How ontology based search can drive information management

Posted by Annabel Snow, Consultant on 21 May 2015

Tags: ,

I wrote in my last blog about why I think a search-driven environment is the way forward for businesses and the wider realm of information management. By moving to a folderless environment we are encouraging a search based system of storage and information management rather than static navigation paths that limit the visibility of connections between knowledge sets. Folderless structures encourage businesses and their employees to think more laterally about the information they’re storing and sharing. One way to support this evolution is through the use of an ontology.

What is an ontology?

An ontology is a domain model made up of distinct facets or classes (usually taxonomies) that have links or connections between them. Put simply, lists of words or ways of describing things that are then interrelated to disambiguate individual terms and give them more meaning.

An ontology based approach to information management not only models the key terms of an organisation but also their context and any interrelated concepts. The ontology can then be used to classify information in a content management system, alongside a search platform or website to assist with the retrieval of structured and unstructured data.

 

Ontology provides rich context and relationships

It is the recognition of the relationships between terms and concepts that gives an ontology its power to improve search functionality. When an ontology is used to auto-classify content, for example cooking recipes, it is doing far more than just text-matching. An ontology-driven system would recognise semantic annotations, picking up on keywords in the ingredients or the method or the two together, to recognise the conceptual nature of the recipe to then accurately auto-classify or tag the content as related to ‘baking’ or ‘Thai food’. An ontology is also smarter than just a controlled vocabulary. This is leveraged by the use of ontology management tools, such as Smartlogic’s Semaphore, which enable grammar, spelling constraints and classes to also be used. By applying this logic to the ontology you are able to ensure that items in the system are uniformly categorised.

Search and retrieval with ontology driven information management

 So what sort of value can you add to content retrieval using this rich contextual categorisation? A common example is type-ahead search suggestions which recognise synonyms and colloquialisms. This enables people to start searching using language that makes the most sense to them, or to see the links between what they are typing in the search box and what is being retrieved.

On shopping websites ontology-driven navigation can be a way of promoting additional related products or shopping categories. If you take a look at a site like ASOS, you can see ontology in action; when you search a generic word like ‘winter’ recommended search terms pop up, and it even has the number of preliminary results per search term, guiding the user on best search practice as they go.

 blog2

Earley & Associates write about the use of taxonomies for “searchandising”; where related topics and recommended products are guiding the user through the site’s navigation. By recognising the consumer’s need to shop online businesses can employ ontology-driven filtering to promote related wares to a customer while they are browsing their search results.  

On government sites, having a richly populated ontology can suggest topics that can help to clarify for people exactly what it is they are looking for. It can also offer multiple language tags, so when a search is conducted the system provides the relevant results no matter the search term language.

Content retrieval: tags, filters, and virtual collections

If search is driven by categorisation and taxonomy, the content that is returned to the user can be much more refined. Categorisation can target the relevant content and bundle it in response to the search query entered, rather than a pre-determined folder of content that may not all be useful or correctly filed in the first place.

Additionally, where users want to filter the results they have got (or create their own virtual collection) this can easily be done using pre-assigned categories or ‘tags’ which they can combine as they see fit. Sites like Flickr and Tumblr are doing this particularly well. Post-search navigation is also best driven by dynamic filters which are tailored depending on the initial search results.                                                         

Data intelligence

By breaking down the traditional paradigm of ‘everything must have one point of access’, we can start to use data, information and knowledge in innovative ways. Enterprise level analysis across both structured and unstructured data is something most organisations are looking to implement in order to manage their rapidly expanding information ecosystem.  The first step in being able to run analytics across your content or to allow it to be dynamically served to audiences, is to know what you’ve got.

Benefits of ontology-driven search

The benefits of ontology-driven search are wide reaching, but here are just a few highlights:

  • Allows access to content based on the user’s perspective not a rigid pre-determined one.
  • The ‘big bucket’ approach to information management means less effort on arrangement up front, and more of a focus on getting users the right information.  Audiences and users are increasingly sophisticated and want their information delivered quickly via multiple channels.
  • A rich pool of metadata can be applied to a document, enabling categorisation for multiple purposes at the same time.
  • Ontology-based search provides a smarter search than full text search does. For example,  a search that returns every document that mentions the word ‘tax’ compared to returning only documents are significantly about ‘tax management’ based on the term ‘tax’, enough related terms and their location within the document.
  • Allows organisations to leverage the things that are important to them (particular entities or document features) in information retrieval.

Further reading

-          How important is metadata to an information management strategy?

 

-          What is the business value of taxonomy?