AMMoRe 2018: First international workshop on analytics and mining of model repositories Önder Babur Michel R. V. Chaudron Loek Cleophas Eindhoven University of Technology Chalmers | University of Gothenburg Eindhoven University of Technology Eindhoven, The Netherlands Gothenburg, Sweden Eindhoven, The Netherlands o.babur@tue.nl chaudron@chalmers.se Stellenbosch University Stellenbosch, Republic of South Africa l.g.w.a.cleophas@tue.nl Davide Di Ruscio Dimitris Kolovos University of L’Aquila University of York L’Aquila, Italy York, United Kingdom davide.diruscio@univaq.it dimitris.kolovos@york.ac.uk ABSTRACT AMMoRe aims to bring together researchers in software mod- Model-based approaches promote the use of models and related elling, model repositories and model analytics, big-data and ma- artifacts (such as metamodels and model transformations) as cen- chine learning, information retrieval and natural language pro- tral elements to tackle the complexity of building systems. Both cessing. Orthogonal to these domains, it welcomes contributions in academia and in industry there is a growing need to efficiently from a wide range of technical spaces to promote cross-fertilization: i) store; ii) analyze; and ii) search & navigate, and iii) curate large Model-Driven Engineering, Systems Engineering, Business Process collections of models. Such collections include for example large Modelling, Software Architecture and so on. The topics of interest sets of software models such as the Lindholmen UML dataset [1], of AMMoRe 2018 include: or of heterogeneous models in large MDE ecosystems and systems- • Industrial reports, empirical studies on model corpora, ap- of-systems, including e.g. software, hardware, and business models. plications of model corpora The workshop Analytics and Mining of Model Repositories (AM- • Repository mining and management for modelling artefacts MoRe) aims to gather modelling researchers and practitioners to • Clone-, pattern-, aspect-mining for modelling artefacts discuss the emerging problems and propose solutions. The scope • Applications of exploratory or descriptive data analytics, ranges from industrial reports and empirical analyses in the prob- predictive analytics, machine learning or deep-learning lem domain to novel cross-disciplinary approaches for large-scale • Large-scale model management and consistency checking analytics and management, e.g. exploiting techniques from data • Natural language processing for modelling analytics, repository mining and machine learning. • Model searching, indexing, retrieval, storage • Linking, enriching and labelling model-repositories 1 OBJECTIVES AND SCOPE • Visualization of (possibly heterogeneous) large sets of mod- elling artifacts Big data approaches are causing large changes in the way we can • Techniques to analyze and automate (co-)evolution in mod- perform science and business. Big Data is also entering the arena of elling software engineering and software modelling. We want to bring to- • Variability mining and management, model-driven software gether the communities of Big Data/Machine Learning and Software product lines Modelling. Various datasets of models have now become available • Distributed computing for modelling, with an eye towards and now our community must learn methods, techniques and tools Big Data for analyzing these large datasets. Many such methods, techniques • Intelligent techniques for automating modelling tasks and tools are known from the Big Data/Machine Learning and Infor- • Building and composing model-analytics workflows (based mation Retrieval/Natural Language Processing communities. How on online services/repositories) they can be adapted and applied to models and model repositories is an open question. Conversely, the insights that come out of this While this is the first edition of this workshop per se, a Dutch may lead to insights for these communities that are usable beyond Symposium on Model Management and Analytics in October 2017 software modelling. (with invited talks) was well-attended, attracting 23 participants Undoubtedly, MODELS is the premier conference series for model- from 6 universities and 7 companies, and 11 talks. The correspond- driven software and systems engineering. It tries to cover all aspects ing website 1 , which is being used for a range of related events on of modeling, yet analytics and mining of model repositories (and similar topics (including AMMoRe), has so far attracted 2k+ views other large collections of models) has not been a prime focus or the from 60+ countries. topic of a MODELS workshop, nor of any other event in the commu- nity so far (to the best of our knowledge). Yet it is an increasingly 1 http://modelanalytics.wordpress.com timely and important topic. 2 WORKSHOP ORGANIZATION automatically fix broken models by using machine learning. Önder Babur, Michel Chaudron, Loek Cleophas, Davide Di Ruscio Since appropriate data sets are seen as lacking, the authors’ and Dimitris Kolovos organized and chaired the program committee prototype tool uses reinforcement learning, which does not (PC) for the first edition of AMMoRe. Each of the submissions was need initial data sets. A case study was performed with this reviewed by at least four PC members and the papers were selected prototype on a set of broken models. based on their relevance to the workshop’s topics and the reviews • Model analytics for feature models: case studies for S.P.L.O.T. provided by PC members. The AMMoRe 2018 PC consisted of: repository (by Önder Babur, Loek Cleophas and Mark van den Brand) adapts the model analytics framework SAMOS, which • Nour Assy (Eindhoven University of Technology, NL) uses information retrieval techniques to compare models, to • Alessandra Bagnato (Softeam, FR) the setting of feature models, and discusses two case studies • Ludovico Iovino (Gran Sasso Science Institute L’Aquila, IT) in which the resulting framework was used to get insight • Yannis Korkontzelos (Edge Hill University, UK) into collections of feature models. • Henrik Leopold (Vrije Universiteit Amsterdam, NL) • Exploring model repositories by means of megamodel-aware • Ivano Malavolta (Vrije Universiteit Amsterdam, NL) search operators (by Francesco Basciani, Juri Di Rocco, Da- • Nicholas Matragkas (University of York, UK) vide Di Ruscio, Ludovico Iovino and Alfonso Pierantonio) • Richard Paige (University of York, UK) aims to improve model artefact reuse by presenting a novel • Alfonso Pierantonio (University of L’Aquila, IT) approach to model search. Repositories are structured into • Ivan Polasek (Slovak University of Technology & Gratex megamodels, with specific search operators allowing effec- International, SK) tive exploration and browsing of the model repositories. A • Gregorio Robles (Universidad Rey Juan Carlos, ES) case study implements this in the MDEForge platform, using • Christoph Seidl (Technische Universität Braunschweig, DE) the Lucene search library. • Matthew Stephan (Miami University, US) • Daniel Strüber (University of Koblenz and Landau, DE) 4 OUTLOOK • Bedir Tekinerdogan (Wageningen University, NL) • Mark van den Brand (Eindhoven University of Technology, With this first edition of AMMoRe, we aim to build and strengthen NL) an audience from various domains and communities working on • Maurice van Keulen (University of Twente, NL) analytics and mining of model repositories. We hope to increase • Barbara Weber (Technical University of Denmark, DK) our effort and organize further iterations of this workshop and • Manuel Wimmer (Vienna University of Technology, AT) other follow-up events, with the goal of attracting more attention to these timely and important topics. 3 PROGRAM AMMoRe attracted 5 submissions on a variety of topics related to ACKNOWLEDGEMENTS the analytics and mining of model repositories. The final program The organizers are very thankful to all PC members for their re- included one invited talk and 3 paper presentations: viewing, an important service to the workshop; and for the quality of their reviews. • Process model management and analytics (invited talk, by Barbara Weber) gives an overview of the invited speaker’s past 10+ years research on business process modeling and REFERENCES [1] Regina Hebig, Truong Ho-Quang, Michel R. V. Chaudron, Gregorio Robles, and model management, including the adaptation and evolution Miguel Angel Fernández. 2016. The quest for open source projects that use UML: of process models, the management of process model reposi- mining GitHub. In Proceedings of the ACM/IEEE 19th International Conference on Model Driven Engineering Languages and Systems, Saint-Malo, France, October tories, as well as model analytics conducted online as needed 2-7, 2016, Benoit Baudry and Benoît Combemale (Eds.). ACM, 173–183. https: for e.g. personalized modeling environments. //doi.org/10.1145/2976767 • Automatic model repair using reinforcement learning (by An- gela Barriga, Adrian Rutle and Rogardt Heldal) proposes to