Future Trends on Software and Systems Modeling
1 like2,689 views
Jordi Cabot discusses future trends in modeling, emphasizing the significance of abstraction in software engineering and the evolving role of modeling in various domains, particularly in the context of low-code development and machine learning. He outlines challenges such as usability, infrastructure, and perceptions of modeling in the industry, advocating for more flexible, user-friendly approaches and the integration of AI to enhance productivity. The document presents various innovative techniques and frameworks for modeling, including the use of textual models and cloud-based solutions, while stressing the need for domain-specific adaptations.
Future Trends on Software and Systems Modeling
- 1. Future Trends on Modeling 18th of June, 2020 Jordi Cabot ICREA Research Professor at UOC Jordi.cabot@icrea.cat @softmodeling / modeling-languages.com #SiriusCon
- 3. I model, therefore I am - René Descartes
- 4. • Grady Booch – history of softwre engineering The entire history of software engineering is that of the rise in levels of abstraction - Grady Booch
- 6. To model, or not to model, this is the WRONG question - Shakespeare
- 7. What/when/how many models ? Depends on: Size, Team, Domain,…. Real question (difficult!) Modeling ROI: Cost of modeling vs Bº of modeling <- Cost depends on tools!
- 9. This is a model (that you all understand)
- 10. Engineering Systems SW Everything is a model SW is probably the worst market for selling modeling
- 11. SysMLe Profile: a SysML Profile for ECAD
- 12. Modeling in…
- 13. Moving towards modeling is the natural evolution in any domain (e.g. Machine Learning)
- 14. Why not more modeling?
- 15. BUT he is saying this since 1985
- 16. History of ML History of Modeling Langs – J.P. Tolvanen
- 17. We have a marketing problem
- 18. Modeling is not sexy (in some communities, e.g. devs) • Low-code is modeling with another name that just sells more • Modeling has bad press (mostly due to salesmen selling UML as a silver bullet) • Exception: Reverse Engineering (OpenAPItoUML, JSONDiscoverer,…) People believe that modeling … • Is not agile • Is just for documentation (and a posteriori) • Is just nice pictures
- 19. VCs believe in “modeling” (when marketing it well!)
- 20. We have a usability problem
- 21. Have you ever found a user that loves her modeling tool? • Too many clicks • Too many options • Eclipse/EMF was a great environment Not a Good first impression • Lack of documentation • Installation/Configuration issues • Generated code not optimized
- 22. We have an infrastructure problem
- 23. Modeling is also: • Collaboration • Security • Confidentiality Adapted to where the users are/want • GitHub (textual modeling) • Online (web-based editors) • Flexible (inconsitent, lightweight MDE)
- 24. And there’s more
- 25. Bucchiarone, Cabot, Paige, Pierantonio: Grand challenges in model-driven engineering: an analysis of the state of the research. Software and Systems Modeling 19(1): 5-13 (2020)
- 26. What about better modeling?
- 27. New Interfaces for Modeling
- 28. Textual models are models • Lower barrier to entry • Easier to integrate in CI pipelines • One DSL -> multiple notations – Killer combination: textual notation to write the model, graphical one to visualize & read it
- 29. E.g. Modeling of chatbots (xatkit.com) Daniel, Cabot, Deruelle, Derras: Xatkit: A Multimodal Low-Code Chatbot Development Framework. IEEE Access 8: 15332-15346 (2020)
- 30. Moving to the cloud • All Programming IDEs are moving to the cloud • Modeling IDEs should follow suit • Multiple JS libraries can be used in the front-end • In the back-end: EMF-REST
- 31. Modeling outside modeling tools
- 32. Modeling with VoiceBots / Chatbots Pérez-Soler, Daniel, Cabot, Guerra, de Lara: Towards Automating the Synthesis of Chatbots for Conversational Model Query. EMMSAD@CAiSE 2020: 257-265
- 33. Openness comes with new challenges • Intellectual Property Protection – Robust Hashing for models • Accountability – Blockchain infrastructure for models • Security – Access-control for models • Scalability – NoSQL backends like NeoEMF
- 35. Smarter tools makes us more productive https://modelia.eu/
- 36. Mendix Assist
- 37. OutSystems.ai
- 38. Model autocompletion from textual data • Previous approaches are based on historical models. Only useful if you have many models • What every project has is lots of documents • We’re developing an NLP-based model autocompletion
- 39. AI for manipulationg models: the MT example Original model … CODE Software code1 refinementst n refinementth Model-to-model Transformation Model-to-text Transformation CODE CODE • Requires learning a new language (the MT Language) • Time consuming • Error prone
- 40. Let’s try to learn the MTs automatically Input Output Training Transforming ML Input OutputML Machine Learning Artificial Neural Networks Deep Networks Recurrent networks LSTM BPMN Petri nets BPMN Petri Net
- 41. Tree-to-Tree Encoder-decoder arquitecture InputModel (preprocessed) InputTree EmbeddingLayer Encoder LSTM network OutputTree ExtractionLayer OutputModel (non-postprocessed) Decoder LSTM network AttentionLayer InputModel OutputModel Preprocessing Postprocessing Burgueño, Cabot, Gérard: An LSTM-Based Neural Network Architecture for Model Transformations. MoDELS 2019: 294-299 We are developed a similar approach for model-to-code transformations. Useful to mimick the company’s programming style.
- 43. We live in a modeling multi-verse We need to develop domain- specific modeling environments to better serve the needs of users in different domains
- 44. Models are not a static, fixed and complete artifact. Rather a partial, dynamic, personal and temporal view of the domain How tools and languages should adapt to better serve this view?
- 45. Customization • It should be easy to adapt tools to your specific needs – Even better: morphing modeling tools that automatically adapt based on what you do. • Also the notations – Personal notations sharing a common abstract syntax
- 46. Views and viewpoints Brunelière, García Perez, Wimmer, Cabot: EMF Views: A View Mechanism for Integrating Heterogeneous Models. ER 2015: 317-325 EMF Views
- 47. Extensible/Composable DSLs • No one size fits all solution. Let’s not reinvent UML • Set of modeling libraries/packages to import into a small core DSL – Uncertainty modeling – Temporal modeling – Spatial modeling • Somehow going back to the idea of (EMF) Profiles
- 51. If you want to keep talking with the speakers of actual talk, you will have to come back to this session by opening the SiriusCon agenda menu in the top left-hand corner Thanks for listening to Jordi Cabot Any questions? Please wait a few seconds before we automatically bring you to the next session Next Talk: Sirius Overview and Roadmap