iLink is a social networking component for building a dynamic topic model for each user that is used by the recommendation engine to suggest related people, or information artifacts. iLink’s FAQtory allows a community to ask and answer questions and to build a FAQ repository.
Social networks and online communities have become ubiquitous forums for people to share knowledge and expertise. However, it is often tedious and time-consuming to find useful answers to questions. Suppose you are a hiker and a member of an online group of hiking enthusiasts and you would like to know how much skill it takes to climb Yosemite’s Half Dome. You might broadcast the question to the entire online hiking group in order to find someone may be able to answer the question. Or, you might pointcast the question to particular individuals who you think might know the answer or know someone who does.
In broadcast systems, it is easy to overwhelm the limited set of users who have expertise, causing them to miss relevant queries. In pointcast systems, the burden is on the inquirer to know who might have the answer. Moreover, these systems do not get systematically smarter over time about where community expertise resides, nor do they help validate the answers. .
iLink is designed to help find your answer without overburdening you or the social network. iLink will first attempt to answer the question by searching for the best matches from a knowledge repository of previous answers captured from interactions in social networks. If the answer is not found, iLink, which keeps track of the expertise of the members of the community, can direct the question to those who are knowledgeable about the subject.
iLink unobtrusively learns profiles of user expertise from a wide variety of behaviors in social networks. It captures new knowledge from interactions between users in the form of a Frequently Asked Questions (FAQ) repository [also known as a Question and Answer (Q&A) repository]. Specifically, iLink addresses the problems of the broadcast and pointcast systems in the following ways:
- iLink supports a narrowcasting model, with messages sent to a restricted group of highly likely respondents who need not be known to the inquirers.
- Answers are verified within the message routing system by the members with the most relevant expertise.
- Answers are stored and matched against new incoming questions, reducing the number of questions that the members have to answer.
When a user posts a question, iLink consults its FAQ repository to retrieve highly rated answers. It also searches for answers to the question on the Web. For the example, iLink might return answers about hiking in Yosemite or mountain climbing in California if it does not have a closer match on climbing Half Dome.
If the user is not satisfied, iLink looks for expert members, as determined from their previous answers, ratings for those answers, and their profiles. These experts are notified of the question and may choose to answer. If no one has an answer, iLink enables routing the question to other people through email. For the example, iLink might route the question to people who have expertise in climbing and knowledge of Yosemite.
When a member answers a question, the user can rate the response. A favorable rating results in the answer being added to the FAQ repository and the responder’s expertise level for associated topics being increased.
iLink uses topic discovery and classification algorithms to create the expertise model and to determine the best-matching answers for a question.
iLink provides a set of web service APIs, based on Representational State Transfer (REST), that support programmatic user actions. Through these APIs, other applications can find iLink recommendations and can create iLink objects for discussion threads, contributions, and users.
- Java 1.6 and above
- Apache Tomcat
- The iLink REST APIs provide programmatic user actions. More detailed iLink operations such as querying a list of existing discussion threads, contributions and members, forums and FAQs are only supported by the iLink web interface.