Seq2Seq Deep Learning Models for Microtext Normalization

Satapathy, Ranjan; Li, Yang; Cavallari, Sandro; Wang, Zhaoxia

doi:10.1109/ijcnn.2019.8851895

Cited by 21 publications

(16 citation statements)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multimodal features have been proposed to recognize human emotions. Commonly used modalities are visual [16], audio [18], text [14], facial action [15], posed versus spontaneous expression and multiple physiological parameters [13]. In the multimodal emotion recognition, audio-visual contents are most studied [12], [19].…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Outlier Processing in Multimodal Emotion Recognition

et al. 2020

View full text Add to dashboard Cite

Automatic emotion recognition plays a key role in human-computer interactions. Multimodal emotion recognition has attracted much attention in recent years. When multimodalities are used, different modalities interact with each other and the obtained results tend to be accurate in general. However, there are also cases of unimodal anomalies. Most of the existing studies do not take into account the existence of outliers in the multimodality, which leads to low accuracy of the prediction results. This paper proposes fuzzy weighted support vector machine for regression (FWSVR) to deal with outliers and prediction errors. We design an automatic affective recognition model structure to analyze continuous dimension emotions based on multimodality (audio and visual). The LIRIS-ACCEDE database is used in this work. Experimental results indicate that the concordance correlation coefficient (CCC) is 0.9456 for arousal and 0.9183 for valence on the test set. The fusion result obtained when using fuzzy weighting is much better than the direct fusion one. INDEX TERMS Multimodal emotion recognition, fuzzy weighted support vector machine for regression (FWSVR), continuous dimension emotional space, outlier processing.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Only when α i = µ i C and α * i = µ i C, the error between the estimated value f (x) and the actual value y i is more than ε. Thus, we can obtain the bias as (14).…”

Section: ∂L ∂Wmentioning

confidence: 99%

Outlier Processing in Multimodal Emotion Recognition

et al. 2020

View full text Add to dashboard Cite

show abstract

“…To factor in such contextual signals, recent advancements in NLP has considered these sequential nature of a written language as well as the long-term dependencies present in sentences. Thus, the research community has proposed different methodologies to perform micro-text normalisation based on deep learning (Min and Mott, 2015;Edizel et al, 2019;Gu et al, 2019;Satapathy et al, 2019). While we address the problem of text normalisation in the NLP context, it has also been adopted as a key component for speech applications (Sproat and Jaitly, 2016;Zhang et al, 2019).…”

Section: Related Workmentioning

confidence: 99%

Neural Multi-task Text Normalization and Sanitization with Pointer-Generator

Nguyen¹,

Cavallari²

2020

Proceedings of the First Workshop on Natural Language Interfaces

Self Cite

View full text Add to dashboard Cite

Text normalization and sanitization are intrinsic components of Natural Language Inferences. In Information Retrieval or Dialogue Generation, normalization of user queries or utterances enhances linguistic understanding by translating non-canonical text to its canonical form, on which many state-of-the-art language models are trained. On the other hand, text sanitization removes sensitive information to guarantee user privacy and anonymity. Existing approaches to normalization and sanitization mainly rely on hand-crafted heuristics and syntactic features of individual tokens while disregarding the linguistic context. Moreover, such context-unaware solutions cannot dynamically determine whether out-of-vocab tokens are misspelt or are entity names. In this work, we formulate text normalization and sanitization as a multi-task text generation approach and propose a neural pointer-generator network based on multihead attention. Its generator effectively captures linguistic context during normalization and sanitization while its pointer dynamically preserves the entities that are generally missing in the vocabulary. Experiments show that our generation approach outperforms both token-based text normalization and sanitization, while the pointer-generator improves the generator-only baseline in terms of BLEU4 score, and classical attentional pointer networks in terms of pointing accuracy.

show abstract

“…'af' -'as fuck', 'kys' -'kill your self ', etc.). A list of microtext 3 from Satapathy et al (2019) was used to normalize those words. Due to the limitation of computational power, we decide not to pre-train BERT model from scratch but fine-tune from the BERT-Large, Uncased (Whole Word Masking) checkpoint.…”

Section: Data Preprocessingmentioning

confidence: 99%

PGSG at SemEval-2020 Task 12: BERT-LSTM with Tweets’ Pretrained Model and Noisy Student Training Method

Pham-Hong¹,

Chokshi²

2020

Proceedings of the Fourteenth Workshop on Semantic Evaluation

View full text Add to dashboard Cite

The paper presents a system developed for the SemEval-2020 competition Task 12 (OffensEval-2): Multilingual Offensive Language Identification in Social Media. We achieve the second place (2nd) in sub-task B: Automatic categorization of offense types and are ranked 55th with a macro F1-score of 90.59 in sub-task A: Offensive language identification. Our solution is using a stack of BERT and LSTM layers, training with the Noisy Student method. Since the tweets data contains a large number of noisy words and slang, we update the vocabulary of the BERT large model pre-trained by the Google AI Language team. We fine-tune the model with tweet sentences provided in the challenge.

show abstract

Seq2Seq Deep Learning Models for Microtext Normalization

Cited by 21 publications

References 41 publications

Outlier Processing in Multimodal Emotion Recognition

Outlier Processing in Multimodal Emotion Recognition

Neural Multi-task Text Normalization and Sanitization with Pointer-Generator

PGSG at SemEval-2020 Task 12: BERT-LSTM with Tweets’ Pretrained Model and Noisy Student Training Method

Contact Info

Product

Resources

About