[Next] [Up] [Previous]
Next: Adapting the Tagger
Up: Probabilistic Tagging in
Previous: The Algorithm
The tagger is implemented in a client-server model. The knowledge base
resides on the server, and the tagging engine establishes a network
connection to the server and retrieves the relevant data for tags and
tag sequence probabilities from it. This has several advantages:
- The resource data is in a well defined place and can be shared between
different invocations of the tagger
- The workload can be distributed across different machines
- The memory usage of the tagger decreases enormously
- The tagger can be used on small machines
- The server can easily be used by other applications as well
The server is implemented in C, while the client is written in Java. That
way it can run on all kinds of machines, provided there is a server available.
Multext-East