Context-aware Selective Label Smoothing For Calibrating Sequence Recognition Model

Abstract

Despite the success of deep neural network (DNN) on sequential data (i.e., scene text and speech) recognition, it suffers from the over-confidence problem mainly due to overfitting in training with the cross-entropy loss, which may make the decision-making less reliable. Confidence calibration has been recently proposed as one effective solution to this problem. Nevertheless, the majority of existing confidence calibration methods aims at non-sequential data, which is limited if directly applied to sequential data since the intrinsic contextual dependency in sequences or the class-specific statistical prior is seldom exploited. To the end, we propose a Context-Aware Selective Label Smoothing (CASLS) method for calibrating sequential data. The proposed CASLS fully leverages the contextual dependency in sequences to construct confusion matrices of contextual prediction statistics over different classes. Class-specific error rates are then used to adjust the weights of smoothing strength

Stats

• citations0
• S2 citations—
• github stars0
• HF likes0
• heat score0.00
• arxiv keyhuang2023context

Related papers

• Training LDCRF Model On Unsegmented Sequences Using Connectionist Temporal Classification (2016)2.26
• Right Label Context In End-to-end Training Of Time-synchronous ASR Models (2025)0.00
• Semantic-aware Confidence Calibration For Automated Audio Captioning (2025)0.00
• Multi-scale Alignment And Contextual History For Attention Mechanism In Sequence-to-sequence Model (2018)7.16
• Homophone-based Label Smoothing In End-to-end Automatic Speech Recognition (2020)0.00
• Supervised Attention In Sequence-to-sequence Models For Speech Recognition (2022)5.84
• Towards Relevance And Sequence Modeling In Language Recognition (2020)9.23
• Multitask Learning With CTC And Segmental CRF For Speech Recognition (2017)0.00