Browsing Search Results via Formal Concept Analysis: Automatic Selection of Attributes

Concept location using formal concept analysis and information retrieval

“…Additionally, an inherent advantage of concept lattices is a set of relations among the concept nodes, which users can navigate in a lattice. Among visible disadvantages of FCA, the size and complexity of lattices depends on the number of objects and attributes and can be computationally demanding as compared to other clustering techniques [Cigarrán et al 2004]. …”

“…There are several published solutions to extract descriptive terms for sub-collections of documents. For example, the okapi weighting scheme and the terminological formula are two of the approaches previously used in free-text IR systems [Cigarrán et al 2004]. We adopt here the technique proposed by Kuhn et al [Kuhn et al 2007], since it was defined in the context of source code to select relevant terms with respect to given clusters of source code elements.…”

“…For the FCA-based approach, we assume that the ideal "user" will navigate the concept lattice starting at the root and investigating every method in the nodes encountered on the shortest path from the root to the target method. This approach is not uncommon, the shortest path was previously used to define the minimal browsing area [Cigarrán et al 2004], which is in fact the measure we use.…”

“…We plan to compare this approach with at least two other different strategies used to rank and select descriptive attributes to build concept lattices, for example, the terminological weighting formula and the Okapi [Cigarrán et al 2004]. We also plan to design a heuristic-based approach to experiment with different strategies for splitting identifiers [Dit et al 2011a;Enslen et al 2009], mining abbreviations [Hill et al 2008] or selecting attributes, which may be specific to source code, for example, selecting only attributes that represent data types or only class or methods names, etc.…”

“…Specifically, we augment an existing information retrieval (IR) based technique for concept location [Marcus et al 2004] with automatic clustering of the search results using formal concept analysis (FCA). This approach is inspired by previous work, which applied FCA to free-text search [Cigarrán et al 2004]. The IR-based concept location technique uses a search engine based on Latent Semantic Indexing (LSI) [Deerwester et al 1990], which allows users to search source code and related textual documentation by writing natural language queries and retrieving a list of source code elements (that is, classes, methods, functions or files), which are ranked based on their similarity to the query.…”