Powerset and Natural Language Search March 7, 2007

Search has become a critical part of our daily lives and the primary gateway to information and services of all kinds. Much of the success has been driven by the growth of the internet and behind-the-scenes improvements in search technology. However, the search user interface, based on keywords and advanced search operators, has changed very little since the earliest information retrieval systems.

In this talk, we discuss the concept of natural language search. Central to this is a new user experience, in which users express queries in natural language and the system responses respect the linguistic information in the query.

To realize this vision at broad scope and scale will require advances in a variety of technology areas, including natural language processing, information extraction, knowledge representation, and large-scale search indexing and retrieval systems.

In addition, it will require innovations in user interface. Issues include changing user behavior, education, supporting users in formulating effective queries, and managing expectations.

With a new search architecture centered on natural language, linguistic and lexical knowledge translate directly into improved capabilities and experiences for end users. This creates a challenge to encode enough knowledge to cross the threshold for a broad application. It also creates an opportunity to establish a framework for introducing new knowledge through automated and community-based approaches.

While the challenges are formidable, the opportunities are great. Natural language search has the potential to transform and improve all aspects of the search experience: precision, recall, power, usability, presentation and interaction.

We will conclude the talk with a discussion of Powerset, a startup company that is tackling these challenges in an attempt to bring natural language search to the world.

Official Website: http://www.ischool.berkeley.edu/about/events/sl03072007