RUMVO : user modeling using runtime verification

Abstract

The World Wide Web (WWW) is an ever-growing source of information. Especially with the introduction of Web 2.0, people are encouraged to feed new data daily, being it blogs, social networks, forums, and so on. This has made the WWW an integral part of our everyday lives, putting it as a primary reference for anyone who wants to be updated on any particular aspect of life. However this has also introduced a new challenge for 福利在线免费 Retrieval (IR) researchers since, together with this exponential growth of the WWW, users' requests have become more sophisticated and demanding especially in today's hectic daily schedules. The challenge is to provide ways how to make the users and the relevant information find each other in the most efficient and effective way possible. This field of study is known as User modeling (UM), where the main task is to produce a representation of the user's likes and retrieve the information accordingly. Several methods have been introduced to accomplish this task. The general idea of tackling UM is to collect user browsing information, either on a long-term or on a short-term period of time, capture the user feedback, either implicitly or explicitly, and try to work out a suggestion based either on the content of the documents viewed or using by comparing a particular user's behavior to other similar users (called collaborative filtering). One drawback encountered was that UM modules lack reusability due to the high level of subjectivity of properties and also because of the fact that different websites have different architectures. Runtime verification (RV) is a method used to analyze the behavior of a program, comparing a single thread of execution to a previously defined formal specification in order to assure correctness for that thread. RV monitors are the programs performing such validation checks. These are in charge of (i) collect and report information about system events (ii) handling violation of properties. Monitors are separate programs to the systems, analyzing correctness externally, however performing actions on the actual system when necessary. RUM VO adapts an RV approach to the UM task. The basic intuition of the solution is to analyze the user's interaction with the use of an RV monitor, and then perform the appropriate operations to produce and apply a UM. Given the problem of lack of code reusability of UM modules, however, this project attempts to provide mechanisms through which developers can define their own properties for a UM feature. These mechanisms are intended to be portable between different systems, and this was accomplished by defining them as RV monitors. The advantages observed in taking such an approach were that a developer would only need to define an access class to the database, and the properties invoking the mechanisms that will produce the UM; all of this without having to perform intense changes to the original system.