2010-07-01

Phonexia added new features to the telephone Czech speech to text transcript demo

The demo www.PrepisReci.cz supports processing of manually uploaded audio files (the most frequent formats) now. The first three minutes of the audio file are processed by the speech recognizer.

Length of recorded speech via telephone was also increased to 3 minutes. Simply call to +420 54114 2070 and speak or read text.

2010-06-25

Phonexia sponsors the Odyssey speaker and language recognition workshop

Phonexia is proud to announce sponsorship of prestigious research event - The Odyssey 2010 Speaker and Language Recognition Workshop.

"This sponsorship goes in line with our strategy of close cooperation with worldwide speech research community. Phonexia already commercialized several results of the BUT Speech@FIT research group and our technology is used by several universities worldwide for education and for academic research. This experience allows Phonexia to provide highly accurate always state-of-the art speech recognition technologies.", said Radim Kudla, Phonexia´s Business Development Manager.

The Odyssey 2010 Speaker and Language Recognition Workshop is hosted by Brno University of Technology between the 28th June and the 1st July in Brno. We invite you to stop at our booth on the 29th Jun to discuss capabilities of our speaker and language recognition technologies.

2010-06-07

Phonexia launched telephone Czech speech to text transcript demo

Simply call to +420 54114 2070 and speak or read text up to 30 sec long. The automatically transcribed text will be available in a minute on www.PrepisReci.cz.

Phonexia´s speech transcript technology achieves great results, where other technologies fail - in spontaneous telephone call transcription (calls with ambient noises, noise, signal failures, poor pronunciation ...). The accuracy of the transcript exceeds 85% in real condition. By the experience of call-centre managers the quality of transcript is sufficient to fully understand the content of phone calls while reading text.

Technology uses a dictionary of more than one million words. It is a general dictionary of both formal and colloquial Czech. The speech transcription system can be easily integrated to any telephone systems by our partner company. The demo is prepared for the Czech language at the moment.

2010-06-01

The speed of Phonexia´s products significantly increased

We are please to introduce you that the speed of our products has been significantly increased:

• Phonexia Language Identification (LID)... speed increased from 54RT to 120RT
• Phonexia Gender Identification (GID)... speed increased from 125RT to 900RT
• Phonexia Speaker Identification (SID)... speed increased from 20RT to 180RT
• note: RT = faster than realtime; the tests were done on 3GHz Intel CPU, one core, 64 bit Linux (12 languages for LID)

What this means?
Now, one instance of the Phonexia Speaker Identification (SID) gives you the possibility to process 180 minutes of speech in one minute of computer time. This makes capacity of 4320 hours (before only 480 hours) processed daily for the speaker identification.

Compared to our old system (running 20RT) you can process 9 times more data daily what means saving 9 times more money for one instance. How much data you process daily? And how much you save if you run more instances of our speaker recognition? Try our the new versions on http://www.phonexia.com/download/

2010-05-31

The Phonexia´s CEO will speak at the ISS 2010 conference

Pavel Matejka (CEO, Phonexia) will speak about the topic “Intelligence Gathering from Voice Records in Interception and Criminal Investigation”. He will describe the benefits of the Phonexia´s speech recognition products:

• Phonexia Speaker Identification (SID),
• Phonexia Gender Identification (GID),
• Phonexia Language Identification (LID),
• Phonexia Keyword Spotting - spoken term detection (KWS),
• Phonexia Speech to Text Transcription (STT/LVCSR).

Pavel will also introduce a new Phonexia´s Speech Intelligence Platform what integrates all the technology mentioned above to a server based solution. The security and defence intelligence agencies can easily use the speech recognition technologies and integrate them into their own solutions. The speech is scheduled for Thursday, 3 June 2010 for 14:30-15:00, Session B. Feel free to visit also our booth.

2010-03-22

Conference ISS World Europe invitation

We wish to invite you to visit our booth at ISS World Europe Conference taking place at the Prague Conference Centre on 2-4 June 2010.

The ISS World Europe is the world's largest gathering of European law enforcement, intelligence, defence, homeland security analysts and telecom operators responsible for lawful interception, electronic investigations and network intelligence gathering.

We will present following speech data mining technologies: Speaker Identification (SID), Gender Identification (GID), Language Identification (LID), Keyword Spotting - spoken term detection (KWS), Speech to Text Transcription (STT/LVCSR).

2010-02-18

Phonexia entered the technology incubator with its inovative ideas

Phonexia has been successful in negotiation with the the South Moravian Innovation Centre (JIC) and Brno University of Technology (BUT). Phonexia has entered to the Technology Incubator II (TI II) with the idea of inovative products for data mining from speech records.

The TI II became home for part of the Phonexia´s team. The project includes futher improvement of the Phonexia Speech Core (basic software library), as well as development of new products. The team will work also on a platform for speech recognition services .

Phonexia was founded in 2006. During our history, Phonexia has build many speech recognition products. The products and services are provided for security/defense agencies, call-centers and multimedia archives in several countries world wide.