Release of Firefox 78.0.1 and update of voice data for Mozilla Common Voice

An emergency patch release has been published Firefox 78.0.1, which fixes a pop-up issue in Firefox 78 problem, causing the disappearance of installed search engines. After updating the browser, the quick access list to search engines for some users was empty, auto-completion in the address bar malfunctioned, and search queries from the search field on the homepage failed to send. The cause of the failure turned out to be the inclusion of the search engine settings synchronization feature in Firefox 78. In Firefox 78.0.1, remote retrieval of settings has been disabled and the local storage method restored.

Also, with a delay of almost a day information was revealed about the vulnerabilities fixed in Firefox 78. Firefox 78 addressed 16 vulnerabilities, 10 of which are marked as critical. Four vulnerabilities, classified under CVE-2020-12426, could potentially allow the execution of arbitrary code by an attacker when opening specially crafted pages. It should be noted that memory issues, such as buffer overflows and accesses to already freed memory areas, have recently been marked as critical, but not critical.

Additionally, announced an update of voice data sets collected as part of an initiative , providing a validated dataset with 780 hours in including pronunciation samples from nearly a hundred thousand people. A total of 7226 hours (with 5591 hours verified) of speech material was collected in 54 languages, 14 of which were proposed for the first time. This includes a set for the Ukrainian language, prepared by the work of 235 participants who dictated 22 hours. For the Russian language, the number of participants increased to
928, and the volume of speech material grew to 105 hours. In comparison, more than 60 thousand people participated in preparing materials in English, dictating 1452 hours of verified speech.

The proposed sets can be used in machine learning systems to build recognition and speech synthesis models. The data has been published is in the public domain (CC0). Let's remember that the Common Voice project aims to organize collaborative efforts to build a database of voice samples that reflect all the diversity of voices and speaking styles. Users are invited to record the phrases displayed on the screen or evaluate the quality of data added by other users. The accumulated database of recordings featuring various pronunciations of standard human speech phrases can be used without restrictions in machine learning systems and research projects.

Among the drawbacks of the Common Voice project, the author of the speech recognition library Vosk called the one-sidedness of the voice material (predominance of male voices aged 20-30 and a lack of material featuring women, children, and elderly voices), the absence of vocabulary variability (repetition of the same phrases), the distribution of recordings in a distortion-prone MP3 format, and the creation of a new project instead of joining an existing one. VoxForge.

Source: opennet.ru

Buy reliable website hosting with DDoS protection, VPS VDS servers 🔥 Buy reliable website hosting with DDoS protection, VPS VDS servers | ProHoster