Blogs

As a target end-user for ALERT, KDE has been involved from the start to help with the specification of the system and to provide insights into the open source way of working. Now, in the final weeks of the initial ALERT Project, we are playing a big role in testing the software, validating its usefulness and providing input for future improvements.
 

During the past months we have been evaluating the recommendation service component and tweaking some of the inner workings. Moreover, we extended its functionality by extracting topic models based on the annotations extracted by the KEUI component.

The ALERT system in already fully functional and we are running the trials in order to test the performance and functionalities of the system in live environments. Our end-users who are participating in the trial are using the system with real data coming from their open source communities.

We have 3 communities participating in the trial, namely 5-5 persons from the KDE/Solid, the OW2/Petals and the Morpheo/Optimis communities each. Instances of ALERT have been installed and sensors for the relevant community knowledge basis have also been implemented.

In order to achieve our goal of improving collaboration between open source developers in real-time, the accessibility to underlying information sources is quite important. As one of the unique features of the ALERT system is the notification component which allows developers to subscribe to complex patterns which rely on heterogeneous data streams, the major design goal of the implementation of sensors was their capability of real-time detection of changes.

Technical evaluation of the integrated ALERT system is one of the challenging ongoing tasks on the project. In order to support this very important task, and concurrently to make it more efficient, CIM development team developed two additional support testing tools:

• Event Simulator
• Event Monitor

One of the most significant features of ALERT is its capacity to merge data retrieved from different sources in a software development project. In this way, ALERT can gather information from version control systems, bugtrackers, forums, wikis and mailing lists, creating a multifaceted overview of the software project and the profiles of developers working on it.

Back when we started the project, we at ALERT designed a product that would be innovating, if not revolutionary, in many ways. Our goal could be synthetized as such: "allow a developer to focus on developing, by handling automatically tedious tasks such as bug triaging, bug assignement and information gathering".

This is accomplished in ALERT through three axes:

• automate bug triaging as much as possible by doing a multi-criteria analysis on new bug reports.

The ontologies allow to represent a knowledge domain in a way that could be understood by machines. This allows performing some kind of reasoning in order to enrich the available information. However, there are a set of domains which constantly changes, new concepts and also relations emerges every day and must be taken into account. For this reason, it is required a mechanism that allow update of ontological resources during his lifecycle, otherwise, the ontological resources will get out of date and eventually will not provide a good enough representation of the domain.

Many open source software projects have a bug tracking system incorporated into their software development process. Bug tracking system (BTS) allows users of the software to report issues they encountered and also to suggest new features. An unfortunate problem with BTS is that users usually don’t spend much time checking if the same problem is already reported but instead create a new issue. Consequently, a significant percentage of the reported issues are duplicates of previously reported issues.

Free software communities, such as those that are the target end users of the ALERT software, are in many ways different to large companies. In a company, there is a hierarchical structure and normally one person you can go to for a definitive answer or to get something done. Within a volunteer community, there is much less hierarchy and things tend to be decided by consensus.