Interdisciplinarity in Sedela
The first SEDELA challenge was to bridge the gap between several domains, Educational Science, Technology Enhanced Learning and Computer Science in order to foster synergies between self-directed learning methods formalization, semantic models design, and trusted collaborative services design. Meeting this challenge was ensured thank different tasks :
i) initial proposal was reworked in order to tighten links between teams, conducting to collective position papers.
ii) a tentative follow up of the project by answering to ANR 2018 project call, provided the opportunity to deepen the interdiscplinary links of the project. Our proposal went to the second turn, but was finally not retained.
iii) project workshops were conducted to encourage exchanges. Out of them, trust was identified as a cornerstone of the project.
iv) the design of the experimental infrastructure allowed us to specify the main issues for the development of autonomy and to focus on two specific service, namely autonomy questionnaire and reflective writing. By adopting a participatory approach, we have been able to associate teachers from the experimental fields and to encourage links with pedagogical experience of the computer science researchers.
v) external workshops with interdisciplinary researchers were also organized to get feedback from the community.
Personal Cloud Infrastructure
Since the begining of the project, the question of personal data privacy has become widespread. GDPR has been adopted by European Union. Recommandations for educational data management have been promoted [Draschler et al. 2016] and yet no solution to a lifelong access for learners has been proposed. We still believe that Semantic Personal Information Managers (SPIMS) is an elegant solution to ensure learner autonomy in an international and inter-institutional environment. We helped to initiate another experimentation conducted by Rennes Academy, Ministry of National Education, that share with us the same initial hypothesis.
However, existing infrastructures are still in infancy. We based our developments on an innovative, and open-source solution proposed by a french startup, namely CozyCloud. Even with close relations with the developers of CozyCloud, we encountered many difficulties. In short, our research needs didn’t met the development line of the startup.
Another candidate infrastructure for SPIMS was released during the time of our project, namely Solid, proposed by Tim Berner’s Lee. This solution seems to be closer to our research hypothesis, as it provides natively a linked data management, and clear policies for external access control of data. We conducted basic tests that confirmed this. Even so, we didn’t migrate to this solution, because lack of time, and because that the infrastructure was not proved to be stable enough.
For stability reasons, our prototype was finally deployed out of any personal cloud infrastructure.
Advances in evaluating Autonomy Development
The CREAD team developped in previous studies a professional development model that includes autonomy. More precisely, the theoretical framework is set in order to determine to what extent learners develop their autonomy capacity to progress in their professional development. This model integrate four dimensions: autonomy, identity, socialization and posture.
The SEDELA project provide the opportunity to operationalize this model in a questionnaire that make possible to collect declarative answers on quantitative Lickert scales. Consistency tests has been conducted and proved that the “professional posture” dimension was not correctly measured. Further studies are needed to provide a fully operational model, based on analysis of reflective writings.
This questionnaire has been proposed three times to six cohorts of students (N=88). This provided data that give some results of student progress during the year, that proves student progress in their professional development and autonomy as well. These results give the opportunity to conduct further studies, especially to assess the effect of professional posture to autonomy development.
Aside these results, this data set was made available to test a machine learning approach to derive learner models out of questionnaire data. We will cover this aspect in the next section.
In parallel, other approaches supporting autonomy development were analyzed. Two papers were published.
Journal paper at The Canadian Journal for the Scholarship of Teaching and Learning paper and one paper under publication
Computable Learner Models
Learner Models may be seen as a pivot of the project. Three challenges were identified. The first challenge is to derive computationable models from experimental data. A second challenge is about opening learner models, i.e. making them scrutable by learners, but was not addressed during this project. Third challenge is to make models and data semantically available.
Two specific services are explored by SEDELA project, namely autonomy questionnaire and reflective writing.
The first service is a questionnaire measuring learner’s autonomy. Litterature depicts that self-regulated learning is one of the operational key of autonomy. Self regulation may be observed along six dimensions. Hence we developped a process deriving those dimensions out of questionnaire answers. This process is based on a bayesian networks analysis, that give to the learner a measure on each of these dimensions. This process was trained with a subset of the available data, and tested on the rest of data.
The second service is reflective writing. An analysis conducted by the CREAD team confirms that many information about autonomy development may be collected out of these writings, and out of exchanges between the learner and his tutors. However, automating data collection will require Natural Language Processing and an extensive study of possible observations. This work is beyond the scope of SEDELA project, but is identified as a possible fruitful continuation of SEDELA project.
Full bayesian model available at github
Trusted Collaborative Services
1- State of the art on data integration techniques and usage policies
This report presents the state of the art of the two main issues to construct the decentralized and semantic learning Infrastructure for lifelong learning, i.e. semantic personal data integration and trusted data sharing.
A short version of the report is published at Atelier Web des Données (AWD) at EGC 2019. Paper.
Presentation at Journée scientifique "Data Science, IA et Education", 2019 Slides.
2- On-demand Semantic integration service
ODMTP (On Demand Mapper with Triple pattern matching) enables triple pattern matching over non-RDF datasources.
Live demos available for:
github for Github (limited to 60 request per hour)
LinkdIn for your Linkedin profile (you will need to login to access your personal LI profile).
Demos presented at:
16th International Semantic Web Conference (ISWC2017), demo paper
atelier Web des Données (AWD) with EGC 2019, demo paper
3- A classification model for RDF datasets licenses
Web applications facilitate combining resources (linked data, web services, source code, documents, etc.) to create new ones. For a resource producer, choosing the appropriate license for a combined resource is not easy. It involves choosing a license compliant with all the licenses of combined resources and analyzing the reusability of the resulting resource through the compatibility of its license. The risk is either, to choose a license too restrictive making the resource difficult to reuse or to choose a not enough restrictive license that will not sufficiently protect the resource. Finding the right trade-off between compliance and compatibility is a difficult process. An automatic ordering over licenses would facilitate this task. Our research question is: given a license li, how to automatically position li over a set of licenses in terms of compatibility and compliance? We propose CaLi, a model that partially orders licenses. Our approach uses restrictiveness relations among licenses to define compatibility and compliance. We validate experimentally CaLi with a quadratic algorithm and show its usability through a prototype of a license-based search engine. Our work is a step towards facilitating and encouraging the publication and reuse of licensed resources in the Web of Data.
Research Paper published at 16th Extended Semantic Web Conference (ESWC2019), paper
Demo paper published at 34ème Conférence sur la Gestion de Données – Principes, Technologies et Applications (BDA 2018). Demo paper.
- Find dataset compatible License :http://cali.priloo.univ-nantes.fr/ld
- Find Respository Compatible License: http://cali.priloo.univ-nantes.fr/rep/
- A video of demonstrations in youtube
4- Recommending Plausible Federated SPARQL queries
Federated SPARQL queries allow to query multiple interlinked datasets hosted by remote SPARQL endpoints. However, finding federated queries over a growing number of datasets is challenging. In this paper, we propose PFed, an approach to recommend plausible federated queries based on real query logs of different datasets. The problem is not to find similar federated queries, but plausible complementary queries over different datasets. Starting with a real SPARQL query from a given log, PFed stretches the query with real queries from different logs. To prune the research space, PFed proposes semantic summary to prune the query logs. Experimental results with real logs of DBpedia and SWDF demonstrate that PFed is able to prune drastically the logs and
recommend plausible federated queries.