FAQs
Does an increasing number of ontology relationships become difficult to manage? What are some guidelines around how many relationships are manageable?
Wednesday, 2 February 2011
Any increase in complexity means an increased management overhead. We try to make the tools easy to use to keep the process of building and maintaining the model possible. In our domain the ontology is often more of a set of “federated taxonomies” – a series of distinct classes or facets of information with links between them. The semaphore model operates at a human interaction/discovery level – so keeping the relationships to an manageable set is important for the users to gain familiarity with the model. We do not build OWL type structures where unique relationships can be created between any target and source. This model is excellent for machine inference engines where traversing complex interactions is the goal – a different purpose then the semantic enhancement for content management and findability that we provide.
How would you scale this technology of intelligent search for the enterprise to the world wide web at large?
Wednesday, 2 February 2011
There are other software vendors and applications trying to generate a “generic” semantic intelligence layer for the world wide web. Our solution is focused on the enterprise, where there is huge value in capturing the subject matter expertise into a taxonomy or ontology and deploying that in search so all members of the enterprise benefit from an improved understanding of the context of the information, increased discovery of new items and improved re-use of existing artifacts. The actual Semaphore engines – Classification Server and Semantic Enhancement Server are extremely efficient and capable of scaling to large throughput and query volumes respectively.
In all search problems, there is a combinatorial explosion of relationships which is exponential. Which tool do you use to handle this?
Wednesday, 2 February 2011
Our models tend to be designed for post co-ordinate indexing. By classifying content along many facets, with potential for many subjects per facet being returned a reasonable spread of meta-data is generated to match any related search. The user experience design is often aimed at filtering a search by entering into a dialogue with the user – i.e. providing intelligent concept mapping suggestions or dynamic filters so the user to refine their initial search. For example by adding 2 filters the combination of possible answers is greatly reduced.
Are you using audio or video content for searching?
Wednesday, 2 February 2011
The images are indexed using the metadata as discussed.
How the audio/video content is automatically taxonomized automatically?
Wednesday, 2 February 2011
Semaphore classification is based on a Natural Language Processing engine. We can extract any text metadata associated with image and video formats that support some form of text property, or often the path and filename can give us an indication. This can be assessed by the rules based engine. A different classification strategy can be employed for images to compensate for the relatively small amounts of text evidence available.
Will your tool work together with Sharepoint 2010?
Wednesday, 2 February 2011
Yes, we provide an integration pack from Semaphore to SharePoint 2010. It is a comprehensive integration, extending the Management Metadata Service and Term Store to provide comprehensive taxonomy development and governance capability and automatic classification of any content (library, blog, wiki, publishing page, etc).
How do you account for the term variation in legacy documents? You are surely only able to capture a fragment of synonyms (and term variation) in the taxonomy tool.
Wednesday, 2 February 2011
A combination of techniques including stemming, sentence rules (reordering the order of terms) text mining and synonym enrichment will help to overcome a significant amount of variations.
What tools have you used to manage the taxonomies etc?
Wednesday, 2 February 2011
Semaphore’s Ontology Manager provides proper governance and control for taxonomies and ontologies. It maintains the integrity of the model, makes it available to all systems and users that need it, maximizes the performance of the model and slashes the time to build and manage taxonomies by using text mining and entity extraction techniques.
Our content on the Intranet is not tagged and is unstructured. What is the best means to inject semantics?
Wednesday, 2 February 2011
This simplest way is to use Semaphore alongside the search engine – so that content is classified automatically and the meta-data is stored in the search engine index. This is easy to implement and will make a huge difference to the search and navigation experience and the findability of content. Alternatively Semaphore can be integrated with the underlying content management system – if there is one.
Does Smartlogic have out of the box healthcare ontologies or are they typically developed in collaboration with the client?
Wednesday, 2 February 2011
There are many healthcare ontologies available commercially and in the public domain. Smartlogic also has a number of its own which are available. However we always suggest understanding the types of users and the nature of the content before choosing which models and resources to re-use.
Page 1 of 5
 |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |  |