71–79 of 79 posts
Mozilla supports censorship and banning for wrong-think. So I don't care what happens to them.
Can you please source at least when you state such things on HN ? this is not reddit
Earlier quoted context omitted.
Plucking a single comment entirely out of context and presenting it as damning evidence against an entire company's competency feels kind of over the top.
The context is that Firefox is not supporting essential OOP features (like private fields) in Javascript, which Chrome has been supporting since April 2019. In the past, Firefox had a lot of experimental features and was often the first browser to support new stuff. This has changed, Firefox is now behind in quite a few areas and if you look at their priorities, management doesn't really seem to care.
I'd dispute private fields being "essential". People have been writing a lot of OOP code in JavaScript without them for a long time now.
I don't understand what Mozilla is. They re-upped the Google deal to the tune of $400-450million/year for the next 3 years, so their revenue is secure. That's great. I can also understand that it makes sense to cut some side-projects that are probably a distraction (and this one may very well fit that category). But if you cut funding to things like MDN, Servo/Rust and move away from Firefox as a core focus (the thin…
Wait, move away from firefox as a core focus ?
Earlier quoted context omitted.
"But I think most would also agree that if all corporate lawyers, bank lobbyists, or marketing gurus were to similarly vanish in a puff of smoke, the world would be at least a little bit more bearable." I remember just yesterday there was a great article on HN that discussed bullshit jobs and counted corporate lawyers towards the "goons". The rationale was that if nobody has them, nobody else needs them. https://www.…
Who exists to enforce the rules? Contracts? Etc. I feel like this sounds great on the surface, but is quite flawed.
Earlier quoted context omitted.
It's supposed to work on sentence long audio (4 - 5 seconds), they suggest chunking your audio first: https://discourse.mozilla.org/t/longer-audio-files-with-deep...
Also there's essentially two parts to this, the neural net is used for speech-to-characters, and then a language model is used to convert the character stream to words. I found that the language model they supplied was trained data that did not contain the words I needed, and got significantly improved results when making my own language model using the kenlm[1] tools. [1]: https://kheafield.com/code/kenlm/
E.g. some very active projects are:
* Kaldi (https://github.com/kaldi-asr/kaldi/) obviously, probably the most famous one, and most mature one. For standard hybrid NN-HMM models and also all their more recent lattice-free MMI (LF-MMI) models / training procedure. This is also heavily used in industry (not just research).
* ESPnet (https://github.com/espnet/espnet), for all kind of end-to-end models, like CTC, attention-based encoder-decoder (including Transformer), and transducer models.
* Espresso (https://github.com/freewym/espresso).
* Google Lingvo (https://github.com/tensorflow/lingvo). This is the open source release of Googles internal ASR system, and used by Google in production (their internal version of it, which is not too much different).
* NVIDIA OpenSeq2Seq (https://github.com/NVIDIA/OpenSeq2Seq).
* Facebook Fairseq (https://github.com/pytorch/fairseq). Attention-based encoder-decoder models mostly.
* Facebook wav2letter (https://github.com/facebookresearch/wav2letter). ASG model/training.
* Vosk (https://github.com/alphacep/vosk-api). Offline lightweight speech recognition API with support for 10 languages.
And there are much more.
Earlier quoted context omitted.
Also there's essentially two parts to this, the neural net is used for speech-to-characters, and then a language model is used to convert the character stream to words. I found that the language model they supplied was trained data that did not contain the words I needed, and got significantly improved results when making my own language model using the kenlm[1] tools. [1]: https://kheafield.com/code/kenlm/
Would it be possible to substitute this for GPT2/BERT? Or is that a different type of language model? Can the pre-trained language model be fine-tuned? I’m using DeepSpeech to transcribe long-form lecture audio, and have just assumed there would be a massive improvement once they noise-harden the models with 1.0.
Earlier quoted context omitted.
Would it be possible to substitute this for GPT2/BERT? Or is that a different type of language model? Can the pre-trained language model be fine-tuned? I’m using DeepSpeech to transcribe long-form lecture audio, and have just assumed there would be a massive improvement once they noise-harden the models with 1.0.
GPT2 is not a good language model but there are things like XLM. Mozilla DeepSpeech doesn't support XLM rescoring, other toolkits do and it gives great improvement in accuracy. If you care about accurate transcription you'd better consider alternatives.