Since it´s establishment, TOP 09 has been always offering a high quality service for journalists and members of the party or its supporters like the video, online broadcasting and also fast transcription of virtually everything what has been said. This information is used not only by the general public (on our websites or social networks) but also by the professionals – journalists and analysts, who get mainly quality information for their work very fast. During last six years a whole range of procedures has been applied. This article deals with experience of using the voice (or speech) recognition and transcription with using the unique ZAV method.
Since the establishing TOP 09 convention at the end of November 2009, two technologies for record and information processing has been used. A typical audio record, which was streaming in the same time on the internet, was completed with the record of speeches (MP3), which served mainly for the transcription. Later, it was used for the double check and proofreading of the text, because two writers were able to catch most of the information in the real time. There was only two exceptions – speeches of the very fast speakers or these who do not deliver at least the concept of the speech before it was given.
In 2013 new technologies were deployed. Thanks to the spreading of the YouTube services it was possible not only to broadcast online, but also freely return in time in the broadcasting itself. This function replaced special MP3 records. The streaming is still backed by the second audio record, but it is done only for unexpected fail of technique. We also started to use the technology for speech recording in 2013. It was backed by quite numerous amount of writers – about 8 persons. These writers were processing text which was written by computer – they have to correct it and do necessary stylistic changes. The original plan was to use four teams of two. When we launched the system, we found this disadvantages:
1) The record which is created by computer is available only after it is done completely. So the situation occured when writers were waiting ten or more minutes till the speech was delivered. It could be probably solved by using more techniques, which would higher the cost for the transcription (techniques and the service).
2) A lots of errors. Voices of participants on the convention are not – from the major part – professional. An amount of mistakes in the record which were done made such a huge requirement on writers, so it would be easier to re-write the text again.
So the voice recording system were start off during the convention, because combination of the circumstances written above caused huge delay in the information processing. That was a big handicap mainly for the professional use. The system of work was changed and writers were doing the transcript of the speech by standard way. Writers who were able to type fast (writing all ten) were writing what they were hearing and the second group were correcting the text of the first group with the help of the record on YouTube and the possibility to move back with the record. Thanks to this change the time delay was reduced and usability of the record was increased.
Two years later, based on the experience from 2013, the cooperation with the team of three writers which were using the ZAV technology. The input data were provided by the TOP 09 YouTube channel. The offer for data processing from different place than on the convention was considered, but was not accepted due to random failure of the internet on the ZAV workplace, but I am convinced that we will use it in future.
A combination of writers writing all ten and differentiation of the roles in quite small team let us to lower the time loss. In 2015 there also was no need of special equipment but stable high speed internet. An example: the hardest speech to transcript was and eight minutes long emotive speech. Despite the high speed it was fully written for publishing twenty minutes after the speaker ended the speech.
I would like to explain that – due to the huge pressure of media on the speed – TOP 09 is publishing the transcription without proofreading and then (in tens of minutes) is publishing the proofread version of the text, similar system is used in the Chamber of Deputies. In this case I am speaking about the final version of the text after corrections.
One of the reasons for the quality results were the professionals in the ZAV team. Not every group of writers can contain the world champions or the word vice champions. The application of several ways of the record and transcriptof information let the organizers of this event to define two conclusions:
1) For the Czech language and non professional speakers the most advantageous is still the all ten writer who also handles the Czech spelling, from the price X achievement point of view.
2) The development of technologies is changing and make cheaper the whole process of transcription of the information. It enables to work from other surroundings. Audio record is more flexible and distinctly cheaper. The text information still offers easier indexation and search.
Jaroslav Poláček
election manager TOP 09
vicepresident Interinfo CZ
