Mobile devices contain more personal data such as GPS location, contacts and music, with which users can create innovative and pragmatic mashup applications for different areas such as social networking, E-commerce, and entertainment. We propose a semantic-based mobile mashup platform which enables users to create mashup applications by simply selecting service nodes, linking them together and configuring some connection parameters. Our platform also offers a recommendation mechanism on linkable services by adding semantic annotation to service description, so that users do not need to read specifications of web services in order to find out linkable ones. Therefore, users can focus more on the innovation and practicability of their mashup applications, which will surely result in the emergence of abundant mobile mashup applications.
Presently, the number of mobile mashup applications grows rapidly with fast mobile
computing development. Current mobile devices’ multimedia (e.g. Camera and media
playback), sensing (e.g. GPS, Bluetooth, RFID and barcode readers) and
communicating (e.g. GPRS, UMTS, Bluetooth, Wi-Fi) capabilities are ideal sentient
intermediaries with which users can obtain information for mashup applications.
Sentient Graffiti (SG) [
2.1
On our platform, users are able to make a mashup application by selecting a few service nodes and connecting them together. The platform is able to find matched services and the inter-connections among them, which allows users to easily create their applications without knowing every single detail about each service. The recommendation mechanism is implemented by adding semantic annotation to each input or output parameter, so that all services have a common interface. The linkability judgment of two services is shown in Fig. 1.
Mobile GPS Service - getLocation lat lon outputs geo.latitude geo.longitude semantic annotation
VirtualEarth Service - addPushpin latitude longitude inputs
Along with this basic parameter-match recommendation theory, we also include
other criteria to filter and rank recommended services. The first criterion is matching
degree (the numbers of parameters matched) of two operations. Required input
parameters must be met, or the service will be filtered out. After that, with bigger
matching degree, the recommended service receives higher rank. The second criterion
is conditional probability calculation [
Besides the mashup constructing UI, our platform also includes a mashup execution engine. The execution result is shown in a new page temporarily with a list view or map view. Moreover, advanced users who expect more complicated functions can also encapsulate services by conforming to a specified standard by themselves. 2.2
The mashup constructing UI is shown in Fig. 2. On the home screen, users can either execute mashups which are already made by themselves or by others or construct a new one by following a few steps. Firstly, users select a node from the service list. Then they will get a list of recommended services and they can add recommended services to the construction panel. After that, users need to do a little configuration work for connections between service nodes. Lastly, users can execute the resulting mashup, and save it if they are satisfied with the result. With our platform, users can get a variety of innovative and pragmatic mashup applications in different areas such as social networking, E-commerce and entertainment. In Fig. 3 we show several mashups obtained with our platform including a Douban1 social mashup, a weather checking mashup, a friends’ location viewing mashup and a mashup which shows disc information and related purchasing information for music in the mobile1 device. Users are able to obtain more diverse mobile mashups by selecting linkable services and constructing a mashup graph. 1 http://www.douban.com/
We have presented a mobile mashup platform with a semantic-based recommendation mechanism and a mashup execution engine. The recommendation mechanism makes mashup construction and execution much easier. By following a few simple steps, users are able to obtain a mashup based on their preferences. For future work, we will improve the platform to deal with more complicated situations. 4
The authors acknowledge the support provided by CCNT Laboratory of Zhejiang University and Nokia Research Center of Nokia Corporation.