The Ingredients of EMF: Analysis of Documentation and Implementation Topics

Status

Submitted for publication. Website created on 7 March.

Authors

Johannes Härtel, Lukas Härtel, and Ralf Lämmel

Abstract

The paper recovers relationships between the implementation of the Eclipse Modeling Framework (EMF) and a major documentation source on the framework, i.e., Steinberg et al.'s textbook on EMF. The underlying method recovers relationships on the grounds of semantic information that can be retrieved from implementation and documentation. To this end, LDA is applied to both implementation and documentation and the resulting topics are compared in a systematic metrics-driven manner, thereby revealing cases of alignment and also cases of misalignment, both of which are to be explained with the help of domain knowledge. Paragraph granularity is used for documentation and method granularity is used for the implementation. Thereby, we abstract from the modular structure of the information sources.

Keywords

LDA, Topic Modeling, EMF

Downloads and links

• Paper: [.pdf]
• Data set supporting paper:
  • Topics documentation: [.csv]
  • Topics implementation: [.csv]
  • Regions, Characteristics and Types: [.csv]
  • Stopword list: [.txt]

The annotated textbook cannot be made available for copyright reasons.

The EMF implementation is available here: [.html] (ecp, edapt and emfstore packages are excluded in our analysis)

Our tools are available on request.

The LDA configuration is described in the paper.

@misc{HaertelHL17,
  author    = {Johannes H\"artel and Lukas H\"artel and Ralf L{\"a}mmel},
  title     = "{The Ingredients of EMF: Analysis of Documentation and Implementation Topics}",
  year      = {2017},
  note      = "16 pages. Submitted for publication."
}