Fragments of the article (translation):

What is the accuracy of Skrybot? “When the system has learned the particular person, it can even reach up to 100 percent,” assures Bosky. “Without learning it reaches, unfortunately, the accuracy of only about 80 percent. The Polish language is complicated, it has a specific inflection, a noun has more than ten versions. For the time being, I cannot imagine that a machine could have high accuracy without learning the voice of a specific person,” adds Bosky.

Skrybot is also trying to make its way on the market – it was founded among others by Paweł Bosky, the owner of Przepisywanie.pl. The test version for decoding text from audio and video recordings can be found on Skrybot.tv website. “We have signed a letter of intent with ESAProjekt concerning the cooperation on speech recognition in courts,” – adds Bosky. It’s a bigger project that is to enable recording of all court hearings. This will be the task for ESAProjekt, Skrybot would play the role of subcontractor and will be responsible for converting of recordings to text.

link to the entire article (in Polish): http://wyborcza.biz/biznes/1,101562,8268832,Sprzet_elektroniczny_wreszcie_zrozumie_polska_mowe_.html

To make the speech recognition process accurate, a voice personalisation is needed – it is free of charge, the instruction how to do it can be found here here:
SkryBot speech personalisation and recognition instructions

SkryBot Home – what is new:

SkryBot speech recognition announcements:

Programmers are invited to cooperate with us in the process of creating the graphical interface versions for Linux and Mac OS systems.

SkryBot Home website: http://skrybot.pl/produkty/rozpoznawanie-mowy-skrybot-domowy

Forum about SkryBot and speech recognition for those interested in these topics: SkryBot.TV/forum

Fragments from the article:

The headline of the article about Tepson with speech recognition.

TEPSON – an interactive system enabling multi-modal communication with humans has been presented in the article. The prototype is a hardware and a software platform which would enable testing of software coming from different producers, in particular of software that can be used to create a social robot.

Software architecture.
The separate subsystems of TEPSON device along with the interfaces between those subsystems and the system of speech stream registration, SkryBot speech recognition and the system of image acquisition are integrated by means of the YARP middleware (…)
(…) acquisition and processing of sound: YARP interface for the sound card (module /audio), integrated with the module for detection and extraction of speech fragments and with /asr – speech recognition module (YARP interface for the SkryBot program).

One of many features of Tepson:
The dialogue system uses the Wirtualny Doradca software, it is equipped with a general knowledge base and a specialised knowledge base indicated by TP SA, the sound is synthesised by the Ivona software and the speech recognition subsystem uses the SkryBot software which responds to a limited number (up to 500) of phrases or words.

Link to the entire article (in Polish):
Artykuł o robocie społecznym Tepson.

Fragments of the article:

I am impressed by this very innovative product. If, in time, its effectiveness is proven, we will be talking about a small revolution. (…) The product has great potential so there is a possibility to earn money. Well done!

Wojciech Przyłęcki, CEO at InQbe technological incubator.

The project is incredibly interesting and technologically advanced. Even though it contains some errors, the home version impresses with its speech recognition effectiveness. Contrary to the social networking start-ups that sprout like mushrooms, SkryBot’s distinctive feature is its innovation level.

Stanisław Rogoziński, Investment Manager at Trinity Adventure technological incubator.

On skrybot.tv website one can find a functioning automatic device recognising text from wav files. Frankly speaking, at this moment the accuracy of speech recognition in case of longer utterances leaves a lot to be desired, but it is still impressive. Speech recognition technologies may have many applications.

Jarosław Obara, President of the Management Board at Ventures Hub technological incubator.

I am happy that we already have systems made in Poland which deal with the Polish language well. SkryBot has developed a whole range of products dedicated to health service, police, telecommunications and to speech recognition online.

Stanisław Tomczak, Investment Manager at SpeedUp Seed Funding Group.

An interesting start-up, anyway it has probably already been awarded and distinguished (as far as I remember, at the Polish edition of Seedcamp), it’s one of the more prospective technologies. The products are interesting, I personally believe in this technology.

Zbigniew Krzewiński, CoWinners consulting firm

Link to the entire article (in Polish) mamybiznes.pl

Link to the software:

sourceforge.net/projects/skrybotdomowy

SkryBot Home innovations:

1. Date and hour of processing is not displayed by default.
2. Colour indicators of icons designing the operating modes of SkryBot waiting/ recording/ decoding have been added.
3. An explanation has been added in configuration/ text insertion.

The new SkryBot Home version 20100612 works stable with Windows XP, Vista, 7; 32-bit or 64-bit.

Enhancing Polish speech recognition software:

We invite you to leave comments, suggestions and, if need be, to report on noticed errors on the following website:
www.skrybot.tv/Forum or by email. In case of finding errors, we ask you to send us screen shots with their description which will make it easier for us to understand the problem.

Programmers are invited to cooperate with us in the process of creating the graphical interface versions for Linux and Mac OS systems.

SkryBot Home website:
http://skrybot.pl/produkty/rozpoznawanie-mowy-skrybot-domowy

Link to the software:

sourceforge.net/projects/skrybotdomowy

SkryBot Home news:

The new version 20100612 works stable with Windows XP, Vista, 7; 32-bit or 64-bit.
Programmers are invited to cooperate with us in the process of creating the graphical interface versions for Linux and Mac OS systems.

Enhancing Polish speech recognition software:

We invite you to leave comments, suggestions and, if need be, to report on noticed errors on the following website:www.skrybot.tv/mantisbt. After choosing the SkryBot Home project and clicking on add, you can leave your comment. In case of finding errors, we ask you to send us screen shots with their description which will make it easier for us to understand the problem.

We have launched a forum for SkryBot users and persons interested in speech recognition under the following address: SkryBot.TV/forum

SkryBot Home website:
http://skrybot.pl/produkty/rozpoznawanie-mowy-skrybot-domowy

The Polish language version, module: general Polish language.

Download SkryBot from here:

sourceforge.net/projects/skrybotdomowy

We have decided not to postpone making SkryBot Home available to all Internet users and even though the first version has some limitations, defects or elements that still need to be worked on, it can already be used.

We invite you to notify us about errors, leave comments and suggestions on our website:
www.skrybot.tv/mantisbt. After choosing the SkryBot Home project and clicking on add, you can leave your comment. In case of finding errors, we ask you to send us screen shots with their description which will make it easier for us to understand the problem.

We have launched a forum for SkryBot users and persons interested in speech recognition under the following address: SkryBot.TV/forum

We are happy to announce that the first version of SkryBot Home speech recognition software with embedded graphical user interface will be available for everybody as freeware already in June.

SkryBot Home software with limited functionalities will be free. The paid version will have attractive functionalities with the possibility to increase the accuracy of recognition.

Functionalities of the free version:

Recognition of speech – spoken to the microphone or recorded in files.

Possibility to change interface colours.

Switch: fast / slow decoding.

Additional functionalities of the paid version:

Personalisation of the user’s voice which will increase the accuracy of speech recognition by several up to a dozen or so percent.

Built-in communication with skrybot.pl server for software updates.

Automatic redaction of text.

Synchronising of a personalised SkryBot Home module with speech recognition online: SkryBot.tv

System requirements:
At least 2GHz processor, dual core processor recommended.
(the better the computer configuration, the faster the speech recognition process)
RAM: at least 1GB; 2GB recommended
Operating system:
Windows XP, Vista, 7 – 32-bit or 64-bit.

We are constantly working to make the speech recognition algorithms faster and we are correcting software errors.

We have launched a speech recognition software forum www.skrybot.tv/forum

Programmers are invited to cooperate with us and to use SkryBot software.
The official website of the project: http://www.skrybotdomowy.sourceforge.net

We present the first version of the user interface of SkryBot Home speech recognition software for personal computers.

SkryBot Home speech recognition - user interface 1.0.

Basic functionalities of SkryBot Home speech recognition for personal computers:

1. Speech recognition with the use of a microphone and speech recognition from previously recorded wav files.
2. Visualisation of the recorded sound.
3. Speech recognition mode setting: fast / slow.
4. Personalisation of speech recognition (increasing of accuracy).
5. Switching of speech recognition for many language versions, for different fields of knowledge and speech characteristics.
6. Entering of text in any text field (among others: Word, Office package, browsers, messengers).
7. Integration of SkryBot Home speech recognition with SkryBot.TV speech recognition online website.
8. Support for disabled persons: reading of the recognised text (built-in speech synthesiser), text edition and steering by voice.
9. Possibility to change skins: backgrounds and fonts.
10. Configurable buttons enabling software control.
11. Recording of voice notes and recorded sound edition.
12. Operating with the systems from the Windows family.
13. Batch mode recognition.

Specifics of the successive versions of SkryBot Home software may be subject to change without previous notice.

New speech recognition models PLPL-v2.1,
WER (Word Error Rate) = 14%
TOTAL Percent correct = 85.74% WER (Word Error Rate) = 16.01% Accuracy = 83.99%.

The accuracy mentioned above can be obtained for recordings having the quality of at least:
16kHz / 16bit mono, wav format, noise smaller than -40dB,
when the vocabulary used in the recordings is similar to the one found in TV and radio programmes.