Tetra, the hottest voice recognition company, won

Speech recognition company Tetra received $1.5 million seed round financing

on August 10, 2017, speech recognition system company Tetra received $1.5 million seed round financing led by amplify partners and co invested by Y combinator. This financing will be used to expand tetra's engineering department and optimize the deep learning method

although there are many speech recognition systems on the market, due to the problems of words and accents when people speak, almost no product can completely achieve accurate recognition. Although this immature technology will be widely used in the future, the competition is also very fierce

tetra co-founder Jon goldsmith believes that the competition in the speech recognition market is indeed very fierce. The internal disassembly of TetR is still a professional after-sales repair of the manufacturer, and its own technology has not reached 100% maturity. However, the speech recognition technology still needs to solve the problem of word and accent recognition through in-depth learning

after installing tetra, its voice recognition and deep learning system will automatically run in the background to do real-time voice recording and text translation when receiving calls. Users can search keywords in the system to find voice and text call records

however, according to the speed of speaking, accent and word use, Tetra's recognition is not perfect. Therefore, Tetra has set up a 24-hour manual paid translation service to meet users' needs for high accuracy. This service can not only meet the accuracy needs of users, but also provide learning data for Tetra's deep learning system

jon gol under normal network conditions, dsmith said that at present, t has benefited from the sharp decline in iron ore and coal prices. ETRA's target customers are those who are very busy every day. These customers hope that their call records with industry experts will be preserved forever. At the same time, traditional enterprises are also the service objects of tetra. Many enterprise level services, such as sales related services, can also be recorded and identified through tetra

in engineering, Tetra provides ready-made APIs to strengthen some infrastructure. The charge for the call recognition system is shown in the figure below:

in this way, Tetra can obtain a large amount of high-frequency learning data, so as to continue to optimize the speech recognition system through deep learning. Compared with simple enterprise level speech recognition services, this data acquisition method costs less and obtains more diverse data

at the same time, the progress of this technology is closely related to which cloud API to use. For example, some npl (natural language processing) service operators have strong speech processing ability, and some have strong digital processing ability. Choosing different NPL may have a great difference in the processing effect of call voice

for voice 1 In the field of coaxiality identification, foreign giants such as Google and Amazon, and domestic giants such as Baidu, Alibaba and Tencent are involved. This market has a wide range of application scenarios, and the technology is not yet mature, so the potential investment and entrepreneurship space is naturally large. In the field of enterprise services, Sogou, iFLYTEK and other enterprises have also launched their voice shorthand systems

however, at present, no company in the field of speech recognition can completely solve the problem of human language diversity mentioned above. How to solve this problem and whether deep learning is the answer still need time to verify

this seed round financing is tetra's first round financing