How artificial intelligence allows you to turn an audio recording of an interview into a finished text in an instant

The address in the navigator, dialed using voice, search engines that process voice queries are just a few examples of how a person turns speech into text and uses speech recognition technology in his daily life.

The business community also quickly enough appreciated the unique data that can be obtained when turning speech into text. The first representatives of the banking community, the retail segment, and, of course, contact centers introduced the voice recognition system. This interest is logical - the amount of interaction with customers is constantly growing, while it is difficult to assess its quality. All business representatives want to know the opinion of the consumer about their services, products and level of service. Here, speech analytics is indispensable: in addition to providing daily feedback from the consumer, it gives an understanding of how company representatives build interaction with the client.

It would seem that the media are far from these processes and technological solutions are not applicable in the context of creative work on preparing texts for publication. However, the Phonemics company ruined this idea by showing us the capabilities of its speech recognition service using the example of processing audio recordings of interviews.

The main task of a journalist is to publish timely and relevant material as soon as possible. The accuracy of the reproduction of comments and the completeness of the information provided are also important. However, decoding audio recordings takes up a significant part of the time and effort. So, on average, turning 15 minutes of audio into text takes about one hour. If the speech recording is not in Russian, then translation into the text will take several times longer. There is not much time and effort left for literary and editorial editing of the text.

For several years, experts at Fonemica have been developing and improving algorithms for recognizing speech and translating it into text. Currently, stereo audio is decrypted with 95% accuracy. Speech can be recorded using any device - camera, voice recorder, mobile or landline phone.

The process of converting speech into text is as follows: the audio recording is uploaded to the server, where the algorithm translates speech into text at a speed of one to one. That is, the processing will take exactly as much time as the audio or video itself. At the same time, there are no restrictions on the volume of material - you can recognize both a short comment and a long speech. Speech recognition can be done at any time of the day and makes it possible to prepare the text for publication within a short time, accelerating the transcription process by 10 times.

One of the significant advantages of the system is its high level of recognition, even if the speech was recorded in a noisy room or under other adverse conditions.

How else can this system help a journalist? Its settings make it possible to recognize text in 34 languages ​​and automatically translate it (for example, into Russian), divide the dialogue into replicas of participants, and analyze the emotional state of speakers.

Turning video and audio recordings into text, the speech recognition system also facilitates the process of finding information in the archive of materials. After a few years, you don’t have to revise or re-listen to all the records to find the right last name, event or date. Automation of the speech recognition process not only saves time, but also allows in this case to process the entire complex of information, even those parts that are not included in the prepared publication.

Artificial intelligence allows the journalist to optimize the preparation of the text. All the mechanical part of the work is carried out by the speech recognition system, while the person uses the time available for literary design and minor editorial editing of the text. The number of prepared materials can be increased several times. And which is much more important - it is always interesting to apply new technologies yourself, to see how they optimize your daily activities and allow you to use temporary and intellectual resources more efficiently.