Researchers to Develop Accent Detection AI

The team of scientists was from Cisco, the Moscow Institute of Physics and Technology and the Higher School of Economics
13 July 2018   179

A team of scientists used machine learning to develop an improved model for speech recognition. This is reported by Venture Beat.

Previously, scientists manually identified phonological similarities between units of language in general American English and the pronunciation dictionary of the Carnegie Mellon University. To create an improved model, they went non-standard way and allowed it to automatically form the rules. Then, it compared the resulting unique list with a set of examples from George Mason University's speech accents archive.

More non-native accented speech data is necessary to enhance the performance of … existing [speech recognition] models. However, its synthesis is still an open problem.


Based on the received examples, the team created a phonetic data set, through which a neural network, often used for speech recognition, was trained. The accuracy of the definition of words, after overcoming the mark of 800,000 examples, was 59%.

The study was called preliminary due to fewer sounds in the Carnegie-Mellon University dictionary. Despite phonetic coincidences in 13 out of 20 dictionary comparisons, scientists managed to increase the data array from 103 thousand phonetic transcriptions with one accent to 1 million samples with several accents.

Samsung to Create DRAM for AI Mobile Apps

The chip called LPDDR5 is created using a 10-nanometer process technology and has a capacity of 8 GB
20 July 2018   114

Samsung has completed the development and testing of a new memory module standard LPDDR5. The chip is created using a 10-nanometer process technology and has a capacity of 8 GB. LPDDR5 DRAM is planned to be used in the production of smartphones with support for mobile communication 5G, on-board electronics of cars, as well as solutions for AI based on mobile platforms.

One of the innovations of LPDDR5 technology is improved energy efficiency. According to the company, the chip consumes 30% less electricity compared to the previous generation of LPDDR4X memory chips. The number of memory banks (subsections of the DRAM cell) is increased from 8 to 16 to provide high speed with low power consumption. In active mode, the module lowers the operating voltage to synchronize with the application processing speed.

Also, the new chip implements an advanced "deep sleep mode" to reduce the power consumption level to half the level of "standby" in LPDDR4X memory modules.

In the new LPDDR5 chip, the data transfer rate is increased to 6400 Mbit / s, which is 1.5 times more than the last LPDDR4X chips. According to Samsung, the module allows to transfer the volume in 51.2 GB of data (14 films with the size of 3.7 GB each as a Full-HD) in just 1 second.

The chip is scheduled to be launched in mass production as soon as the LPDDR5 specification is approved.