SIMRE

*SimRE is a Python-based tool for automatically identifying requirements' similarity in SPL projects.

Parameters description

• You need to enter seven parameters. These parameters are the following:
  • nameNewReq: name of the CSV file that contains the list of new requirements.
  • nameFeatures: name of the XML or UVL file that contains the existing requirements.
  • nameReqDescription: name of the JSON file that contains the requirements description.
  • language: 'en' for English and 'es' for Spanish
  • listModels: array with the models. optional. The default model is 1. The models are the following:
    • 1:Model multilingual MiniLM-L12-v2
    • 2:Model multilingual distiluse-cased-v2
    • 3:Model multilingual mpnet-base-v2
    • 4:'Model word2vec
    • 5:'Model fastText
  • threshold: optional. The default value is 0.7.
  • preprocess: optional. The values are: True to allow the pre-processing (default value), and False for without pre-processing

Installation by pip

• Its necessary to have installed al least Python 3.8.10 and pip 23.1.2
• The tool can be used by installing the library or downloading the code.
• After download the library using pip, use the following code to download several pre-trained models and put them on caché.

from simre import init_models
init_models.main()

from simre import similarity
models = similarity.load_models()

• This process may take several minutes depending on the processor's capacity and memory. We recommend a RAM of at least of 16 GB.
• When the process finishes, you should confirm that a "fileserver" folder has been created and contains four files.

Usage

• The method similarity_process perform the similarity process.

similarity.similarity_process(nameNewReq, nameFeatures, nameReqDescription, 'en',models) 

• In the examples folder are some files to make a test.

similarity.similarity_process('newRequirements.csv', 'featureModel.xml', 'descRequirements_en.json', 'en',models) 

• With all the parameters.

similarity.similarity_process('newRequirements.csv', 'featureModel.xml', 'descRequirements_en.json', 'en',models,[1,2,3],0.6,False) 

• The results will be provided in a CSV file ('Similarity List.csv').

Installation by code

• Download the code

• Install the required libraries. All necessary libraries are listed in the requirements.txt file. To install them, execute the following command pip install -r ./requirements.txt

• It is necessary to download several pre-trained models. This can be done automatically or manually. To download the models automatically, execute the following command: python process.py init. To download the models manually, follow these steps:

• a. Download the models of spacy for Spanish and English: python -m spacy download es_core_news_sm and python -m spacy download en_core_web_sm

• b. Download the models of fastText for Spanish and English: cc.es.300.bin and cc.es.300.bin from https://fasttext.cc/docs/en/crawl-vectors.html

• c. Download the word2vec-based models for Spanish and English: SBW-vectors-300-min5.bin.gz from https://crscardellino.github.io/SBWCE/ and GoogleNews-vectors-negative300.bin.gz from https://code.google.com/archive/p/word2vec/

• d. At the same directory level as the src folder, create a new folder named fileserver. Place all the pre-trained models into this folder.

Usage

• To execute the tool without optional parameters, use the following command: python process.py nameNewReq nameFeatures nameReqDescription language. Example:: python process.py newRequirements.csv featureModel.xml descRequirements_en.json en.
• This is an example using all the parameters: python process.py newRequirements.csv featureModel.xml descRequirements_en.json en 1,2,3,5 0.7 False
• The results will be provided in a CSV file on the fileserver folder ('Similarity List.csv').

Formats

*XML file:

<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<featureModel chosenLayoutAlgorithm="4">
  <struct>
    <and mandatory="true" name="GEMA_SPL">      
      <and name="UserManagement">
        <or name="UM_Registration">
          <feature mandatory="true" name="UM_R_ByAdmin"/>
          <feature mandatory="true" name="UM_R_Anonymous"/>
        </or>              
      </and>      
    </and>
  </struct>
  <featureOrder userDefined="false"/>
</featureModel>

*UVL file:

features
	UserManagement 
		optional
			UM_Registration 
				or
					UM_R_ByAdmin 
					UM_R_Anonymous

*JSON file:

{
  "UserManagement": { 
      "label": "User Management",
      "desc": "User Management" 
  },
  "UM_Registration": { 
      "label": "User Registration",
      "desc": "User Registration" 
  },
  
}