Comprehensive Personal Home Page, CNBC, Zhiyu, Editor: Xiaoqin
[Introduction to Xin Zhiyuan]Speech recognition Daniel Dan Povey is John for student farce
Dan Povey, a voice guru and developer of open source speech recognition system kaldi, was fired by Johns Hopkins University (JHU).
Dan PoveyFacebook immediately extended an olive branch to Dan Povey, and Dan accepted it. But Facebook says it's only for Dan six months.
Dan refused and satirized Facebook's actions to make him feelTo work for a Chinese company, and.。 As soon as the news came out, the domestic academic circles were very excited and speculated about where the God would go.
What made the voice community angry and ran away. What happened at Johns Hopkins University? All this, starting with Dan Povey.To recapture the server
Professor broke in late at night just to recapture the server
It happened at midnight on May 7, when Dan Povey, with a broken clamp, tried to break into the administrative building occupied by the protested students and regain control of the server maintained by CLSP (Language and Voice Processing Center). Prior to that, Garland Hall, the executive building, had been occupied by sit-in students for 34 days. They sealed the windows and chained the doors.
Garland Hall Exterior Wall Occupied by Students
First, students protested against the establishment of Private Police, which they said was aimed at strengthening campus security, which they believed would exacerbate racial conflicts. Second, students protested against the training contract signed by the school and the Immigration and Customs Enforcement Agency (ICE). The school said that this was mainly an emergency medical training signed with the Medical College. Contracts with leadership education.
The principal said that he would not terminate the contracts and that they would soon expire. In a fierce confrontation between the two sides, students chained Garland Hall's gates to protest in a sit-in manner. It was alleged that the school had agreed to sit down with the protesting students for a meeting, only because the notification time was only 12 hours, and the students refused to talk.
Inside the building
But Dan maintains the CLSP server in the building.After unsuccessfully trying to apply to the school for access to the building maintenance server, he decided to stop the protest.Dan is famous for being a workaholic, and his student Guoguo Chen said in Zhihu:One is time, the other is server..
Guoguo Chen said:
Dan feels that data and computing resources, for the entire CLSP, are very important assets and need to be assured that every one of the CLSP can use it. I believe that's why Dan, in this event, will try to retake the server and avoid the loss of data and its resources.
According to local newspapers, that night, in the other six.
Video footage shows Dan being pushed out of the building by students
But he failed. A physical clash ensued. Dan was pushed out of the building by the students, while another man with him was accused of beating and protesting against the students. Dan said the students scratched him when they took him out of the building with a big scratch on his back; they also punched the people they brought, and they showed admirable restraint, although one of them responded with a punch.
It is alleged that a few hours before he broke into the building, he took it with him
Dan explained on his website that the forcible entry to the building was due to a server failure for research, which was in the building occupied by students. Hours after Dan tried to enter the occupied building, on the morning of May 8, nearly 100 Baltimore police surrounded the building, cleaned it up and arrested seven protestors, but the city's top prosecutor later dropped all charges against them.
Dan's dismissal: I still have dignity and independence
Dan Povey, a prominent voice guru, has been an assistant research professor at Johns Hopkins University since 2012, according to LinkedIn. Before joining Hopkins University, he worked as a researcher at Microsoft and IBM.
But Dan said he'd leave early. He defended on his website:Maybe I lost my job, but at least I still have dignity and independence..
He said he had many ways out:
Finally, I leave a few words of Bob Dylan's lyrics:
I don't regret anything I do.
I'm glad I fought. I just hope we win.
Dan is scheduled to work for Facebook speech recognition on the 19th of this month at the Facebook Seattle office. He has said he wants to play a new role, but it's not surprising if he can't get rid of the controversy, he said:
Last Friday, Dan updated his profile on his personal website:
I will no longer work for Facebook. I was supposed to start working full-time on Monday the 19th; they told me yesterday, Thursday the 15th, that wasn't the case. Technically, they did not withdraw their offer; I would be allowed to work for them as full-time contractors for up to six weeks, and (with my cooperation) they would conduct independent investigations into what had happened. Whether I eventually get a full-time offer will depend on the results of the survey. I rejected the offer; the whole thing gave me a sense of deja vu.
Of course, I know that keeping this page will almost certainly end my work on Facebook. Everyone around me is asking me to take it off, but I say, how can I ask someone to be a man and then become a coward?
If someone is curious, my plan is to work in a Chinese company and maybe find a part-time job in a Chinese university.
Dan, the legendary voice legend, develops Kaldi to support half the sky for open source voice recognition software
Non-phonetic readers may not be familiar with Dan Povey, in the phonetic world, Dan Povey can be said to be a legend. A brief glance at the discussion on Zhihu shows that Professor Dan has a very high reputation among his students.
His former students said:
WorkaholicDan is a workaholic. It is said that he works more than 14 hours a day. Besides answering emails, he is writing code.
enthusiasticCare for students, take the initiative to maintain servers, come out for meetings and share hotels to help poor students, and even often give money to homeless people;
pureHis ideal is
Dan Povey's most famous job is to lead developmentKaldiThis is an open source voice recognition toolkit that is available free of charge under Apache license. Kaldi aims to provide flexible and scalable software for speech recognition developers. It supports linear transforms, MMI, boosted MMI and MCE discriminant training, feature space discriminant training and deep neural network. It has also provided TensorFlow integration.
As for the origin of Kaldi's name, it is said that Kaldi is the legendary god of coffee. Dan hopes that the toolkit will be as easy, convenient and popular as coffee.
Now, Kaldi has become one of the most widely used open source speech recognition toolkits, and many speech recognition teams rely on Kaldi. Since its release in 2011, the toolkit has been downloaded more than 20,000 times, and co-authored papers have been cited more than 1,000 times. It can be said that Kaldi supports half the sky of open source speech recognition software.
The influence of Kaldi is huge. It is the first open source software of speech recognition based on weighted finite state and theory. Its modularization and high extensibility design, detailed description documents, complete and open tutorials are also one of the main reasons why it is popular with the majority of developers.
At present, it is widely used as a standard tool in the industry, including MIT, CMU, GHU, Cambridge, Tsinghua University, Shanghai Jiaotong University and so on. The industry includes Microsoft, Google, IBM, Facebook and so on. Its launch also greatly promotes the development of the whole speech recognition field.
Interestingly, Dan Povey is also one of the authors of the Cambridge University Speech Recognition Toolkit HTK, a student of Professor Phil Woodland who led the development of the HTK Toolkit. After several years of development, Kaldi integration technology has been more than HTK. Teachers have developed the first generation of open source speech recognition software, and students have developed the second generation of open source speech recognition software.
Finally, welcome Dan to China. I don't know which company or school will eventually find the baby.
Knowing and discussing: