Native Language Identification in Very Short Utterances Using Bidirectional Long Short-Term Memory Network

Abstract

Native language identification (NLI) is the task of identifying the first language of a user based on their speech or written text in a second language. In this paper, we propose the use of spectrogramand cochleagram-based features extracted from very short speech utterances (0.8 s on average) to infer the native language of an Urdu speaker. The bidirectional long short-term memory (BLSTM) neural networks are adopted for the classification of utterances among the native languages. A set of experiments is carried out for the network architecture search and the system's accuracy is evaluated on the validation data set. Overall accuracy of 74.81% and 71.61% is achieved using the Mel-frequency cepstral coefficients (MFCC) and Gammatone frequency cepstral coefficients (GFCC), respectively. Moreover, the optimized MFCC featurebased BLSTM network and GFCC feature-based BLSTM network are merged together to take advantage of both the feature sets. The experiments show that the performance of the merged network surpasses the individual BLSTM networks and accuracy of 75.69% is achieved on the evaluation data. The effect of test data duration is also analyzed (from 0.27 s to 1.5 s); in addition, it is observed that with very short duration as 0.4 s, an accuracy of over 50% can be achieved.

A Deep Learning Approach for Identifying and Discriminating Spoken Arabic Among Other Languages

“…Table 1 presents a summary of the considered literature review. Initial quarrying experiments were executed in order to uncovered suitability of different deep learning architectures, hence; BLSTM was found to outperform all others [33], [34]. As a consequence in this paper, we propose using the BLSTM architecture for Arabic language identification.…”

Improving Indian Spoken-Language Identification by Feature Selection in Duration Mismatch Framework

“…The modern endto-end language recognition models based on deep learning (DL) algorithms improves the performance by increasing the data set requirement and do not perform well for small data set. In duration-matched condition, DL algorithms like bidirectional long short term memory (BLSTM) and recurrent neural network (RNN) reported excellent performance for short duration language identification [23]. In duration mismatched condition, the performance of these algorithms degrade.…”

HMM Based Language Identification from Speech Utterances of Popular Indic Languages Using Spectral and Prosodic Features

“…Prosody features also play a vital role in recognizing language from speech signal [1,2]. Prosodic features like pitch, energy, stress are different in tonal language compared to non-tonal languages [3,4]. Adeeba and Hussain [3] used BLSTM method to identify the Urdu language with short duration of speech signal.…”

“…Prosodic features like pitch, energy, stress are different in tonal language compared to non-tonal languages [3,4]. Adeeba and Hussain [3] used BLSTM method to identify the Urdu language with short duration of speech signal. The duration of speech signal is an average 8s.…”

“…Language identification was also done by considering the log likelihood by considering the average time for some selected speech samples [3]. LID based on Recurrent neural networks by using prosodic features explained [4].…”