On Compression of Parse Trees

We consider methods for compressing parse trees, especially techniques based on statistical modeling. We regard a sequence of productions corresponding to a suffix of the path from the root of a tree to a node Ü as the context of a node Ü. The contexts are augmented with branching information of the nodes. By applying the text compression algorithm PPM on such contexts we achieve good compression results. We compare experimentally the PPM approach with other methods.

• Type: Conference
• Year: 2001
• Where: SPIRE
• Authors: Jorma Tarhio
• Keywords: Compression Algorithm Ppm, Information Management, Node Ü, Parse Trees, SPIRE 2001