From 05e8a671c1caf54a0ec8f8e0db4d786e87d24055 Mon Sep 17 00:00:00 2001
From: yemaozi88 <428968@gmail.com>
Date: Sun, 13 Jan 2019 23:36:02 +0100
Subject: [PATCH] forced alignment by novoapi is performed.

---
 .vs/acoustic_model/v15/.suo         | Bin 84480 -> 71168 bytes
 acoustic_model/check_novoapi.py     |  69 +++++++++++++++++++++++++++-
 acoustic_model/novoapi_functions.py |  21 ++++++++-
 3 files changed, 87 insertions(+), 3 deletions(-)

diff --git a/.vs/acoustic_model/v15/.suo b/.vs/acoustic_model/v15/.suo
index 9fe1f2136ee18d47f8e6c82b824e73835098274b..f5fd7886def8bd17ad70bdd8dbb5b128a60a9480 100644
GIT binary patch
delta 4289
zcmeH}YfzNe8OPt}WkD8LSp@;*w!i`_thmcXR<g+ABJmzp!Aom`sCY#{T?h~_t0tML
z*JYp7H?^rvH<`8*O`^P0O`9fJk))YK4YnEMv^E(#`H;-mjyl>VHcJ0z*(FW1ZStYh
zFFiZIJ<qwH_qm*NJWoY>j%mF1adDC)MS#0QLqo`da>@x;syy1e^z`=*9(|t{qSn*E
zJ#r5<@GW_WJbWu6v!M*t7!yo2kmrF|p#7)%WXLl@<p+?*gF6^80r^-E1GWM&$r$7l
zdtH;$w}mQ1<!H!ekOUIJ^j=r;j&Tqs0Vk*cqGLFXN4XNr2UVfA>QG*aJP+7`1{8wu
zIFZjmxef?bi1ZNT*}w%#z$}33u|xF`8o=V=ilG&~uEHHFQCJ1mhAM+o;BOtu>%j)l
z_O-TllsADMum!k5ClF)1L%EQ5Kz<AqfGJ=z5cOiG3ovV&<mORw^JwY$7wuk{Y%<!v
zVzMT_H7b8cdT5vtdDv72$XP&`N=&IkE=(n4k%Xy4E==_Z5atV-3bG!gf=rMLGC&j%
ziZubTy?o@^piz=`NJS9Rz&&O~ozSiYSV1u;2V!AiX^{&{+rU&X4TLQ%+JvPYKv=pK
zge@)F!s$Wri8IkS7vz9&!vgpeE#p~{Io9Q<ei(%Hya=-Rh-T#B?jm1}@*2?6>q^jW
zgwP5)dR?j7M<IwFUHnJwhQu5i9zPCw637L-EXK0K7-|S=2(k%G1O=cFn874a1SSIu
zu!3Sx0!l#{m;%ay4NL_U-~nI<m0%jE0uE3OrUNIzt7{<40JY#jFcZuIv%wrN7l_Tx
zNB$6~1Fle=s2`!X>8LNxE!Puw#%-4u68{z1Jd||v*r_Xvwuu7<$rm&8|LQ1<EE4Kq
zt0O_xJAO@85%3*A900;~x_}6Lq25=KXN6S09kMtC9#`)qL}>g5nM)n|;1tE*dJ0Ls
zJHC_T(XdhK9%VrhIaEhRfchr5!<Bv0+(ez+QMH%-t8J>y_%?l6<ufJH9{9a=FZQz!
z><0(HLEr_wzy}yS1-=gsfx9P~eoVdX(f$IBf3EDe$i6%yg(~g-Qca)PEz1i??W{dV
ze5RyQ^%Q%^XFsE*Xp9~O_`nGh-?Z0h#5uv9GBdj?_x0JFq3%<=i1O6WQ$~?*nUkHx
zm|>ZhRQ>X6<Xi95N+Oz<hNe;7?#E?H;Llr!qQz+`l7Epb)KRP$v@~Ume$vc@*FI@@
zd+Fb2HXhHZS}oz5d0*Z6ya?Nsh0~HgY+Mz0v+=v%c~zYIBZotGq$K?THftTl9+#)D
zXF!f#jAj=IZr<b7sGn_(rr~woGSFz%u{J>d7TLc<_HUL?P9qN~0XsPqPbE3}cHxs=
z=kDFJBS~C1!Ee6#M6#x~ytqb<fATu<!~3<W_Q2mFeMRn|@{0C^CVwS~4Cd~>?UiYf
z%Aa&;Pe1ccWlkODkQOK4GLpaCGh8j0)a@&Fk+0Y}qI<5MQ^V$lM4q!ImfsspXV2^a
z9~?;MgPweDNl)b8tV!afzc1(gYf}A(qkQ#&WXnj}w$*AgTej9VwQOnL&2Oy9K+kNw
zyLpSdvAMN#-R8E|jm@oYu1+7%js{Phut7W!R|zk_kU8#<L{eK(dv^=2ajL1X;P+Um
zKaD0}g-Ii;K-ji*N~yWL1S@f%el>g2J>S4*Ry&AaS>xnmm-KviZ(h{a(h{?^3{}<s
zXJrn|nHpDYDL31OxBn!zZ)eNdH2&evTpiXc34fReEmq%74<A35$p<`H)Wy}ARerzR
zx5q&V!CgWRQ!L^B+t9z7U%L>?vFPmO>f9>DQ$>Nya90}^EgoLg-rCX9A}%_ikt=7i
z4po!aI?wrLS?&id=|`>}xs<#rOD`J3_WL*s_ZqI>-MD&9IA#jUB<ZJOiGLp<9_l;!
zsfKT?NP+LROWn-Bs$j>_K;KvqR|YG)6Y4cf#3m4ZQHgNwvyytdF-pEqOs(HX{$sMQ
z^|N~;>tDyh{hT~j#yXe9*J0hEm!IS8;~CEdj#G>{KjFW4zj4#xl=xN6tD8Dn+>P#y
z?e~O4gzt=r54|}W9S2Cuje{T1EWUnvlHVuum5Zrz!Az8F+;&0FZ(p#n<6>-*8eCBP
znF8nV4~<AQZ*kMby)eI4t@vFo>BJi*oNQ{!C6(e9T_4wY{TFYDXHPtDI`3ugfRp_n
zJ;L5VBp(^j@#`N=ACc3~J9Tr=^NXO}@&oUvn?BLd*x{GAm^kF?HC$bq!WlQ#vg4|Q
zv(l5eZom=T?VUjr|LF7$-=Bx5hz+F%e!1VE2Cg~ikeveTx?=U6i9b|{XL|gZ@vIrh
z*CpUPL~v>G;C90WOYi{v1Pb1QxBAD(-LaC?kM?a)?i<KA>QI*1nEj06a8M&>=49y?
zqi@*L`mLYN<+>aTTLaDb)>QUf?TP7Xvd&*19$ymYUbaUsc~86;QCN4`#-3{?Q~gTU
z)Q&D^`=Yt+r7eXUEQKvI>ZX@Xe)Pj%nx?;0`+h%tJpP^F#Mdt$_ur7&*E^sN9JWXD
z*c)!BATa`m183ejK?jtqLh9z*8Hu<PQn18Jd|_xm&p%;6mk2&L^dcQrs&i?JQXrF=
zb`8sU$)u;Tm|%mlB!X(`DP?C2ZBh(6O5^h_QJ4yfkUB9{89#m16tPlj<E$TSQrr8y
zqqu3ESUwh`GZlRdEfP0`J-m?$V<Il7M2z-HA)ED_x~8RKQg+4CF(o^emUH%<w9rT$
zM)m}k>B!V`<epM2E#Ac84dJ>|F6yX2aT#gW$o_c43){aMqpIW86%paMh!xMQ;kY;l
zcP$IL{=0{TBa&8a9Q+9d{aZXK#M1)L^1!>0#ohFu{ChK<Or)O^`Gmq2vwqM%CKtn`
zZ|U@DJm!T73NEEQdOB!8B|C|Xv}i;VPn7FVl_t|!nRosr|2`2C9d?+n$Awxq5KlPC
z%MldD;h;Eu+VN#TP`@J>4b}Zd90h6ry=d6k-!9_wq=9e;-_+9Ph>cR4GA4$`!=qLp
zex#UKaeDCXYn3DMp;(O+3=sb9^C|kT2ZLI7{kW_gF;Y@UYkb?1p=2AWOL>+^bE+Vp
zS_#gz_Gim2ME|=`QCt&ghVs0XmML}PNF$CdWzY;yt}s)YVkji9Qg5Km$`=K&xj$(C
sQ)oXQ+`DpjG6M4PQ0UE9ypt)4)(Ur9$P2SYl;;bFQ)ZY@5&qVH07$}MVgLXD

delta 7179
zcmeHLdr*|u760zX@{)(kONl6mqN2QZmwh5C3(JDIz%DNz1#wxHmkW#R3lPGZVltyt
zVUKa6O&hyuZQ9Ai%92=3s#|nAHdUjUO#et?VrHBqnv4}kjcKa%T-c4o#))ZTY9D9k
zH}~9o&pr2^d(Phb-D7`|W8dXJm>rYAFpMAK&iME^7888rK}{SRK9_m%pzv33kv!VG
z61v%LG9PZR<D?UA_$Bns$5tFOK|<ukQ|f76ekPW|UOsh7|Ewwe&%AsDmYH6jgXI!L
zC}K9^HqH`{Wg;ROv6x}{n1#qpNBkTy2k{sp1F--hK%^k(>Xu{a^71RN3`bCFrC4Tr
z+oQcDm8X>Bt<S?!fe7-}X`d2#9)tTvD+O37M5v|+(Dt$+%}i1SD}(3vRby8TVm-ow
z*nrR?^av{IqXw)uA~xRBZoqmgq7Bh;PoD|v)D}uMZ=VRu$M4F)C&4ie))4NA3A}kf
zcLpw+aHr3WbXNlq)Qxl}eQu;Pg<+k#gVHy-k;?eo7=q(y{T??m(<a;d#<|Gc>qa`@
z;kwc1MKtk|cVkU7bpS;SVVD^R%0u-u9#n+9IMja95o%_;8L);v&UP_y6AL5O!ixO~
zjbE^fQbTKzu~2^^Ii?l<>l#C*XmpQ*g1~W>59X&UI8`_W?D%DJa5S!Tgih$b9p%Sy
z<lZ9T?$2`r$?$A5VPC}+&e`VbBrtYlG1#|;!`TnRp~kor>|J(eStQAE=fr!6E0#@S
zn0HXq7)IvDSl&dNE@Jr^M$jnLj)%oD29i9_a^E|gocogUtx<CEvDZiWmsGIz_-6ll
z>zeL*Q2(CSS7y}|3LC^JhQ{LjGFw4sky_H&k#F22Gc~oyD!Wsv3ffy&HVdt5bhi3#
zgHEq*>{iJw>(tHV1vaTzR@9i%Y~vSOZ6cmgRiLvNn<V@$X%oNDmdh_PNh=$asy17>
zuB0uuseWyhYPG40ztXl!w5C{8P^h)ZEX3kUVjY5{;BC0*?TEo4O_cu-GS3WY;==YI
zvlsDW#Iuknm>G<QGtVOy9n_eJBRQH3pxojgVapE0PJ{!2p)f&!{4T`Ph}{S$>WG`r
z@f_9`xf2CHB(5bKN0yoGWZc%QZN$Z69oc!1r&#8;XWNNO$QjPbBf*5K<KY<UM7LSq
zPrQC}EI&Ts<m=1ToIu)-fd%Fj1H<#ogqXoHe>(^xxo&RcYb1<DC_VJ72pjMtqoNYg
zXS9_bQGJMhR4~q8@a=a4EisW{shOR$<23owg4ukLxK=2ZNo3M=u|y{k==kDReMh4{
z-DELBggqISwgux#2aWM?>{2w?+dRbVXj71GZ&N(5c1fUAn9A!>YF+89V@4roIj)uN
z1ej-3z{%KDxIUvam@LEA6<_wuR0q*bqc;W>4Y`g?F9OM_i6ohD=KUbZDoKynxGAA>
zbM@d~ZVVX@&Ppm8VlV_?pybTLoxO%X58e`2KOebJ#oqq#DLRy@q*RQ28;4>3#sPb&
z{U3b<X8C}_>fsnndcD_>>9DFx1k2lIL5|HHnvM$S*-p=3IQMe=j3xA9Gp4t98PEu#
z$l~E|v(TktU29lJ=RU5WHr>&)_Ky|N_u0`##;^~5)2aTt|EWWP;ggNwqZlM^uIegr
zDOvj86~L^kb0c<Hn|H_7?EFI)+4`r)LN~anu|YU#s#Y(MxF2tMg4`W%C&N8C1m3qw
z;jEDd8Ev6($~p~{z02V0ZGTq<>nO|%CackawbY1-$SfQnt}511bc=+*=SAZr-Tirf
zAnRK9wX*%}q*Nf?3?%n>DPf1zC)l7hSkoXToNTY>xNnsO5l7jHukQ$|Z8KVp+V*Dn
zz?eYS8NwG&gbVKe7Vqv0;JGb{&}$3~;Pa)ad_FoNm#uDdY_Y==_8k0<k|(SfUf)T`
zE6`P2<j+*n+36|G;(z2_tWf$o0k~c}lOq!fr0(e(#z=U*zCmj-m}||g?E;C=9lqfj
z8Sc*^;k`ZlyNkLT+ZSG*4i}BHanT`U5abJ1xawF}U4W~e1G_ejfL+gZpJ>k};Lx9-
zj6clPZ*XsTbDzJfgB>2c#KP|!v4nxM7h(c=_$7n{M;-IrIUjgfmzj0!cR}6$_~$+Q
z3CM-95Pu>W=51<sx4$_U=x|)R@8CMWdxuS<7gZFh4`e`dWdP=m1j9Ca)V!X}(h1TD
z(=x>kvNWMkl$j>uX|-trS%X$sFVMqtUnD~7m3E97F6=pw3_Ffo1JC6>y0?8no~g*o
zmkVX`G^tV{NfQc0qBOZckeMcuV2?r|l8N$py*=edgT>rzY@55%tZy}KexO0{!I3z4
z$tZzee6$Gm99smQn|klPCW1XT2SS?68*8nm+KpO6J?#4^6_P(}nP&bUv2?G`iEF>v
zi6RM4+#AN^rj=Jr9!!~1iLQ<>txps4(cwZo<kMsVkt9u~)nzuw1UyMyoxWFrzCH;%
zFU62zh;XaO5*WKQ7(p-9NlNt4YcLgxuMGOulKg;ciGwZnI8uRX<rs&N?2>XH3&N4S
z8T9coNqmpo%PB@-=aKE+FqFahk6wqNnnVnT>AuMy`bJ!fpbrNhrBdc07@y^v0Ly&D
zBE&x!|3)+u%Mf<*88kT{(X-jT<8lXqr6bvW^yS7>kHyrEy?l(mJLo;7_#XbUJ$YHx
zl_ouWd?ecU&n3}#|3nitXb#N6Ht}5!o+(N0<5B8quF5Sfv+-6HG#cedCFPY(`IU{x
z<<hELlP6zG%LZwNtcf!6H6`U`yp{3}lTatt(1x-$O(h?t^Ez8>R_R(zC2cIy<u|R<
z+2nGCsTivgg<Mx}-Bdsa6za<>%W|u@iY9e?RdJ`5Us9%2OZ0WnKbjq+Xs*m#wJNs-
zPL4`*q^%l-rnsp|*|JKMQCugp<{Ql`6&)61zDa4c=66?F*VU=Dt5p>`jm*-S*IuP+
z7aCGF@=^@>>ReHV4YqqkU^%}1(fMxfv+z{<^1R~u`gQV5fuUQ{-BPOQEY{_At>Txf
zD>F+o%39jA&4PxFEhXL343SDAsp!%aW#E2_C90B!OpSgu-ClXVL0Pe}+@fjCn@DZY
zp!3mWc+zhqA4Pf@#c6lhl~gkEbefc+FJvz*<NiO0|1*L(<y$=dVD6y?SbjM#tl8&h
zhVOYDc3j?d?;RdCeafZY3@?3>9eQuWSB7YZ3&M)1Qv2j#SK(grTMmI6uWsv8Ka3~m
zga32X-}PU|U3^7Az<${7<VBKLh#jp9&|?~_8b5Sj9mi8O@sR$z+T#>_8{y@qcx`DA
zv*rz&SYn(9qeucYj7A6OYfXA{y5o5A(=U)%x97DSvY#dFY(uqHXSA5i&AQrFW4+!`
zJ;iTuA2&t3vasR%xux#6Z~p3UZ7CEBL-AX@2|S0R;X>(LCyzrmz+dB|@dY1?+FpQj
zr{iGxc@^9|rG)-7JcvADeo*0G!`ag#<O!$TpO|3b7}uYkntLa!BNYq%7-kxfb3Y+>
zCpmgnNvt@f2qPldPsbSHfY^d^!EpNYp?_Wz)KtczK{VO81LRk`AS;O)7(HR&mUlP#
z9^m-wrD;+J<9(uZyRW_@AY>h`NCJ{A^Kd1zAmRK0%m~mmS`xV6-!DW>G#mW_P)+;Z
z%LdN54%8?BZ6%%M`=0gf(~;C+?)9G*5y-jvQ{updnPB$SB+NYfmc3QNbqmK%v2bk+
zA90@+ezWB8oo!Ue>MgsSgB-Hkuaz-6$Nk96Ln9obBF+I0+545Yiv%j}h~duL3a2NK
z<iPqdrIQULin~5{D#D2pg*u!9E=h9orjtRxeT)L+hqpqIn7WPLs4nWZLua|<3?WTu
zPq}kRG?_*Ey|Gf_JR3y<rYH@FB1(LW8qde=4<rPqRnU2=cL$JX{(TF=Ph97mXp#tx
zp0$sBHqvp_1NNF6=bPaqi~4Q`sE_f$Q<F0%#D6kh6pUGVn#NjyAITl+4r4>JLB9ar
z@XQ~swiCyb9?(50hMS(n@V+N8p+}hM%Y#fLU(v1EZ5D%p=FgB|%NL1`rv|w{IM`Ns
oYA`C`KuCUGX~R%3T&Ro*HDgw&XYw_+pC%@f{%cackMJh{9q>r7NB{r;

diff --git a/acoustic_model/check_novoapi.py b/acoustic_model/check_novoapi.py
index 178706b..1814a6b 100644
--- a/acoustic_model/check_novoapi.py
+++ b/acoustic_model/check_novoapi.py
@@ -139,6 +139,71 @@ df_per_word = pd.DataFrame(index=[], columns=df_samples.keys())
 
 for word in word_list:
 	df_samples_ = df_samples[df_samples['word']==word]
-	df_samples_ = df_samples_[df_samples_['frequency']>1]
+	df_samples_ = df_samples_[df_samples_['frequency']>2]
 	df_per_word = df_per_word.append(df_samples_, ignore_index=True)
-df_per_word.to_excel(os.path.join(default.stimmen_dir, 'pronunciation_variants_novo70.xlsx'), encoding="utf-8")
\ No newline at end of file
+#df_per_word.to_excel(os.path.join(default.stimmen_dir, 'pronunciation_variants_novo70.xlsx'), encoding="utf-8")
+
+
+## ===== forced alignment =====
+if forced_alignment:
+	Results = pd.DataFrame(index=[],
+		columns=['filename', 'ipa', 'word', 'result_ipa', 'result_novo70', 'llh'])
+	for word in word_list:
+	#for word in ['Oor']:
+		# pronunciation variants top 3
+		df_per_word_ = df_per_word[df_per_word['word']==word]
+		df_per_word_ = df_per_word_.sort_values('frequency', ascending=False)
+		if len(df_per_word_) < 3: # pauw, rozen
+			pronunciation_ipa = list(df_per_word_['ipa'])
+		elif word=='Reuzenrad':
+			pronunciation_ipa = [
+				df_per_word_.iloc[0]['ipa'], 
+				df_per_word_.iloc[1]['ipa'], 
+				df_per_word_.iloc[2]['ipa'],
+				df_per_word_.iloc[3]['ipa']]
+		else:
+			# oog, oor, reus, roeiboot
+			pronunciation_ipa = [
+				df_per_word_.iloc[0]['ipa'], 
+				df_per_word_.iloc[1]['ipa'], 
+				df_per_word_.iloc[2]['ipa']]
+		#print("{0}: {1}".format(word, pronunciation_ipa))
+
+		# samples for the word
+		df_ = df[df['word']==word]
+
+		# samples in which all pronunciations are written in novo70.
+		samples = df_.query("ipa in @pronunciation_ipa")
+
+		results = pd.DataFrame(index=[],
+			columns=['filename', 'ipa', 'word', 'result_ipa', 'result_novo70', 'llh'])
+	
+		#j = 0
+		for i in range(0, len(samples)):
+			sample = samples.iloc[i]
+			wav_file = os.path.join(default.stimmen_wav_dir, sample['filename'])
+			if os.path.exists(wav_file):
+				#j += 1
+				#print('{0} - {1}'.format(word, i))
+				pronunciation_ipa_ = [ipa.replace(':', 'ː') for ipa in pronunciation_ipa]
+				result = novoapi_functions.forced_alignment(wav_file, word, pronunciation_ipa_)
+				result_ipa, result_novo70, llh = novoapi_functions.result2pronunciation(result, word)
+				result_ = pd.Series([
+					sample['filename'],
+					sample['ipa'],
+					sample['word'],
+					' '.join(result_ipa),
+					' '.join(result_novo70),
+					llh
+					], index=results.columns)
+				results = results.append(result_, ignore_index = True)
+				print('{0}/{1}: answer {2} - prediction {3}'.format( 
+			 i+1, len(samples), result_['ipa'], result_['result_ipa']))
+
+		if len(results) > 0:
+			Results = Results.append(results, ignore_index = True)
+		Results.to_excel(os.path.join(default.stimmen_dir, 'Results.xlsx'), encoding="utf-8")
+else:
+	Results_xlsx = pd.ExcelFile(os.path.join(default.stimmen_dir, 'Results.xlsx'), encoding="utf-8")
+	R = pd.read_excel(Results_xlsx, 'Sheet1')
+
diff --git a/acoustic_model/novoapi_functions.py b/acoustic_model/novoapi_functions.py
index bbf44ff..c8640c0 100644
--- a/acoustic_model/novoapi_functions.py
+++ b/acoustic_model/novoapi_functions.py
@@ -36,10 +36,29 @@ def load_phonset():
 				phoneset_novo70.append(novo70)
 				translation_key_ipa2novo70[ipa] = novo70
 				translation_key_novo702ipa[novo70] = ipa
+
+	# As per Nederlandse phoneset_aki.xlsx recieved from David
+	# [ɔː] oh / ohr # from ipa->novo70, only oh is used.
+	# [ɪː] ih / ihr # from ipa->novo70, only ih is used.
+	# [iː] iy
+	# [œː] uh
+	# [ɛː] eh
+	# [w] wv in IPA written as ʋ. 
+	extra_ipa	 = ['ɔː', 'ɪː', 'iː', 'œː', 'ɛː', 'ʋ']
+	extra_novo70 = ['oh', 'ih', 'iy', 'uh', 'eh', 'wv'] 
+	for ipa, novo70 in zip(extra_ipa, extra_novo70):
+		phoneset_ipa.append(ipa)
+		phoneset_novo70.append(novo70)
+		translation_key_ipa2novo70[ipa] = novo70
+		translation_key_novo702ipa[novo70] = ipa
+
+	translation_key_novo702ipa['ohr'] = 'ɔː'
+	translation_key_novo702ipa['ihr'] = 'ɪː'
+
 	phoneset_ipa    = np.unique(phoneset_ipa)
 	phoneset_novo70 = np.unique(phoneset_novo70)
 
-	return 
+	return phoneset_ipa, phoneset_novo70, translation_key_ipa2novo70, translation_key_novo702ipa
 
 
 def multi_character_tokenize(line, multi_character_tokens):