users to follow as well as the interesting tweet messages to read. The performance study with real-life data sets confirms the effectiveness of the proposed model and the accuracy of our personalized recommendations.     

Mining User Preferences,Page Content and Usage to Personalize Website Navigation

The growing availability of information on the Web has raised a challenging problem: can a Web-based information system tailor itself to different user requirements with the ultimate goal of personalizing and improving the users' experience in accessing the contents of a website? This paper proposes a new approach to website personalization based on the exploitation of user browsing interests together with content and usage similarities among Web pages. The outcome is the delivery of page recommendations which are strictly related to the navigational purposes of visitors and their actual location within the cyberspace of the website. Our approach has been used effectively for developing a non-invasive system which allows Web users to navigate through potentially interesting pages without having a basic knowledge of the website structure.     

一种基于常用搜索引擎的智能信息检索系统

针对用户利用常用搜索引擎查询信息时，搜索引擎返回海量杂乱、无序的网页，用户难以从中快速、准确地获得真正关心的信息的现状，从Internet用户的兴趣度出发，设计了一种基于近似网页聚类算法的智能搜索系统。该系统在用户利用常用搜索引擎系统进行信息检索时，消除搜索引擎返回的重复页，对剩余页面进行聚类，返回给用户聚类后的网页簇，这样用户就可以选择浏览自己感兴趣的页面，从而大大提高了信息检索的查准率；实验证明该系统在保证查全率和查准率的基础上大大提高了搜索效率。     

基于活动的社交网络中的群组推荐算法设计

在基于活动的社交网络（EBSN）中,群组中聚集了具有相似兴趣的用户,并为用户组织并举办线下活动,在社区的发展中起到了至关重要的作用,因而理解用户加入群组的原因和群组形成的过程在社交网络的研究中是一个重要的议题.本文通过基于活动的社交网络中的一些相关内容信息,比如社交网络中的标签信息和地理位置信息,来辅助推荐系统更好地为用户预测对于群组的偏好.本文提出了SEGELER （pair-wiSE Geo-social Event-based LatEnt factoR）模型,并使用这些社交网络中的信息,来为用户的兴趣进行预测.通过在真实的EBSN数据集上进行实验与验证,本文的模型不仅可以有效提升对于用户偏好的预测,也可以缓解冷启动问题.     

基于地理近邻关系的微博系统朋友推荐

近年来,微博的蓬勃发展吸引了大量网络用户,用户所发海量微博呈现的大数据环境成为理解用户行为的重要资源。目前,大量在线朋友推荐研究通过对微博内容分析推断用户的兴趣和喜好以进行朋友推荐,但大多数已有研究忽略了用户位置和兴趣之间的潜在关系。事实上,多数情况下用户真正感兴趣的还是他周围的人。为此,提出了基于地理近邻关系的朋友推荐方法,通过把所处位置周围兴趣爱好相似的微博用户彼此推荐,为用户提供了与周围可能感兴趣的人联系的独特渠道。仿真分析证明,与传统朋友推荐方法相比,基于地理近邻的朋友推荐具有较高的推荐性能。     

UbiShop: Commercial item recommendation using visual part-based object representation

With the popularity of online shopping, people have used to shop commercial items on the online shopping websites for convenience. However, based on traditional text search methods, people usually can not find the interesting commercial item they want if they do not know its detailed information, e.g., the name and the seller. Therefore, a more convenient method to help people find the commercial item they want is desired. In this work, we develop a practical system, UbiShop, on mobile phones, whereby users can timely get the related information of interesting commercial items by taking pictures of them. Users can also obtain recommendations on visually similar commercial items to help their buying selections. With the observation that people’s preferences on commercial items usually simply depend on their partial visual styles, we propose a novel representation, Visual Part-based Object Representation (VPOR), for commercial item images. The concept of VPOR is to decompose an item image into a set of disjointed partitions, with each of them represents a meaningful semantic parts. User can thus assign non-uniform preferences on the different parts of the commercial item to obtain a personalized recommended results. The experimental results verify our observation and show that the proposed VPOR based commercial item recommendation can achieve better performance than existing text-based and visual-based methods according to the user study.     

UGC网站用户画像研究

近几年,社交网络的高速发展使人们的工作、生活、学习方式发生了重大改变,人们获取知识的方式呈现明显的网络化趋势.人们通过网络获取信息的同时,也在其上留下了个人的痕迹,考虑到现实中获取个人信息成本高昂,捕捉其在网络中留下的痕迹,研究其在网络社会中的“映射”,不失为一种可行的方法.用户画像作为真实用户的虚拟代表,是建立在一系列真实数据之上的用户模型.通过对“知乎”网站的深入挖掘,构建了基于用户基本属性、社交属性、兴趣属性和能力属性四个维度的动态用户画像模型,并对“知乎”网站PM 2.5话题下1303位用户进行实证分析.     

Modeling software architectures with goals in virtual university environment

Recently, the coupling of goal-based and user-centered approaches has resulted in a tremendous impact on the research of software engineering. However, there is no systematic way in the extant approaches to handling the effects of requirements on the structuring of software architectures. As an attempt towards the investigation of the interactions among goals, scenarios, and software architectures, we proposed, in this paper, a goal-driven architecture trade-off analysis method to analyze and construct software architectures in an incremental manner. We also identified criteria for architecture evaluation and verification and explored the possible types of realization of software architectures for goals. The proposed approach is illustrated using the problem domain of virtual university environment.     

基于标签卷积神经网络的文本推荐算法

在面向用户的文章收集系统中,用户会将自己喜欢的文章收集起来构成自己的偏好文章集合,理解用户为何喜欢特定文章、如何精确的找到用户喜欢的文章目前成为了一个重要的研究课题.本文通过基于面向用户的文章收集系统中的一些相关信息,比如文本信息、标签等,来辅助推荐系统更好的进行文章的推荐.文中提出了基于标签卷积神经网络的文本推荐算法,结合神经网络和协同过滤算法的同时,将标签加入到神经网络的设计中.通过在真实的citeulike数据集进行的实验和验证,使用本文的模型可以有效的提高对用户偏好文章预测的准确性.     

基于用户兴趣子类的协作推荐算法

随着电子商务规模的进一步扩大,用户数目和文档资源急剧增加,导致用户数据的极端稀疏性.传统协作推荐算法都无法很好地解决数据稀疏性问题.本文提出一种基于兴趣子类的协作推荐算法,通过子类处理思想的引入,使得某两个用户即使整体不相似而因为“局部点”的相似产生有用的推荐,“最近邻居”的发现变得更容易更准确.实验结果表明,该算法能有效地解决用户数据的极端稀疏问题,在同等条件下,相对于传统协作推荐算法有更好的推荐质量.     

Supporting serendipity: Using ambient intelligence to augment user exploration for data mining and web browsing

Serendipity is the making of fortunate discoveries by accident, and is one of the cornerstones of scientific progress. In today's world of digital data and media, there is now a vast quantity of material that we could potentially encounter, and so there is an increased opportunity of being able to discover interesting things. However, the availability of material does not imply that we will be able to actually find it; the sheer quantity of data mitigates against us being able to discover the interesting nuggets.This paper explores approaches we have taken to support users in their search for interesting and relevant information. The primary concept is the principle that it is more useful to augment user skills in information foraging than it is to try and replace them. We have taken a variety of artificial intelligence, statistical, and visualisation techniques, and combined them with careful design approaches to provide supportive systems that monitor user actions, garner additional information from their surrounding environment and use this enhanced understanding to offer supplemental information that aids the user in their interaction with the system.We present two different systems that have been designed and developed according to these principles. The first system is a data mining system that allows interactive exploration of the data, allowing the user to pose different questions and understand information at different levels of detail. The second supports information foraging of a different sort, aiming to augment users browsing habits in order to help them surf the internet more effectively. Both use ambient intelligence techniques to provide a richer context for the interaction and to help guide it in more effective ways: both have the user as the focal point of the interaction, in control of an iterative exploratory process, working in indirect collaboration with the artificial intelligence components.Each of these systems contains some important concepts of their own: the data mining system has a symbolic genetic algorithm which can be tuned in novel ways to aid knowledge discovery, and which reports results in a user-comprehensible format. The visualisation system supports high-dimensional data, dynamically organised in a three-dimensional space and grouped by similarity. The notions of similarity are further discussed in the internet browsing system, in which an approach to measuring similarity between web pages and a user's interests is presented. We present details of both systems and evaluate their effectiveness.     

Improving sparsity and new user problems in collaborative filtering by clustering the personality factors

In collaborative filtering recommender systems, items recommended to an active user are selected based on the interests of users similar to him/her. Collaborative filtering systems suffer from the ‘sparsity’ and ‘new user’ problems. The former refers to the insufficiency of data about users’ preferences and the latter addresses the lack of enough information about the new-coming user. Clustering users is an effective way to improve the performance of collaborative filtering systems in facing the aforementioned problems. In previous studies, users were clustered based on characteristics such as ratings given by them as well as their age, gender, occupation, and geographical location. On the other hand, studies show that there is a significant relationship between users’ personality traits and their interests. To alleviate the sparsity and new user problems, this paper presents a new collaborative filtering system in which users are clustered based on their ‘personality traits’. In the proposed method, the personality of each user is described according to the big-5 personality model and users with similar personality are placed in the same cluster using K-means algorithm. The unknown ratings of the sparse user-item matrix are then estimated based on the clustered users, and recommendations are found for a new user according to a user-based approach which relays on the interests of the users with similar personality to him/her. In addition, for an existing user in the system, recommendations are offered in an item-based approach in which the similarity of items is estimated based on the ratings of users similar to him/her in personality. The proposed method is compared to some former collaborative filtering systems. The results demonstrate that in facing the data sparsity and new user problems, this method reduces the mean absolute error and improves the precision of the recommendations.