Hanyu Lai, Xiao Liu, Junjie Gao, Jiale Cheng, Zehan Qi, Yifan Xu, Shuntian Yao, Dan Zhang, Jinhua Du, Zhenyu Hou, Xin Lv, Minlie Huang, Yuxiao Dong, Jie Tang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Neural Theory-of-Mind (N-ToM), machine's ability to understand and keep track of the mental states of others, is pivotal in developing socially intelligent agents.However, prevalent N-ToM benchmarks have several shortcomings, including the presence of ambiguous and artificial narratives, absence of personality traits and preferences, a lack of questions addressing characters' psychological mental states, and limited diversity in the questions posed.In response to these issues, we construct Open-ToM, a new benchmark for assessing N-ToM with (1) longer and clearer narrative stories, (2) characters with explicit personality traits, (3) actions that are triggered by character intentions, and (4) questions designed to challenge LLMs' capabilities of modeling characters' mental states of both the physical and psychological world.Using OpenToM, we reveal that state-of-the-art LLMs thrive at modeling certain aspects of mental states in the physical world but fall short when tracking characters' mental states in the psychological world. 1

Graph Neural Networks (GNNs) have achieved great success in Knowledge Graph Completion (KGC) by modelling how entities and relations interact in recent years. However, the explanation of the predicted facts has not caught the necessary attention. Proper explanations for the results of GNN-based KGC models increase model transparency and help researchers develop more reliable models. Existing practices for explaining KGC tasks rely on instance/subgraph-based approaches, while in some scenarios, paths can provide more user-friendly and interpretable explanations. Nonetheless, the methods for generating path-based explanations for KGs have not been well-explored. To address this gap, we propose Power-Link, the first path-based KGC explainer that explores GNN-based models. We design a novel simplified graph-powering technique, which enables the generation of path-based explanations with a fully parallelisable and memory-efficient training scheme. We further introduce three new metrics for quantitative evaluation of the explanations, together with a qualitative human evaluation. Extensive experiments demonstrate that Power-Link outperforms the SOTA baselines in interpretability, efficiency, and scalability. The code is available at https://github.com/OUTHIM/power-link

Difficulty in recognizing complex entities and scarcity of annotated corpus have become the two major difficulties constraining effective information extraction from electronic medical records. Recent research focuses on utilizing complex feature engineering or large-scale pre-trained models to indirectly reduce the degradation of entity recognition caused by these two difficulties. In an attempt to directly address the difficulties, we propose KrNER, a new healthcare-oriented approach to named entity recognition. It utilizes knowledge enhancement and remote supervision techniques to effectively deal with the problem of recognizing complex entities, while guaranteeing better recognition results even in corpus-scarce contexts. Experiments show that KrNER is very effective and significantly outperforms the state-of-the-art Chinese medical named entity recognition models.

The explosive growth of medical data has perfected the establishment of patients' personal health records and provided favorable conditions for smart healthcare, but its fragmentation also brings challenges to patient management. Mainstream research focuses on utilizing medical data to construct disease knowledge graphs to assist patient management, but does not effectively manage massive patient data. In order to make full use of patient data and facilitate the circulation of patient data elements, we propose a new patient sketching technique, KLDP. it constructs knowledge graphs through pre-training techniques, effectively manages patient data based on patients' personal health records and medical history information throughout the treatment cycle, and elementalizes patient data, providing new ideas and implementation solutions for patient management.

While widely used in industry, recurrent neural networks (RNNs) are known to have deficiencies in dealing with long sequences (e.g. slow inference, vanishing gradients etc.). Recent research has attempted to accelerate RNN models by developing mechanisms to skip irrelevant words in input. Due to the lack of labelled data, it remains as a challenge to decide which words to skip, especially for low-resource classification tasks. In this paper, we propose Knowledge-AwareLeap-LSTM (KALL), a novel architecture which integrates prior human knowledge (created either manually or automatically) like in-domain keywords, terminologies or lexicons into Leap-LSTM to partially supervise the skipping process. More specifically, we propose a knowledge-oriented cost function for KALL; furthermore, we propose two strategies to integrate the knowledge: (1) the Factored KALL approach involves a keyword indicator as a soft constraint for the skip-ping process, and (2) the Gated KALL enforces the inclusion of keywords while maintaining a differentiable network in training. Experiments on different public datasets show that our approaches are1.1x~2.6x faster than LSTM with better accuracy and 23.6x faster than XLNet in a resource-limited CPU-only environment.

Recurrent neural networks (RNNs) suffer from well-known limitations and complications which include slow inference and vanishing gradients when processing long sequences in text classification.Recent studies have attempted to accelerate RNNs via various ad hoc mechanisms to skip irrelevant words in the input.However, word skipping approaches proposed to date effectively stop at each or a given time step to decide whether or not a given input word should be skipped, breaking the coherence of input processing in RNNs.Furthermore, current methods cannot change skip rates during inference and are consequently unable to support different skip rates in demanding real-world conditions.To overcome these limitations, we propose Pointer-LSTM, a novel LSTM framework which relies on a pointer network to select important words for target prediction.The model maintains a coherent input process for the LSTM modules and makes it possible to change the skip rate during inference.Our evaluation on four public data sets demonstrates that Pointer-LSTM (a) is 1.1x∼3.5xfaster than the standard LSTM architecture; (b) is more accurate than Leap-LSTM (the state-of-the-art LSTM skipping model) at high skip rates; and (c) reaches robust accuracy levels even when the skip rate is changed during inference.

Yuyun Huang, Jinhua Du. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.

Attention mechanisms are often used in deep neural networks for distantly supervised relation extraction (DS-RE) to distinguish valid from noisy instances.However, traditional 1-D vector attention models are insufficient for the learning of different contexts in the selection of valid instances to predict the relationship for an entity pair.To alleviate this issue, we propose a novel multi-level structured (2-D matrix) self-attention mechanism for DS-RE in a multi-instance learning (MIL) framework using bidirectional recurrent neural networks.In the proposed method, a structured word-level self-attention mechanism learns a 2-D matrix where each row vector represents a weight distribution for different aspects of an instance regarding two entities.Targeting the MIL issue, the structured sentence-level attention learns a 2-D matrix where each row vector represents a weight distribution on selection of different valid instances.Experiments conducted on two publicly available DS-RE datasets show that the proposed framework with a multi-level structured self-attention mechanism significantly outperform state-of-the-art baselines in terms of PR curves, P@N and F1 measures.

Syntactic reordering on the source side has been demonstrated to be helpful and effective for handling different word orders between source and target languages in SMT. In this article, we focus on the Chinese (DE) construction which is flexible and ubiquitous in Chinese and has many different ways to be translated into English so that it is a major source of word order differences in terms of translation quality. This article carries out the Chinese “DE” construction study for Chinese--English SMT in which we propose a new classifier model---discriminative latent variable model (DPLVM)---with new features to improve the classification accuracy and indirectly improve the translation quality compared to a log-linear classifier. The DE classifier is used to recognize DE structures in both training and test sentences of Chinese, and then perform word reordering to make the Chinese sentences better match the word order of English. In order to investigate the impact of the DE classification and reordering in the source side on different types of SMT systems (namely PB-SMT, hierarchical PB-SMT (HPB-SMT) as well as the syntax-based SMT (SAMT)), we conduct a series of experiments on NIST 2005 and 2008 test sets to verify the effectiveness of our proposed model. The experimental results show that the MT systems using the data reordered by our proposed model outperform the baseline systems by 3.01% and 4.03% relative points on the NIST 2005 test set, 4.64% and 4.62% relative points on the NIST 2008 test set in terms of BLEU score for PB-SMT and HPB-SMT respectively. However, the DE classification method does not perform significantly well for SAMT. Additionally, we also conducted some experiments to evaluate our DE classification and reordering approach on the word alignment and phrase table in terms of these three types of SMT systems.

Modern speech synthesis systems attempt to produce\nspeech utterances from an open domain of words. In some situations, the synthesiser will not have the appropriate units to pronounce some words or phrases accurately but it still must attempt to pronounce them. This paper presents a hybrid machine translation and unit selection speech synthesis system. The machine translation system was trained with English as the source and target language. Rather than the synthesiser only saying the input text as would happen in conventional synthesis systems, the synthesiser may say an alternative utterance with the same\nmeaning. This method allows the synthesiser to overcome the\nproblem of insufficient units in runtime.

With the amount of data continuing to grow, extracting "data of interest" is becoming popular, pervasive, and more important than ever. Data mining, as this process is known as, seeks to draw meaningful conclusions, extract knowledge, and acquire models from vast amounts of data. These compute-intensive data-mining applications, where thread-level parallelism can be effectively exploited, are the design targets of future multi-core systems. As a result, future multi-core systems will be required to process terabyte-level workloads. To understand the memory system performance of data-mining applications, this paper presents the use of hardware-software co-simulation to explore the cache design space of several multi-threaded data mining applications. Our study reveals that the workloads are memory intensive, have large working-set sizes, and exhibit good data locality. We find that large DRAM caches can be useful to address their large working-set sizes