Over the past decades, information retrieval has been widely studied due to its importance in people’s everyday life. However, how to train students with assignments using large- scale real-world datasets is a significant challenge. In this thesis, we address this challenge by proposing CLAIRE, a novel cloud-based virtual lab which is beneficial to Information Retrieval (IR) educators, learners and researchers. Following the common routine of learning or conducting research on IR models, CLAIRE provides support in three key phases. In the phase of implementation, CLAIRE provides an interactive way to create new IR models from scratch as well as from existing parameterized models. It then offers the environment for evaluating these IR models over real-world datasets without having to move around the large- size datasets. With the tightly connected implementation and evaluation phases, CLAIRE further enables the analysis of performances of different IR models, including parameter sensitivity analysis and query-wise comparison. Models can also be instantly turned into a search engine application. Leveraging the scaling power of CLaDS[1], CLAIRE is further capable of supporting tasks that involve large datasets conducted by a large number of users at a relatively low cost.
【 预 览 】
附件列表
Files
Size
Format
View
CLAIRE: A cloud-based lab for information retrieval
PDF
download
878987797
028-85220240
