NWU-MuST/za_lex

Lexical pronunciation resources for TTS in South African languages ZA_LEX: lexical resources for South African languages ===================================================== This repository contains lexical pronunciation resources and modules for use in text-to-speech (TTS) systems. Specifically, it was originally set up to track work on updating and enhancing existing resources for the NTTS project funded by the _Department of Arts and Culture_ (DAC) of the Government of South Africa. The copyright and licence information for scripts in `./scripts/` can be found in `./COPYRIGHT` and `./LICENCE-APACHE`/`./LICENCE-MIT`. This repository also contains data from various sources under different licences in the `./data/*` directories. Copyright and licence information for data and third-party components is contained in each individual sub-directory or source file. For more information contact: _Daniel van Niekerk_ (nwu.ac.za). ## Software dependencies - OpenFST 1.5.0 or higher with Python bindings. - PyICU. ## Description of contents The top level directory structure is summarised as follows: ``` . |-- data | |-- afr | |-- eng | |-- sot | |-- tsn | |-- xho | `-- zul |-- examples |-- scripts |-- COPYRIGHT |-- LICENCE-APACHE |-- LICENCE-MIT `-- README.md ``` * The `data` directory contains core language resources organised by language, each associated with its own LICENCE and README. * The `examples` directory contains some example outputs when running scripts as described below. * The `scripts` directory contains implementations and UNIX tools for grapheme-to-phoneme (G2P) conversion, syllabification, word decompounding and morphological analysis (some usage examples are given below). ## Usage examples #### Decompounding The decompounder `decomp_simple.py` requires a word list and can be run for example on the Afrikaans data as follows: ```bash cut -d " " -f 1 data/afr/pronundict.txt | scripts/decomp_simple.py examples/afr.words5.txt > examples/afr.decomp.txt ``` #### Morphological analysis …