Search

The online question answering (QA) community has been popular in recent years. In this paper, we focus on the online health question answering (HQA) community. The HQA community provides a platform for health consumers to inquire about health information. There are two ways to use this platform. One is to post a question and wait for answers to be provided by authenticated doctors. The other is to search for relevant questions with answers. For the latter, health consumers may prefer an accepted answer marked by the previous health consumer. However, there is a large proportion of questions without an accepted answer and it is inconvenient for people who want to search for relevant questions. To address this issue, we aim to select high-quality answers from the answers without marked accepted answers. We propose a deep learning approach to achieve this goal. To train the model for the prediction of answer quality, we first view the accepted answer as the positive answer and propose a method to label the negative answer. Next, we capture the semantic information on the question and the answer by the deep learning structure. We then combine the information to predict the quality score of the answer. We collect data from one of the biggest Chinese HQA community and divide them into groups by the medical departments for detailed analysis. Finally, we conduct experiments to show the effectiveness of categorization and the labeling method. The results show that our approach outperforms other studies and we further research into the differences among the results of different categories.

Recommendation Systems (RSs) are effective tools to find interesting items for users in explosive information on the network. However, traditional RSs still suffer from limitations. The first problem is feature extraction; lack of item information leads to the difficulty of semantic discrimination between items. The second one is the cold-start problem that happens when we do not have much information related to user ratings. This paper aims to alleviate the drawback by exploiting benefits from LOD - a data source using structured data presents its data. The method is tested on MovieLens 1M dataset and DBpedia - the subset of LOD to enhance semantically. In contrast to previous work, we use PathSim to capture semantic between items. Furthermore, we use five-star rating that uses a 5-star scale instead of binary rating that uses 0 and 1 to present user preference for LambdaMart model. Overall, our approach outperforms SPrank, a hybrid recommendation algorithm proposed in the literature.

The prediction of future events has great importance in many applications. The prediction is based on episode rules which are composed of events and two time constraints which require all the events in the episode rule and in the predicate of the rule to occur in a time interval, respectively. In an event stream, a sequence of events which matches the predicate of the rule satisfying the specified time constraint is called an occurrence of the predicate. After finding the occurrence, the consequent event which will occur in a time interval can be predicted. However, the time intervals computed from some occurrences for predicting the event can be contained in the time intervals computed from other occurrence and become redundant. As a result, how to design an efficient and effective event predictor in a stream environment is challenging. In this paper, an effective scheme is proposed to avoid matching the predicate events corresponding to redundant time intervals for prediction. Based on the scheme, we respectively consider two methodologies, forward retrieval and backward retrieval, for the efficient matching of predicate events over event streams. The approach based on forward retrieval construct a queue structure to incrementally maintain parts of the matched results as events arrive, and thus it avoids backward scans of the event stream. On the other hand, the approach based on backward retrieval maintains the recently arrived events in a tree structure. The matching of predicate events is triggered by identifiable events and achieved by an efficient retrieval on the tree structure, which avoids exhaustive scans of the arrived events. By running a series of experiments, we show that each of the proposed approaches has its advantages on particular data distributions and parameter settings.

Mining frequent sequences in large databases has been an important research topic. The main challenge of mining frequent sequences is the high processing cost due to the large amount of data. In this paper, we propose a novel strategy to find all the frequent sequences without having to compute the support counts of non-frequent sequences. The previous works prune candidate sequences based on the frequent sequences with shorter lengths, while our strategy prunes candidate sequences according to the non-frequent sequences with the same lengths. As a result, our strategy can cooperate with the previous works to achieve a better performance. We then identify three major strategies used in the previous works and combine them with our strategy into an efficient algorithm. The novelty of our algorithm lies in its ability to dynamically switch from a previous strategy to our new strategy in the mining process for a better performance. Experiment results show that our algorithm outperforms the previous ones under various parameter settings.

It is foreseen that more and more music objects in symbolic format and multimedia objects, such as audio, video, or lyrics, integrated with symbolic music representation (SMR) will be published and broadcasted via the Internet. The SMRs of the flowing songs or multimedia objects will form a music stream. Many interesting applications based on music streams, such as interactive music tutorials, distance music education, and similar theme searching, make the research of content-based retrieval over music streams much important. We consider multiple queries with error tolerances over music streams and address the issue of approximate matching in this environment. We propose a novel approach to continuously process multiple queries over the music streams for finding all the music segments that are similar to the queries. Our approach is based on the concept of n-grams, and two mechanisms are designed to reduce the heavy computation of approximate matching. One mechanism uses the clustering of query n-grams to prune the query n-grams that are irrelevant to the incoming data n-gram. The other mechanism records the data n-gram that matches a query n-gram as a partial answer and incrementally merges the partial answers of the same query. We implement a prototype system for experiments in which songs in the MIDI format are continuously broadcasted, and the user can specify musical segments as queries to monitor the music streams. Experiment results show the effectiveness and efficiency of the proposed approach.

In this paper, we propose a novel algorithm for mining frequent sequences from transaction databases. The transactions of the same customers form a set of customer sequences. A sequence (an ordered list of itemsets) is frequent if the number of customer sequences containing it satisfies the user-specified threshold. The 1-sequence is a special type of sequences because it consists of only a single itemset instead of an ordered list, while the k-sequence is a sequence composed of k itemsets. Compared with the cost of mining frequent k-sequences (k???2), the cost of mining frequent 1-sequences is negligible. We adopt a two-phase architecture to find the two types of frequent sequences separately in order that the discovery of frequent k-sequences can be well designed and optimized. For efficient frequent k-sequence mining, every frequent 1-sequence is encoded as a unique symbol and the database is transformed into one constituted by the symbols. We find that it is unnecessary to encode all the frequent 1-seqences, and make full use of the discovered frequent 1-sequences to transform the database into one with a smaller size. For every k???2, the customer sequences in the transformed database are scanned to find all the frequent k-sequences. We devise the compact representation for a customer sequence and elaborate the method to enumerate all distinct subsequences from a customer sequence without redundant scans. The soundness of the proposed approach is verified and a number of experiments are performed. The results show that our approach outperforms the previous works in both scalability and execution time.

Querying polyphonic music from a large data collection is an interesting topic. Recently, researchers have attempted to provide efficient methods for content-based retrieval in polyphonic music databases where queries are polyphonic. However, most of them do not work well for similarity search, which is important to many applications. In this paper, we propose three polyphonic representations with the associated similarity measures and a novel method to retrieve k music works that contain segments most similar to the query. In general, most of the index-based methods for similarity search generate all the possible answers to the query and then perform exact matching on the index for each possible answer. Based on the edit distance, our method generates only a few possible answers by performing the deletion and/or replacement operations on the query. Each possible answer is then used to perform exact matching on a list-based index, which allows the insertion operations to be performed. For each possible answer, its edit distance to the query is regarded as a lower bound of the edit distances between the matched results and the query. Based on the kNN results that match a possible answer, the possible answers that cannot provide better results are skipped. By using this mechanism, we design a method for efficient kNN search in polyphonic music databases. The experimental results show that our method outperforms the previous methods in efficiency. We also evaluate the effectiveness of our method by showing the search results to the musician and nonmusician user groups. The experimental results provide useful guidelines on the design of a polyphonic music database.

As the population of web users grows, the variety of user behaviors on accessing information also grows, which has a great impact on the network utilization. Recently, many efforts have been made to analyze user behaviors on the WWW. In this paper, we represent user behaviors by sequences of consecutive web page accesses, derived from the access log of a proxy server. Moreover, the frequent sequences are discovered and organized as an index. Based on the index, we propose a scheme for predicting user requests and a proxy-based framework for prefetching web pages. We perform experiments on real data. The results show that our approach makes the predictions with a high degree of accuracy with little overhead. In the experiments, the best hit ratio of the prediction achieves 75.69%, while the longest time to make a prediction only requires 2.3 ms.

Ridesharing has the great opportunity to reduce the consumption of energy and the emission of harmful gases, and to let people share the traffic costs with others. Most of the current ridesharing systems simply provide a number of candidates for users to choose. Time-consuming negotiation often discourages people from ridesharing. We propose a novel approach that assigns users to form ridesharing groups according to their routes and payments. Given a driver, our goal is to find a group of passengers who will pay the driver the most. Under the payment scheme, the passengers who share rides on the same route will equally share the expense with the driver. For the prompt response to an online system, our approach aims for the near-optimal group, where the available seats on the driver route are occupied by passengers as many as possible. Compared with the previous methods, the experiment results show that our approach incurs a little overhead but obtains answers of good quality, measured by the driver's saving, under various parameter settings.

Event prediction over event streams is an important problem with broad applications. For this problem, rules with predicate events and consequent events are given, and then current events are matched with the predicate events to predict future events. Over the event stream, some matches of predicate events may trigger duplicate predictions, and an effective scheme is proposed to avoid such redundancies. Based on the scheme, we propose a novel approach CBS-Tree to efficiently match the predicate events over event streams. The CBS-Tree approach maintains the recently arrived events as a tree structure, and an efficient algorithm is proposed for the matching of predicate events on the tree structure, which avoids exhaustive scans of the arrived events. By running a series of experiments, we show that our approach is more efficient than the previous work for most cases.

The one-shot shortest path query has been studied for decades. However, in the applications on road networks, users are actually interested in the path with the minimum travel time (the fastest path), which varies as time goes. This motivates us to study the continuous evaluation of fastest path queries in order to capture the dynamics of road networks. Repeatedly evaluating a large number of fastest path queries at every moment is infeasible due to its computationally expensive cost. We propose a novel approach that employs the concept of the affecting area and the tolerance parameter to avoid the reevaluation while the travel time of the current answer is close enough to that of the fastest path. Furthermore, a grid-based index is designed to achieve the efficient processing of multiple queries. Experiments on real datasets show significant reduction on the total amount of reevaluation and therefore the cost for reevaluating a query.

In some applications, data may possess both location-dependent and location-independent attributes. For example, in a job database, each job can be associated with both location-dependent attributes, e.g., the location of the work place, and location-independent ones, e.g., the salary. A person who uses this database to find a job may prefer not only a shorter distance between his/her house and the work place but also a higher salary. Therefore, a query with both concepts of "shorter distance" and "higher salary" should be considered to meet the user's needs. We call it the heterogeneous k-nearest neighbor (HkNN) query in distinction from the traditional k-nearest neighbor (kNN) query in the spatial domain, which only concerns location-dependent attributes. To our knowledge, this paper is the first work proposing a generic framework for solving the HkNN query. We propose an efficient approach based on the bounding property for the HkNN query evaluation. Furthermore, we provide an update mechanism for continuously monitoring the HkNN queries in a dynamic environment. Experimental results verify that the proposed framework is both efficient and scalable.

Since tree-structured data such as XML files are widely used for data representation and exchange on the Internet, discovering frequent tree patterns over tree-structured data streams becomes an interesting issue. In this paper, we propose an online algorithm to continuously discover the current set of frequent tree patterns from the data stream. A novel and efficient technique is introduced to incrementally generate all candidate tree patterns without duplicates. Moreover, a framework for counting the approximate frequencies of the candidate tree patterns is presented. Combining these techniques, the proposed approach is able to compute frequent tree patterns with guarantees of completeness and accuracy.

Mining frequent itemsets has been widely studied over the last decade. Past research focuses on mining frequent itemsets from static databases. In many of the new applications, data flow through the Internet or sensor networks. It is challenging to extend the mining techniques to such a dynamic environment. The main challenges include a quick response to the continuous request, a compact summary of the data stream, and a mechanism that adapts to the limited resources. In this paper, we develop a novel approach for mining frequent itemsets from data streams based on a time-sensitive sliding window model. Our approach consists of a storage structure that captures all possible frequent itemsets and a table providing approximate counts of the expired data items, whose size can be adjusted by the available storage space. Experiment results show that in our approach both the execution time and the storage space remain small under various parameter settings. In addition, our approach guarantees no false alarm or no false dismissal to the results yielded.

Pattern extraction from music strings is an important problem. The patterns extracted from music strings can be used as features for music retrieval or analysis. Previous works on music pattern extraction only focus on exact repeating patterns. However, music segments with minor differences may sound similar. The concept of the prototypical melody has therefore been proposed to represent these similar music segments. In musicology, the number of music segments that are similar to a prototypical melody implies the importance degree of the prototypical melody to the music work. In this paper, a novel approach is developed to extract all the prototypical melodies in a music work. Our approach considers each music segment as a candidate for the prototypical melody and uses the edit distance to determine the set of music segments that are similar to this candidate. A lower bounding mechanism, which estimates the number of similar music segments for each candidate and prunes the impossible candidates is designed to speed up the process. Experiments are performed on a real data set and the results show a significant improvement of our approach over the existing approaches in the average response time.

In this paper, we introduce a novel algorithm for mining sequential patterns from transaction databases. Since the FP-tree based approach is efficient in mining frequent itemsets, we adapt it to find frequent 1-sequences. For efficient frequent k-sequence mining, every frequent 1-sequence is encoded as a unique symbol and the database is transformed into one in the symbolic form. We observe that it is unnecessary to encode all the frequent 1-seqences, and make full use of the discovered frequent 1-sequences to transform the database into one with a smallest size. To discover the frequent k-sequences, we design a tree structure to store the candidates. Each customer sequence is then scanned to decide whether the candidates are frequent k-sequences. We propose a technique to avoid redundantly enumerating the identical k-subsequences from a customer sequence to speed up the process. Moreover, the tree structure is designed in a way such that the supports of the candidates can be incremented for a customer sequence by a single sequential traversal of the tree. The experiment results show that our approach outperforms the previous works in various aspects including the scalability and the execution time.

Mining sequential patterns in large databases is an important research topic. The main challenge of mining sequential patterns is the high processing cost due to the large amount of data. We propose a new strategy called direct sequence comparison (abbreviated as DISC), which can find frequent sequences without having to compute the support counts of nonfrequent sequences. The main difference between the DISC strategy and the previous works is the way to prune nonfrequent sequences. The previous works are based on the antimonotone property, which prune the nonfrequent sequences according to the frequent sequences with shorter lengths. On the contrary, the DISC strategy prunes the nonfrequent sequences according to the other sequences with the same length. Moreover, we summarize three strategies used in the previous works and design an efficient algorithm called DISC-all to take advantages of all the four strategies. The experimental results show that the DISC-all algorithm outperforms the PrefixSpan algorithm on mining frequent sequences in large databases. In addition, we analyze these strategies to design the dynamic version of our algorithm, which achieves a much better performance.

In this paper, we propose a novel representation of sequences based on the structural information of the sequences. A sequence is represented by a set of rules, which are derived from its subsequences. There are two types of subsequences of interest. One is called frequent pattern, which is a subsequence appearing often enough in the sequence. The other is called correlative pattern, which is a subsequence composed of highly correlated elements. The rules derived from the frequent patterns are called frequent rules, while the ones derived from the correlative patterns are called correlative rules. By considering music objects as sequences, we represent each of them as a set of rules and design a similarity function for effective music retrieval. The experimental results show that our approaches outperform the approaches based on the Markov-model on the average precision.

With the popularity of multimedia applications, a large amount of music data has been accumulated on the Internet. Automatic classification of music data becomes a critical technique for providing an efficient and effective retrieval of music data. In this paper, we propose a new approach for classifying music data based on their contents. In this approach, we focus on monophonic music features represented as rhythmic and melodic sequences. Moreover, we use repeating patterns of music data to do music classification. For each pattern discovered from a group of music data, we employ a series of measurements to estimate its usefulness for classifying this group of music data. According to the patterns contained in a music piece, we determine which class it should be assigned to. We perform a series of experiments and the results show that our approach performs on average better than the approach based on the probability distribution of contextual information in music.

Querying polyphonic music from a large data collection is an interesting and challenging topic. Recently, researchers attempt to provide efficient techniques for content-based retrieval in polyphonic music databases where queries can also be polyphonic. However, most of the techniques do not perform the approximate matching well. In this paper, we present a novel method to efficiently retrieve k music works that contain segments most similar to the user query based on the edit distance. A list-based index structure is first constructed using the feature of the polyphony. A set of candidate approximate answers is then generated for the user query. A lower bounding mechanism is proposed to prune these candidates such that the k answers can be obtained efficiently. The efficiency of the proposed method is evaluated by real data set and synthetic data set, reporting significant improvement over existing approaches in the response time yielded.

The dramatic growth of the Web has brought about the rapid accumulation of data and the increasing possibility of information sharing. As the population on the Web grows, the analysis of user interests and behaviors will provide hints on how to improve the quality of service. We define user interests and behaviors based on the documents read by the user. A method for mining such user interests and behaviors is then presented. In this way, each user is associated with a set of interests and behaviors, which is stored in the user profile. In addition, we define six types of user profiles and a distance measure to classify users into clusters. Finally, three kinds of recommendation services using the clustered results are realized. For performance evaluation, we implement these services on the Web to make experiments on real data/users. The results show that the average acceptance rates of these services range from 71.5% to 94.6%.

Searching information from the WWW efficiently and effectively has become a very important issue. We apply the information filtering concept to finding good matches of Web pages to what the user needs. Furthermore, we propose four new methods for indexing user information needs and show their efficiency in comparison with two well-known approaches.

Building a system for the users on the Internet to search interested resources or to provide guided browsing services has formed an important trend as a result of the rapid accumulation of information, the booming development of information system and the increasing requirement of resource sharing. In this paper, we present the Integrated resource Query and guidedBrowsing System (IQBS), which has been implemented at our laboratory, to achieve this goal. The scheme we proposed takes advantage of the strong power of the database system to structure and query data, and greatly promote the qua!ity and performance of searching resources in Internet. In addition, mechanisms are designed to guide the users on the Internet to conveniently browse resources with a reduced traffic cost. Keywords: Internet resource discovery, World wide web, Query processing, Rough set, Guided browsing, Supervised learning, Hypertext management.