Alive?

[danijel3]

@AdolfVonKleist

Are you still interested in developing this project?

I've finally forced myself and made a web-based Phonetisaurus model training tool: https://github.com/danijel3/PhonetisaurusWebApp

During this, I've also managed to set up a Docker setup to compile and run Phonetisaurus and managed to get SRILM to work (cause MITLM isn't supported much). I'd gladly contribute these changes to this project, but I'm not sure if anyone is interested?

[AdolfVonKleist]

Wow that is really cool!

I try to address issues when they pop up, and merge contributions. But unfortunately I do not have a lot of free time to work on it these days.

There have also been quite a few advances in this area in recent years, and I think there are other interesting, more advanced projects; end-to-end approaches in ASR are also bringing into question whether there is really a continuing need for it in that context.

I'm not against merging it - actually I always wanted to build something like this for it - but are you sure you want to?

[danijel3]

So, I'm not gonna comment on the end-to-end approaches in ASR (it's not all rosy), but there are other uses for G2P outside of pure ASR. The reason I was forced to develop it is because many researchers (linguists, phoneticians, sociologists) use these tools to analyze speech data. If nothing else, people like to perform speech alignment on the phonetic level and this requires G2P (especially if you want a custom phonetic alphabet).

I'm not sure if Your users are interested in any of those features, but I reckon I'd get slightly more exposure through your repository than if I did everything on my own, so if you don't mind, I'll plan out a few PRs in the upcoming days/weeks. Is that okay?

[AdolfVonKleist]

I'm sure there would be some interest. I'm happy to merge it if you plan out the PRs as you say.

[danijel3]

BTW, do you have a dockerhub account? Do you mind if we make a "phonetisaurus" organization on dockerhub to store the compiled images of the project?

[AdolfVonKleist]

I have an associated dockerhub account, and some rather old containers there:

• https://hub.docker.com/r/bkbroiler/phonetisaurus

there's no organization, but we could add one - or the other way around.

[danijel3]

An org should make things a bit easier, since a couple of people can manage the same set of repositories. I assume that "bkbroiler" is your account, so I added it to the org:

https://hub.docker.com/orgs/phonetisaurus

I'll create a Dockerfile and configure a repo now.

[AdolfVonKleist]

thanks that works fine.

[danijel3]

Ok, I've added the PR #65 with the Docker setup and uploaded one image to the repo manually.

Next, you will need to configure the repository on the Dockerhub to turn on automatic builds. All you need to do is enter here: https://hub.docker.com/repository/docker/phonetisaurus/phonetisaurus/builds

Link to Github and then select this Phonetisaurus repository (unfortunately I can't do this step), leaving everything on default, eg:
https://imgur.com/lzYtCjl

Once you click "Save and build" the images will be created automatically whenever you make a Push to this repository. Pretty neat feature, IMO.

If you're happy with this, so far, you can close this issue and I'll work on some more PRs later.

[AdolfVonKleist]

I'll try to setup the build configuration at the weekend and let you know if I run into any trouble.

[AdolfVonKleist]

Ok it is done. Sorry it took a bit longer than planned. It looks like the setup worked, it is trying to build a container.

[sventech]

Thank you for this -- tools like this, as @danijel3 said, are quite necessary both for linguistic research and for building the latest neural text to speech + ASR models for new languages. Would it be easy to enable Docker builds for linux/aarch64 for Raspberry Pi and Apple M1 users?