Tekom: Profanity Finder

Main article: Speech recognition (technology, market)

2022: Profanity Finder View

On May 19, 2022, Tekom presented a decision on the detection of swear words in media content. The program is based on a neural network previously trained on a certain language material. The dictionary for training includes the main roots of obscene words from the list of Roskomnadzor and lexical units formed from them. As a result of the check, the user receives a marked txt file with specific words and their corresponding timecodes. An advanced version of the report is also available, in which the obscene word is given along with the speech context. In addition to detecting forbidden vocabulary, Profanity Finder can mask the mat, replacing it with an audio signal.

As of May 2022, the accuracy of word detection from the basic lexical set using Profanity Finder is 94%. In addition to the preset dictionary, the solution implements a user dictionary. This feature allows you to add user-relevant lexical units that need to be detected additionally.

Profanity Finder supports verification of video files in MP4, M4A, 3GP formats. A further increase in the number of content formats available for analysis is envisaged. The solution from Tekom analyzes video for prohibited words three times faster than real time.

Tekom also began active work on the search for sound mentions and visual images of Meta services, since from March 2022 the holding's activities in Russia are considered illegal. This will help media companies detect and hide from the content the logos of social networks Facebook and Instagram (recognized extremist organizations and banned in Russia).