Notices

Ok, I'll add a language selector. Saera will be able to respond in Spanish but you'll have to talk to her in English, because I haven't separated the keywords out into their own module yet.

Everyone: New version of Saera with much better accuracy!!
Gory details: It's using a much smaller model, compiled from a list of sentences I hand-typed that I imagine one might say. It will recognize other sentences than the ones I typed, but only ones with the same words (because the dictionary is generated from the corpus). It would be quite a help if other people could create other sentences to expand the model (see corpus.txt for example/what I have so far).

Also: could someone point me to a good packaging tutorial? I tried the dh_make route but got hopelessly entangled and lost.

While this may be great for general use, it may be worth keeping both modes depending on what you're doing. Having a limited vocabulary based on input commands is great (better if it parses those and auto-generates the library at startup if it's changed); but if someone wants to use it for entry for say a text message, it may be desirable to have more words and less accuracy.

I'm working on it. Pocketsphinx allows for context switching of models as I understand it, so it should be possible; i.e. you say "Text John Smith let's go to Pizza Hut", and the limited model recognizes "Text John Smith" and then switches to the extended model to parse the text itself.

python saera.py
saera.py:186: SyntaxWarning: name 'able_to_listen' is used prior to global declaration
global able_to_listen
INFO: cmd_ln.c(512): Parsing command line:
gst-pocketsphinx \
-samprate 8000 \
-cmn prior \
-nfft 256 \
-fwdflat no \
-bestpath no \
-maxhmmpf 1000 \
-maxwpf 10

Current configuration:
[NAME] [DEFLT] [VALUE]
-agc none none
-agcthresh 2.0 2.000000e+00
-alpha 0.97 9.700000e-01
-ascale 20.0 2.000000e+01
-backtrace no no
-beam 1e-48 1.000000e-48
-bestpath yes no
-bestpathlw 9.5 9.500000e+00
-bghist no no
-ceplen 13 13
-cmn current prior
-cmninit 8.0 8.0
-compallsen no no
-debug 0
-dict
-dictcase no no
-dither no no
-doublebw no no
-ds 1 1
-fdict
-feat 1s_c_d_dd 1s_c_d_dd
-featparams
-fillprob 1e-8 1.000000e-08
-frate 100 100
-fsg
-fsgusealtpron yes yes
-fsgusefiller yes yes
-fwdflat yes no
-fwdflatbeam 1e-64 1.000000e-64
-fwdflatefwid 4 4
-fwdflatlw 8.5 8.500000e+00
-fwdflatsfwin 25 25
-fwdflatwbeam 7e-29 7.000000e-29
-fwdtree yes yes
-hmm
-input_endian little little
-jsgf
-kdmaxbbi -1 -1
-kdmaxdepth 0 0
-kdtree
-latsize 5000 5000
-lda
-ldadim 0 0
-lextreedump 0 0
-lifter 0 0
-lm
-lmctl
-lmname default default
-logbase 1.0001 1.000100e+00
-logfn
-logspec no no
-lowerf 133.33334 1.333333e+02
-lpbeam 1e-40 1.000000e-40
-lponlybeam 7e-29 7.000000e-29
-lw 6.5 6.500000e+00
-maxhmmpf -1 1000
-maxnewoov 20 20
-maxwpf -1 10
-mdef
-mean
-mfclogdir
-mixw
-mixwfloor 0.0000001 1.000000e-07
-mllr
-mmap yes yes
-ncep 13 13
-nfft 512 256
-nfilt 40 40
-nwpen 1.0 1.000000e+00
-pbeam 1e-48 1.000000e-48
-pip 1.0 1.000000e+00
-pl_beam 1e-10 1.000000e-10
-pl_pbeam 1e-5 1.000000e-05
-pl_window 0 0
-rawlogdir
-remove_dc no no
-round_filters yes yes
-samprate 16000 8.000000e+03
-seed -1 -1
-sendump
-senmgau
-silprob 0.005 5.000000e-03
-smoothspec no no
-svspec
-tmat
-tmatfloor 0.0001 1.000000e-04
-topn 4 4
-topn_beam 0 0
-toprule
-transform legacy legacy
-unit_area yes yes
-upperf 6855.4976 6.855498e+03
-usewdphones no no
-uw 1.0 1.000000e+00
-var
-varfloor 0.0001 1.000000e-04
-varnorm no no
-verbose no no
-warp_params
-warp_type inverse_linear inverse_linear
-wbeam 7e-29 7.000000e-29
-wip 0.65 6.500000e-01
-wlen 0.025625 2.562500e-02

INFO: feat.c(848): Initializing feature stream to type: '1s_c_d_dd', ceplen=13, CMN='prior', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean[0]= 12.00, mean[1..12]= 0.0
ERROR: "acmod.c", line 84: Must specify -mdef or -hmm
Segmentation fault


what can i do?

Somone else had this problem earlier. You are missing the packages pocketsphinx-lm-en-hub4wsj and pocketsphinx-hmm-en-hub4wsj.

That sounds like a good idea.

Would it be possible for the limited model to be compiled on the user's device? I ask because I think the limited model should be based on the commands that are standard+locally configured, and the contact names in the user's contacts.

I believe so. I have to do something like that for contacts at least anyway, although I should include app names as well.

Just tell me when it is ready and I will translate the keywords.

Ok.
In the GTK version first, the QT version doesn't support showing markers & routing result at the moment. There are also actually no versions, just different GUIs with a shared backend. So you can either get the package from Extras-Devel and ignore the Qt dependencies or just clone the github repo and run modRana with:
Well, once I'm done with the CLI stuff anyway.

Ok, good idea. I'll keep it in mind while implementing it.

BTW, modRana has a built-in POI database - the points are stored in categories and each has a name and a description. You mentioned parsing contacts - maybe the POIs would be also usable for something similar ?

Do POIs usually have names that are regular words? (I haven't actually used POIs in ModRana before.) If so, that should be pretty easy to deal with, but otherwise it will have to prompt for pronunciation.