next up previous
Next: Relations Up: THE ARCHITECTURE OF THE Previous: Multi-level data structures


This section presents the archiecture used in the Festival speech synthesis system, version 1.3 [3]. The most significant difference between the Festival architecture and those using MLDS is that Festival does not constrain linguistic items to be in linear lists: any graph structure is allowed. Additionally, items can be contained in more than one structure, which leads to more efficient representations. We have also generalised the content of an item such that information of arbitrary complexity can be easily stored.

Figure 1: An example representation of an utterance structure. This example shows the word relation and the syntax relation. The syntax relation (shown on top) is a tree with links connecting the nodes, shown as black circles. The word relation (shown on the bottom) is a list. The items contain the actual linguistic information and are shown in the rounded boxes. The dotted lines show the connections between the nodes and items.

Alan W Black