WikiSpeech – enabling open source text-to-speech for Wikipedia

Abstract

We present WikiSpeech, an ambitious joint project aiming to (1) make open source text-to-speech available through Wikimedia Foundation's server architecture; (2) utilize the large and active Wikipedia user base to achieve continuously improving text-to-speech; (3) improve existing and develop new crowdsourcing methods for text-to-speech; and (4) develop new and adapt current evaluation methods so that they are well suited for the particular use case of reading Wikipedia articles out loud while at the same time capable of harnessing the huge user base made available by Wikipedia. At its inauguration, the project is backed by The Swedish Post and Telecom Authority and headed by Wikimedia Sverige, STTS and KTH, but in the long run, the project aims at broad multinational involvement. The vision of the project is freely available text-to-speech for all Wikipedia languages (currently 293). In this paper, we present the project itself and its first steps: requirements, initial architecture, and initial steps to include crowdsourcing and evaluation. Index Terms: speech synthesis, accessibility, crowdsourcing, multilingual 2 https://meta.wikimedia.org/wiki/Wiki_Project_Med and

Controlling Prominence Realisation in Parametric DNN-Based Speech Synthesis

“…This work is part of the Wikispeech project [15]. The objective of Wikispeech is to deliver freely available, Wikipediaoptimised text-to-speech through Wikimedia Foundation's server architecture.…”

PROMIS: a statistical-parametric speech synthesis system with prominence control via a prominence network

“…Finally, in this work, we also aim for societal impact. PROMIS was developed as part of the Wikispeech project [18] in which the general objective is to deliver freely available, Wikipedia-optimised text-to-speech through Wikimedia Foundation's server architecture. A TTS system with improved expressiveness and prominence control would find good use in synthesising Wikipedia: the input are long, domainhomogenous texts in which modeling information beyond the current, short utterance is especially desirable [19].…”