|
Tuesday, March 1, 2005 |
|
|
|
|
Natural Language Information Processing: Words and Tasks, Statistics and Models |
|
|
I will review the development of some key ideas in natural language information processing (NLIP), showing how statistical approaches were established for text retrieval and have spread to many other tasks like summarising. Some simple statistical techniques have proved very effective, and are now widely applied in operational systems as well as research. These techniques should be motivated by underlying models for the NLIP tasks. I will consider how different models apply to a single task, retrieval, and how a single model applies to very different tasks, illustrating the impact of the single-model approach on current research. |
|