Richard Caley

Heterogeneous relation graphs as a formalism for representating linguistic information

Abstract Heterogeneous relation graphs (HRGs) can be used to represent arbitrary linguistic information. Originally designed for use in speech synthesis, HRGs can be used for speech annotation purposes also. In the HRG formalism, atomic linguistic entities such as words, syllables and phones are represented by attribute value matrices known as linguistic items. Using attribute value matrices for items allows them to contain any type or amount of linguistic information. Items are organised into linguistic relations, which take the form of lists, trees or other structures. Items can belong to more than one relation, which allows words to appear in the word relation and, say, the syntax relation. The ability to have items in multiple relations, along with a function ability which can calculate certain values on the fly, eliminates much of the redundancy present in simpler systems. The HRG formalism is not tied to any particular linguistic theory, nor does it impose any preset ideas about what sort of format syntax, prosody or phonology information should have. This paper explains the HRG formalism in detail, and shows why we think this is superior to the types of “multi-level” formats normally used in speech synthesis and database annotation.

Evaluation of speech synthesis techniques in a comprehension task

Abstract Six types of speech synthesis were evaluated for comprehensibility: standard linear predictive coding analysis/ resynthesis; pitch synchronous analysis/resynthesis; pitch synchronous multi-pulse analysis/resynthesis; and three PSOLA (pitch synchronous overlap-and-add) techniques. The relative comprehensibility of the synthesis types was tested by using the synthesised speech to convey information that subjects needed in order to perform a diagram-based multiple-choice task.