ESSV Konferenz Elektronische Sprachsignalverarbeitung

Title: Constructing Petri net transducers with PNTooL

Authors: Markus Huber, Robert Lorenz, Daniel Straßner

Abstract:

Petri net transducers (PNTs) were developed for the translation of utterances into meanings within a hierarchical cognitive dynamic speech signal processing system [9, 10]. PNTs are a natural generalisation of weighted finite state transducers (FSTs) for the translation of partial languages consisting of partial words (with a partial order on their symbols) instead of (linear) words (having a total order on their symbols). PNT(ooL) is a python library for the modular construction of PNTs through composition operations. Constructed PNTs can be exported in an XML-format which is based on the standard PNML format developed for basic Petri net variants [16]. Moreover, PNTs can be visualised and pictures can be exported in all standard formats. PNT(ooL) serves as a basis for the implementation and evaluation of algorithms for analysis, simulation and optimisation of PNTs. PNT(ooL) is implemented within the framework SNAKES aiming at the quick prototyping of new Petri net classes [12, 13]. Its basic functionality was developed in the bachelor thesis [15].


Year: 2014
In session: Sprachdialog
Pages: 110 to 117