Automating Feature Modeling in Product Line Engineering for

Systems Engineering: The Application of Natural Language

Processing

José Lameh

1,2 a

, Alexandra Dubray

and Marija Jankovic

Université Paris Saclay - CentraleSupélec, Laboratoire Genie Industriel, Gif-sur-Yvette, France

Renault Group - Ampère, Technocentre, 1 Av. du Golf 78288 Guyancourt, France

Keywords: Product Line Engineering, Feature Models, MBSE, Modeling, Artificial Intelligence, Natural Language

Processing.

Abstract: This paper explores the integration of Artificial Intelligence (AI), particularly Natural Language Processing

(NLP), with feature modeling (FM) in Product Line Engineering (PLE) for Systems Engineering. By

leveraging AI to formalize and model variability, the study proposes an algorithm to assist subsystem owners

in describing variability, generating prompts, and producing feature models. The results demonstrate AI’s

ability to detect and resolve common modeling issues, such as dead features, false optional features, and

constraint inconsistencies, while enhancing model validation and anomaly detection. Although the approach

is promising, limitations in scalability, conflict resolution, and integration across subsystems highlight the

need for future research to establish a comprehensive and scalable methodology. This work underscores AI's

potential to streamline feature modeling and improve the consistency and efficiency of variability

management in complex systems.

1 INTRODUCTION

Feature modeling is the major mean of representing

variability in Product Line Engineering (Oliinyk et

al., 2017). The hierarchical structure and constraints

of Feature Models (FM) effectively capture the

diverse configurations of complex systems, enabling

systematic variability management (Krueger &

Clements, 2017). In our previous work, we proposed

a novel approach to integrate systems engineering

principles into product line engineering (PLE)

(Lameh et al., 2025). This was based on two studies

done: Systematic Literature Review (Lameh et al.,

2024a) and Interviews (Lameh et al., 2024b). This

integration resulted in a multi-layered PLE

framework, where FMs serve as a central modeling

artifact for representing variability across multiple

domains.

The increasing complexity of modern systems has

underscored the need for advanced techniques to

manage variability. Artificial Intelligence (AI), and

https://orcid.org/0000-0001-9762-663X

https://orcid.org/0000-0002-3870-0331

particularly Natural Language Processing (NLP), has

emerged as a promising approach to enhance

variability management processes, where FMs are

foundational (Felfernig et al., 2024). AI methods,

especially NLP, can be leveraged to automate and

augment various aspects of FM creation, analysis, and

maintenance, offering significant improvements in

scalability and precision (Benavides et al., 2010). The

challenges that we aim to address in this paper include

detecting anomalies in FM, as well as managing

complexity in large-scale systems (Felfernig et al.,

2024). By leveraging generative AI technologies, this

study demonstrates how NLP used for automation can

streamline variability modeling, reduce error-prone

manual tasks, and enable faster, more reliable model

generation. The central research question is: How can

AI-driven approaches enhance the efficiency,

accuracy, and scalability variability modeling while

considering SE’s viewpoints (Lameh et al., 2025)?

After the current introduction, section 2 presents a

literature review on automated analysis and AI-driven

approaches in feature modeling. Section 3 outlines

450

Lameh, J., Dubray, A. and Jankovic, M.

Automating Feature Modeling in Product Line Engineering for Systems Engineering: The Application of Natural Language Processing.

DOI: 10.5220/0013442900003896

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 13th International Conference on Model-Based Software and Systems Engineering (MODELSWARD 2025), pages 450-457

ISBN: 978-989-758-729-0; ISSN: 2184-4348

the methodology employed, and Section 4 discusses

the results and identifies challenges and opportunities

for further enhancement of AI-based FM automation.

2 BACKGROUND

The integration of AI into Model-Based Systems

Engineering (MBSE) has garnered significant

attention in recent years, aiming to enhance system

design, analysis, and decision-making processes

(Schneider et al., 2022). AI's capabilities in handling

complex datasets and automating intricate tasks align

seamlessly with the objectives of MBSE, which

focuses on using models to support system

requirements, design, analysis, verification, and

validation activities throughout the system lifecycle.

Recent studies have explored various AI

applications within MBSE. For instance, AI-based

assistants have been developed to support MBSE

adoption in practice, providing an overview of

existing and potential application areas for AI in

MBSE (Anacker et al., 2024). These assistants can

augment human decision-making and improve the

overall efficiency of the MBSE process. Machine

learning algorithms, in particular, have been applied

to analyze large amounts of data generated during

system development, offering insights that can

optimize system design and performance (Visure

Solutions, 2023).

The convergence of MBSE and AI has also been

recognized as a platform for unlocking the power of

systems thinking throughout systems design,

increasing the ability to manage disruptive and

emergent system behaviors. Generative AI tools, such

as large language models, are impacting the systems

engineering lifecycle, serving as platforms for

innovation and understanding through model-based

systems engineering standardization and artificial

intelligence (Aerospace America, 2023).

Concerning feature modeling, AI-driven

approaches have demonstrated significant potential.

Feature models are essential in representing

variability and commonality within software product

lines, facilitating the configuration of diverse system

variants from a shared set of features. The integration

of AI methods with feature modeling has been

explored to enhance design, analysis, and application

processes (Lopez-Herrejon et al., 2023). An open

access book provides a basic introduction to feature

modeling and analysis, as well as the integration of

AI methods with feature modeling, serving as an

introduction for researchers and practitioners new to

the field (Felfernig et al., 2024). AI-driven

approaches, particularly those utilizing machine

learning and recommender systems, have shown great

promise in feature modeling. These approaches assist

human decision-making during the analysis phase,

effectively detecting anomalies, proposing solutions,

and generating configurations that satisfy a given set

of constraints. Such methods can significantly reduce

manual effort while improving the reliability of the

models. For example, AI can assist in anomaly

detection, solver support for satisfiability checking,

and the generation of consistent configurations.

Although full automation in modeling is challenging

due to the need for human oversight, AI's role in

analysis and validation is particularly noteworthy

(Sundermann et al., 2024). In this context, the focus

is on AI's application to the modeling and analysis

phases, rather than configuration generation.

Furthermore, AI aspects such as knowledge

representation, reasoning, explainable AI, and

machine learning have been linked to feature model-

related tasks, including modeling, analysis, and

configurators. This linkage underscores AI's potential

in automating model generation and analysis,

enhancing the efficiency and effectiveness of feature

modeling processes (Felfernig et al., 2024).

In summary, the integration of AI into MBSE and

feature modeling presents a promising avenue for

enhancing system engineering processes. AI-driven

approaches can automate and improve various

aspects of modeling and analysis, leading to more

efficient and reliable system development. As

research and development in this area continue to

evolve, the collaboration between AI and MBSE is

expected to yield innovative solutions to complex

engineering challenges.

3 METHODOLOGY

Our methodology utilized the Feature IDE tool, an

open academic software platform for feature

modeling. We integrated an NLP-based AI model,

specifically ChatGPT, to automate the generation of

feature models. The technology already exists, and

the goal was not to create something new but to make

effective use of it. It wasn’t just about using ChatGPT

directly; instead, we provided ChatGPT with our

specific modeling approach. The aim was to use

ChatGPT to connect the answers to the questions and

leverage its existing capabilities to formalize the

entire process. The process involved:

Formulating Variability: Variability was

described based on the input provided by subsystem

owners and the feature descriptions in our previous

Automating Feature Modeling in Product Line Engineering for Systems Engineering: The Application of Natural Language Processing

451

work. Our approach focuses on capturing all

variability as described by the subsystem owner,

transforming this input into a structured feature

model. Currently, this process involves manual

meetings between system engineers and domain

experts to extract variability information. By

leveraging AI, we propose automating this

interaction. The AI system would engage directly

with stakeholders through guided questioning,

helping to formalize their inputs into structured

variability descriptions. Once the information is

captured, the AI would process it to automatically

generate a feature model, reducing reliance on

manual interpretation and ensuring a more precise

and efficient modeling process.

AI Prompting and Output with Iterative

Refinement: Initial prompts were formulated to

describe the system’s variability. As the AI-generated

models occasionally included errors or

inconsistencies (e.g., special characters incompatible

with FeatureIDE), iterative refinements were applied.

This included avoiding parentheses in feature names,

clarifying constraints, and ensuring the parent-child

hierarchy was accurately represented. The prompt

was constructed using a structured framework,

integrating key elements such as system context,

variability dimensions, and expected outputs. To

ensure its quality, the prompt underwent iterative

refinement based on subsystem owner feedback and

trial runs. The GenAI model used was ChatGPT-4,

chosen for its advanced language comprehension,

context retention, and capacity to handle complex

prompts effectively.

In our methodology, we build upon the example

developed in our previous work, which focused on

Advanced Driver Assistance Systems (ADAS),

specifically the Park Assist feature. In that study, we

demonstrated how to formalize and model an FM in

the context of PLE for SE. This approach emphasized

maintaining the three essential SE perspectives:

operational, functional, and organic (constructional).

By structuring the variability model around these

viewpoints, we ensured that the FM accurately

captured the system's mission diversity (operational),

functional variations (functional), and constructional

components (organic). This example serves as a

foundation for illustrating how AI-driven methods

can further enhance the formalization and modeling

processes, providing a structured approach to

managing variability while aligning with SE

principles.

Validation was performed by comparing AI-

generated models with manually constructed feature

models for the same subsystem. The intuitive nature

of the GenAI-driven process, particularly its ability to

capture implicit variability details was highlighted.

Suggestions for improvement included enhancing AI

explanations for identified constraints and anomalies,

which will guide future iterations of the approach.

4 RESULTS

In this section, we present the outcomes of applying

the proposed algorithm for detecting and formalizing

variability using AI. This algorithm serves as a

structured framework to guide subsystem owners in

articulating variability and ensures that the captured

information can be systematically translated into a

feature model. By focusing on variability detection

and formalization, the algorithm reduces ambiguity

and bridges the gap between informal descriptions

and formalized outputs. We begin by introducing the

algorithm designed to detect and formalize variability

in subsystem descriptions. The algorithm employs a

question-driven approach, structured around

variability dimensions such as operational,

functional, and component diversity. It incorporates

mechanisms to validate the necessity of component

variability, prompting subsystem owners to justify

distinctions based on operational requirements or

functional differences. This systematic process

ensures that only relevant variability is modeled,

avoiding unnecessary complexity. The results are

then organized into three main parts.

4.1 Proposed Algorithm

This section provides a detailed breakdown of the

proposed algorithm for capturing variability in

feature modeling with component rationalization.

The following algorithm ensures that component

diversity is justified by identifying whether

variability arises from operational or functional

differences, avoiding unnecessary complexity. Key

steps are outlined to enhance clarity and

reproducibility. The section addresses challenges

such as inconsistency detection, conflict resolution,

and scalability, demonstrating how the approach

automates manual tasks.

This algorithm was proposed after working on

several projects at Renault. Through these projects,

we refined the questions by applying the model and

improving it based on interactions. The goal was to

replace manual meetings with subsystem teams by

using AI to make the process easier and more

efficient. The usual process involved many

interactions and several meetings, which took a lot of

MBSE-AI Integration 2025 - 2nd Workshop on Model-based System Engineering and Artiﬁcial Intelligence

452

time. This approach was applied to over 25

perimeters, and we found that the questions were

mostly the same. Based on this, we formalized the

algorithm to simplify and standardize the process.

Step 1: Identify Mission-Level Variability

1. Ask: What are the different missions or services

proposed to the client? List distinct options or

variations in the missions offered.

2. For each mission: Are there optional extensions or

customizations? Record additional mission-

specific options.

Step 2: Capture Functional Variability

1. For each mission: What are the core functions

required to achieve this mission? Focus on

variable functions only.

2. Ask: Are there alternative ways to implement any

function? Document functional diversity and

optional implementations.

3. Ask: Are there extra or optional features offered

for any function? Note additional capabilities as

optional features.

Step 3: Analyze Component-Level Variability

1. For each function: Are there variable components

or configurations used to deliver this function?

Focus only on components with variability.

2. Rationalize Component Variability: Ask: Why do

we need this component diversity if it performs

the same function? If a cheaper alternative exists

and performs the same, avoid adding variability.

Ask: Does the difference indicate operational or

functional variability instead? If so, reclassify as

operational or functional variability and update

the model.

Step 4: Formalize Constraints and Relationships

1. For each variability point: Define constraints

(e.g., "if mission X, then mission Y must exist").

Map dependencies (e.g., "function A requires

mission B").

2. Validate: Check for redundancies, false options,

or unnecessary conflicts.

Step 5: Review and Simplify

1. Review: Does the variability model accurately

reflect client needs? Ensure all variability adds

value and aligns with operational or functional

requirements.

2. Verify: Is the variability clear, justified, and cost-

effective? Remove unjustified diversity or

redundancies.

This algorithm, should give us as an output, a refined

variability model structured as:

- Mission-Level Variability: Client-focused

operational differences.

- Functional Variability: Alternative

implementations and extra features.

- Component-Level Variability: Rationalized

with clear justification or reclassified if

operational or functional.

- Constraints and Dependencies: Rules ensuring

consistency and reducing complexity.

This approach minimizes unnecessary variability,

ensuring the model is both practical and cost-

effective.

4.2 Input for Modeling

The proposed algorithm systematically retrieves and

formalizes variability information from subsystem

owners, ensuring alignment with PLE and feature

modeling principles. Its design reflects a structured

approach, leveraging NLP capabilities for variability

extraction while adhering to the operational,

functional, and organic SE perspectives. The

rationale stems from the need to streamline the

elicitation process and minimize variability errors,

which are common challenges in PLE. Completeness

was achieved through iterative GenAI interactions,

employing the “5 Whys” technique to probe deeper

into responses. A checklist of mandatory variability

dimensions (e.g., operational constraints, feature

dependencies) ensured no critical information was

omitted.

Using the proposed algorithm, we captured the

variability as described by the subsystem owner. This

input reflects the subsystem’s missions, functional

diversity, and specific features offered to the client.

The structured representation highlights how the

algorithm transformed a potentially vague and

unstructured description into a comprehensive and

clear variability framework. This input forms the

foundation for generating the feature model.

Automating Feature Modeling in Product Line Engineering for Systems Engineering: The Application of Natural Language Processing

453

PROMPT:

According to feature modeling rules,

create a feature model for ADAS Park

Assist System. This is the description

of variability: ADAS Park Assist system

can offer different park assist

missions: (i) Ultrasonic Park Assist

(UPA): This can exist in two variants:

rear-only or rear-and-front. These

variants are formed by incremental

elementary missions, (ii) Camera Park

Assist (CPA): It can come in two

variants: rear-view camera or around-

view camera, where only one variant is

installed on the vehicle, representing

alternative missions, and (iii)

Automatic Park Assist (APA): This can

also exist in two variants: hands-free

parking and/or remote parking. A vehicle

could have one or both variants,

representing different elementary

missions that form the overall mission.

As an example of constraints, the

dependency of APA REMOTE on CPA

AROUNDVIEW for optimal operation is a

technical constraint, modeled as a

logical condition to ensure system

compatibility. In contrast, the decision

to avoid offering UPA FRONT without UPA

REAR, while technically feasible, is a

marketing constraint defined in the

product structure.

Once done, create an .xml file so it can

be used for FeatureIDE tool.

End.

Algorithm 1: AI prompt used.

4.3 Output of the AI-Driven Process

The variability input was processed by an AI-driven

prompt to produce a detailed feature model. This

model encapsulates the identified variability and

associated constraints, ensuring consistency and

alignment with the subsystem owner’s requirements.

Logical constraints, such as the dependency of certain

features on others, were accurately included in the

output. Additionally, the feature model was exported

as an .xml script for direct use in FeatureIDE. This

script was validated for compatibility and correctness

within FeatureIDE, showcasing the seamless

integration of the generated model into existing feature

modeling tools. Note that, before presenting the script

and feature model, we also provide a preparatory

document explaining the fundamentals of feature

modeling and the functionalities of FeatureIDE. This

ensures that readers unfamiliar with these tools can

fully understand the process and outcomes.

ADAS Park Assist System

├── Ultrasonic Park Assist (UPA)

[Optional]

│ ├── Rear-Only [Mandatory Child]

│ │ └── Rear-and-Front [Optional]

├── Camera Park Assist (CPA) [Optional]

│ ├── Rear-View Camera [Alternative]

│ └── Around-View Camera [Alternative]

└── Automatic Park Assist (APA)

[Optional]

├── Hands-Free Parking [OR Group]

└── Remote Parking [OR Group]

Constraints:

UPA implies UPA Rear-Only.

Algorithm 2: FM model output generated by AI.

Figure 1: FM model generated by AI.

MBSE-AI Integration 2025 - 2nd Workshop on Model-based System Engineering and Artiﬁcial Intelligence

454

The generated feature model closely resembled the

one we created manually, demonstrating comparable

accuracy while significantly reducing the time

required for modeling. Although establishing the

process initially required effort, we believe that once

fully implemented, it will save substantial time by

minimizing meetings and automating repetitive tasks,

allowing engineers to focus on higher-value

activities.

These results highlight the algorithm’s

effectiveness in detecting and formalizing variability,

demonstrating how AI can streamline the creation of

feature models while maintaining accuracy and

relevance.

<?xml version="1.0" encoding="UTF-8"

standalone="no"?>

<and name="Ultrasonic Park

Assist UPA">

<mandatory name="UPA Rear-

Only"/>

<optional name="UPA Rear-

and-Front"/>

</and>

<alt name="Camera Park Assist

CPA">

<feature name="CPA Rear-View

Camera"/>

<feature name="CPA Around-

View Camera"/>

</alt>

<or name="Automatic Park Assist

APA">

<feature name="APA Hands-

Free Parking"/>

<feature name="APA Remote

Parking"/>

</or>

</and>

</struct>

<!-- APA Remote Parking requires CPA

Around-View Camera -->

<rule>

<imp>

<var>APA Remote

Parking</var>

<var>CPA Around-View

Camera</var>

</imp>

</rule>

</constraints>

</featureModel>

Algorithm 3: .xml of FM model output generated by AI.

5 DISCUSSION

The integration of Natural Language Processing

(NLP) with feature modeling has provided significant

insights and revealed both opportunities and

challenges. This approach demonstrated that AI can

effectively assist in analyzing and refining feature

models, but it also underscored areas where

improvements are needed to enhance the overall

process.

5.1 Insights and Identified Issues

During the modeling process, several key issues were

detected and addressed. These include but are not

limited to:

i) Dead Features: Features that could not

participate in any valid configuration were

identified, highlighting the importance of

systematic validation during model creation.

ii) False Optional Features: Features

incorrectly marked as optional but required

for consistency were flagged, emphasizing

the necessity of logical verification.

iii) Constraint Inconsistencies: Logical errors

in constraints, such as unsatisfiable or

contradictory rules, were detected, ensuring

model coherence.

iv) Redundant Constraints: Unnecessary or

duplicate constraints were identified and

removed, streamlining the model and

improving its efficiency.

v) Conflict Detection: Faulty constraints and

interdependencies leading to conflicts were

flagged, with the potential for conflict

aggregation and resolution proposed.

These detections not only validated the model but

also provided opportunities for improving its

robustness by addressing errors such as wrong

cardinalities and identifying anomalies. Additionally,

AI-assisted processes could verify product validity

and estimating the number of possible configurations,

further underscoring their utility in model analysis.

5.2 Opportunities for Improvement

While the AI demonstrated substantial promise, the

current approach highlighted several areas for

enhancement:

i) Advanced Anomaly Detection:

Incorporating more sophisticated AI

techniques could enable the identification of

Automating Feature Modeling in Product Line Engineering for Systems Engineering: The Application of Natural Language Processing

455

subtle and complex issues beyond the

current capabilities.

ii) Dynamic Conflict Resolution: Future

development could focus on AI-driven

methods for resolving detected conflicts,

providing practical recommendations for

engineers.

iii) Scalability: Ensuring that the approach is

scalable to accommodate large and complex

feature models remains an essential goal for

broader adoption.

5.3 Limitations and Future Directions

This study's reliance on a direct interaction with an AI

tool, such as ChatGPT, without establishing a

formalized process or methodology, is a noted

limitation. Developing a structured framework for

NLP-driven feature modeling would enhance its

effectiveness and allow for deployment at larger

scales. Additionally, the focus on a limited perimeter

presents challenges in integrating constraints across

multiple subsystems, particularly when features are

interdependent. This highlights the need for a

continuous process where AI not only models’

variability but also adapts dynamically to evolving

system constraints and interactions. While the

integration of NLP into feature modeling is

promising, further advancements are needed to

establish a comprehensive, scalable, and automated

methodology that can be widely applied in PLE. The

other main limits:

i) Data Protection Concerns: Utilizing GenAI

systems like ChatGPT raised questions about

data confidentiality, especially when dealing

with sensitive system requirements. Future

implementations must integrate secure, on-

premise AI models to safeguard proprietary

information.

ii) Variability of Outputs: While the AI

demonstrated consistency in generating

feature models, slight variations were

observed across iterations. These variations

are due to the AI’s process of searching for

additional information to enhance its

responses. To ensure consistent results,

additional tuning and domain-specific

adjustments should be implemented, focusing

on aligning the AI's outputs with predefined

parameters and minimizing unnecessary

deviations.

iii) Generalization of deployment: Applying

this approach to other systems is needed for

further validation. This highlights the need for

customizable templates and modular

algorithms that can generalize across multiple

domains. This approach would allow the

model’s applicability to be extended across

various sectors, insuring integration into

heterogeneous environments.

6 CONCLUSION

This paper demonstrated the application of NLP-

based AI to automate feature modeling in PLE. By

leveraging ChatGPT for model generation and

analysis, we reduced manual effort and improved

accuracy. However, further advancements are needed

to address existing challenges and fully realize AI’s

potential in this domain. Future research will focus on

enhancing anomaly detection, conflict resolution, and

the integration of AI-driven methods into the broader

systems engineering process.

REFERENCES

Aerospace America. (2023, novembre). Accelerating

innovation and understanding through model& # x2d ;

based systems engineering standardization and artificial

intelligence. Aerospace America. https://aerospace

america.aiaa.org/year-in-review/accelerating-innovatio

n-and-understanding-through-model-based-systems-en

gineering-standardization-and-artificial-intelligence/

Anacker, H., Wecker, S., & Harting, K. (2024). Artificial

Intelligence (AI) in Model-based Systems Engineering.

Systems Engineering - Innovationen. Mit System.

https://www.selive.de/ai-in-mbse/

Benavides, D., Segura, S., & Ruiz-Cortés, A. (2010).

Automated analysis of feature models 20 years later : A

literature review. Information Systems, 35(6), 615‑636.

https://doi.org/10.1016/j.is.2010.01.001

Felfernig, A., Falkner, A., & Benavides, D. (2024). Feature

Models : AI-Driven Design, Analysis and Applications.

Springer Nature.

Krueger, C., & Clements, P. (2017). Enterprise Feature

Ontology for Feature-based Product Line Engineering

and Operations. Proceedings of the 21st International

Systems and Software Product Line Conference -

Volume A, 227‑236. https://doi.org/10.1145/310619

5.3106218

Lameh, J., Dubray, A., & Jankovic, M. (2024a, submitted).

A Systematic Literature Review on Product Line

Engineering Approaches in Systems Engineering :

From Models to Applications (in press). submitted to

INCOSE Systems Engineering Journal (October 2024).

Lameh, J., Dubray, A., & Jankovic, M. (2024b). Variability

in complex product/system design : Case study in

automotive industry. Proceedings of the Design

MBSE-AI Integration 2025 - 2nd Workshop on Model-based System Engineering and Artiﬁcial Intelligence

456

Society, 4, 2635‑2644. https://doi.org/10.1017/pds.20

24.266

Lameh, J., Dubray, A., & Jankovic, M. (2025). Modeling

Variability in Product Line Engineering for Systems

Engineering (in press). submitted to ICED25.

Lopez-Herrejon, R. E., Martinez, J., Guez Assunção, W. K.,

Ziadi, T., Acher, M., & Vergilio, S. (Éds.). (2023).

Handbook of Re-Engineering Software Intensive

Systems into Software Product Lines (1st ed. 2023).

Springer International Publishing. https://doi.org/10.10

07/978-3-031-11686-5

Oliinyk, O., Petersen, K., Schoelzke, M., Becker, M., &

Schneickert, S. (2017). Structuring automotive product

lines and feature models : An exploratory study at Opel.

Requirements Engineering, 22(1), 105‑135.

https://doi.org/10.1007/s00766-015-0237-z

Schneider, B., Riedel, O., & Bauer, W. (2022). Review :

Model-based Systems Engineering and Artificial

Intelligence for Engineering of Sustainable Systems –

What contribution can systems engineering and

artificial intelligence provide for the engineering of

sustainable systems as of today? In P. Plapper (Éd.),

Digitization of the work environment for sustainable

production (p. 37‑59). GITO Verlag.

https://doi.org/10.30844/WGAB_2022_3

Sundermann, C., Kuiter, E., Heß, T., Raab, H., Krieter, S.,

& Thüm, T. (2024). On the benefits of knowledge

compilation for feature-model analyses. Annals of

Mathematics and Artificial Intelligence, 92(5),

1013‑1050. https://doi.org/10.1007/s10472-023-09906-6

Visure Solutions. (2023, janvier). Best ALM Tools with

Round Trip integrations with Word & ; Excel. Visure

Solutions. https://visuresolutions.com/mbse-guide/ai-

in-mbse/

Automating Feature Modeling in Product Line Engineering for Systems Engineering: The Application of Natural Language Processing

457