A semantic-based approach to component retrieval

Abstract

There continues to be a great deal of pressure to design and develop information systems within a short period of time. This urgency has reinvigorated research on software reuse, particularly in component based software development. One of the major problems associated with component-based development is the difficulty in searching and retrieving reusable components that meet the requirement at hand. In part, this problem exists because of the lack of sophisticated query methods and techniques. In this research, a semantic-based approach to component retrieval is presented as a solution to this problem. This approach makes use of domain models containing the objectives, processes, actions, actors, and, an ontology of domain terms, their definitions, and relationships with other domain-specific terms. A reuse repository is developed that contains the components relevant for the creation of new applications, along with their attributes and methods. The natural language interface, domain model, and reusable repository are implemented in a prototype that uses Web and JavaBeans technologies. A sample session is provided for an online auction application to illustrate the usefulness of the proposed approach.

Semantics-based code search

“…Early work here demonstrated that keywords from comments and variable names were often sufficient for finding reusable routines [12,22]. Later work here did query refinement either directly [33], by looking at what the programmer was doing [39,40], using an appropriate ontology [38], using learning techniques [10], using natural language [8], or using collaborative feedback [35]. Recent approaches, such as Assieme [17], Sorcerer [1], and Codifier [4], incorporate program structure and semantics as a search basis.…”

Search in Source Code Based on Identifying Popular Fragments

“…In [12] authors present approach, where queries and restrictions can be formulated in natural language, for example, "Give me details about the bidding process". The presented approach is based on the use of domain models containing the objectives, processes, actions, actors and a domain ontology, their definitions, and relationships with other domain-specific terms.…”

Finding a Needle in the Haystack: A Technique for Ranking Matches Between Components

“…The keyword-based approach [7] is simple and flexible as users simply specify the query as a set of keywords representing the component requirements in which they are interested. This approach while simple is also prone to low accuracy resulting in either too many or too few hits, or in some cases even completely unrelated hits [15]. The faceted approach [12] classifies components based on predefined taxonomies.…”

“…While this approach provides a better description of components than a pure keyword-based approach, users must be familiar with the classification scheme to effectively retrieve a needed component. Moreover, it is often hard to manage classification schemes when domain knowledge evolves and as a result the component falls into two or more categories [15]. Signature matching approaches [21] decide the match between two given components, the query and library components, based on the signatures of the methods in the two components.…”