Source: eKapija | Tuesday, 28.08.2018.| 12:30

Daktilograf startup soon to develop software for automatic speech-to-text transfer – First two apps in spring 2019

Illustration (Photo: FeudMoth/shutterstock.com)
The founders of the Daktilograf startup, which is developing the eponymous piece of software for real-time speech-to-text transfer, are likely to become favorites of their journalist colleagues. If they realize their ideas successfully, they are also likely to be held in high regard by clerks, judges, lawyers, translators and others who spend a lot of time performing a very non-creative task – listening to and transcribing audio recordings.

According to the development director, Mirsad Hadzalic, also the president of the NGO Super from Pljevlja, which is working on the startup's registration and further development in cooperation with Om3ga Solutions, Daktilograf is now developing a prototype applicable to South Slavic languages, but the plan is to also develop modules for English and Spanish.

– We are currently in the phase of a functional prototype and we have obtained an investment for the further development. The software is very simple to use, and our technology is unique and different from the competition in the region, which develops dictionary-based solutions, as well as from global companies, which also use AI. Our speech-to-text system recognizes words as audio signals and also features already memorized words, and since we don't use AI, we need far less processing power. The results are much more predictable than with AI systems too – Hadzalic explains in his interview with eKapija.

He notes that Daktilograf is more reliable than global competitors such as Apple and Microsoft when it comes to South Slavic languages.

– The prototype handles Serbo-Croatian and Slovenian for now, and we've started developing support for Macedonian, Bulgarian, Russian and other Slavic languages. We also have an algorithm for “Western” languages, and the development of the module for English and Spanish is planned for the second quarter of 2019 – our interviewee says.

Hadzalic points out that four people took part in the preparation of the prototype, mostly “IT enthusiasts”, and that the plan is to expand the team and employ experts for the processing of audio signals, linguists and from several other professions. The work on the project started in early 2018, when, during the development of a previous innovation, Mirsad got the idea to use a part of the code for a unique speech-to-text engine. He insisted on having that part of the code rearranged and that's how the development of Daktilograf started.

– In September, we will start developing products for the market, and in six months we should come out with the first two applications. These are Snimac/Konvertor (Recorder/Converter), meant for media, press agencies and similar companies, and Stenogram – meant for state administrations, courts, assemblies, media archives and so on. Thanks to the flexibility of the code, we can also develop any kind of app based on speech-to-text conversion per order, and it can also be applied in the industry of voice commands, for subtitles, robotics, education and in many other industries. We are therefore planning to develop RoboTeacher, which will be primarily used by educational institutions and companies which make smart education products, whereas Titler is meant for preparing subtitles for movies and TV shows...

(Photo: mpaniti/shutterstock.com)
As eKapija's interviewee points out, the solution is equally efficient for all Serbo-Croatian speaking countries, and when it gets developed further, it will be the most advanced tool for the translation from the local languages to the most important languages of the world.

Plan to conquer new markets

Around seven months of work and around EUR 10,000 were invested in the development of the software, primarily due to the servers needed for the functioning of the prototype.

– We are more than satisfied with the development of the project so far, considering how fast we obtained the investment and the fact that our investors are the two most important regional Venture Capital companies which invest in startups – ICT Hub Belgrade and South Central Ventures from Slovenia – Hadzalic says with pride.

Daktilograf is now signing pre-contracts with early clients, which they say will be given big benefits after the development of the project is finished, while they will in turn support the development of the company in the meantime.

– The use of Daktilograf will save a lot of time and provide a number of new possibilities. For example, media will no longer have to send journalists to events, but be able to report in real time thanks to the direct connection of our app with the CMS portal of the medium in question. In parliaments, MPs and administration services will get transcripts in real time, judges will dictate verdicts, lawyers will dictate claims and other documents and get documents immediately, and there are numerous similar examples.

Mirsad says that the main challenge in the upcoming period will be the thorough digitization that the entire economy and the society of the Balkans is looking at, because it will be necessary to “explain to clients the necessity of switching to new technologies and win over the skepticism and the prejudices”.

– Still, we expect to develop a range of good digital products, successfully initiate the sale of apps and multiply the value of the company. We believe that we can become the most successful speech-to-text company in the region and expand to foreign market, and the reason we are optimistic is that there's already a great interest from investors and future clients – Hadzalic says.

Marko Andrejic
Only logged-in users can comment.