Stacking non-linear layers allows deep neural networks to model complicated functions, and including residual connections in Transformer layers is beneficial for convergence and performance. However, residual connections may make the model “forget” distant layers and fail to fuse information from previous layers effectively. Selectively managing the representation aggregation of Transformer layers may lead to better performance. In this paper, we present a Transformer with depth-wise LSTMs connecting cascading Transformer layers and sub-layers. We show that layer normalization and feed-forward computation within a Transformer layer can be absorbed into depth-wise LSTMs connecting pure Transformer attention layers. Our experiments with the 6-layer Transformer show significant BLEU improvements in both WMT 14 English-German / French tasks and the OPUS-100 many-to-many multilingual NMT task, and our deep Transformer experiments demonstrate the effectiveness of depth-wise LSTM on the convergence and performance of deep Transformers.

This article presents the solution of Qufu Normal University for the Multimodal Sentiment Cause Analysis competition in SemEval2024 Task 3.The competition aims to extract emotion-cause pairs from dialogues containing text, audio, and video modalities. To cope with this task, we employ a hybrid pre-train model based approach. Specifically, we first extract and fusion features from dialogues based on BERT, BiLSTM, openSMILE and C3D. Then, we adopt BiLSTM and Transformer to extract the candidate emotion-cause pairs. Finally, we design a filter to identify the correct emotion-cause pairs. The evaluation results show that, we achieve a weighted average F1 score of 0.1786 and an F1 score of 0.1882 on CodaLab.

There has been a growing interest in developing conversational recommendation system (CRS), which provides valuable recommendations to users through conversations. Compared to the traditional recommendation, it advocates wealthier interactions and provides possibilities to obtain users’ exact preferences explicitly. Nevertheless, the corresponding research on this topic is limited due to the lack of broad-coverage dialogue corpus, especially real-world dialogue corpus. To handle this issue and facilitate our exploration, we construct E-ConvRec, an authentic Chinese dialogue dataset consisting of over 25k dialogues and 770k utterances, which contains user profile, product knowledge base (KB), and multiple sequential real conversations between users and recommenders. Next, we explore conversational recommendation in a real scene from multiple facets based on the dataset. Therefore, we particularly design three tasks: user preference recognition, dialogue management, and personalized recommendation. In the light of the three tasks, we establish baseline results on E-ConvRec to facilitate future studies.

We present a new human-human dialogue dataset - PhotoChat, the first dataset that casts light on the photo sharing behavior in online messaging. PhotoChat contains 12k dialogues, each of which is paired with a user photo that is shared during the conversation. Based on this dataset, we propose two tasks to facilitate research on image-text modeling: a photo-sharing intent prediction task that predicts whether one intends to share a photo in the next conversation turn, and a photo retrieval task that retrieves the most relevant photo according to the dialogue context. In addition, for both tasks, we provide baseline models using the state-of-the-art models and report their benchmark performances. The best image retrieval model achieves 10.4% recall@1 (out of 1000 candidates) and the best photo intent prediction model achieves 58.1% F1 score, indicating that the dataset presents interesting yet challenging real-world problems. We are releasing PhotoChat to facilitate future research work among the community.

Pretrained language models like BERT have achieved good results on NLP tasks, but are impractical on resource-limited devices due to memory footprint. A large fraction of this footprint comes from the input embeddings with large input vocabulary and embedding dimensions. Existing knowledge distillation methods used for model compression cannot be directly applied to train student models with reduced vocabulary sizes. To this end, we propose a distillation method to align the teacher and student embeddings via mixed-vocabulary training. Our method compresses BERT-LARGE to a task-agnostic model with smaller vocabulary and hidden dimensions, which is an order of magnitude smaller than other distilled BERT models and offers a better size-accuracy trade-off on language understanding benchmarks as well as a practical dialogue task.

Current state-of-the-art supervised word sense disambiguation (WSD) systems (such as GlossBERT and bi-encoder model) yield surprisingly good results by purely leveraging pre-trained language models and short dictionary definitions (or glosses) of the different word senses. While concise and intuitive, the sense gloss is just one of many ways to provide information about word senses. In this paper, we focus on enhancing the sense representations via incorporating synonyms, example phrases or sentences showing usage of word senses, and sense gloss of hypernyms. We show that incorporating such additional information boosts the performance on WSD. With the proposed enhancements, our system achieves an F1 score of 82.0% on the standard benchmark test dataset of the English all-words WSD task, surpassing all previous published scores on this benchmark dataset.

Tokenization is a fundamental preprocessing step for almost all NLP tasks. In this paper, we propose efficient algorithms for the WordPiece tokenization used in BERT, from single-word tokenization to general text (e.g., sentence) tokenization. When tokenizing a single word, WordPiece uses a longest-match-first strategy, known as maximum matching. The best known algorithms so far are O(nˆ2) (where n is the input length) or O(nm) (where m is the maximum vocabulary token length). We propose a novel algorithm whose tokenization complexity is strictly O(n). Our method is inspired by the Aho-Corasick algorithm. We introduce additional linkages on top of the trie built from the vocabulary, allowing smart transitions when the trie matching cannot continue. For general text, we further propose an algorithm that combines pre-tokenization (splitting the text into words) and our linear-time WordPiece method into a single pass. Experimental results show that our method is 8.2x faster than HuggingFace Tokenizers and 5.1x faster than TensorFlow Text on average for general text tokenization.

Generating long and informative review text is a challenging natural language generation task. Previous work focuses on word-level generation, neglecting the importance of topical and syntactic characteristics from natural languages. In this paper, we propose a novel review generation model by characterizing an elaborately designed aspect-aware coarse-to-fine generation process. First, we model the aspect transitions to capture the overall content flow. Then, to generate a sentence, an aspect-aware sketch will be predicted using an aspect-aware decoder. Finally, another decoder fills in the semantic slots by generating corresponding words. Our approach is able to jointly utilize aspect semantics, syntactic sketch, and context information. Extensive experiments results have demonstrated the effectiveness of the proposed model.

Incompleteness is a common problem for existing knowledge graphs (KGs), and the completion of KG which aims to predict links between entities is challenging. Most existing KG completion methods only consider the direct relation between nodes and ignore the relation paths which contain useful information for link prediction. Recently, a few methods take relation paths into consideration but pay less attention to the order of relations in paths which is important for reasoning. In addition, these path-based models always ignore nonlinear contributions of path features for link prediction. To solve these problems, we propose a novel KG completion method named OPTransE. Instead of embedding both entities of a relation into the same latent space as in previous methods, we project the head entity and the tail entity of each relation into different spaces to guarantee the order of relations in the path. Meanwhile, we adopt a pooling strategy to extract nonlinear and complex features of different paths to further improve the performance of link prediction. Experimental results on two benchmark datasets show that the proposed model OPTransE performs better than state-of-the-art methods.

Person-job fit has been an important task which aims to automatically match job positions with suitable candidates. Previous methods mainly focus on solving the match task in single-domain setting, which may not work well when labeled data is limited. We study the domain adaptation problem for person-job fit. We first propose a deep global match network for capturing the global semantic interactions between two sentences from a job posting and a candidate resume respectively. Furthermore, we extend the match network and implement domain adaptation in three levels, sentence-level representation, sentence-level match, and global match. Extensive experiment results on a large real-world dataset consisting of six domains have demonstrated the effectiveness of the proposed model, especially when there is not sufficient labeled data.