This paper describes the experience of researching and teaching the conceptual and practical basis for the specification, modelling and design of an ontology-based news authoring environment for the Semantic Web, that takes into account the construction and use of an ontology of the Zika disease. It has been said that CMSs are being adapted in order to receive semantic features, such as automatic generations of keywords, semantic annotation and tagging, content reviewing etc. We present here the infrastructure designed to foster research on semantic CMSs as well as semantic web technologies that can be integrated into an ontology-based news authoring environment.
Nowadays, text authoring can be seen as a similar practice to those taken 100 years ago, with a slight difference: we have shifted from the hand-pen-paper model in cellulose (that still exists), to the digital finger-keyboard-cursor-white page. In the support level a lot has changed - such as making links to other documents; making and sending as many copies as desired - as we can see from the development of editing resources, which were in the past restricted to editing houses and their complex software. In the syntactic level, we can benefit from searching and ordering key words. However, in the semantic level, text production is the same as before: it depends on the writer´s ability to associate his contents to existing formal concepts structures (links with other documents, links to webpages, associating text to dictionaries, terminologies, taxonomies, indexes, etc).
In the Semantic Web, we are facing a new opportunity to use concept referencing ability of a text – and not only its objects and components such as summaries, images, links, descriptive terms and their meanings. The main problem we are facing today is that the available content on the Web is generated by one person, indexed by another and retrieved by computers that do not make a difference between variant terms.
Based on previous studies
Particularly, when it comes to preparing a journalistic text, users of CMSs – Content Management Systems – in newspapers newsrooms, they count with a blank screen to insert texts with basic formatting options that current editors offer. However, the problem is that these tools limit the use of correct terms, by not giving the author the awareness of using the best term to identify a certain subject as well as its variations. To identify the best keywords to label the subject, to produce tags that are semantically linked, other than hanging loose and ambiguous. This happens to be the case of the subject Zika, disease or virus. The impact of this problem is related to the news production: they may contain useful information but they were not well represented via keywords or hash tags. Our question is: why don´t we use a text authoring solutions that is based in the use of ontologies in the production of news articles, and that is able to make a liaison between the editor´s page and the ontology, thus allowing the author to link a term with a class or instance of an ontology, in search of is localization in the scheme of things?
In this paper we intend to present the context of the creation of the solutions, its motivation and proposals, by indicating a semantically computational platform designed to receive the solution; the Zika Ontology construction process; a general model of the architecture and the support of the authoring environment via a semantic CMS.
This research started with a general specification of the environment, by using as
a starting point the general requirement of ontology based authoring environments
In a nutshell, the proposal of this tool is to annotate terms and concepts used by writers/journalists and to relate these terms with the ontology of the subject, to create links with other information resources about that subject: existing news pages or any other page selected by the writer. Regarding the users side, text can be produced by many journalists and go under different review processes, but tagging and terminology consistency will be provided and supported by the ontology that will guide users in the task of choosing a term to be used and making the links between this term and its related concepts (synonyms, hyperonyms, antonyms, related subjects, etc).
Since the start of this course, due to the teams know-how, we decided to use Python to
manipulate RDF models. Considering that RDF was created to describe resources on the
Web, the resource description framework is of great importance to help find a way to
extract relevant resources. Therefore, we have decided to used RDF/XML due to its
simplicity and as a first formalized serialization, according to Gandon&Schreiber (2014),
as a working model to represent the base ontology, initially constructed in OWL format.
We intend to browse the structure of the ontology and to recover classes and instances
more relevant to the specific contexts about which are being written. Also by identifying
the relations between classes and instances that were listed. Python with RD
Python is a small scale language (RD
The semantic annotation steps are as follows
The RDFS, the RDF Schema (
To perform the search in the database, we used Sparql, which is a standard search language and a data access protocol. It means that Sparql allows more than access to the RDF triples – subject-predicate-object – or graphs but also to any data sources that can be mapped in the RDF. It allows the extraction of semi or fully structured data, to explore data used in unknown relationships queries, makes complex combinations of heterogeneous databases with simple queries, converts RDF data from a vocabulary to another and constructs new RDF graphs from queries in other vocabularies.
During the setting up of the authoring environment, a specific working group responsible for the ontology had meetings with a team of medical staff from the Health Department of Brasília, in order to generate a conceptual map of the Zika Disease (Figure 1). The meetings lasted 10h total approximately and the medical team informed all their knowledge about the virus and the situation of related diseased. As a firs artifact produced, there came a conceptual map for the understanding of the ontology domain. The results were homologated by the teams after the edition and there were some adjustments necessary for the construction of the ontology.
We used the 101 Methodology
When considering reuse in the 101 Methodology, we identified that the term Zika is considered an instance within the ontology of diseases - Diseases (RDO:0000001) and Zika Virus Infection (RDO:0016040) – as presented in Figure 2.
The main definition of Zika was founded in RGD [13]: A viral disease transmitted by the bite of AEDES mosquitoes infected with ZIKA VIRUS. Its mild DENGUE-like symptoms include fever, rash, headaches and ARTHRALGIA. The viral infection during pregnancy, however, may be associated with other neurological and autoimmune complications (e.g., GUILLAIN-BARRE SYNDROME; and MICROCEPHALY).Zika virus (C0318793). In the Snomed CT [17] we found only a description of the Zika Virus as a final instance of a Flavivirus family. Thus we concluded the absence of a specific ontology about Zika, which allowed us to build a new one. After the definition of classes, relationships, properties and restrictions, we conducted the construction of the ontology itself with the support of Protegé.
Thus, after following all the steps and activities provided by the 101 Methodology, we generated the ontology, represented here by the graph on Figure 3. This ontology was then validated by the group responsible for its construction, including the medical staff and other specialists in the subject area. We chose the reasoned Hermit to validate the Zika ontology to do its consistency checking capacities and to help review the general structure of the ontology.
This project is an experiment to develop an ontology application in which undergraduate and master students learned the process and the contents via a PBL – Project Based Learning approach at the University of Brasília, UnB. Throughout the context of application done in the area of journalistic texts production about Zika, we needed to apply a flexible software development process along with the detailed documentation of all its guidelines.
For this purpose a unified and hybrid development process was defined, that was a combination of the RUP processes and the application of the Scrum methodology using Kanban. From the traditional process we used general artifacts (architecture documents, glossary, iteration plan, vision document) and the phases (conception, preparation, construction and transition) were incorporated the Scrum methodology inside the 5 sprints planed for the project.
The vision document was the starting point to have a better notion of the scope of the product, however only in the architecture report the scope was finally defined and all the details designed and documented. The glossary was useful to a better understanding of the specific terms that involves the context of the project. Finally, the iteration plan documented all the used methodology, as well as the division of the Sprints phases, as well as the definition of the platform used.
The iteration plan aims to give information of how the project organization was designed, including the integration of its activities, resources, deadlines, technologies and all the relevant information. A general scope, a selected development process and their guidelines were presented as a result.
Considering that the project is about a new theme to the majority of the students, with the constant need to research the areas that involves or results, it was necessary to apply an agile method, thus, the Scrum methodology was chosen. There was also the need of another more precise documentation process and well defined phases (figure 4). For all these reasons and also for the demand of an agile methodology used, the development process chosen for the project was a mixture of the traditional and the agile into a hybrid software development process .
As showed in Figure 4, the goal of each phase inside the sprint is arranged as follows: Initiation (or conception) with its focus on the understanding of the system; Elaboration: with its focus on the definition of the solution; Construction: with its focus on the implementation and to have the solution tested; and Transaction: with its focus on the implementation of the system in the context itself.
When trying to create a semantic authoring environment, we faced some problems, starting with the lack of literature describing this very objective. We searched for recommendations from theW3C about ontology driven applications as well as other books. Some of our main questions surrounded the following issues: RDF is a kind of a set of individual data saved in a schema based on an OWL graph. The retrieval of this data is made via Sparql, however there are some limitations: How can we write an RDF at each new register of an application? How can we modify this same RDF? Will it be necessary to create and RDF every time the new register is filled in? Is it possible to convert OWL to a structure model in SQL? Then we defined some sceneries: today´s use of structural data bases, as showed in figure 5 below, is a structured application that works in 3 main areas: logic and processing, data bases and visual. The user visualizes the screens, makes requests to the program and process the requests for data from the base, which returns the data that are interpreted by the application and shown to the user.
We show a general view of a system architecture that uses a series of different architectural views to illustrate the different aspects of the system. The intention is to capture and transmit the main decisions that were taken in regards to the system, from an architectural point of view.
In a simplified manner, we present here how the project implementation was modeled, by showing each tool and language used. In order to implement the project, we found two viable paths: one using Python, according to the views of Jakub Talas, Tomás Gregar, and Tomás Pitner (2011) and another one based in the Java platform, as suggested by Seiji Isotani and IgIbert Bittencourt (2015). This research was conducted in an undergraduate course research, so we were free to try both paths.
The Python platform is composed of machines based in an Intel X86 Architecture, Linus O.S. Ubuntu, We Apache Server, applications in language Python and libraries in Python RDFlib for the data treatment in the RDF and Django formats for the CMS. Thus the text authoring environment interface can present 3 worlds of functions: writing, ontology and semantic search engine: 1) we use Django to present a window of the document being edited, 2) another one with an RDF graph, corresponding to the semantic document annotation of the text being edited, and in a 3) third window the returned documents fro the semantic search engine too that uses the RDF graph of the document to search for semantically related document.
The Java platform was set up with the following specifications. In a machine with
and Ubuntu S.O. of Unix distribution (Linux), we installed the Apache Tomcat software
so that is tis possible to manage a local sever based on Java servlets and supported by a
Semantic CMS, here suggested the Apache Stanbol, from
Figure 9 shows a summary of the solution components model, where the
application database comes from the domain ontology defined. While editing a text in a
wiki (i.e. MediaWiki) environment, the platform recognizes the text edited via annotation
with a Sesame RDF4J Joint and Jena application, and via servlets Java, interacts with the
Apache Tomcat and Apache Stanbol,
We presented in this paper the experience taken in the Laboratory of Special Projects with students of the Department of Computer Science and Software Engineering of the University of Brasília with the challenge to understand and apply Semantic Web technologies to enhance the semantic capacity of CMSs. The main results of the experience, related in the paper, was the construction of an OWL Zika Ontology, the modelling of the authoring environment and the implementation of the database search mechanism.
The architecture model for a prototype of a semantic CMS was described and implemented in the lab. This model represents the effort and practice of the students who showed advanced abilities to deal with semantic web challenging issues in a computer science environment, even though these students were not familiar with these technologies before.
We brought here the following contributions to the area: the specification, modelling of an authoring environment for a text editor supported by a semantic and lexical interpreter for the edition of news articles about Zika, supported by a specific ontology created by the students. The following steps in this research is to implement the authoring environment in full, allowing real-time concept recognition from text annotation with the ontology.
This experience resulted in an environment that allows the use of a text editor, integrated to a semantic CMS, in which terms can be typed and in parallel be automatically recognized and associated to classes and instances of the Zika ontology. From the relationships created between the ontology and the text, one is able to obtain for this annotation a list of keywords and conceptual #tags that identify specific subjects of the text, the scope of the article in relation to the general context of the Zika ontology. It also correlates the text with already existing texts and articles or pages so that they can be interconnected via non ambiguous semantic relationships.
This work shows the feasibility in the use of ontologies during the moment of text production, that is, during the moment authors are deciding which terms to use in the text, in order to enhance information representation. The difference from other approaches is that the use ontologies mostly for post-publication or for information retrieval purposes. We also showed that it is possible to implement the solutions, not yet identified in existing CMSs available in the market, which yet do not benefit from ontology-based solutions that enhance knowledge representation capabilities.
Apache Jena. A free and open source Java framework for building Semantic Web and
Linked Data applications. http://jena.apache.org/index.html> Access in May 2016. Apache Jena. Configuring Fuseki. https://jena.apache.org/documentation/fuseki2/fusekiconfiguration.html Access in April 2016.
/50471002 Access in May 2016.
http://purl.bioontology.org/ontology/SNOMEDCT RDFLIB. rdflib 4.2.2-dev. https://rdflib.readthedocs.io/en/latest/> Access in May 2016.
Access in