Dongli Han

Efficient Event Matching in Publish/subscribe: Based on Routing Destination and Matching History

Efficient event matching in a content-based publish/subscribe system is an important problem. Most existing matching solutions focus on subscription relations, such as cover and merge. We observe that event matching can leverage matching history and routing destination as well. Once one of the subscriptions associated with a destination is matched, undecided subscriptions associated with the destination need not be evaluated. Since each different subscription has a different possibilities to match an event, different matching order of subscriptions can result in different matching time. We propose a new efficient event matching approach based on matching order, cover relation, matching history (matching possibility) and routing destination. Our solution indexes subscriptions in an efficient way. Based on event matching history, cover relation and routing destination, our matching approach can cause efficient matching with a special match order, and can easily register or cancel subscriptions. We also propose a highly efficient index structure for numeric filters, which can evaluate N numeric filters with O(logN) time complexity. Quantity analysis of our solution is given. We carry out detailed experimentations to verify the effectiveness of our approach. The results show that our approach achieves high matching efficiency.

Structural Analysis of Compound Words in Japanese Using Semantic Dependency Relations

Structural analysis of compound words is a necessary and important process in natural language processing. Proposed here is a corpus- and statistics-based method for the structural analysis of compound words in Japanese. We determine the structure of a compound word by using an Internet corpus and calculating the strength of word association among its constituent words. Experiments with 5, 6, 7, and 8 kanji compound words show that our method works well and its performance is better than those of other comparable studies.

A New Approach for Overload Management in Content-based Publish/Subscribe

Overload management is of vital importance in wide-area publish/subscribe systems, yet current solutions are best-effort. In this paper, we present an admission control scheme for overload management in large-scale and scalable content-based publish/subscribe systems. We analyze the stumbling block for implementing admission control in publish/subscribe systems, and point out how it differs from admission control schemes in other research areas. We propose a cover relation based algorithm to compute subscription resource requirements and an admission control algorithm based on subscription routing. The scheme ensures time, space and flows decoupling without sacrificing scalability of publish/subscribe systems. Finally, we conduct experiments to verify the effectiveness of the scheme.

Topic control in a free conversation system

Conversation systems can be divided into two types. Task-oriented ones and non-task-oriented ones. Most recent researches on the latter one have focused on how to make the user continue the conversation without getting bored or tired. Some studies propose several ways to enrich the expressive forms of words or sentences. Others make efforts on topic association. Both solutions try to preserve the conversation interesting so that the user would like to continue talking with the computer. In this paper, we talk about the significance of providing topics, and give an approach to provide and control topics for a free conversation system. Specifically, the system extracts topics from the Web according to the users interest automatically, and keeps the conversation going on by providing the fabricated topics to the user. Also the system constantly detects the timing when the user might have lost interests in the current topic, and then changes the topic to another one obtained from a topic website. Experiments and questionnaires show that our method works well.

Automatic Utterance Generation by Keeping Track of the Conversation’s Focus within the Utterance Window

The insufficiency in methods for generating utterances still remains as a critical issue unsolved in the community of non-task-oriented conversation. Previous studies provide various strategies to enrich the methods for generating utterances, thus making the conversation systems or agents appear more interesting. However, none of them could escape from the fact that they all generate utterances depending mainly on some particular kinds of templates or augmented templates. We propose here in this paper a thorough modification to a preceding work to address this problem. Specifically, we first introduce a concept Utterance Window to strengthen the association between continuous utterances, and then employ a Two-starting-word Markov connection to cope with the ease of losing focus of the current utterance. In addition, we try to keep tract of user’s interests and reflecting them in the process of topic-word extraction and utterance generation as well. The experimental results show the effectiveness of our method.

Recognition and Utilization of Clausal Relations in Complex Sentences for Improving the Performance of Machine Translation Systems

Machine translation systems are inefficient when translating complex sentences. An important reason for this, among others, comes from the fact that the systems translate a complex sentence without taking into account the clausal relation between main and subordinate clauses. We devise a method for recognizing and utilizing the clausal relation to make the machine translation systems self-improve their performance. First, we attempt to find the clausal relation between the two clauses in a complex sentence through a machine learning mechanism. Second, utilizing this information, we try to pre-edit or modify the sentence so that the clausal relation is expressed overtly in the modified sentence. Experiments with a machine translation system have shown that our idea and methodology improve the performance of the system.

Comparison of multiple recommendation methods of similar onomatopoeia

Onomatopoeia is a generic name for onomatopoeia and mimetic words. Using onomatopoeia can express the behavior and state of things in more detail, widening the range of communication. However, learning onomatopoeia has been a difficult task for Japanese learners. There are several existing studies aiming at a support with onomatopoeia learning, while no platform is available to help learners find similar onomatopoeia based on different criteria. In this paper, we have developed a system that proposes similar onomatopoeia for an input in three manners: one with a dictionary, and two based on statistics. A comparison with an existing system shows the effectiveness of our approach and exposes some future issues.

Efficiency improvement of literature survey based on citation-reason visualization

Literature survey is the first step of scientific research. However, this process could be quite time-consuming. Previous works aiming to automatically address this issue employ textual similarity or reference-relation between papers, while neither of which is flexible enough for context-specific demanded In this paper, we have proposed an idea to narrow down the search range for relevant literatures based on citation-reason visualization. Practicality and efficiency improvement of literature survey has been confirmed according to the results of the evaluation experiments.

Acquisition of Scientific Literatures based on Citation-reason Visualization

When carrying out scientific research, the first step is to acquire relevant papers. It is easy to grab vast numbers of papers by inputting a keyword into a digital library or an online search engine. However, reading all the retrieved papers to find the most relevant ones is agonizingly time-consuming. Previous works have tried to improve paper search by clustering papers with their mutual similarity based on reference relations, including limited use of the type of citation (e.g. providing background vs. using specific method or data). However, previously proposed methods only classify or organize the papers from one point of view, and hence not flexible enough for user or context-specific demands. Moreover, none of the previous works has built a practical system based on a paper database. In this paper, we first establish a paper database from an open-access paper source, then use machine learning to automatically predict the reason for each citation between papers, and finally visualize the resulting information in an application system to help users more efficiently find the papers relevant to their personal uses. User studies employing the system show the effectiveness of our approach.

A case study on experimental-data validation for natural language processing

Text data randomly extracted from a particular corpus are usually employed by NLP (Natural language processing) systems as experimental data. However, it is hard to determine whether the experimental data is appropriate without a reasonable validation process. We in this paper describe a data validation approach for a NLP-based e-learning system we have built before.