The development of complex systems involves the collaboration of many stakeholders. In order to design the system, they produce many artefacts i.e., requirements and models that are correlated with each other’s and which evolve constantly. In such a volatile environment, there is a critical need to manage the impact of the different changes occurring during the project lifetime. Traceability, as defined by Edwards and Howell, is “A technique used to provide a relationship between the requirements, the design and the final implementation of the system.”

In complex systems engineering, establishing such traceability involves dealing with a large volume of requirements and models. For example, the full specification of an aircraft includes about 10,000 requirements and a subway line of about 6,000. And modelling an aircraft can lead to hundreds of thousands of elements in hundreds of different models. Dealing manually with traceability issues in such a context is obviously unbearable.

In the literature, many works propose to automate the identification of traceability links. In the Information Retrieval (IR) community, approaches have been used to recover traceability links between artefacts. However, traceability identification in these approaches is still complex and error prone. Due to that limited accuracy, candidate links are systematically checked by an analyst which manually classifies them.

Thus, we investigate the benefits of the latest advances in semi-supervised approaches and NLP approaches in order to improve the performances of IR techniques. We aim to enhance candidate traceability links generation and suggestion to the analysts. Thus, we propose an approach, called ATLaS (Aggregation Trace Links Support). It is based on the clustering hypothesis that combines different strategies of IR and NLP techniques (i.e. word-embeddings and sentence embeddings) to improve the accuracy of IR techniques.

• In Step 1 “Compute similarities”, we perform pre-processing of the input data which consists of the tokenization and Stopwords removal.

• In step 2 “Build training set”, we propose a heuristic to label some pairs of artefacts as related or non-related.

• In step 3 “Classify Links”, we use the descriptor matrix and the labels vector as inputs.

Aggregation Trace Links Support (ATLaS) version 1.0.

This program is a prototype which role is to recover links between artefact (requirement, models elements) produce in the differents tools of the collaborative space in order to check the consistency of data. It uses a semi-supervised method with a combination of similarities measure defined by IR techniques (VSM, LSI, LDA).

This is the code for the paper

The code is written in python and requires numpy, pandas, pickle, sklearn, csv, gensim, keras, sklearn, matplotlib, stop_words, math, scipy, and nltk.

It also require some file available online :

-glove  download the file glove.840B.300d.txt and put it on ATLaS\SIFmaster\data available on https://nlp.stanford.edu/projects/glove/
	    download the file glove.6B.300d.txt and put it on ATLaS\Helper  available on https://nlp.stanford.edu/projects/glove/
	
-google download the file GoogleNews-vectors-negative300.bin.gzword2vec and put it on ATLaS\Helper on available on https://code.google.com/archive/p/word2vec/

-ATLaS\SIFmaster is the code of Arora Sanjeev for the paper "A Simple but Tough-to-Beat Baseline for Sentence Embeddings".
 Some functions/classes are based on this code. It requires numpy, scipy, pickle, sklearn, theano and the lasagne library. 

The inputs data sets are also available on online as describe below.

-ARC-IT 1 & 2 : The Architecture Reference for Cooperative and Intelligent Transportation (ARC-IT) includes a set of interconnected components that are organized into four views that focus on four different architecture perspectives. available on https://local.iteris.com/arc-it/html/architecture/architecture.html

-CM1-NASA : Data is provided courtesy of NASA collated by Jane Huffman Hayes. available on http://coest.org/

-EASYCLINIC : small student created dataset in english, italian containing diverses artifacts. Includes use case, intercation diagrams, test cases and class description. available on http://coest.org/

-HIPAA : HIPAA (Healthcare Insurance Portability and Accountability Act). Provides traceability from 10 HIPAA Technical safeguards to requirements in 10 different EHR systems. available on http://coest.org/

-ICEBREAKER : IBS Ice Breaker System. Provides traceability form High level requirements to UML Classes. available on http://coest.org/

execute the main.py file.

The code is separated into the following parts:

• aggregation : it provides the code for the semi-supervised training

• SIFmaster : this is the code of the code of Arora Sanjeev for the paper "A Simple but Tough-to-Beat Baseline for Sentence Embeddings".

• feature : it provides the code for extraction of words, noun and verbal pharses and the computation of the defined similarity scores.

• helpers : it provides dictionaries, stopword files and the architeture of VSM in tracelab.

• inputs : it provides all the evaluated datasets.

• outputs : it provides the results of the inputs data.

• irmodels : it provides all the information retrieval and machine learning models. VsmTracelab.py is the python version of the SEMERU Component Library.

• utils : it provides utilisty functions

For technical details and full experimental results, see the papers.

-Semi-supervised Approach for Recovering Traceability Links in Complex Systems, 
Conference Proceedings, 2018 23rd International Conference on Engineering of Complex Computer Systems (ICECCS), 
Emma Effa Bella, Marie-Pierre Gervais, Reda Bendraou, Laurent Wouters and Ali Koudri.

-ATLaS: A Framework for Traceability Links Recovery Combining Information Retrieval and Semi-supervised Techniques, 
Conference Proceedings, 2019 23rd ieee international edoc conference - the enterprise computing conference, 
Emma Effa Bella, Stephen Creff, Marie-Pierre Gervais and Reda Bendraou.

This research work has been carried out in the framework of the Technological Research Institute SystemX, and therefore granted with public funds within the scope of the French Program "Investissements d’Avenir".