Search

In recent years, with the development of social media platforms, more and more people express their emotions online through short messages. It is quite valuable to detect emotions and relevant topics from such data. However, the feature sparsity of short texts brings challenges to joint topic-emotion models. In many cases, it is necessary to know not only what people think of specific topics, but also which individuals have similar feedback, and what characteristics of these users have. In this paper, we propose a user group based topic-emotion model named UGTE for emotions detection and topic discovery, which can alleviate the above feature sparsity problem of short texts. Specifically, the characteristics of each user are used to discover groups of individuals who share similar emotions, and UGTE aggregates short texts within a group into long pseudo-documents effectively. Experiments conducted on a real-world short text dataset validate the effectiveness of our proposed model.

This book constitutes the refereed proceedings of the 15th Information Retrieval Technology Conference, AIRS 2019, held in Hong Kong, China, in November 2019.The 14 full papers presented together with 3 short papers were carefully reviewed and selected from 27 submissions. The scope of the conference covers applications, systems, technologies and theory aspects of information retrieval in text, audio, image, video and multimedia data.

Citation evaluation has become one of the important indexes to evaluate the paper influence. In the academic area, paper influence is quite important to evaluate the author’s academic level, in addition to help researchers to identify important academic literatures. The traditional citation evaluation methods are mainly developed based on the citation frequency. Nevertheless, other citation characteristics, such as the nature, depth, and distribution are also valuable. Citation nature includes positive, neutral, and negative citations. The depth can be divided into deep, medium, and shallow citations, which could reveal the correlation and the degree of paper reference between citing and cited papers. The citation distribution contains three types: forepart, middle, and tail citations. The data sample of the empirical study is collected from highly cited papers in Chinese and English core journals. Firstly, by using content analysis methods, we analyze the citation evaluation from three dimensions of Chinese papers. Secondly, we compare the similarities and differences between Chinese and English papers. Thirdly, we show a series of results to verify the accuracy and effectiveness of our method.

Contrary to traditional Web information retrieval methods that can only return a ranked list of Web pages and only allow search terms in the query, we have developed a novel learning framework for retrieving precise information blocks from Web pages given a query, which may contain some search terms and prior information such as the layout format of the data. There are two challenging sub-tasks for this problem. One challenge is information block detection, where a Web page is automatically segmented into blocks. Another challenge is to find the information blocks relevant to the query. Existing page segmentation methods, which make use of only visual layout information or only content information, do not consider the query information, leading to a solution having conflict with the information need expressed by the query. Our framework aims at modeling the query and the block features to capture both keyword information and prior information via a probabilistic graphical model. Fisher Kernel, which can effectively incorporate the graphical model, is then employed to accomplish the two sub-tasks in a unified manner, optimizing the final goal of block retrieval performance. We have conducted experiments on benchmark datasets and read-world data. Comparisons between existing methods have been conducted to evaluate the effectiveness of our framework.

With the explosive growth of user-generated multimedia resources in the big data era (e.g., video, audio, text or even their combinations), bridging the semantic gap between low-level features and high-level semantics in multiple modes of data is a critical and indispensable issue. In this paper, we exploit a popular type of metadata called collaborative tags to address this issue. Specifically, we identify basic level concepts and then construct ontologies from the collaborative tags. The generated ontologies can be employed to organize and index user-generated multimedia data. Existing research lacks a principle to supervise the ontology extraction from a human perspective. In contrast, we borrow the idea of a family of concepts called basic level concepts from cognitive psychology. These basic level concepts are frequently used by people in their daily life and when organizing human knowledge. In this paper, we extract ontologies with basic level concepts from collaborative tags. Furthermore, we model the effect of context and present a method for context-aware basic level concepts detection for ontology learning. To the best of our knowledge, this is the first work on context-aware basic level concept detection in collaborative tagging for the construction of ontologies. To evaluate the proposed method, experiments were conducted on real datasets using the Open Directory Project (ODP) as a benchmark. The experimental results illustrate that ontologies extracted by our approach are more rational and human-oriented.

Top-N personalized recommendation has been extensively studied in assisting learners in finding interesting courses in MOOCs. Although existing Top-N personalized recommendation methods have achieved comparable performance, these models have two major shortcomings. First, these models seldom learn an explicit representation of the structural relation of items. Second, most of these models typically obtain a user’s general preference and neglect the recency of items. This paper proposes a Top-N personalized Recommendation with Graph Neural Network (TP-GNN) in the Massive Open Online Course (MOOCs) as a solution to tackle this problem. We explore two different aggregate functions to deal with the user’s sequence neighbors and then use an attention mechanism to generate the final item representations. The experiments on a real-world course dataset demonstrated that TP-GNN could improve the performances. Furthermore, the system developed based on our method obtains positive feedback from the participants, which denotes that our method effectively predicts learners’ preferences and needs.

With the rapid development of Web 2.0 communities, there has been a tremendous increase in user-generated content. Confronting such a vast volume of resources in collaborative tagging systems, users require a novel method for fast exploring and indexing so as to find their desired data. To this end, contextual information is indispensable and critical in understanding user preferences and intentions. In sociolinguistics, context can be classified as verbal context and social context. Compared with verbal context, social context requires not only domain knowledge to build pre-defined contextual attributes but also additional user data. However, to the best of our knowledge, no research has addressed the issue of irrelevant contextual factors for the verbal context model. To bridge this gap, the dominating set obtained from verbal context is proposed in this paper. We present (i) the verbal context graph to model contents and interrelationships of verbal context in folksonomy and thus capture the user intention; (ii) a method of discovering dominating set that provides a good balance of essentiality and integrality to de-emphasize irrelevant contextual factors and to keep the major characteristics of the verbal context graph; and (iii) a revised ranking method for measuring the relevance of a resource to an issued query, a discovered context and an extracted user profile. The experimental results obtained for a public dataset illustrate that the proposed method is more effective than existing baseline approaches.

We present a framework SenticRank to incorporate sentiment for personalized search.Content-based and collaborative sentiment ranking methods are proposed.We compare the proposed sentiment-based search with baselines experimentally.We study the influence of sentiment corpora by using some sentiment dictionaries.Sentiment-based information can significantly improve performance in folksonomy. In recent years, there has been a rapid growth of user-generated data in collaborative tagging (a.k.a. folksonomy-based) systems due to the prevailing of Web 2.0 communities. To effectively assist users to find their desired resources, it is critical to understand user behaviors and preferences. Tag-based profile techniques, which model users and resources by a vector of relevant tags, are widely employed in folksonomy-based systems. This is mainly because that personalized search and recommendations can be facilitated by measuring relevance between user profiles and resource profiles. However, conventional measurements neglect the sentiment aspect of user-generated tags. In fact, tags can be very emotional and subjective, as users usually express their perceptions and feelings about the resources by tags. Therefore, it is necessary to take sentiment relevance into account into measurements. In this paper, we present a novel generic framework SenticRank to incorporate various sentiment information to various sentiment-based information for personalized search by user profiles and resource profiles. In this framework, content-based sentiment ranking and collaborative sentiment ranking methods are proposed to obtain sentiment-based personalized ranking. To the best of our knowledge, this is the first work of integrating sentiment information to address the problem of the personalized tag-based search in collaborative tagging systems. Moreover, we compare the proposed sentiment-based personalized search with baselines in the experiments, the results of which have verified the effectiveness of the proposed framework. In addition, we study the influences by popular sentiment dictionaries, and SenticNet is the most prominent knowledge base to boost the performance of personalized search in folksonomy.

In recommender systems, personal characteristic is possessed by not only users but also displaying products. Users have their personal rating patterns while products have different characteristics that attract users. This information can be explicitly exploited from the review text. However, most existing methods only model the review text as a topic preference of products, without considering the perspectives of users and products simultaneously. In this paper, we propose a user-product topic model to capture both user preferences and attractive characteristics of products. Different from conventional collaborative filtering in conjunction with topic models, we use non-negative matrix tri-factorization to jointly reveal the characteristic of users and products. Experiments on two real-world data sets validate the effectiveness of our method in Top-N recommendations.

The rapid development of Internet, social media, and news portals has provided a large amount of information in various aspects. Confronting such plenty of resources, it is valuable to develop effective clustering approaches. However, performance of traditional clustering models on web resources is not good enough due to the high dimension. In this paper, we propose a clustering model based on topic model and density peaks. Our model combines biterm topic model and clustering by fast search of density peaks, which firstly extract a set of features with the co-occurrence of two words from the original documents, followed by clustering analysis via topical features. Web resources are translated from raw data into clusters, and evaluation on clustering results of center part verifies the effectiveness of the proposed method.

Stimulated by the arrival of the big data era, various and heterogeneous data sources such as data in social networks, mobile devices and sensor data for users have emerged, mirroring characteristics and preferences of data owners. These data sources are often used to construct user profiles so as to facilitate personalized services like recommendations or personalized data access. In the context of second language learning, learner data involve learning logs, standard test results, and individual learning preferences and styles. Given its attribute of reflecting the characteristics of learners, such data can be exploited to build the learner profiles. However, these data sources possibly include noises or bias, and hence influence the reliability of the correspondingly constructed learner profiles. Consequently, the inaccurate profiles may result in ineffective learning tasks that are generated by e-Learning systems. To tackle this issue, it is significant and critical to evaluate the accuracy of learner profiles. In a response to this call, we propose a novel metric named “profile mean square error” to examine the accuracy of learner profiles founded upon diverse sources. We also demonstrate how to construct various learner profiles though applying different data sources such as learning logs, standard test results, and personal learning preferences in e-Learning systems and pedagogical activities. Moreover, we conduct an experimental study among some second language learners, the results of which illustrate that the most accurate profiles are generated from multiple data sources if they are integrated in a rational way.

As a large amount of learning materials available in e-learning, automatic summarization has been utilized in digital library for e-learning. Multi-document summarization extracts information from multiple texts documents. It improves the effectiveness of retrieval and accessibility of learning materials in e-learning. Related study has shown that hierarchical summarization is a promising technique. Hierarchical summarization has been extended to summarization of multiple documents. This paper investigates the impact of organization of documents on hierarchical summarization. Automatic document clustering technique is developed for organization of learning materials. A multi-document hierarchical summarization system is developed based on these techniques. The system provides an essential tool for e-learning.

Multiple document summarization technique can be utilized for improving the effectiveness of retrieval and accessibility of learning objects in e-learning. Related study has shown that hierarchical summarization is a promising technique which can effectively extract the most important information from the source documents. Hierarchical summarization has been extended to summarization of multiple documents. Three hierarchical structures were proposed to organize a set of related documents. This paper investigates the impact of document structure on hierarchical summarization.

We develop an unsupervised learning framework for tackling the problem of automatic site-specific ontology discovery from multiple pages of a Web site. To harness the uncertainty involved, our framework is designed based on a generative model which models the generation of text fragments contained in the pages of a Web site. One characteristic of our framework is that we consider clues from multiple pages collected from the Web site. Another characteristic is that we learn the regularities of the layout format to discover the site-specific ontology via stochastic grammatical inference. To accomplish the goal of ontology discovery, the ontology information blocks of a Web page are identified by making use of the site invariant information. We have conducted extensive experiments using real-world Web sites. Comparisons between existing methods and our framework have been carried out to demonstrate the effectiveness of our framework.

Counterterrorism is one of the major challenges to the society. In order to flight again the terrorists, it is very important to have a through understanding of the terrorism incidents. However, it is impossible for a human to read all the information related to a terrorism incident because of the large volume of information. Summarization technique is urgently required for analysis of terrorism incident. In this work, we propose a multi-document summarization system to extract the critical information from terrorism incidents. News stories of a terrorism incident are organized into a hierarchical tree structure. Fractal summarization model is employed to generate a summary for all the news stories. Experimental results show that our system can effectively extract the most important information for the incident.

Recommender systems make suggestions to users. Collaborative filtering techniques make the predictions by using the ratings on items of other users. In this paper, we have studied item-based and user-based collaborative filtering techniques. We identify the shortcomings of current filtering techniques. The performance of recommender systems was deeply affected by user's rating behavior. We propose some improvements to overcome this limitation. User evaluation has been conducted. Experiment results show that the new algorithms improve the performance of recommender systems significantly.

As a result of the recent information explosion, there is an increasing demand for automatic summarization, and human abstractors often synthesize summaries that are based on sentences that have been extracted by machine. However, the quality of machine-generated summaries is not high. As a special application of information retrieval systems, the precision of automatic summarization can be improved by user relevance feedback, in which the human abstractor can direct the sentence extraction process and useful information can be retrieved efficiently. Automatic summarization with relevance feedback is a helpful tool to assist professional abstractors in generating summaries, and in this work we propose a relevance feedback model for fractal summarization. The results of the experiment show that relevance feedback effectively improves the performance of automatic fractal summarization.