https://github.com/disa-lab/SANER2021-DocSmell

https://github.com/disa-lab/SANER2021-DocSmell

Last synced: 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/disa-lab/SANER2021-DocSmell
• Owner: disa-lab
• Created: 2021-01-05T16:53:22.000Z (about 4 years ago)
• Default Branch: main
• Last Pushed: 2021-01-06T14:52:03.000Z (about 4 years ago)
• Last Synced: 2024-08-04T00:11:33.498Z (5 months ago)
• Language: Python
• Size: 1.28 MB
• Stars: 3
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-software-engineering-research - DocSmell Benchmark

README

        
# SANER2021-DocumentationSmell-ReplicationPackage

## Automatic Detection of Five API Documentation Smells: Practitioners’ Perspectives

## Documentation Smell

Documentation smells can be described as bad documentation styles that do not necessarily make a documentation incorrect but make it difficult to understand and use.

## Types

We present 5 types of documentation smells. They are:

* Bloated: too lengthy and verbose.

* Excess Structural Info: too many structural syntax or information

* Tangled: too complex to read and understand

* Fragmented: scattered over multiple pages or sections

* Lazy: does not provide extra info other than the method prototype

## Survey

We conducted a survey of 21 software developers to validate these documentation smells. The survey questionnaire and responses can be found in the 'Survey' folder. 

## Benchmark Dataset

We created a benchmark dataset of 1000 documentations with these 5 types of smells. The benchmark dataset (with the features) can be found in the 'Benchmark Dataset' folder.

## Automatic Detection of Documentation Smells

We employed rule-based, shallow, and deep machine learning techniques to automatically detect documentation smells.

## Codes

Codes of our rule-based, shallow, and deep learning classifiers are available at three different subfolders under the 'Codes' folder. The codes are organized in a sequential manner. All file&folder names are self-explanatory. We have run the deep learning classifiers (i.e., Bi-LSTM, BERT) on Google Colab. You can run the corresponding python notebooks on Google Colab by uploading them on your google drive with the benchmark dataset.