Research Area

Integrating DB & IR for End Users

Structured queries are a powerful tool to precisely describe a user's informational need and retrieve the intended information from a database. However, manual creation of a structured query is a labour-intensive and error-prone task. This task requires exact knowledge of the database schema as well as proficiency in a query language, which are typically beyond the expertise of end users. Keyword search and natural language based QA on the other hand can be performed efficiently by novice users, as it requires neither apriori schema knowledge nor query construction skills. However, keyword search and QA lacks expressiveness to precisely describe a user's informational need, and may return irrelevant or incomplete results. The aim of our research is to take advantage of both, i.e., expressiveness of structured queries and usability of keyword search. Please check out our demo at http://iqp.l3s.uni-hannover.de .

 

Diversified Information Seeking

People's information needs are very diverse. To Web search engines, users' queries are classified into informational queries, navigational queries and transactional queries. This categorization is in our opinions neither complete nor precise. In real world, an information seeker sometimes needs an overview, sometimes she needs very focused results, and sometimes she wants compare people's opinions to avoid bias, etc. Such diverse needs cannot be satisfied by today's search engines, which rely on a uniform procedure to answer users' queries. Our research aims to provide different information seeking tools to meet users' different needs. To have an impression, check out http://www.whatdoestheinternetthink.net. In fact, there are many many different possibilites.

 

High Performance DBMS

In our High Performance DBMS Lab, we are working on various technologies to improve the performance of OLTP and OLAP. The technologies include Main Memory DB, Parallel DB, MapReduce, Flash Memory. We aim to create very efficient data processors that can handle the ever-growing data volumes in real world. A part of our research results have been directly adopted by Kingbase, one of the largest local DBMS vendors in China.