Add README

README.md

@@ -0,0 +1,20 @@
+# Czech Kaldi Profile
+
+A [Rhasspy](https://github.com/rhasspy/rhasspy) profile for Czech (`cs`).
+
+Includes:
+
+* A [Kaldi nnet3](https://kaldi-asr.org/doc/dnn3.html) speech to text model
+    * See files in `acoustic_model/`
+    * Recipe created with [ipa2kaldi](https://github.com/rhasspy/ipa2kaldi)
+    * Trained on:
+        * [Vystadial VOIP](http://www.openslr.org/6/) (18.2 hours)
+        * [Common Voice](https://commonvoice.mozilla.org) (26.6 hours)
+* An [IPA](https://en.wikipedia.org/wiki/International_Phonetic_Alphabet) pronunciation lexicon pulled from [Wiktionary](https://www.wiktionary.org/)
+    * See `base_dictionary.txt.gz`
+* A [phonetisaurus](https://github.com/AdolfVonKleist/Phonetisaurus) grapheme to phoneme model for predicting word pronunciations
+    * See `g2p.fst.gz`
+    * Trained on `base_dictionary.txt.gz`
+* A tri-gram [ARPA language model](https://cmusphinx.github.io/wiki/arpaformat/)
+    * See `base_language_model.txt.gz`
+    * Text from audio transcriptions and [Universal Dependencies](https://universaldependencies.org/)