SlideWiki.org is a Web-based OpenCourseWare (OCW) authoring system that enables educators and learners to collaborate on creating, sharing, re-using and re-purposing multi-lingual open educational content. The SlideWiki platform allows people to author FAIR (Findable, Accessible, Interoperable, and Reusable) educational content. SlideWiki supports many features to semantically-enrich educational content to support FAIR authoring. In this paper we will present those features of the platform such as Linked Data interface, manual and automatic content annotation as well as content linking and metadata.
A major obstacle to increase the e ciency, e ectiveness and quality of
education is the lack of widely available, accessible, multilingual, timely, engaging and
high-quality educational material (i.e. OpenCourseWare). The creation of
comprehensive OpenCourseWare (OCW) is tedious, time-consuming, and expensive.
Courseware employed by educators is therefore often incomplete, outdated, dull,
and inaccessible to learners with disabilities. With the open-source SlideWiki
platform (available at SlideWiki.org) the e ort in creating, translating, and
evolving FAIR (Findable, Accessible, Interoperable, and Reusable) OCW can
be widely shared (i.e. crowdsourced). Similarly to Wikipedia for encyclopaedic
content, SlideWiki [
Figure 1 depicts the 3-tier technical architecture of the SlideWiki platform where data, service and user interaction concerns are decoupled as individual stand-alone components. Our contribution to semantically enrich educational content in SlideWiki, touches upon all those 3 layers: RDF and Linked Data version of content are provided in the data layer, NLP (Natural Language Processing) services are exposed for automatic content annotation, and user interfaces for manual content annotation together with inline metadata to increase 1 see http://slidewiki.eu for more information the ndability of content are provided in the interaction layer. In the following subsections we brie y introduce the SlideWiki semantics-related features: 1.1
An RDF data model allows easy exposure and integration of educational data
and the interlinking of data across system boundaries. In order to foster the
generation of RDF from legacy systems based in relational databases, there is a
stack of technologies and standards such as a) R2RML (W3C recommendation2)
to generate RDF from relational databases, b) RML (an extension to R2RML3)
to generate RDF from JSON, XML and CSV datasets, c) XR2RML [
To generate the RDF version of SlideWiki, we employed Morph (MorphXR2RML4) that implements the standard XR2RML. Morph is a java program that reads two les: a con g le and a mapping le. Although both les can be modi ed by the developers, in practical terms the con g le is xed (stable in time) and the real e ort is focused on the mapping le. A mapping le establishes how the elements in the SlideWIki MongoDB instance are converted. The result of SlideWiki RDF conversion is exposed as a Linked Data interface on top of an open SPARQL endpoint5. 2 https://www.w3.org/TR/r2rml/ 3 http://rml.io/RML_R2RML.html 4 https://github.com/frmichel/morph-xr2rml/tree/morph-xr2rml-1.0 5 available at http://slidewiki.oeg-upm.net/sparql
Presentations get more visibility when they are tagged with words or entities about their topic and contents. Annotations make the contents of a deck more meaningful to users as tags tell a user what the deck is about. An author can manually set tags for his/her presentation. To support the user in this process, tag recommendations are calculated via a Natural Language Processing (NLP) service (based on calculation of important words and named entities in the presentation) and presented to the user. The user can select/approve recommended tags to improve the semantic enrichment of his/her presentation (cf. Figure 2).
Another content annotation method is via in-slide annotation. Users select phrases in a slide and then annotate these phrases to be instances of an ontology class. We will provide both our own SlideWiki speci c ontology (partially based on existing OCW and OER ontologies) a user-created or uploaded ontology, and the existing ontologies or vocabularies used in knowledge bases accessible on the internet via a SPARQL endpoint. We also call DBpedia spotlight to give suggestions to annotate phrases based on the DBpedia knowledge base (linking to a DBpedia instance or instantiating as DBpedia concept). Presentations are considered semantically enriched when they are linked to external knowledge bases. The in-slide annotation (cf. Figure 3) is still under development.
There are also automatic processes in SlideWiki for content annotation: In each slide DBPedia Spotlight entities are identi ed by the NLP Analysis from the NLP service (See https://nlpstore.experimental.slidewiki.org/documentation) including a DBPedia URI. Named Entity Recognition is used to recommend tags via the NLP service. The calculated results of the NLP service are stored in the NLPstore in order to provide annotation and tag suggestions to the user in a fast manner by querying pre-processed recommendations in the future.
The main NLP API called nlpForDeck performs several NLP steps encapsulated in one service. It can be divided in 3 main parts: 1) Preprocessing (like html to text, automatic language detection, tokenization) 2) Automatically retrieve entities used in the slides via DBpedia spotlight (use for in-slide annotation and semantic enrichment) 3) Identi cation of important words and entities useful for tagging the presentation in the platform6.
Presentations are considered semantically enriched when linked to presentations with similar content. The user can dive deeper into the given topic and get a better understanding of the topic by viewing related presentations. This can be either done by showing presentations with the same tags or by linking to presentations with similar content (even if they do not share the same tags). The latter is performed by the deck recommendation based on the content of the presentation.
As most web search engines incorporate semantic data in their search, semantic enrichment of the educational resources will result in a higher ranking and better visibility. This is particularly important for improving search engine performance for content within SlideWiki. SlideWiki will make use of metadata HTML tags as well as embedded Microdata and JSONLD description of educational resources to support SEO (Search Engine Optimization). 6 The importance of the words and entities is determined via TFIDF (term frequencyinverse document frequency) ranking.