Abstract:
|
This document first introduces general notions about ILP (inductive logic programming), including a basic vocabulary of ILP, a typology of ILP systems and a description of the main techniques in ILP. It discusses the application of one particular ILP system, FOIL, to the problem of chunking (segmenting) time expressions occurring in natural language text. We employ a propositional knowledge representation that considers features of the individual tokens plus the tokens in a context window of limited size. We trained three rule-based classifiers with FOIL to learn to recognize time expressions using IOB tags, using annotated data from the ACE 2005 corpus. The evaluation methodology and the results of our experiments are reported in this document. |