Liệt kê các phương pháp đo lường kết quả trong các bài báo quốc tế về ASR (Automatic Speech Recognition).

Danh sách:

Tên bài	Phương pháp	Model liên quan	năm & link	Ảnh
SMALL-FOOTPRINT KEYWORD SPOTTING USING DEEP NEURAL NETWORK AND CONNECTIONIST TEMPORAL CLASSIFIER	False Reject Rate - False Alarm Rate	Baseline Deep-KWS (3x128), DNN(3x128)+CTC, DNN(3x256)+CTC, DNN(4x256)+CTC, DNN(6x512)+CTC	12 Sep 2017
END-TO-END MODELS WITH AUDITORY ATTENTION IN MULTI-CHANNEL KEYWORD SPOTTING	False Reject Rate - False Alarm Rate	Baseline, Attention, Attention_map, Transfer, Transfer_map, Tran_Multi_map	3 Nov 2018
HIERARCHICAL NEURAL NETWORK ARCHITECTURE IN KEYWORD SPOTTING	Recall rate – Miswake/hour	HNN1/2/3 (allbn, 1bn), baseline	6 Nov 2018
EFFICIENT KEYWORD SPOTTING USING DILATED CONVOLUTIONS AND GATING	False Reject Rate - False Alarm Rate per hour	CNN, LSTM, WaveNet	19 Nov 2018
QUERY-BY-EXAMPLE KEYWORD SPOTTING USING LONG SHORT-TERM MEMORY NETWORKS	False Reject Rate - False Alarm Rate	Phone DNN/LSTM+DTW, LSTM Feat Extrator	2015
Wav2Letter: an End-to-End ConvNet-based Speech Recognition System	LER- train set size	MFCC, MFCC+AUG, POW, POU+AUG	13 Sep 2016
minimum-word-error-rate-training-for-attention-based-sequence-to-sequence-models	WER-Training Eporchs	Lamda=0/0.01/0.1, Bi-LAS, +MWER, Uni-LAS, MWER, CD-phone (CE + sMBR)	Dec 05, 2017
EESEN: END-TO-END SPEECH RECOGNITION USING DEEP RNN MODELS AND WFST-BASED DECODING	WER, #params~8.5M	Eesen RNN,Hybrid HMM/DNN dùng LM: Lexicon, trigram	2015
Low latency acoustic modeling using temporal convolution and LSTMs	WER	TDNN-D, LFR-LSTM, LFR-BLSTM, MFR-LSTM, MFR-BLSTM	2018	Stacking LSTMs over time-delay neural network (TDNN)
Long Short-Term Memory Recurrent Neural Network Architectures for Large Scale Acoustic Modeling			interspeech_2014	Baseline model