This demonstration paper introduces a diversity-aware extension for the content management system Drupal. It shows how different aspects, such as automatically recognized entities, topics and sentiment scores can be leveraged in a Web user interface. We introduce a coherent approach that enables readers to navigate to further related articles. In particular, we demonstrate new ways to quickly grasp what the articles' sentiments are and which topics they cover before the actual click.
The Diversity Enricher module provides several functionalities that present and
process information that can be considered to enrich Drupal articles with more
diverse information.
One of the most important things about the Drupal extension is the fact that no
user interaction is needed to extract the necessary diversity information. This
information is generated by the Enrycher service3 which is publicly available.
Enrycher utilizes natural language processing techniques to extract diversity
information such as topics, sentiments, sentiment scores or named entities
captured by the article text. This information is then described by using SIOC [
Links to Related Articles and Topics Figure 1 shows an overview on the extension. On the left hand side the original article is presented, whereas on the right hand side the main functionality of the extension is located. There, related articles within the Drupal database are listed, split up according to their extracted overall sentiment. An article is considered to be related, if it has at least one topic in common with the currently shown one and is located in the same cluster of the Diversity-Aware Ranking Service5.
The topics of the related articles can be shown by clicking on the + button
in front of the article titles. In addition, tags extracted from the currently shown
article are presented in a tag cloud below the related article’s tree. The size of
the respective tag is determined by its number of occurrences in the triple store.
Named entities are recognized within the text and get marked. It is possible to
click on all tags, named entities, and topics in order to get articles with the same
tag/topic (see Figure 2). As a further diversity feature, each article’s sentiment
is displayed between the title and the actual text.
3 Enrycher – http://enrycher.ijs.si
4 KDO – http://kdo.render-project.eu/
5 Diversity-Aware Ranking Service – http://ranking.render-project.eu
The diversity data produced by the Enrycher service can be exported in the
following formats: RDF+XML, JSON, and Turtle.
Another important function of the extension is the ability to import additional
articles into the Drupal database, if they are stored in an Sesame store and are
described with the SIOC [
The Drupal extension makes use of several technologies and tools. This section describes how the main parts of the Diversity Enricher Drupal extension interact.
Diversity Mining Web Services (Enrycher) The main functionality of the Enrycher service has already been described in Section 2.1. However, the Enrycher service could be replaced by any Web service that is SIOC and KDO compliant. This means, that the service has to support a subset of the SIOC and KDO functionalities - namely the extraction and proper output of topics, sentiments and sentiment scores.
Sesame triple store Sesame is used as data store back-end. All other components operate on the Sesame store by using SPARQL queries. The Enrycher service returns RDF data which can be directly submitted to the Sesame store. The Diversity-Aware Ranking Service component and the Drupal tool read from the store using a set of predefined queries. Diversity-Aware Ranking Service This service is used to retrieve related articles with differing sentiments. It operates on a Sesame triple store. The core of the ranking service is a clustering algorithm that operates using a distance metric based on topics and sentiment scores. Articles that have at least one topic in common with the current article are preselected and then clustered by topic. All articles that are in the same cluster as the currently browsed one are then marked as related.
Drupal Integration The tool is connected to Drupal with so-called hooks. An implemented hook is called each time a certain event occurs. The hooks of the Diversity Enricher module are – New Article Created: As soon as an article is created, the raw text data is submitted to Enrycher, which extracts the diversity information. This information is then stored to the local Sesame store. – Article Viewed: If the article is viewed the first time and it has been in the database before the Drupal extension has been activated, this hook acts the same as in the case for New Article Created. Additionally, the information needed to present is generated (by using ranking, SPARQL queries) and then presented beside the raw article text. – Article Changed: If an article is changed, Enrycher is again asked for diversity information and the store is updated with the new enrichment. – Article Deleted: If the article is removed, all links to the diversity information is deleted from the Drupal database. 4
The integration of semantic technologies into CMSs brings clear benefits especially for improving search, integration and intelligent management of the content. During the last years several approaches have been published on how semantics can be used within CMSs in general and Drupal in particular.
Since version 7, Drupal natively supports RDF representation of posts, making use of vocabularies like SIOC, FOAF, Dublin Core, and SKOS. Although the new RDF module in Drupal easily enables publishing LOD, it does not provide means for the automatic creation of links to relevant LOD resources.
The approaches described in [
VIE.js6 is a JavaScript-based semantic interaction framework. It facilitates annotation and interaction with textual and RDFa-annotated content on Web pages. It is used in combination with Apache Stanbol7 that supports the extension of CMSs with semantic services. Another annotation framework is given by the OpenCalais8 Drupal extension that uses the OpenCalais API of Thomson Reuters to annotate posts with named entities, facts, and events.
While the above approaches focus on the named entity or topic aspects, we introduce a new dimension given by the active utilization of automatic sentiment extraction. Eventually, this is expected to support the content creation and perception process (given a more fine-grained sentiment and opinion extraction). Also, in contrast to the above approaches, our approach focuses on providing a complete and fully automatic cycle to support the management of diversity; from text analysis and annotation to different visualization methods within Drupal. 5
We developed a diversity-aware Drupal extension coined Diversity Enricher. The module is currently available at http://drupal.org/sandbox/sti-innsbruck/ 1991696. As of the time of writing (i.e., August 12, 2013) the extension is within a review process to achieve “full project status” within the http://drupal.org Web portal. Amongst our next steps will be the qualitative evaluation of the Diversity Enricher Drupal module.
Acknowledgment This research was partly funded by the European Union’s Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 257790 (RENDER project).