Cheng Wang


2022

pdf bib
Calibrating Imbalanced Classifiers with Focal Loss: An Empirical Study
Cheng Wang | Jorge Balazs | György Szarvas | Patrick Ernst | Lahari Poddar | Pavel Danchenko
Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: Industry Track

Imbalanced data distribution is a practical and common challenge in building production-level machine learning (ML) models in industry, where data usually exhibits long-tail distributions. For instance, in virtual AI Assistants, such as Google Assistant, Amazon Alexa and Apple Siri, the “play music” or “set timer” utterance is exposed to an order of magnitude more traffic than other skills. This can easily cause trained models to overfit to the majority classes, categories or intents, lead to model miscalibration. The uncalibrated models output unreliable (mostly overconfident) predictions, which are at high risk of affecting downstream decision-making systems. In this work, we study the calibration of production models in the industry use-case of predicting product return reason codes in customer service conversations of an online retail store; The returns reasons also exhibit class imbalance. To alleviate the resulting miscalibration in the production ML model, we streamline the model development and deployment using focal loss (CITATION).We empirically show the effectiveness of model training with focal loss in learning better calibrated models, as compared to standard cross-entropy loss. Better calibration, in turn, enables better control of the precision-recall trade-off for the models deployed in production.

pdf bib
Deploying a Retrieval based Response Model for Task Oriented Dialogues
Lahari Poddar | György Szarvas | Cheng Wang | Jorge Balazs | Pavel Danchenko | Patrick Ernst
Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing: Industry Track

Task-oriented dialogue systems in industry settings need to have high conversational capability, be easily adaptable to changing situations and conform to business constraints. This paper describes a 3-step procedure to develop a conversational model that satisfies these criteria and can efficiently scale to rank a large set of response candidates. First, we provide a simple algorithm to semi-automatically create a high-coverage template set from historic conversations without any annotation. Second, we propose a neural architecture that encodes the dialogue context and applicable business constraints as profile features for ranking the next turn. Third, we describe a two-stage learning strategy with self-supervised training, followed by supervised fine-tuning on limited data collected through a human-in-the-loop platform. Finally, we describe offline experiments and present results of deploying our model with human-in-the-loop to converse with live customers online.

2021

pdf bib
Learning Slice-Aware Representations with Mixture of Attentions
Cheng Wang | Sungjin Lee | Sunghyun Park | Han Li | Young-Bum Kim | Ruhi Sarikaya
Findings of the Association for Computational Linguistics: ACL-IJCNLP 2021

2018

pdf bib
LRMM: Learning to Recommend with Missing Modalities
Cheng Wang | Mathias Niepert | Hui Li
Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing

Multimodal learning has shown promising performance in content-based recommendation due to the auxiliary user and item information of multiple modalities such as text and images. However, the problem of incomplete and missing modality is rarely explored and most existing methods fail in learning a recommendation model with missing or corrupted modalities. In this paper, we propose LRMM, a novel framework that mitigates not only the problem of missing modalities but also more generally the cold-start problem of recommender systems. We propose modality dropout (m-drop) and a multimodal sequential autoencoder (m-auto) to learn multimodal representations for complementing and imputing missing modalities. Extensive experiments on real-world Amazon data show that LRMM achieves state-of-the-art performance on rating prediction tasks. More importantly, LRMM is more robust to previous methods in alleviating data-sparsity and the cold-start problem.

2016

pdf bib
Punctuation Prediction for Unsegmented Transcript Based on Word Vector
Xiaoyin Che | Cheng Wang | Haojin Yang | Christoph Meinel
Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)

In this paper we propose an approach to predict punctuation marks for unsegmented speech transcript. The approach is purely lexical, with pre-trained Word Vectors as the only input. A training model of Deep Neural Network (DNN) or Convolutional Neural Network (CNN) is applied to classify whether a punctuation mark should be inserted after the third word of a 5-words sequence and which kind of punctuation mark the inserted one should be. TED talks within IWSLT dataset are used in both training and evaluation phases. The proposed approach shows its effectiveness by achieving better result than the state-of-the-art lexical solution which works with same type of data, especially when predicting puncuation position only.

2015

pdf bib
Co-training for Semi-supervised Sentiment Classification Based on Dual-view Bags-of-words Representation
Rui Xia | Cheng Wang | Xin-Yu Dai | Tao Li
Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers)