Subscribe

Make the connection with social networks

Social networks are all around us. People interact with each other, information is shared and recommended from person to person, and communities of interest grow and ebb as priorities and projects change.
By Garth Wittles, District manager for Verity South Africa
Johannesburg, 13 Aug 2004

There is a demand for vendors to provide solutions that take advantage of social networks to boost company productivity by providing employees with useful recommendations, identifying experts, and personalising the information discovery process - because efficiency is about getting the right information at the right time to the right person.

Social networks are formed when people interact with information and each other. Employees engage in water-cooler discussions, recommend best business practices, and form communities around common interests. In the process, important relationships between people and information are formed, but they are not necessarily used to help improve corporate productivity. Typically, some of these relationships, such as information taxonomies and organisation charts, are well known. However, other attributes of these social networks, such as communities of interest and subject experts, may be more difficult to recognise and locate.

This is complicated by the fact that these relationships are constantly changing and evolving. Social network technology requires a recommendation engine that analyses the knowledge entities within organisations and how they relate to one another. These knowledge entities include products, documents, users and queries, which should then be represented as tensors in a hyper-dimensional tensor space. By representing all entities in a common tensor space, arbitrary correlations across disparate entities can be discovered. And both implicit and explicit relationships can be mined and used to model the social networks within organisations.

Social network and portal infrastructure

Social network technology requires a recommendation engine that analyses the knowledge entities within organisations and how they relate to one another.

Garth Wittles, district manager, Verity

Products in this space should take advantage of social network technology to automatically provide users with personalised discovery features. Through automatic analysis of the way users create, modify, locate and retrieve information, a model of the entire community can be built and used to power the following capabilities:

* Adaptive ranking: Traditionally, the scoring and ranking of documents has depended only on the content in documents matching search queries. The simplest application of social networks is to re-rank the results of a search based not only on document content, but also on the historical behaviour of users who have issued similar searches and selected particular documents to view. In simple terms, the more relevant other users have found a particular document, the higher in results lists it appears.

* Document/product recommendations: Through analysis of social networks, accurate document and product recommendations can be made. The system can analyse not only which documents each user has accessed and what searches he has issued, but also which department he belongs to and who his colleagues are. It can then generate accurate recommendations for documents related to the user`s complete context and profile. Context-based personalisation is more powerful than simple, rule-based personalisation, since contexts are dynamic, and human administration is not required.

* Expert location: Expert location takes the recommendation of documents to the next logical step. Once the system adopts a uniform view of entities within the organisation (documents, queries, users, user contexts, categories, an so on), the recommendation process identifies the closest entities in that space. For instance, users can ask the engine for documents (one of the types of entities the tensor space contains) that are close to specific queries or users (other entities in the tensor space). Similarly, they can ask for other users that are close to particular documents in the tensor space. Once these expert users are located, the user can be presented with their e-mail addresses, locations and other information.

* Community: Just as experts can be located using social networks, communities of interest can also be identified and defined. A recommendation engine can locate users in the tensor space that have similar interests based on the queries they have issued and the documents they have retrieved in the past. By identifying, or recommending, these people to each other, a community is created. As employees are recommended to one or more communities, their reuse of best practices tends to increase, and duplication of effort is reduced as they begin sharing knowledge. It is possible to further connect these communities using chat groups, e-mail lists and bulletin boards. As the media that the community uses to communicate changes, a recommendation engine can track and respond to the different ways users are connecting with each other.

* Auto-initialisation and profile management: Just as you can start categorising documents immediately by deploying a comprehensive industry-specific taxonomy, you can jump-start user profiles by importing information into your recommendation engine from various sources of information, including authored documents, public e-mail forums in the organisation, customer relationship management systems and Web server logs.

* Deploying social network technology: Social networks already exist in every organisation. It`s up to the organisation to recognise and utilise them. Technology can give corporations the ability to easily harness the power of these social networks. The key requirements for taking advantage of social networks include providing the recommendation engine with appropriate information, enhancing the tensor model through integration with other models, and presenting the analysed results to users. The recommendation engine must also allow for explicit feedback, so users can manually indicate their relationship to another entity to provide the system with the best possible data.

Valuable user data can be imported into the engine in order to jump-start the user experience. In addition, third-party applications and custom interfaces can make use of transaction logging functions, files and APIs to feed in additional information. This level of extensibility makes it possible to track virtually every electronic interaction in an organisation in order to model the social networks as accurately as possible.

By building a comprehensive and accurate model of social networks in organisations, and integrating these with other systems, this type of technology can build a complete model of how employees interact with various entities to provide them with the best adaptive ranking, document recommendations, expert location and community identification. The end result is that their information discovery process is dynamically personalised to the changing ways in which they use information from day-to-day, and project-to-project.

Verity is a co-sponsor of ITWeb`s software industry portal, which updates readers on new releases, product reviews, and breaking news and trends in SA and overseas markets. Columnists report back on strategic business issues, market moves and vendor announcements, while the Industry Solutions feature carries relevant case studies.

Share