Zaizi recently participated in the 7th IKS workshop that took place in Salzburg, Austria on 12-13 June. We joined the IKS community to share our experience in the use of Semantic Technologies within the enterprise and how IKS technology stack can enhance the functionality of Enterprise Content Management systems and make it easier to interact with and find the right content. We saw great demonstrations from IKS early adopters showing how to build semantic real-world enterprise applications using Stanbol and VIE.
It was a very interesting conference where we had the chance to network, knowledge share and go back with a lot of new ideas for our own products and solutions. For us, it was a pleasure to present our Semantic Search Tool in Alfresco that also uses Stanbol as Semantic Services provider.
We are currently developing a complete Semantic Search Tool in Alfresco. Our main goal is to improve the search and the user experience within Alfresco installations with high volume of information. As the volume of content increase, it becomes more and more difficult to find the right content. Even if content are organised hierarchically, navigating deep number of folders or categories isn’t easy. Also, ECM systems don’t provide tools to automatically extract the underlying knowledge within the content and help the user to find content based on concepts and knowledge.
Our semantic search solution involves distinct search techniques using semantic data:
- Ontologies and thesauri: custom vocabularies are very importants within enterprises. Each organization usually handles a list of terms and concepts specific to a particular domain. Automatically incorporate this knowledge to semantic indexes and the ability to extract them from unstructured content is critical to rapidly adapt the search tool to any domain.
- Data Relationships: the benefits of the LinkedData paradigm in the Web are well known as well as its uses. But LinkedData is also very important in the local or closed scope of an enterprise. Semantically connect related entities from the content allows for example to generate graphs of related documents and explode this relationships to provide advanced search capabilities to our users.
- Faceted Search: as a way to better organize the search results, we can use distinct types of semantic facets using entities as well as any existing metadata.
- Social Recommendations: we use recommendations to improve the user experience in Alfresco. Recommendations are a great resource for users to take advantage of the work of others users in a site. We have developed a Recommender Engine using Apache Mahout which enables Alfresco to provide recommendations. For example, we can recommend content to a user document based on what other users have also viewed. This is similar functionality to how Amazon is recommending content.
- Intelligent Search: in addition to the use of semantic data, we are improving the user interface to improve classical search over a collection of documents. Resources like click through data and techniques like language models, collocations, etcetera are useful in order to boost a typical information retrieval tool.
Stanbol is a perfect tool to achieve most of the above features. In fact, almost each of them could be covered with the Stanbol components. We are gradually integrating Stanbol’s components to support all these features in our Search Tool in Alfresco. Also we are putting Stanbol in the center as a semantic services provider in our architecture. All our knowledge extractors and content enhancers are being deployed as Stanbol engines in the pipeline. So now, for example, we are able to extract entities from the documents and link them not only with DBPedia or others configured sites in Stanbol, also with custom sites like a LDAP Server.
Below you can find the slides of our presentation at IKS Workshop and a screen cast of the current demo:
Using VIE in Alfresco Share
We have extended annotate.js default functionalities, developing some new features, like adding you own selected entities linking them properly and organizing all entities in a list grouping them by type as you can see in the following screencast:
The Complete Presentation
You can view the full presentation from the conference below.