Slide
Centre Interdisciplinaire
de Recherche et d’Innovation
en Cybersécurité et Société
de Recherche et d’Innovation
en Cybersécurité et Société
1.
Ngouanfouo, C.; Davoust, A.
Detecting Machine-Generated Text using Grammatical Features Proceedings Article
In: Proc. Int. Conf. Tools Artif. Intell. ICTAI, pp. 843–848, IEEE Computer Society, 2025, ISBN: 10823409 (ISSN); 979-833154919-0 (ISBN).
Abstract | Links | BibTeX | Tags: AI Text Detection, CNN, Computational grammars, Detection methods, Language model, Machine-generated texts, Natural language generation, Natural language processing systems, Neural encoding, Neural modelling, Part Of Speech, Part-of Speech, Speech communication, Text detection, Written texts
@inproceedings{ngouanfouo_detecting_2025,
title = {Detecting Machine-Generated Text using Grammatical Features},
author = {C. Ngouanfouo and A. Davoust},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-105031903675&doi=10.1109%2FICTAI66417.2025.00123&partnerID=40&md5=5783b8797a3425f9dfa737343ee757d2},
doi = {10.1109/ICTAI66417.2025.00123},
isbn = {10823409 (ISSN); 979-833154919-0 (ISBN)},
year = {2025},
date = {2025-01-01},
booktitle = {Proc. Int. Conf. Tools Artif. Intell. ICTAI},
pages = {843–848},
publisher = {IEEE Computer Society},
abstract = {Large Language Models (LLMs) have advanced natural language generation but pose ethical and practical challenges, making it crucial to detect machine-generated texts. Traditional detection methods rely on complex, hard-to-interpret neural encodings and model-specific features like perplexity. This study explores whether grammatical patterns-specifically sequences of parts of speech (POS), including punctuation and symbols-can distinguish machine-written texts from human ones. Using a CNN classifier on POS sequences, the approach achieves nearly 90 % accuracy on a benchmark dataset. Combining POS-based features with neural embeddings improves performance, and the model shows robustness against adversarial attacks, though it is less effective on short texts. © 2025 IEEE.},
keywords = {AI Text Detection, CNN, Computational grammars, Detection methods, Language model, Machine-generated texts, Natural language generation, Natural language processing systems, Neural encoding, Neural modelling, Part Of Speech, Part-of Speech, Speech communication, Text detection, Written texts},
pubstate = {published},
tppubtype = {inproceedings}
}
Large Language Models (LLMs) have advanced natural language generation but pose ethical and practical challenges, making it crucial to detect machine-generated texts. Traditional detection methods rely on complex, hard-to-interpret neural encodings and model-specific features like perplexity. This study explores whether grammatical patterns-specifically sequences of parts of speech (POS), including punctuation and symbols-can distinguish machine-written texts from human ones. Using a CNN classifier on POS sequences, the approach achieves nearly 90 % accuracy on a benchmark dataset. Combining POS-based features with neural embeddings improves performance, and the model shows robustness against adversarial attacks, though it is less effective on short texts. © 2025 IEEE.
2.
Gagnon, S.; Azzi, S.
Semantic Annotation of Parliamentary Debates and Legislative Intelligence Enhancing Citizen Experience Journal Article
In: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13429 LNCS, pp. 63–76, 2022, ISSN: 03029743, (ISBN: 9783031126727).
Abstract | Links | BibTeX | Tags: Analytic tools, Core functionality, Digital solutions, Knowledge graph, Language processing, Laws and legislation, Legislative intelligence, Natural language processing systems, Natural languages, Parliamentary debate, Parliamentary proceedings, Semantic annotations, Semantic-analytics, Semantics
@article{gagnon_semantic_2022,
title = {Semantic Annotation of Parliamentary Debates and Legislative Intelligence Enhancing Citizen Experience},
author = {S. Gagnon and S. Azzi},
editor = {Kotsis G. Francesconi E. Ko A.},
url = {https://www.scopus.com/inward/record.uri?eid=2-s2.0-85135782631&doi=10.1007%2f978-3-031-12673-4_5&partnerID=40&md5=0765bacc7d38f77896bd9adf402268b9},
doi = {10.1007/978-3-031-12673-4_5},
issn = {03029743},
year = {2022},
date = {2022-01-01},
journal = {Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)},
volume = {13429 LNCS},
pages = {63–76},
publisher = {Springer Science and Business Media Deutschland GmbH},
abstract = {The concept of “Legislative Intelligence” (LegisIntel) refers to Artificial Intelligence (AI) and semantic analytics tools implemented in parliaments to enhance citizen experience in monitoring complex interrelations among various contents of parliamentary proceedings. The integration of a suite of digital solutions can build upon the core functionality of Semantic Annotation of Parliamentary Debates. Using well-established Natural Language Processing (NLP) technologies, linked to ontologies and Knowledge Graphs (KG), it can help identify the concepts and entities throughout texts, and index sentences and summaries as per a citizen’s customized knowledge base. These annotations can then be leveraged to recommend relevant text excerpts end-users could build upon, within teams if they chose to do so, and possibly compose and customize legislative critiques and recommendations thoroughly tested for coherence, accuracy, and evidence. The present study proposes an international open-source initiative among parliaments to ensure the launch and viability of a suite of LegisIntel solutions. It reports on the completed initial phase of this initiative, aiming to prepare discussions in launching an international consultation among peers. The goals of this phase are to document the core functionality of LegisIntel solutions and formulate a proposed architecture that may serve to generate ideas from various developer communities. The Action Design Research (ADR) methodology is used in this process, with results focused on system artefacts such as an interface mockup, a functional design, and a model of infrastructure components. The conclusion addresses risks and outlines the next steps of this initiative. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.},
note = {ISBN: 9783031126727},
keywords = {Analytic tools, Core functionality, Digital solutions, Knowledge graph, Language processing, Laws and legislation, Legislative intelligence, Natural language processing systems, Natural languages, Parliamentary debate, Parliamentary proceedings, Semantic annotations, Semantic-analytics, Semantics},
pubstate = {published},
tppubtype = {article}
}
The concept of “Legislative Intelligence” (LegisIntel) refers to Artificial Intelligence (AI) and semantic analytics tools implemented in parliaments to enhance citizen experience in monitoring complex interrelations among various contents of parliamentary proceedings. The integration of a suite of digital solutions can build upon the core functionality of Semantic Annotation of Parliamentary Debates. Using well-established Natural Language Processing (NLP) technologies, linked to ontologies and Knowledge Graphs (KG), it can help identify the concepts and entities throughout texts, and index sentences and summaries as per a citizen’s customized knowledge base. These annotations can then be leveraged to recommend relevant text excerpts end-users could build upon, within teams if they chose to do so, and possibly compose and customize legislative critiques and recommendations thoroughly tested for coherence, accuracy, and evidence. The present study proposes an international open-source initiative among parliaments to ensure the launch and viability of a suite of LegisIntel solutions. It reports on the completed initial phase of this initiative, aiming to prepare discussions in launching an international consultation among peers. The goals of this phase are to document the core functionality of LegisIntel solutions and formulate a proposed architecture that may serve to generate ideas from various developer communities. The Action Design Research (ADR) methodology is used in this process, with results focused on system artefacts such as an interface mockup, a functional design, and a model of infrastructure components. The conclusion addresses risks and outlines the next steps of this initiative. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.