LexicalNode Struct Reference

A lexical node is a single node of the lexical tree or Pronounciation Prefix Tree (PPT). It contains token, AM and word information. More...

List of all members.


Public Attributes

TokenType ** tokenSeq
 Normally of length NONSIL_NUMBER_OF_STATES (3).
LexicalNodenextTree
 If wordID != -1, this points to the start node of the next tree (for LVCSR back to the top).
LexicalNodenext
 Next node. Multiple 'next' nodes are linked together (in the next node) with 'parallel'.
LexicalNodeparallel
 Parallel node.
TokenType ** inputToken
int modelID
int contextPrev
 Previous modelID (left context).
int contextNext
 Next modelID (right context).
int contextKey
 For speed-up reasons pre-calculated: contextPrev * numberOfPhones + contextNext.
int wordID
 -1 if not an end node. Otherwise the ID of the word.
int depth
int tokenSeqLength
int compressedTreeIndex
bool nodeIsActive
 One or more states of this node are being used.
bool toBeDeletedFromList
 When this switch is TRUE. The node will be deleted from the 'active list'.

Detailed Description

A lexical node is a single node of the lexical tree or Pronounciation Prefix Tree (PPT). It contains token, AM and word information.

Because we use a static PPT (no copies are made), for each HMM state in a node, it is possible that there are multiple valid tokens linked to it.

The modelID points to the PhoneModel object (AM model). If wordID is not -1, this node is a word-end node.

Each node can have a following node (next) in the tree. It may also have a parallel node. Finaly it is possible that a node points to another lexical tree structure. This typically happens with alignment- or grammar-tasks.


Member Data Documentation