Native Language Identification Using Spectral and Source-Based Features

Abstract

The task of native language (L1) identification from nonnative language (L2) can be thought of as the task of identifying the common traits that each group of L1 speakers maintains while speaking L2 irrespective of the dialect or region. Under the assumption that speakers are L1 proficient, non-native cues in terms of segmental and prosodic aspects are investigated in our work. In this paper, we propose the use of longer duration cepstral features, namely, Mel frequency cepstral coefficients (MFCC) and auditory filterbank features learnt from the database using Convolutional Restricted Boltzmann Machine (ConvRBM) along with their delta and shifted delta features. MFCC and ConvRBM gave accuracy of 38.2% and 36.8%, respectively, on the development set provided for the ComParE 2016 Nativeness Task using Gaussian Mixture Model (GMM) classifier. To add complementary information about the prosodic and excitation source features, phrase information and its dynamics extracted from the log(F 0) contour of the speech was explored. The accuracy obtained using score-level fusion between system features (MFCC and ConvRBM) and phrase features were 39.6% and 38.3%, respectively, indicating that phrase information and MFCC capture complementary information than ConvRBM alone. Furthermore, score-level fusion of MFCC, ConvRBM and phrase improves the accuracy to 40.2%.

A review of social background profiling of speakers from speech accents

“…To capture the temporal structure in the short-term features, short term features are usually appended with Delta features, where delta features for each time frame represent the local slope of coefficients over a small number of neighbouring time frames found by least-squares ( Rajpal et al, 2016 ).…”

“…Short term Filter-Bank energies have been proven effective for several speech accent classification tasks ( Rajpal et al, 2016 ; Sailor & Patil, 2016 ; Shon, Ali & Glass, 2018 ). Besides, MFCC coefficients have little correlation, which makes them suitable as input for classification models.…”

Combining Textual and Speech Features in the NLI Task Using State-of-the-Art Machine Learning Techniques

“…The participants were provided with the audio files (amplitude normalized) and were also pointed to the toolkit that was used to extract the audio features for the baseline system provided by the sub-challenge organizers. It is obvious that the extracted features did not reflect only the actual content of the utterances but also -and possibly more prominently -the System UAR (%) 1 (Abad et al, 2016) 84.6 2 (Shivakumar et al, 2016) 78.6 3 (Gosztolya et al, 2016) 70.7 4 (Huckvale, 2016) 69.8 5 (Senoussaoui et al, 2016) 68.4 6 (Keren et al, 2016) 61.5 7 (Jiao et al, 2016) 52.2 8 (Rajpal et al, 2016) 39.8 baseline 45.1 acoustic properties of the speech that are supposedly and significantly influenced by the speaker's native language. Given the usual background of the INTERSPEECH attendees, it is only natural that most participants of the sub-challenge had a strong background in speech signal processing and (at least the top teams) concentrated on their own sophisticated methods for feature extraction.…”

Deep neural architectures for dialect classification with single frequency filtering and zero-time windowing feature representations

“…Studies by Hansen and Liu (2016) have shown that acoustic variations are more prominent than the linguistic variations [acoustic models performed better than linguistic models by 15.8% absolute unweighted average recall (UAR)] for major dialects of English. The acoustic variations among dialects include segmental and supra-segmental features, and they can be extracted directly from the speech signal (Behravan et al, 2016;Bougrine et al, 2018;DeMarco and Cox, 2012;Rajpal et al, 2016;Rouas, 2007) or they can be modelled indirectly from the phonetic information derived from the speech signal (Chen et al, 2011;Chen et al, 2014;Najafian et al, 2018;Shon et al, 2018a).…”