Stephan Raaijmakers is an expert in natural language processing with machine learning/AI methods. He works on various topics in security, health and government. He holds a chair on Communicative AI at Leiden University, focusing on natural language-based dialogues of humans with AI, using deep learning techniques.
The chair addresses communicative AI: AI that engages in communicative interfaces with humans. The chair focuses in particular on deep learning and natural language processing. It addresses the implementation of explanatory, natural language-based facilities in AI, and explicitly seeks connection with neurocognition, neuro/-psycholinguistics and general linguistics - disciplines that resort under LUCL.
Further, the chair enforces the connection with LIACS (Leiden Institute of Advanced Computer Science) and the digital humanities community in Leiden (LUCDH, Leiden University Center for Digital Humanities) and the Netherlands. The professorship reflects TNO's orientation on trustworthy, transparent and controllable AI and will contribute to a better knowledge position for TNO on human-AI interaction.
In 2021, the chair focused on
- The acquisition of collaborative conversational AI research projects with TNO and stakeholders. Stephan has initiated a Horizon Europe proposal (SAFECHAT, submitted November 2021) together with LUCL, TNO and a slate of other stakeholders and parties.
The proposal addresses the role of conversational AI in detecting grooming behavior by paedophiles targeting juveniles, and its dual use as a resilience building tool (pending). TNO, TUD, UvA and other parties submitted a NWO-ORC proposal on cultural heritage of water management, in which conversational AI was embedded.
This proposal is under re-submission in 2022. An ELSA lab proposal on the value base of communicative AI was developed with TNO, Leiden University and (a.o.) LUMC; due to last minute internal disruption of one of the crucial organizations involved, it was decided to not submit the proposal; it is currently being transformed into a NWA-ORC proposal (to be submitted in 2022).
Stephan was involved as advisor in the application of a NWO-SSH proposal by Leiden University on setting up the infrastructure for NLP for legal engineering, which was granted. Stephan participated in preparatory work for the Growth Fund on AI and NLP (the NAIN initiative, a ‘chain project’), together with TNO colleagues (to be submitted most likely in 2022).
Stephan was involved in the application of NWA-ORC InDeep (explainable deep learning-based NLP), together with the UvA a.o., which was granted. His chair contributes to a limited extent to this project. Together with two colleagues from Leiden, a successful NWO SSH submission has been prepared, resulting in the funding for an NLP infrastructure for legal engineering.
- Teaching and carrying out joint research with PhD students. Two courses on deep learning for NLP were taught (BA and MA level). Collaboration with LIACS was initiated, resulting in an on-site (LIACS) course on deep learning (BA) for 2022. See (4).
- The organization of SAILS workshops and hackathons, with anticipated participation of TNO.
- Developing a new master track on computational linguistics. A new master track was developed by Stephan, bringing together his core courses on deep learning for NLP with other relevant core and optional courses for computational linguistics and machine learning. The program will be effective as of 2022-2023.
- Setting up a Special Issue on communicative AI for Frontiers in Artificial Intelligence. A call for papers was prepared, and the issue will appear early 2022, with a collection of selected papers and an editorial written by Stephan and co-organizing colleagues (from TNO, Leiden University and Tilburg University). One the papers submitted was written by one of the PhD students of Stephan, with Stephan as one of the co-authors.
- Developing a novel method for adaptive conversational AI, using deep learning-based Transformers. Results and publication will become available in 2022.
- Active participation in the Leiden SAILS committee (the Leiden University AI program, spanning faculties).
- Contributing to the Leiden University Centre for Digital Humanities lab investment plan.
The inaugural lecture for the chair has been prepared and was originally scheduled for December 2020, but was unfortunately cancelled due to covid restrictions. It has been re-scheduled for November 2022.
- A PhD student to be supervised (as promotor) by the chair, under the SAILS program, has been hired; supervision starts in January 2021. Co-supervision with LIACS and FSW (Psychology).
- A new PhD student (TNO colleague) has been successfully enlisted, starting in 2022.
- A prospective PhD student has been supervised for a PhD grant application.
- Brewster, C., Nouwt, B., Raaijmakers, S., & Verhoosel, J. (n.d.). “Ontology-based access control for FAIR data”. Data Intelligence, 2(2020), 66–77.
- Stephan Raaijmakers. “Deep learning for natural language processing”. Manning publications, prepared for paper edition in 2021, published in paper Spring 2022.
- Raaijmakers, S., & Brewster, C. “Exploiting Ontologies for Deep Learning: A Case for Sentiment Mining”. In Proceedings of the 14th ACM International Conference on Semantic Systems, Posters and demo track, 2018.
- Raaijmakers, S., Sappelli, M., & Kraaij, W. “Investigating the interpretability of hidden layers in deep text mining”. In Proceedings of the 13th ACM International Conference on Semantic Systems, pp. 177-180, 2017.