Making Personalized Retrieval a Reality in Knowledge Enterprises, Dr. Robert Hoffman
“Information needs vary from person to person,” says Dr. Thomas Hofmann, founder and chief scientist at RecomMind, an information management and retrieval company in Berkeley, CA. “Until now, tools have been built for the enterprise that focus on managing content but little has been done to understand the needs of the user. Two people in your enterprise with different needs type in the same query, why should they retrieve the same results? This illustrates the fundamental problem in information retrieval and management.” Until now, enterprises have spent significant resources on the content side of the equation: loading, parsing, understanding, retrieving, filtering, and delivering content but have ignored the most important component in the enterprise: the user.
Personalized RetrievalThe next evolution in enterprise information retrieval and management will come from tools that incorporate context into information retrieval. Context comes from understanding the user and the value of the information sources they seek. By understanding and interpreting information from personal profiles, RecomMind solves this problem by tailoring information results to the end user. “Everybody is used to seeing long lists of results but what people want are results that are relevant to them,” adds Dr. Jan Puzicha, RecomMind’s CTO. “Without understanding and incorporating user data into search, there is little context for the results.”
User-based RecommendationsWhen searching for information, most people in an enterprise pick up the phone and start dialing until they find the person who points them to the source. Picking up the phone works because the value of information is defined by users, not by how many key words are in a document. Incorporating this knowledge into search practices is the key to improving search results. RecomMind does this by first linking an individual with similar people who share a common interest, and then incorporating this data into information retrieval and management. For each search, the user also retrieves results that others have found useful.
Even though we know that information retrieval should focus on the user, the industry has not made strides to incorporate user context into information retrieval and management. Until now, the focus has been purely on content. The next generation of tools for enterprise users will be focused on better understanding those users and meeting their individual needs.
Dr. Hofmann is the Chief Executive Officer and Chief Scientist of Recommind. He is one of the leading experts in machine learning and has a strong background in artificial intelligence and computational statistics. Dr. Hofmann has pioneered statistical methods for information retrieval, text mining, and user modeling. He can be contacted at thomas@recommind.com and welcomes feedback and conversation