Assisted Mashup Development: On the Discovery and Recommendation of Mashup Composition Knowledge

Abstract

Over the past few years, mashup development has been made more accessible with tools such as Yahoo! Pipes that help in making the development task simpler through simplifying technologies. However, mashup development is still a difficult task that requires knowledge about the functionality of web APIs, parameter settings, data mappings, among other development efforts. In this work, we aim at assisting users in the mashup process by recommending development knowledge that comes in the form of reusable composition knowledge. This composition knowledge is harvested from a repository of existing mashup models by mining a set of composition patterns, which are then used for interactively providing composition recommendations while developing the mashup. When the user accepts a recommendation, it is automatically woven into the partial mashup model by applying modeling actions as if they were performed by the user. In order to demonstrate our approach we have implemented Baya, a Firefox plugin for Yahoo! Pipes that shows that it is indeed possible to harvest useful composition patterns from existing mashups, and that we are able to provide complex recommendations that can be automatically woven inside Yahoo! Pipes' web-based mashup editor.

Incorporating OAuth Protocol into Existing Information Systems

“…[15]- [17] proposed the reusable design which can be applied in the OAuth context but the work focuses on the OAuth client party. None of them focuses on integrating the OAuth server to the existing resource server as the OAuth client, scope, token, and OAuth user are highly coupled, making changes to one of them would require updating for the rest of the system.…”

Crowd-Based Mining of Reusable Process Model Patterns

“…Given a set of crowd-mined patterns M P , accepting or rejecting our hypotheses then further requires comparing the quality of M P with that of patterns that are mined automatically (we use for this purpose our algorithm described in [11]). …”

“…To study the described Crowd algorithm, we performed a set of experiments with CrowdFlower and compared the results with those obtained by running our original automated pattern mining algorithm [11] with different minimum support levels and dataset sizes. We refer to this latter as to the M achine algorithm.…”

“…The Machine algorithm is based on a frequent sub-graph mining algorithm described in [11] and implemented in Java. The core parameter used to fine-tune the algorithm is the minimum support that the mined sub-graphs must satisfy; we therefore use this variable to test and report on different test settings.…”

“…A click on a recommended pattern automatically weaves the pattern into the partial pipe in the modeling canvas. Patterns are mined from a repository of freely accessible pipes models [11]. We mined patterns from a dataset of 997 pipes taken from the "most popular" category, assuming that popular pipes are more likely to be functioning and useful.…”