Hang Li

Yu Kang, Tianqiao Liu, Hang Li et al.

Multimodal pre-training for audio-and-text has recently been proved to be effective and has significantly improved the performance of many downstream speech understanding tasks. However, these state-of-the-art pre-training audio-text models work well only when provided with large amount of parallel audio-and-text data, which brings challenges on many languages that are rich in unimodal corpora but scarce of parallel cross-modal corpus. In this paper, we investigate whether it is possible to pre-train an audio-text multimodal model with extremely low-resource parallel data and extra non-parallel unimodal data. Our pre-training framework consists of the following components: (1) Intra-modal Denoising Auto-Encoding (IDAE), which is able to reconstruct input text (audio) representations from a noisy version of itself. (2) Cross-modal Denoising Auto-Encoding (CDAE), which is pre-trained to reconstruct the input text (audio), given both a noisy version of the input text (audio) and the corresponding translated noisy audio features (text embeddings). (3) Iterative Denoising Process (IDP), which iteratively translates raw audio (text) and the corresponding text embeddings (audio features) translated from previous iteration into the new less-noisy text embeddings (audio features). We adapt a dual cross-modal Transformer as our backbone model which consists of two unimodal encoders for IDAE and two cross-modal encoders for CDAE and IDP. Our method achieves comparable performance on multiple downstream speech understanding tasks compared with the model pre-trained on fully parallel data, demonstrating the great potential of the proposed method. Our code is available at: \url{https://github.com/KarlYuKang/Low-Resource-Multimodal-Pre-training}.

Geri Skenderi, Hang Li, Jiliang Tang et al.

Joint-Embedding Predictive Architectures (JEPAs) have recently emerged as a novel and powerful technique for self-supervised representation learning. They aim to learn an energy-based model by predicting the latent representation of a target signal y from the latent representation of a context signal x. JEPAs bypass the need for negative and positive samples, traditionally required by contrastive learning while avoiding the overfitting issues associated with generative pretraining. In this paper, we show that graph-level representations can be effectively modeled using this paradigm by proposing a Graph Joint-Embedding Predictive Architecture (Graph-JEPA). In particular, we employ masked modeling and focus on predicting the latent representations of masked subgraphs starting from the latent representation of a context subgraph. To endow the representations with the implicit hierarchy that is often present in graph-level concepts, we devise an alternative prediction objective that consists of predicting the coordinates of the encoded subgraphs on the unit hyperbola in the 2D plane. Through multiple experimental evaluations, we show that Graph-JEPA can learn highly semantic and expressive representations, as shown by the downstream performance in graph classification, regression, and distinguishing non-isomorphic graphs. The code is available at https://github.com/geriskenderi/graph-jepa.

Xiaoying Zhang, Junpu Chen, Hongning Wang et al.

Off-policy learning, referring to the procedure of policy optimization with access only to logged feedback data, has shown importance in various real-world applications, such as search engines, recommender systems, and etc. While the ground-truth logging policy, which generates the logged data, is usually unknown, previous work simply takes its estimated value in off-policy learning, ignoring both high bias and high variance resulted from such an estimator, especially on samples with small and inaccurately estimated logging probabilities. In this work, we explicitly model the uncertainty in the estimated logging policy and propose a Uncertainty-aware Inverse Propensity Score estimator (UIPS) for improved off-policy learning, with a theoretical convergence guarantee. Experiment results on synthetic and three real-world recommendation datasets demonstrate the advantageous sample efficiency of the proposed UIPS estimator against an extensive list of state-of-the-art baselines.

Xiaoying Zhang, Hongning Wang, Hang Li

Accuracy and diversity have long been considered to be two conflicting goals for recommendations. We point out, however, that as the diversity is typically measured by certain pre-selected item attributes, e.g., category as the most popularly employed one, improved diversity can be achieved without sacrificing recommendation accuracy, as long as the diversification respects the user's preference about the pre-selected attributes. This calls for a fine-grained understanding of a user's preferences over items, where one needs to recognize the user's choice is driven by the quality of the item itself, or the pre-selected attributes of the item. In this work, we focus on diversity defined on item categories. We propose a general diversification framework agnostic to the choice of recommendation algorithms. Our solution disentangles the learnt user representation in the recommendation module into category-independent and category-dependent components to differentiate a user's preference over items from two orthogonal perspectives. Experimental results on three benchmark datasets and online A/B test demonstrate the effectiveness of our solution in improving both recommendation accuracy and diversity. In-depth analysis suggests that the improvement is due to our improved modeling of users' categorical preferences and refined ranking within item categories.

Hang Li, Wei Jin, Geri Skenderi et al.

Denoising Diffusion Probabilistic Models (DDPMs) represent a contemporary class of generative models with exceptional qualities in both synthesis and maximizing the data likelihood. These models work by traversing a forward Markov Chain where data is perturbed, followed by a reverse process where a neural network learns to undo the perturbations and recover the original data. There have been increasing efforts exploring the applications of DDPMs in the graph domain. However, most of them have focused on the generative perspective. In this paper, we aim to build a novel generative model for link prediction. In particular, we treat link prediction between a pair of nodes as a conditional likelihood estimation of its enclosing sub-graph. With a dedicated design to decompose the likelihood estimation process via the Bayesian formula, we are able to separate the estimation of sub-graph structure and its node features. Such designs allow our model to simultaneously enjoy the advantages of inductive learning and the strong generalization capability. Remarkably, comprehensive experiments across various datasets validate that our proposed method presents numerous advantages: (1) transferability across datasets without retraining, (2) promising generalization on limited training data, and (3) robustness against graph adversarial attacks.

Hang Li, Yu Kang, Yang Hao et al.

The quality of vocal delivery is one of the key indicators for evaluating teacher enthusiasm, which has been widely accepted to be connected to the overall course qualities. However, existing evaluation for vocal delivery is mainly conducted with manual ratings, which faces two core challenges: subjectivity and time-consuming. In this paper, we present a novel machine learning approach that utilizes pairwise comparisons and a multimodal orthogonal fusing algorithm to generate large-scale objective evaluation results of the teacher vocal delivery in terms of fluency and passion. We collect two datasets from real-world education scenarios and the experiment results demonstrate the effectiveness of our algorithm. To encourage reproducible results, we make our code public available at \url{https://github.com/tal-ai/ML4VocalDelivery.git}.

Hang Li, Wenbiao Ding, Zhongqin Wu et al.

Speech emotion recognition is a challenging task because the emotion expression is complex, multimodal and fine-grained. In this paper, we propose a novel multimodal deep learning approach to perform fine-grained emotion recognition from real-life speeches. We design a temporal alignment mean-max pooling mechanism to capture the subtle and fine-grained emotions implied in every utterance. In addition, we propose a cross modality excitement module to conduct sample-specific adjustment on cross modality embeddings and adaptively recalibrate the corresponding values by its aligned latent features from the other modality. Our proposed model is evaluated on two well-known real-world speech emotion recognition datasets. The results demonstrate that our approach is superior on the prediction tasks for multimodal speech utterances, and it outperforms a wide range of baselines in terms of prediction accuracy. Further more, we conduct detailed ablation studies to show that our temporal alignment mean-max pooling mechanism and cross modality excitement significantly contribute to the promising results. In order to encourage the research reproducibility, we make the code publicly available at \url{https://github.com/tal-ai/FG_CME.git}.

Hang Li, Tianlong Xu, Jiliang Tang et al.

Knowledge concept tagging for questions plays a crucial role in contemporary intelligent educational applications, including learning progress diagnosis, practice question recommendations, and course content organization. Traditionally, these annotations have been conducted manually with help from pedagogical experts, as the task requires not only a strong semantic understanding of both question stems and knowledge definitions but also deep insights into connecting question-solving logic with corresponding knowledge concepts. In this paper, we explore automating the tagging task using Large Language Models (LLMs), in response to the inability of prior manual methods to meet the rapidly growing demand for concept tagging in questions posed by advanced educational applications. Moreover, the zero/few-shot learning capability of LLMs makes them well-suited for application in educational scenarios, which often face challenges in collecting large-scale, expertise-annotated datasets. By conducting extensive experiments with a variety of representative LLMs, we demonstrate that LLMs are a promising tool for concept tagging in math questions. Furthermore, through case studies examining the results from different LLMs, we draw some empirical conclusions about the key factors for success in applying LLMs to the automatic concept tagging task.

Hang Li, Tianlong Xu, Jiliang Tang et al.

Knowledge tagging for questions plays a crucial role in contemporary intelligent educational applications, including learning progress diagnosis, practice question recommendations, and course content organization. Traditionally, these annotations are always conducted by pedagogical experts, as the task requires not only a strong semantic understanding of both question stems and knowledge definitions but also deep insights into connecting question-solving logic with corresponding knowledge concepts. With the recent emergence of advanced text encoding algorithms, such as pre-trained language models, many researchers have developed automatic knowledge tagging systems based on calculating the semantic similarity between the knowledge and question embeddings. In this paper, we explore automating the task using Large Language Models (LLMs), in response to the inability of prior encoding-based methods to deal with the hard cases which involve strong domain knowledge and complicated concept definitions. By showing the strong performance of zero- and few-shot results over math questions knowledge tagging tasks, we demonstrate LLMs' great potential in conquering the challenges faced by prior methods. Furthermore, by proposing a reinforcement learning-based demonstration retriever, we successfully exploit the great potential of different-sized LLMs in achieving better performance results while keeping the in-context demonstration usage efficiency high.

Hang Li, Yu Kang, Tianqiao Liu et al.

Existing audio-language task-specific predictive approaches focus on building complicated late-fusion mechanisms. However, these models are facing challenges of overfitting with limited labels and low model generalization abilities. In this paper, we present a Cross-modal Transformer for Audio-and-Language, i.e., CTAL, which aims to learn the intra-modality and inter-modality connections between audio and language through two proxy tasks on a large amount of audio-and-language pairs: masked language modeling and masked cross-modal acoustic modeling. After fine-tuning our pre-trained model on multiple downstream audio-and-language tasks, we observe significant improvements across various tasks, such as, emotion classification, sentiment analysis, and speaker verification. On this basis, we further propose a specially-designed fusion mechanism that can be used in fine-tuning phase, which allows our pre-trained model to achieve better performance. Lastly, we demonstrate detailed ablation studies to prove that both our novel cross-modality fusion component and audio-language pre-training methods significantly contribute to the promising results.

Jiahao Chen, Hang Li, Wenbiao Ding et al.

In this paper, we propose a simple yet effective solution to build practical teacher recommender systems for online one-on-one classes. Our system consists of (1) a pseudo matching score module that provides reliable training labels; (2) a ranking model that scores every candidate teacher; (3) a novelty boosting module that gives additional opportunities to new teachers; and (4) a diversity metric that guardrails the recommended results to reduce the chance of collision. Offline experimental results show that our approach outperforms a wide range of baselines. Furthermore, we show that our approach is able to reduce the number of student-teacher matching attempts from 7.22 to 3.09 in a five-month observation on a third-party online education platform.

Hang Li, Zhiwei Wang, Jiliang Tang et al.

Classroom activity detection (CAD) aims at accurately recognizing speaker roles (either teacher or student) in classrooms. A CAD solution helps teachers get instant feedback on their pedagogical instructions. However, CAD is very challenging because (1) classroom conversations contain many conversational turn-taking overlaps between teachers and students; (2) the CAD model needs to be generalized well enough for different teachers and students; and (3) classroom recordings may be very noisy and low-quality. In this work, we address the above challenges by building a Siamese neural framework to automatically identify teacher and student utterances from classroom recordings. The proposed model is evaluated on real-world educational datasets. The results demonstrate that (1) our approach is superior on the prediction tasks for both online and offline classroom environments; and (2) our framework exhibits robustness and generalization ability on new teachers (i.e., teachers never appear in training data).

Shaohua Zhang, Haoran Huang, Jicong Liu et al.

Spelling error correction is an important yet challenging task because a satisfactory solution of it essentially needs human-level language understanding ability. Without loss of generality we consider Chinese spelling error correction (CSC) in this paper. A state-of-the-art method for the task selects a character from a list of candidates for correction (including non-correction) at each position of the sentence on the basis of BERT, the language representation model. The accuracy of the method can be sub-optimal, however, because BERT does not have sufficient capability to detect whether there is an error at each position, apparently due to the way of pre-training it using mask language modeling. In this work, we propose a novel neural architecture to address the aforementioned issue, which consists of a network for error detection and a network for error correction based on BERT, with the former being connected to the latter with what we call soft-masking technique. Our method of using `Soft-Masked BERT' is general, and it may be employed in other language detection-correction problems. Experimental results on two datasets demonstrate that the performance of our proposed method is significantly better than the baselines including the one solely based on BERT.

Hang Li, Wenbiao Ding, Zitao Liu

With the rapid emergence of K-12 online learning platforms, a new era of education has been opened up. It is crucial to have a dropout warning framework to preemptively identify K-12 students who are at risk of dropping out of the online courses. Prior researchers have focused on predicting dropout in Massive Open Online Courses (MOOCs), which often deliver higher education, i.e., graduate level courses at top institutions. However, few studies have focused on developing a machine learning approach for students in K-12 online courses. In this paper, we develop a machine learning framework to conduct accurate at-risk student identification specialized in K-12 multimodal online environments. Our approach considers both online and offline factors around K-12 students and aims at solving the challenges of (1) multiple modalities, i.e., K-12 online environments involve interactions from different modalities such as video, voice, etc; (2) length variability, i.e., students with different lengths of learning history; (3) time sensitivity, i.e., the dropout likelihood is changing with time; and (4) data imbalance, i.e., only less than 20\% of K-12 students will choose to drop out the class. We conduct a wide range of offline and online experiments to demonstrate the effectiveness of our approach. In our offline experiments, we show that our method improves the dropout prediction performance when compared to state-of-the-art baselines on a real-world educational dataset. In our online experiments, we test our approach on a third-party K-12 online tutoring platform for two months and the results show that more than 70\% of dropout students are detected by the system.

Hang Li, Yu Kang, Wenbiao Ding et al.

Classroom activity detection (CAD) focuses on accurately classifying whether the teacher or student is speaking and recording both the length of individual utterances during a class. A CAD solution helps teachers get instant feedback on their pedagogical instructions. This greatly improves educators' teaching skills and hence leads to students' achievement. However, CAD is very challenging because (1) the CAD model needs to be generalized well enough for different teachers and students; (2) data from both vocal and language modalities has to be wisely fused so that they can be complementary; and (3) the solution shouldn't heavily rely on additional recording device. In this paper, we address the above challenges by using a novel attention based neural framework. Our framework not only extracts both speech and language information, but utilizes attention mechanism to capture long-term semantic dependence. Our framework is device-free and is able to take any classroom recording as input. The proposed CAD learning framework is evaluated in two real-world education applications. The experimental results demonstrate the benefits of our approach on learning attention based neural network from classroom data with different modalities, and show our approach is able to outperform state-of-the-art baselines in terms of various evaluation metrics.

Jiahao Chen, Hang Li, Wenxin Wang et al.

Online 1 on 1 class is created for more personalized learning experience. It demands a large number of teaching resources, which are scarce in China. To alleviate this problem, we build a platform (marketplace), i.e., \emph{Dahai} to allow college students from top Chinese universities to register as part-time instructors for the online 1 on 1 classes. To warn the unqualified instructors and ensure the overall education quality, we build a monitoring and alerting system by utilizing multimodal information from the online environment. Our system mainly consists of two key components: banned word detector and class quality predictor. The system performance is demonstrated both offline and online. By conducting experimental evaluation of real-world online courses, we are able to achieve 74.3\% alerting accuracy in our production environment.

Xiaoying Zhang, Hong Xie, Hang Li et al.

Contextual bandit algorithms provide principled online learning solutions to balance the exploitation-exploration trade-off in various applications such as recommender systems. However, the learning speed of the traditional contextual bandit algorithms is often slow due to the need for extensive exploration. This poses a critical issue in applications like recommender systems, since users may need to provide feedbacks on a lot of uninterested items. To accelerate the learning speed, we generalize contextual bandit to conversational contextual bandit. Conversational contextual bandit leverages not only behavioral feedbacks on arms (e.g., articles in news recommendation), but also occasional conversational feedbacks on key-terms from the user. Here, a key-term can relate to a subset of arms, for example, a category of articles in news recommendation. We then design the Conversational UCB algorithm (ConUCB) to address two challenges in conversational contextual bandit: (1) which key-terms to select to conduct conversation, (2) how to leverage conversational feedbacks to accelerate the speed of bandit learning. We theoretically prove that ConUCB can achieve a smaller regret upper bound than the traditional contextual bandit algorithm LinUCB, which implies a faster learning speed. Experiments on synthetic data, as well as real datasets from Yelp and Toutiao, demonstrate the efficacy of the ConUCB algorithm.

Yaohua Tang, Fandong Meng, Zhengdong Lu et al.

In this paper, we propose phraseNet, a neural machine translator with a phrase memory which stores phrase pairs in symbolic form, mined from corpus or specified by human experts. For any given source sentence, phraseNet scans the phrase memory to determine the candidate phrase pairs and integrates tagging information in the representation of source sentence accordingly. The decoder utilizes a mixture of word-generating component and phrase-generating component, with a specifically designed strategy to generate a sequence of multiple words all at once. The phraseNet not only approaches one step towards incorporating external knowledge into neural machine translation, but also makes an effort to extend the word-by-word generation mechanism of recurrent neural network. Our empirical study on Chinese-to-English translation shows that, with carefully-chosen phrase table in memory, phraseNet yields 3.45 BLEU improvement over the generic neural machine translator.

Jiatao Gu, Zhengdong Lu, Hang Li et al.

We address an important problem in sequence-to-sequence (Seq2Seq) learning referred to as copying, in which certain segments in the input sequence are selectively replicated in the output sequence. A similar phenomenon is observable in human language communication. For example, humans tend to repeat entity names or even long phrases in conversation. The challenge with regard to copying in Seq2Seq is that new machinery is needed to decide when to perform the operation. In this paper, we incorporate copying into neural network-based Seq2Seq learning and propose a new model called CopyNet with encoder-decoder structure. CopyNet can nicely integrate the regular way of word generation in the decoder with the new copying mechanism which can choose sub-sequences in the input sequence and put them at proper places in the output sequence. Our empirical study on both synthetic data sets and real world data sets demonstrates the efficacy of CopyNet. For example, CopyNet can outperform regular RNN-based model with remarkable margins on text summarization tasks.

Mingxuan Wang, Zhengdong Lu, Hang Li et al.

We propose a novel convolutional architecture, named $gen$CNN, for word sequence prediction. Different from previous work on neural network-based language modeling and generation (e.g., RNN or LSTM), we choose not to greedily summarize the history of words as a fixed length vector. Instead, we use a convolutional neural network to predict the next word with the history of words of variable length. Also different from the existing feedforward networks for language modeling, our model can effectively fuse the local correlation and global correlation in the word sequence, with a convolution-gating strategy specifically designed for the task. We argue that our model can give adequate representation of the history, and therefore can naturally exploit both the short and long range dependencies. Our model is fast, easy to train, and readily parallelized. Our extensive experiments on text generation and $n$-best re-ranking in machine translation show that $gen$CNN outperforms the state-of-the-arts with big margins.

Mingxuan Wang, Zhengdong Lu, Hang Li et al.

Many tasks in natural language processing, ranging from machine translation to question answering, can be reduced to the problem of matching two sentences or more generally two short texts. We propose a new approach to the problem, called Deep Match Tree (DeepMatch$_{tree}$), under a general setting. The approach consists of two components, 1) a mining algorithm to discover patterns for matching two short-texts, defined in the product space of dependency trees, and 2) a deep neural network for matching short texts using the mined patterns, as well as a learning algorithm to build the network having a sparse structure. We test our algorithm on the problem of matching a tweet and a response in social media, a hard matching problem proposed in [Wang et al., 2013], and show that DeepMatch$_{tree}$ can outperform a number of competitor models including one without using dependency trees and one based on word-embedding, all with large margins

Zhaopeng Tu, Baotian Hu, Zhengdong Lu et al.

We propose a novel method for translation selection in statistical machine translation, in which a convolutional neural network is employed to judge the similarity between a phrase pair in two languages. The specifically designed convolutional architecture encodes not only the semantic similarity of the translation pair, but also the context containing the phrase in the source language. Therefore, our approach is able to capture context-dependent semantic similarities of translation pairs. We adopt a curriculum learning strategy to train the model: we classify the training examples into easy, medium, and difficult categories, and gradually build the ability of representing phrase and sentence level context by using training examples from easy to difficult. Experimental results show that our approach significantly outperforms the baseline system by up to 1.4 BLEU points.

Fandong Meng, Zhengdong Lu, Mingxuan Wang et al.

The recently proposed neural network joint model (NNJM) (Devlin et al., 2014) augments the n-gram target language model with a heuristically chosen source context window, achieving state-of-the-art performance in SMT. In this paper, we give a more systematic treatment by summarizing the relevant source information through a convolutional architecture guided by the target information. With different guiding signals during decoding, our specifically designed convolution+gating architectures can pinpoint the parts of a source sentence that are relevant to predicting a target word, and fuse them with the context of entire source sentence to form a unified representation. This representation, together with target language words, are fed to a deep neural network (DNN) to form a stronger NNJM. Experiments on two NIST Chinese-English translation tasks show that the proposed model can achieve significant improvements over the previous NNJM by up to +1.08 BLEU points on average

Jingbo Shang, Tianqi Chen, Hang Li et al.

Many tasks in data mining and related fields can be formalized as matching between objects in two heterogeneous domains, including collaborative filtering, link prediction, image tagging, and web search. Machine learning techniques, referred to as learning-to-match in this paper, have been successfully applied to the problems. Among them, a class of state-of-the-art methods, named feature-based matrix factorization, formalize the task as an extension to matrix factorization by incorporating auxiliary features into the model. Unfortunately, making those algorithms scale to real world problems is challenging, and simple parallelization strategies fail due to the complex cross talking patterns between sub-tasks. In this paper, we tackle this challenge with a novel parallel and efficient algorithm for feature-based matrix factorization. Our algorithm, based on coordinate descent, can easily handle hundreds of millions of instances and features on a single machine. The key recipe of this algorithm is an iterative relaxation of the objective to facilitate parallel updates of parameters, with guaranteed convergence on minimizing the original objective function. Experimental results demonstrate that the proposed method is effective on a wide range of matching problems, with efficiency significantly improved upon the baselines while accuracy retained unchanged.