o iE@shddlZddlZddlZddlmZddlmZmZmZm Z ddl Z ddl Z ddl m Z ddlmZgdZdeefdd Zdeefd d Zdeefd d ZdeefddZdeefddZd)dededeefddZddZdeefddZdeefddZGdddZGdd d ZGd!d"d"ZGd#d$d$ZGd%d&d&Z Gd'd(d(eZ!dS)*N)Path)IterableListOptionalUnion)check_argument_types) AbsTokenizer)Ng2p_eng2p_en_no_space pyopenjtalkpyopenjtalk_kanapyopenjtalk_accentpyopenjtalk_accent_with_pausepyopenjtalk_prosody pypinyin_g2ppypinyin_g2p_phoneespeak_ng_arabicespeak_ng_germanespeak_ng_frenchespeak_ng_spanishespeak_ng_russianespeak_ng_greekespeak_ng_finnishespeak_ng_hungarianespeak_ng_dutchespeak_ng_english_us_vitsespeak_ng_hindig2pk g2pk_no_space korean_jasokorean_jaso_no_spaceg2p_isreturncCs2d|vr|dd}dd|dDS|dS)Nz z cSg|]}|ddqS) replace.0cr+R/home/ubuntu/.local/lib/python3.10/site-packages/espnet2/text/phoneme_tokenizer.py .z"split_by_space..r%)r'split)textr+r+r,split_by_space+s  r1cCs$ddl}|j|dd}|d}|S)NrFkanar%)r g2pr/)r0r phonesr+r+r,pyopenjtalk_g2p3s r6cCsjddl}ddl}g}||dD]!}|d|}t|dkr2||dd|dd|ddg7}q|S)Nr-\-(.*?)\+.*?\/A:([0-9\-]+).*?\/F:.*?_([0-9]+))rer run_frontendfindalllenr0r:r r5labelspr+r+r,pyopenjtalk_g2p_accent<s  &rAcCsddl}ddl}g}||dD]5}|dddddkr'|dg7}q|d|}t|dkrF||dd|dd|ddg7}q|S)Nrr7-+paur8r9)r:r r;r/r<r=r>r+r+r,!pyopenjtalk_g2p_accent_with_pauseIs   &rEcCsddl}|j|dd}t|S)NrTr2)r r4list)r0r kanasr+r+r,pyopenjtalk_g2p_kanaYsrHTr0drop_unvoiced_vowelscCsddl}||d}t|}g}t|D]}||}td|d}|r.|dvr.|}|dkrg|dks>||dks>J|dkrH|dn||dkrft d|} | dkr]|d n | dkrf|d q|d krq|d q||t d |} t d|} t d|} t d|} t d||d}| dkr|dkr|dvr|dq| dkr|| dkr| | kr|dq| dkr|dkr|dq|S)u1Extract phoneme + prosoody symbol sequence from input full-context labels. The algorithm is based on `Prosodic features control by symbols as input of sequence-to-sequence acoustic modeling for neural TTS`_ with some r9y9's tweaks. Args: text (str): Input text. drop_unvoiced_vowels (bool): whether to drop unvoiced vowels. Returns: List[str]: List of phoneme + prosody symbols. Examples: >>> from espnet2.text.phoneme_tokenizer import pyopenjtalk_g2p_prosody >>> pyopenjtalk_g2p_prosody("こんにちは。") ['^', 'k', 'o', '[', 'N', 'n', 'i', 'ch', 'i', 'w', 'a', '$'] .. _`Prosodic features control by symbols as input of sequence-to-sequence acoustic modeling for neural TTS`: https://doi.org/10.1587/transinf.2020EDP7104 rNr7z \-(.*?)\+AEIOUsil^z!(\d+)_$?rD_z/A:([0-9\-]+)\+z \+(\d+)\+z\+(\d+)/z /F:(\d+)_ aeiouAEIOUNcl#]r9[) r r;r=ranger:searchgrouplowerappend_numeric_feature_by_regex)r0rIr r?Nr5nlab_currp3e3a1a2a3f1a2_nextr+r+r,pyopenjtalk_g2p_prosody`sJ                rdcCs&t||}|dur dSt|dS)Nir7)r:rUintrV)regexsmatchr+r+r,rYs rYcCs,ddlm}m}dd|||jdD}|S)NrStylepinyincSsg|]}|dqS)rr+)r)phoner+r+r,r-z pypinyin_g2p..style)pypinyinrjrkTONE3r0rjrkr5r+r+r,rsrcsBddlm}m}ddlmmfdd|||jdD}|S)Nrri get_finals get_initialscsg|]G}|ddd|ddr%|ddddd|ddn|ddr5|dddn|dfD]}t|dkr:|s:|q:qS)rT)strictN)isdigitisalnumr=)r)rlr@rsr+r,r-s&  z&pypinyin_g2p_phone..rn)rprjrkpypinyin.style._utilsrtrurqrrr+rsr,rs   rc@s2eZdZdZd defddZdeefddZd S) G2p_enzOn behalf of g2p_en.G2p. g2p_en.G2p isn't pickalable and it can't be copied to the other processes via multiprocessing module. As a workaround, g2p_en.G2p is instantiated upon calling this class. Fno_spacecCs||_d|_dSN)r|r4)selfr|r+r+r,__init__ zG2p_en.__init__r"cCs:|jdur t|_||}|jrttdd|}|S)NcS|dkSNr%r+rgr+r+r,z!G2p_en.__call__..)r4r G2pr|rFfilter)r~r0r5r+r+r,__call__s   zG2p_en.__call__N)F) __name__ __module__ __qualname____doc__boolrrstrrr+r+r+r,r{sr{c@s.eZdZdZ d ddZdeefddZdS) G2pkzOn behalf of g2pk.G2p. g2pk.G2p isn't pickalable and it can't be copied to the other processes via multiprocessing module. As a workaround, g2pk.G2p is instantiated upon calling this class. FcCs"||_||_||_||_d|_dSr}) descritive group_vowelsto_sylr|r4)r~rrrr|r+r+r,rs  z G2pk.__init__r"cCsT|jdurddl}||_t|j||j|j|jd}|jr(ttdd|}|S)Nr) descriptiverrcSrrr+rr+r+r,r rzG2pk.__call__..) r4rrrFrrrr|r)r~r0rr5r+r+r,rs  z G2pk.__call__N)FFFF)rrrrrrrrr+r+r+r,rs   rc@seZdZdZdZdddeddDZddded d DZdd ded d DZ eee eeZ dddZ de de e fddZddZde e fddZdS)Jasoz !'(),-.:;?r%cCg|]}t|qSr+chrr)rOr+r+r,r-rmzJaso.iicCrr+rrr+r+r,r-rmiaivcCrr+rrr+r+r,r-rmiiFcCs||_||_dSr} space_symbolr|)r~rr|r+r+r,rrz Jaso.__init__liner"cCstt|}|Sr})rFjamohangul_to_jamo)r~rjasosr+r+r, _text_to_jasoszJaso._text_to_jasocsfdd|D}|S)Ncsg|] }|jvr|qSr+) VALID_CHARS)r)tokenr~r+r,r-#sz6Jaso._remove_non_korean_characters..r+)r~tokens new_tokensr+rr,_remove_non_korean_characters"sz"Jaso._remove_non_korean_characterscsPdd|D}|}jrttdd|}|Sfdd|D}|S)NcSsg|]}|qSr+r+r)xr+r+r,r-'sz!Jaso.__call__..cSrrr+rr+r+r,r+rzJaso.__call__..csg|] }|dkr |njqS)r%)rrrr+r,r--s)rrr|rFr)r~r0 graphemesr+rr,r&s z Jaso.__call__N)r%F)rrrPUNCSPACEjoinrT JAMO_LEADS JAMO_VOWELS JAMO_TAILSrrrrrrrr+r+r+r,rs rc @sTeZdZdZ     ddeedeedeedefd d Zd eefd d Z dS) Phonemizera5Phonemizer module for various languages. This is wrapper module of https://github.com/bootphon/phonemizer. You can define various g2p modules by specifying options for phonemizer. See available options: https://github.com/bootphon/phonemizer/blob/master/phonemizer/phonemize.py#L32 Nr%Fword_separatorsyllable_separatorphone_separatorsplit_by_single_tokenc Ksjddlm}ddlm} | |||d|_td} | tj||di|d| i|_ ||_ ||_ dS)Nr)BACKENDS) Separator)wordsyllablerl phonemizerloggerr+) phonemizer.backendrphonemizer.separatorr separatorlogging getLoggersetLevelERRORrstripr) r~backendrrrrrphonemizer_kwargsrrrr+r+r,r<s      zPhonemizer.__init__r"cCs:|jj|g|j|jddd}|js|Sdd|DS)Nr7)rrnjobsrcSr#)r%r$r&r(r+r+r,r-fr.z'Phonemizer.__call__..)r phonemizerrrr/)r~r0rr+r+r,rZszPhonemizer.__call__)NNr%FF) rrrrrrrrrrr+r+r+r,r1s"  rc @sFeZdZdZ    ddedededefd d Zd eefd d ZdS)IsG2pzMinimal wrapper for https://github.com/grammatek/ice-g2p The g2p module uses a Bi-LSTM model along with a pronunciation dictionary to generate phonemization Unfortunately does not support multi-thread phonemization as of yet standardT,dialect syllabifyword_sepuse_dictcCs8||_||_||_ddlm}||jdd|dd|_dS)Nr) Transcriber.T)r syllab_symbol stress_labelr lang_detect)rrrice_g2p.transcriberr transcriber)r~rrrrrr+r+r,rqs  zIsG2p.__init__r"cCs|j|Sr})r transcriber/)r~r0r+r+r,rszIsG2p.__call__N)rTrT) rrrrrrrrrr+r+r+r,ris   rc @sxeZdZ   ddedefdeeeeefdedefdd Zd d Z d ed e efddZ deed efddZ dS)PhonemeTokenizerNr$Fg2p_typenon_linguistic_symbolsrremove_non_linguistic_symbolsc Cs^tsJ|durt|_n@|dkrtdd|_n4|dkr&tdd|_n(|dkr/t|_n|dkr8t|_n|dkrAt|_n |d krJt|_n|d krRt|_n|d krZt |_n|d krbt |_n|d krpt ddddd|_n|dkr~t ddddd|_n|dkrt ddddd|_n|dkrt ddddd|_n|dkrt ddddd|_n|dkrt ddddd|_n|dkrt ddddd|_n|dkrt ddddd|_n||dkrt d dddd|_nn|d!krt d"dddd|_n`|d#krt dd|_nU|d$krt dd|_nI|d%krt d&ddddd'd(dd)|_n6|d*kr%t |dd+|_n)|d,kr1t dd|_n|d-kr;t|_n|d.krGtd/d0|_ntd1|||_||_|dur^t|_nLt|ttfrt|}z$|jd2d3d4}td5d6|D|_Wdn 1swYWntyt|d7t|_Ynwt||_||_dS)8Nr F)r|r Tr r r rrrrrarespeak)languager with_stresspreserve_punctuationrderzfr-frresrrurelrfirhurnlrhirrrzen-usr%r)rrrrrrrrrrr r! g2p_is_northnorth)rzNot supported: g2p_type=rzutf-8)encodingcss|]}|VqdSr})rstrip)r)rr+r+r, sz,PhonemeTokenizer.__init__..z doesn't exist.)rr1r4r{r6rHrArErdrrrrrrNotImplementedErrorrrsetr isinstancerropenFileNotFoundErrorwarningswarnr)r~rrrrfr+r+r,rs                            zPhonemeTokenizer.__init__cCs&|jjd|jd|jd|jdS)Nz (g2p_type="z", space_symbol="z", non_linguistic_symbols="z")) __class__rrrrrr+r+r,__repr__s zPhonemeTokenizer.__repr__rr"cCsg}t|dkr@|jD]}||r*|js ||dt||t|d}nq |d}|||dd}t|dksd|}||}|S)Nrr7r)r=r startswithrrXrr4)r~rrwtr+r+r, text2tokens!s       zPhonemeTokenizer.text2tokensrcCs d|S)Nr)r)r~rr+r+r, tokens2text3s zPhonemeTokenizer.tokens2text)Nr$F) rrrrrrrrrrrrrr+r+r+r,rs"   r)T)"rr:rpathlibrtypingrrrrr r typeguardrespnet2.text.abs_tokenizerr g2p_choicesrr1r6rArErHrrdrYrrr{rrrrrr+r+r+r,s2     O&!8