# machine-learning
- t
The Wiggles and High School D×D
08/21/2021, 2:40 PMIn other events, my Google Colab is not powerful enough to run all of this and I'm curious to see if anyone here would be willing to give it a shot - t
The Wiggles and High School D×D
08/21/2021, 2:40 PMI can store it all but I can't actually execute most of the commands without limiting myself - u
user
08/21/2021, 3:11 PMHmm - u
user
08/21/2021, 3:12 PMIf you could link me to the files I might be able to (I got a local runtime set up so I can run colab 24/7 without having to worry about disconnecting after 6-12 hours) - t
The Wiggles and High School D×D
08/21/2021, 3:24 PMI can send you the code too - t
The Wiggles and High School D×D
08/21/2021, 3:24 PMThat I wrote - t
The Wiggles and High School D×D
08/21/2021, 3:29 PMI removed my personal touches from here. Basically I had to replace characters in the dictionary to accompany the translation. - j
jens
08/21/2021, 3:30 PMthis sounds really good - t
The Wiggles and High School D×D
08/21/2021, 3:30 PM - t
The Wiggles and High School D×D
08/21/2021, 3:30 PMAnd this is the repository I used - t
The Wiggles and High School D×D
08/21/2021, 3:31 PMYou need to edit the init.py file in the arpabetandipaconvertor folder if you want to be able to translate dictionaries that aren't English - t
The Wiggles and High School D×D
08/21/2021, 3:32 PM - t
The Wiggles and High School D×D
08/21/2021, 3:33 PMI got my dictionary from here. I recommend you get it formatted the way GitHub displays a raw txt file because the available versions to download aren't formatted quite the same way. - t
The Wiggles and High School D×D
08/21/2021, 3:33 PMFor me, I just copied and pasted into my own txt file - c
Couch
08/21/2021, 3:34 PMthere are dictionaries in different languages now i feel inclined to create a tool that converts sentences in a language to arpa to feed as synthesis input to tt2 models - t
The Wiggles and High School D×D
08/21/2021, 3:55 PMI don't think you can make a universal model - t
The Wiggles and High School D×D
08/21/2021, 3:55 PMToo many differences in how IPA is utilised for other languages - h
Holos
08/21/2021, 4:03 PMThe main edge case in ARPA is finding words that are new (like new slang) and aren’t in the dictionary. I think if you encounter a word you haven’t seen before you could get an API that gets you a phoneme not present in the dictionary - h
Holos
08/21/2021, 4:03 PMI know there’s Oxford English Dictionary but even that doesn’t have every word. You need to establish a phoneme rule set of some kind which is what 15 does - h
Holos
08/21/2021, 4:04 PMIf he sees a word that he hasn’t seen he “sounds” it out - t
The Wiggles and High School D×D
08/21/2021, 4:10 PMThat's why there's a converter between IPA and ARPA that goes both ways - t
The Wiggles and High School D×D
08/21/2021, 4:10 PMWhat language do you want to do? - c
Couch
08/21/2021, 4:12 PMin the cmu website theres actually a tool that generates arpa by rule http://www.speech.cs.cmu.edu/tools/lextool.htmlCopy codeBear in mind that this is a dictionary. If your word is not in it (or was misspelled) nothing will be returned. This applies to items such as numbers. This tool will try to come up with pronunciations for words not in the dictionary. - t
The Wiggles and High School D×D
08/21/2021, 4:13 PMAre you using this for English specifically? - c
Couch
08/21/2021, 4:13 PMyes - u
user
08/21/2021, 4:14 PM@The Wiggles and High School D×D I'll check it out - h
Holos
08/21/2021, 5:18 PMJust English but I’ve seen the case of words not in the dictionary and I’m curious as to how to fix it I’ll look at the link that @Couch gave - h
Holos
08/21/2021, 5:19 PM15 used the phonological rules from the LibriTTS dataset which I find interesting - t
The Wiggles and High School D×D
08/21/2021, 5:37 PMI don't know if I can help with that issue in particular - j
JacobLenstar
08/21/2021, 6:27 PMdamn strong stonks with Liane Cartman right off the bat