_/_/_/_/ _/_/ _/_/_/_/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/_/_/_/ _/_/_/_/ _/_/_/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/ _/_/_/_/ _/ _/ _/_/_/_/ BAVARIAN ARCHIVE FOR SPEECH SIGNALS University of Munich, Institut of Phonetics Schellingstr. 3/II, 80799 Munich, Germany bas@phonetik.uni-muenchen.de COPYRIGHT Florian Schiel, University of Munich 1998. All rights reserved. This corpus and software may not be disseminated further - not even partly - without a written permission of the copyright holders. Additional Copyright Holders ---------------------------------------------------------------------- Munich AUtomatic Segmentation (MAUS) BAS Distribution Package MAUS 2003-08-19 / 2019-02-19 ---------------------------------------------------------------------- This package contains all necessary files to setup and run the 'Munich AUtomatic Segmentation' (MAUS) system for automatic segmentation of spoken German developed at the Bavarian Archive for Speech Signals (BAS) at the University of Munich, Germany. MAUS is capable to automatically segment a recorded speech signal into a set of phonemic classes. It uses a automatically derived, statistical rule set about pronunciation to calculate a vast number of hypothesis about the given utterance and then aligns the signal to this hypothesis space finding the phonemic transcript together with the segmentation that yield the highest combined probability of acoustical match and statistical prediction. For more details about the MAUS principle please refer to the publications given in hdl.handle.net/11858/00-1779-0000-0028-421B-4 MAUS is no longer distributed with full language support (only German), since the web-based MAUS system is now operational (https://clarin.phonetik.uni-muenchen.de/BASWebServices/) for several years - and also we encountered a number of illegal re-distributions of resources contained in the package by third parties. ---------------------------------------------------------------------- To use MAUS you need a PC running Linux (any variant will probably do it), the HTK software package from Cambridge, UK (publicly available at htk.eng.cam.ac.uk), standard tools csh, sox, awk, xpath, ffmpeg and this package. There are several technical constraints on the usage of MAUS that are listed together with a short user manual in the file USAGE. Hints for the setup of MAUS are given in the file INSTALL. ---------------------------------------------------------------------- Some explanations to important files in this package. Files not listed here are probably helper scripts called by the main programs: Subdir DOCU/ : development history of maus and other tools and other helpful texts DOCU/HISTORY : main version history for maus and helpers DOCU/USAGE : some useful tips for usage/adaptation of maus DOCU/CORPUSREQUIREMENTS : the requirements for a speech corpus to be used as MAUS trainings corpus (e.g. for a new language in MAUS) DOCU/WebMAUSInfo.txt : infos to the web application WebMAUS and webservices DOCU/SYMBOLSETS : rules about symbols sets of maus Subdir EXAMPLES/: data to test MAUS (see CHECK/) Main Tools (marked with '*' are needed for webservices) : ShowLattice : visualization tool for the SLF files (not needed by the main script but useful to see what MAUS is hypothesizing) maus(*) : the main script; read the help output produced when started without options carefully maus.corpus : wrapper to perform maus on a whole corpus of signals maus.iter : iterative maus (see HISTORY.ITERATIVE for details) maus.trn(*) : wrapper to exploit a chunk segmentationen coded in the TRN tier maus.web : wrapper to use webservice runMAUS instead of local installation this is useful if you cannot install MAUS but want to use the script txt2par : simple tool to create a primitive BPF input file from a table Chunker(*) : Nina Poerner's chunker service Anonymizer(*) : service to anonymize media files and annotations Subtitle(*) : Nina Poerner's sutitling service AudioEnhance(*) : signal process media files (for optimal use in services) Pipeline(*) : the BAS WebServices pipeline scripts AnnotConv(*) : universal converter service for annotation files EMUMagic(*) : pipeline scripts for the EMUMagic Service: produces an optimized emuDB from arbitrary input Asr(*) : Automatic Speech recognition scripts that call third-party services SpeakDiar(*) : the speaker diarization service testEnhance(*) : normalize *.txt files for optimal input to G2P Helper tools (usually not used alone) : HVite,HCopy,HHEd : selected HTK tools pre-compiled for Ubuntu word_var : helper tool to generate the statistical weighted search graph C+++ sources in word_var.src wav2trn : helper program to pre-segment a signal by leading and trailing silence to improve the MAUS result rec2mau.awk : helper to convert HTK output to CSV table in BPF style (MAU tier) kan2mlf.awk : helper to convert a BPF tier KAN in HTK style MLF kan2hmm.awk : helper to convert a BPF tier KAN in HTK style HMM names mau2Text2Grid : helper program to convert a MAU tier in a BPF file to a TextGrid file mausbpfDB2emuRDB: helper program to convert a collection of MAUS created BPFs into emu DB mausbpf2emuR : helpers for mausbpfDB2emuRDB: converts BPFs *.par to emu DB *_annot.json mausbpf2csv : helper to convert BPFs *.par to spread sheet table *.csv mausbpf2eaf : helper to convert BPFs *.par to ELAN compatible *.eaf mausbpf2exb : helper to convert BPFs *.par to EXMARaLDA compatible *.exb mausbpf2tei : helper to convert BPFs *.par to Iso TEI compatible XML (*.tei) par2TextGrid : simple converter from BPF to praat Textgrid par2emu : simple converter of BPF to legacy Emu *.hlb + *.phonetic files xpath : simple XML parser check_param_sets: checks the language parameter sets for syntac errors Libraries/parameter files: ipkclib/ : C library used by word_var PARAM/ : parameter set for classical MAUS PARAM.MAN/ : parameter set for phonological rules PARAM. : parameter set for language PRECONFIGWAV : HTK config file for front end processing HMM/ : tools to build new HMM sets for new languages based on existing HMMs (not a real training!) ---------------------------------------------------------------------- DISCLAIMER This package is public domain and provided 'as is' with no further support or warranty. It is a scientific, experimental system and must not be used for any other purposes than scientific investigations and education. If you intend to use this package or parts of it in a commercial enterprise, please contact the copyright holders. This package or parts of it may not be given to any third parties without the explicit consent of the copyright holders. ---------------------------------------------------------------------- Binaries word_var, wav2trn are pre-compiled for usage on Linux. To use on other OS see sources in *.src and the library ipkclib