Module tree_kernel 

Tree kernels for structured data similarity

This module provides kernel functions for tree-structured data, particularly useful for measuring similarity between hierarchical expressions like TLExpr.

Tree Representations

Trees are represented as labeled nodes with children, where each node has:

• A label (string identifier)
• A list of child nodes

Kernel Types

• SubtreeKernel: Counts common subtrees between two trees
• SubsetTreeKernel: Counts common tree fragments (allows gaps)
• PartialTreeKernel: Partial subtree matching with decay factors

References

• Collins & Duffy (2001): “Convolution Kernels for Natural Language”
• Moschitti (2006): “Making Tree Kernels Practical for Natural Language Learning”

Structs

PartialTreeKernel

Partial tree kernel - allows partial subtree matching

PartialTreeKernelConfig

Configuration for partial tree kernel

SubsetTreeKernel

Subset tree kernel - allows gaps in tree fragments

SubsetTreeKernelConfig

Configuration for subset tree kernel

SubtreeKernel

Subtree kernel - counts common subtrees

SubtreeKernelConfig

Configuration for subtree kernel

TreeNode

A tree node with label and children