![Our approach
11
Model Recommendation Engine
NLP method for
word embeddings
train
A.2
Morphological
analysis &
lemmatization
NLP models
contextual
knowledge
general
knowledge query
B.2
B.1
B.2
B.3
B.4
B.5
C.2
uses
uses
Partial
model
Text preprocessing
algorithm
preprocess
A.1
Domain corpus
of text
Domain
docs
Slice 1: Flight → […]
Slice 2: Pilot → […]
Slice 3: Pilot → […]
Slice 4: Pilot, Flight → […]
Slice 5: Pilot, Flight → […]](https://image.slidesharecdn.com/presentationcaise21-210630165242/85/An-NLP-based-architecture-for-the-autocompletion-of-partial-domain-models-11-320.jpg)
![Our approach
12
Model Recommendation Engine
NLP method for
word embeddings
train
A.2
Morphological
analysis &
lemmatization
NLP models
contextual
knowledge
general
knowledge query
B.2
B.1
B.2
B.3
B.4
B.5
C.2
uses
uses
Partial
model
Text preprocessing
algorithm
preprocess
A.1
Domain corpus
of text
Domain
docs
Slice 1: Flight → [flights,
plane,
pilots,
pilot,
flying,
fly,
airline,
airlines,
airplane,
jet]](https://image.slidesharecdn.com/presentationcaise21-210630165242/85/An-NLP-based-architecture-for-the-autocompletion-of-partial-domain-models-12-320.jpg)
![Our approach
13
Model Recommendation Engine
NLP method for
word embeddings
train
A.2
Morphological
analysis &
lemmatization
NLP models
contextual
knowledge
general
knowledge query
B.2
B.1
B.2
B.3
B.4
B.5
C.2
uses
uses
Partial
model
Text preprocessing
algorithm
preprocess
A.1
Domain corpus
of text
Domain
docs
Slice 1: Flight → [plane,
airline,
airplane,
jet,
flying,
fly]](https://image.slidesharecdn.com/presentationcaise21-210630165242/85/An-NLP-based-architecture-for-the-autocompletion-of-partial-domain-models-13-320.jpg)
An NLP-based architecture for the autocompletion of partial domain models
- 1. An NLP-based architecture for the autocompletion of partial domain models Loli Burgueño1, Robert Clarisó1, Sébastien Gérard2, Shuai Li2, Jordi Cabot2,3 1 Open University of Catalonia, Barcelona, Spain 2 CEA LIST, Paris, France 3 ICREA, Barcelona, Spain June 30th, 2021
- 2. Domain modeling Informal descriptions Structured and unambiguous representation leaving out superfluous details … using a concrete notation 2 Introduction
- 3. Motivation To promote and facilitate the creation and manipulation of domain models → Broad variety of • languages • tools Still, they are typically created by hand!!! DSLER Motivation
- 4. The role of data 4 Contextual knowledge General knowledge
- 5. 5 may have
- 6. 6 Artificial Intelligence Machine Learning Natural Language Processing Domain Modeling assist Natural Language Processing (NLP)
- 7. Our approach 7 Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs
- 8. Our approach 8 Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge NLP components
- 9. Our approach 9 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs
- 10. Model Recommendation Engine query B.2 B.2 B.3 B.4 B.5 C.2 uses uses Our approach 10 NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs B.1 Slice 1 Slice 2 Slice 3 Slice 4 Slice 5
- 11. Our approach 11 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge query B.2 B.1 B.2 B.3 B.4 B.5 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs Slice 1: Flight → […] Slice 2: Pilot → […] Slice 3: Pilot → […] Slice 4: Pilot, Flight → […] Slice 5: Pilot, Flight → […]
- 12. Our approach 12 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge query B.2 B.1 B.2 B.3 B.4 B.5 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs Slice 1: Flight → [flights, plane, pilots, pilot, flying, fly, airline, airlines, airplane, jet]
- 13. Our approach 13 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge query B.2 B.1 B.2 B.3 B.4 B.5 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs Slice 1: Flight → [plane, airline, airplane, jet, flying, fly]
- 14. Our approach 14 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge query B.2 B.1 B.2 B.3 B.4 B.5 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs Add class named Plane Add class named Airline Add class named Airplane …
- 15. Our approach 15 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge update C.1 query B.2 B.1 B.2 B.3 B.4 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs B.5 Add class named Plane Add class named Airline Add class named Airplane …
- 16. Our approach 16 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge update C.1 query B.2 B.1 B.2 B.3 B.4 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs B.5
- 17. Our approach 17 Model Recommendation Engine NLP method for word embeddings train A.2 Morphological analysis & lemmatization NLP models contextual knowledge general knowledge update C.1 query B.2 B.1 B.2 B.3 B.4 C.2 uses uses Partial model Text preprocessing algorithm preprocess A.1 Domain corpus of text Domain docs B.5
- 18. RQ1. Recall RQ2. Precision Preliminary validation 18 • Industrial project (from 2015) - water supply and sewage • Introduce of a notice management system for incidents • Documentation: • a presentation (21 slides) • paper forms • software requirement specification document (7,675 words) • The developers produced manually the domain model RQ3. Source of accepted suggestions RQ4. Performance
- 19. Preliminary validation 19 Automated the reconstruction process • Simulate the behavior of a designer automatically • Process: 1. We have the final model 2. automatically accept/reject suggestions based on whether they appear in the final model or not • This evaluation can be regarded as a worst-case scenario
- 20. RQ2. Precision ✓Precision of 4.46% 15.67 queries (20 suggestions per source of knowledge each) 626.7 suggestions received 25.67 accepted ✓Suggestions accepted per query is 1.79 Preliminary validation results 20 Model reconstructed starting from an emtpy class Notice RQ1. Recall ✓62% of the model reconstructed on average 34 out of the 55 elements
- 21. RQ2. Precision ✓Precision of 4.46% - 15.67 queries - 626.7 suggestions received - 25.67 accepted ✓Suggestions accepted per query is 1.79 Preliminary validation results RQ1. Recall ✓62% of the model reconstructed on average 34 out of the 55 elements RQ3. Source of accepted suggestions RQ4. Performance ✓ All the components of the MR engine take under one second ✓ Querying the NLP models - Training, loading and querying the NLP contextual model: a few milliseconds - Loading the NLP general model: 32 sec - Querying the NLP general model: a few secs 21 ✓ 85.7% from the contextual source ✓ 14.3% from the general source
- 22. • Model recommendation engine • fed with textual descriptions of a domain • generates autocompletions for models under development • First step towards a more general modeling assistant that effectively helps specify better models faster • Integrate other types of information (e.g., past models, ontologies,…) • Work on the usability aspect & empirical study • Refine the techniques presented in this paper • Application on other types of models and modeling languages • Exploitation of other NLP models Conclusions & Future Work 22
- 23. An NLP-based architecture for the autocompletion of partial domain models Loli Burgueño1, Robert Clarisó1, Sébastien Gérard2, Shuai Li2, Jordi Cabot2,3 1 Open University of Catalonia, Barcelona, Spain 2 CEA LIST, Paris, France 3 ICREA, Barcelona, Spain June 30th, 2021 @LolaBurgueno [email protected]