braindead_in

That's great. We have really struggled with diarization. None of the systems out there actually work! We get close but still mess it up from time to time.

Here are the results on the other files.

  4333.mp3: Oh yeah, it's still pretty tight, though. It's very challenging. They actually pull everything out of your.

  7510.mp3: Demons on TV like that. And for people to expose themselves to being rejected on TV or humiliated by fear factor or.

  8036.mp3: Well, I feel like as far as... as far as cursing and language, because I feel like as long as it's not necessarily in context, but.

  8522.mp3: Stuff to you, so you don't have to spend any body. He.

Yeah, I would agree that the academic datasets are not representative, especially for STT. Our models do around 8.7% on LibrisSpeech Clean. But on our internal dataset PaddlePaddle's WER is 29% whereas we do around 15%. And we regularly see higher WER's in production, especially for accented and noisy files. Hopefully continuous re-training will help improve the generalization. Here's the output from our model the 6063 file.

  0:00:00.7 S1: I'd say there's a such thing as eating too much, but I just have a massively fast the table of them and so, I constantly eating so.

Do you do diarisation and punctuations as well?

Any benchmark on LibrisSpeech or TED dataset?

We actually started our experiments with Kaldi and even built a dataset out of our files to train on. But we found that Kaldi required a lot of data-prep and a long lead time. Our internal dataset is quite large and data prep is quite easy compared to Kaldi.

Try choosing Auto transcribe from the dropdown menu next to the Order Transcript button

We do have provide an option to get the transcripts corrected manually by our transcribers. Would that work for you?

The dropdown menu next to the Edit Transcript button has the delete option. It works best for North American files with clean audio. We will eventually get it as good for British, Australian and all other accents as well. We build a new model every month almost, based on the corrections our transcribers make.

I think I made a mistake there. It should have been 92% accuracy. Can't find the edit button now.

Real time transcription is not something we are looking at right now. Our goal with this is to see if we can assist our transcribers and improve the efficiency of our system. So our focus is offline transcription for now.

Done. Thanks!

We are pretty much on par with Trint. They use the API from Speechmatics and by our benchmark, we are better than Speechmatics on conversational audio. We will be posting a proper benchmark numbers soon. We are building a podcast dataset for testing right now.

Yeah, the predictions are bad around the speaker turns right now. We are working on a better turns model and that should fix this issue.

We are planning to do a benchmark with Google Web Speech once it adds support for Multi Speaker files. We once tried with our internal test set on Google Web Speech when we started building this and the WER came to be around 18%.

The WER on LibriSpeech clean test set is around 0.087 and the CER is 0.030. We trained on around a 5000 hours dataset which included LibriSpeech train.

Here's a transcript, if you want to skim through it.

https://scribie.com/transcript/290d2d5d863e4353b3843816aade9...

Do try out https://scribie.com/transcription/free as well once. Our diarisation system is around 90% accurate on longer paragraphs and will be out this week.

Scribie | Full Stack Developer | Bangalore, India | Full time, one-site

At Scribie, we are building tools which help improve the lives of transcribers worldwide. Our speech recognition engine is best in class, built specifically for offline transcription. We tackle hard engineering problems on a day to day basis and offer a competitive package and equity.

Must have's: Node.js, PHP, HTML/CSS/JS, MySQL, Python, Shell scripts, AWS hands-on experience

Good to have: A bachelor's degree, 3-5 years of experience, comfortable with Agile methodology, believer in TDD, love technical challenges, ability to work independently without supervision, high levels of self-motivation

Apply on https://angel.co/scribie/jobs/235433-full-stack-developer

Do you plan to release an implementation? I would definitely like to play around with it.

Here's an automated transcript of the podcast referenced in the post, in case you want to skim through it.

https://scribie.com/transcript/236de3526ba3499a9259bd9fecb9e...

Automated transcripts are free. That's the pricing for our manual transcription.

https://scribie.com/transcription/free

Here's the automated transcript from our system. I checked around a minute or so and looks like accuracy is around 80%.

https://scribie.com/transcript/c5ae0522950c421c9d939dc8a17f4...

Where's the audio? We also provide automated transcripts and I'd love to compare it.

Here's a automated transcript in case you want to skim through it.

https://scribie.com/transcript/89ba55c337fb4bf6b8d75a1443e2d...

In case you don't want to watch the video, there's a free automated transcript

https://scribie.com/transcript/7aaafeae86da402d8f7b4bfc68ea8...

Here's a free automated transcript. No typing required.

https://scribie.com/transcript/37316433662740289509d4234c14f...

Here's a free automated transcript if you don't want to watch

https://scribie.com/transcript/e32d5f58327045a18842116330a91...

Here's a transcript, if you don't want to watch it.

https://scribie.com/transcript/061a38e3523445819f18bcae9b55e...

Here's a automated transcript in case you don't want to wait for the edited one.

https://scribie.com/transcript/26bbb21e33ff4989b9494b496cf25...

We instead built a alternative to Watson Text to Speech! https://scribie.com/transcription/free

Here's a rough automated transcript in case anyone's interested https://scribie.com/transcript/1c42571af44d45aea69d600f9a5a7...