Vision-and-Language Pretrained Models: A Survey

Pretrained models have produced great success in both Computer Vision (CV) and Natural Language Processing (NLP). This progress leads to learning joint representations of vision and language pretraining by feeding visual and linguistic contents into a multi-layer transformer, Visual-Language Pretrained Models (VLPMs). In this paper, we present an overview of the major advances achieved in VLPMs for producing joint representations of vision and language. As the preliminaries, we briefly describe the general task definition and genetic architecture of VLPMs. We first discuss the language and vision data encoding methods and then present the mainstream VLPM structure as the core content. We further summarise several essential pretraining and fine-tuning strategies. Finally, we highlight three future directions for both CV and NLP researchers to provide insightful guidance.

Sequential Attention Module for Natural Language Processing

Recently, large pre-trained neural language models have attained remarkable performance on many downstream natural language processing (NLP) applications via fine-tuning. In this paper, we target at how to further improve the token representations on the language models. We, therefore, propose a simple yet effective plug-and-play module, Sequential Attention Module (SAM), on the token embeddings learned from a pre-trained language model. Our proposed SAM consists of two main attention modules deployed sequentially: Feature-wise Attention Module (FAM) and Token-wise Attention Module (TAM). More specifically, FAM can effectively identify the importance of features at each dimension and promote the effect via dot-product on the original token embeddings for downstream NLP applications. Meanwhile, TAM can further re-weight the features at the token-wise level. Moreover, we propose an adaptive filter on FAM to prevent noise impact and increase information absorption. Finally, we conduct extensive experiments to demonstrate the advantages and properties of our proposed SAM. We first show how SAM plays a primary role in the champion solution of two subtasks of SemEval'21 Task 7. After that, we apply SAM on sentiment analysis and three popular NLP tasks and demonstrate that SAM consistently outperforms the state-of-the-art baselines.

RefBERT: Compressing BERT by Referencing to Pre-computed Representations

Recently developed large pre-trained language models, e.g., BERT, have achieved remarkable performance in many downstream natural language processing applications. These pre-trained language models often contain hundreds of millions of parameters and suffer from high computation and latency in real-world applications. It is desirable to reduce the computation overhead of the models for fast training and inference while keeping the model performance in downstream applications. Several lines of work utilize knowledge distillation to compress the teacher model to a smaller student model. However, they usually discard the teacher's knowledge when in inference. Differently, in this paper, we propose RefBERT to leverage the knowledge learned from the teacher, i.e., facilitating the pre-computed BERT representation on the reference sample and compressing BERT into a smaller student model. To guarantee our proposal, we provide theoretical justification on the loss function and the usage of reference samples. Significantly, the theoretical result shows that including the pre-computed teacher's representations on the reference samples indeed increases the mutual information in learning the student model. Finally, we conduct the empirical evaluation and show that our RefBERT can beat the vanilla TinyBERT over 8.1\% and achieves more than 94\% of the performance of $\BERTBASE$ on the GLUE benchmark. Meanwhile, RefBERT is 7.4x smaller and 9.5x faster on inference than BERT$_{\rm BASE}$.

Progressive Open-Domain Response Generation with Multiple Controllable Attributes

It is desirable to include more controllable attributes to enhance the diversity of generated responses in open-domain dialogue systems. However, existing methods can generate responses with only one controllable attribute or lack a flexible way to generate them with multiple controllable attributes. In this paper, we propose a Progressively trained Hierarchical Encoder-Decoder (PHED) to tackle this task. More specifically, PHED deploys Conditional Variational AutoEncoder (CVAE) on Transformer to include one aspect of attributes at one stage. A vital characteristic of the CVAE is to separate the latent variables at each stage into two types: a global variable capturing the common semantic features and a specific variable absorbing the attribute information at that stage. PHED then couples the CVAE latent variables with the Transformer encoder and is trained by minimizing a newly derived ELBO and controlled losses to produce the next stage's input and produce responses as required. Finally, we conduct extensive evaluations to show that PHED significantly outperforms the state-of-the-art neural generation models and produces more diverse responses as expected.

PALI at SemEval-2021 Task 2: Fine-Tune XLM-RoBERTa for Word in Context Disambiguation

This paper presents the PALI team's winning system for SemEval-2021 Task 2: Multilingual and Cross-lingual Word-in-Context Disambiguation. We fine-tune XLM-RoBERTa model to solve the task of word in context disambiguation, i.e., to determine whether the target word in the two contexts contains the same meaning or not. In the implementation, we first specifically design an input tag to emphasize the target word in the contexts. Second, we construct a new vector on the fine-tuned embeddings from XLM-RoBERTa and feed it to a fully-connected network to output the probability of whether the target word in the context has the same meaning or not. The new vector is attained by concatenating the embedding of the [CLS] token and the embeddings of the target word in the contexts. In training, we explore several tricks, such as the Ranger optimizer, data augmentation, and adversarial training, to improve the model prediction. Consequently, we attain first place in all four cross-lingual tasks.

Sattiy at SemEval-2021 Task 9: An Ensemble Solution for Statement Verification and Evidence Finding with Tables

Question answering from semi-structured tables can be seen as a semantic parsing task and is significant and practical for pushing the boundary of natural language understanding. Existing research mainly focuses on understanding contents from unstructured evidence, e.g., news, natural language sentences, and documents. The task of verification from structured evidence, such as tables, charts, and databases, is still less explored. This paper describes sattiy team's system in SemEval-2021 task 9: Statement Verification and Evidence Finding with Tables (SEM-TAB-FACT). This competition aims to verify statements and to find evidence from tables for scientific articles and to promote the proper interpretation of the surrounding article. In this paper, we exploited ensemble models of pre-trained language models over tables, TaPas and TaBERT, for Task A and adjust the result based on some rules extracted for Task B. Finally, in the leaderboard, we attain the F1 scores of 0.8496 and 0.7732 in Task A for the 2-way and 3-way evaluation, respectively, and the F1 score of 0.4856 in Task B.

MagicPai at SemEval-2021 Task 7: Method for Detecting and Rating Humor Based on Multi-Task Adversarial Training

This paper describes MagicPai's system for SemEval 2021 Task 7, HaHackathon: Detecting and Rating Humor and Offense. This task aims to detect whether the text is humorous and how humorous it is. There are four subtasks in the competition. In this paper, we mainly present our solution, a multi-task learning model based on adversarial examples, for task 1a and 1b. More specifically, we first vectorize the cleaned dataset and add the perturbation to obtain more robust embedding representations. We then correct the loss via the confidence level. Finally, we perform interactive joint learning on multiple tasks to capture the relationship between whether the text is humorous and how humorous it is. The final result shows the effectiveness of our system.

Emotion Dynamics Modeling via BERT

Emotion dynamics modeling is a significant task in emotion recognition in conversation. It aims to predict conversational emotions when building empathetic dialogue systems. Existing studies mainly develop models based on Recurrent Neural Networks (RNNs). They cannot benefit from the power of the recently-developed pre-training strategies for better token representation learning in conversations. More seriously, it is hard to distinguish the dependency of interlocutors and the emotional influence among interlocutors by simply assembling the features on top of RNNs. In this paper, we develop a series of BERT-based models to specifically capture the inter-interlocutor and intra-interlocutor dependencies of the conversational emotion dynamics. Concretely, we first substitute BERT for RNNs to enrich the token representations. Then, a Flat-structured BERT (F-BERT) is applied to link up utterances in a conversation directly, and a Hierarchically-structured BERT (H-BERT) is employed to distinguish the interlocutors when linking up utterances. More importantly, a Spatial-Temporal-structured BERT, namely ST-BERT, is proposed to further determine the emotional influence among interlocutors. Finally, we conduct extensive experiments on two popular emotion recognition in conversation benchmark datasets and demonstrate that our proposed models can attain around 5\% and 10\% improvement over the state-of-the-art baselines, respectively.

KGSynNet: A Novel Entity Synonyms Discovery Framework with Knowledge Graph

Entity synonyms discovery is crucial for entity-leveraging applications. However, existing studies suffer from several critical issues: (1) the input mentions may be out-of-vocabulary (OOV) and may come from a different semantic space of the entities; (2) the connection between mentions and entities may be hidden and cannot be established by surface matching; and (3) some entities rarely appear due to the long-tail effect. To tackle these challenges, we facilitate knowledge graphs and propose a novel entity synonyms discovery framework, named \emph{KGSynNet}. Specifically, we pre-train subword embeddings for mentions and entities using a large-scale domain-specific corpus while learning the knowledge embeddings of entities via a joint TransC-TransE model. More importantly, to obtain a comprehensive representation of entities, we employ a specifically designed \emph{fusion gate} to adaptively absorb the entities' knowledge information into their semantic features. We conduct extensive experiments to demonstrate the effectiveness of our \emph{KGSynNet} in leveraging the knowledge graph. The experimental results show that the \emph{KGSynNet} improves the state-of-the-art methods by 14.7\% in terms of hits@3 in the offline evaluation and outperforms the BERT model by 8.3\% in the positive feedback rate of an online A/B test on the entity linking module of a question answering system.

Automatic Intent-Slot Induction for Dialogue Systems

Automatically and accurately identifying user intents and filling the associated slots from their spoken language are critical to the success of dialogue systems. Traditional methods require manually defining the DOMAIN-INTENT-SLOT schema and asking many domain experts to annotate the corresponding utterances, upon which neural models are trained. This procedure brings the challenges of information sharing hindering, out-of-schema, or data sparsity in open-domain dialogue systems. To tackle these challenges, we explore a new task of {\em automatic intent-slot induction} and propose a novel domain-independent tool. That is, we design a coarse-to-fine three-step procedure including Role-labeling, Concept-mining, And Pattern-mining (RCAP): (1) role-labeling: extracting keyphrases from users' utterances and classifying them into a quadruple of coarsely-defined intent-roles via sequence labeling; (2) concept-mining: clustering the extracted intent-role mentions and naming them into abstract fine-grained concepts; (3) pattern-mining: applying the Apriori algorithm to mine intent-role patterns and automatically inferring the intent-slot using these coarse-grained intent-role labels and fine-grained concepts. Empirical evaluations on both real-world in-domain and out-of-domain datasets show that: (1) our RCAP can generate satisfactory SLU schema and outperforms the state-of-the-art supervised learning method; (2) our RCAP can be directly applied to out-of-domain datasets and gain at least 76\% improvement of F1-score on intent detection and 41\% improvement of F1-score on slot filling; (3) our RCAP exhibits its power in generic intent-slot extractions with less manual effort, which opens pathways for schema induction on new domains and unseen intent-slot discovery for generalizable dialogue systems.