Multiobjective optimization (MOO) is important for deep learning applications with multiple conflicting objectives. Pareto front learning (PFL) methods learn a single model conditioned on the preference of objectives and can be applied to any preference at inference time. However, existing PFL methods use predefined strategies (e.g., uniform sampling) to generate preferences, which could result in unevenly spaced solutions since the shape of Pareto front is largely ignored. In this article, we propose a lightweight and model-agnostic method to train a preference generator for a given PFL model, which learns to generate proper preferences from uniformly sampled ones, such that the resulting solutions are evenly spaced on the Pareto front. Compared to previous works, our method enables a more rational allocation of preferences, which can either be utilized to enhance a pretrained PFL model or be seamlessly integrated into the PFL training process to improve efficiency. We apply our method to state-of-the-art PFL methods with various backbones (e.g., multilayer perceptron, convolutional neural network, transformer) and validate the significance of preference generation across various tasks, from multitask supervised learning to multiobjective reinforcement learning-based neural combinatorial optimization. Experimental results show that our method improves the backbone algorithm in most settings, showing its effectiveness and general applicability.

To understand a document with multiple events, event-event relation extraction (ERE) emerges as a crucial task, aiming to discern how natural events temporally or structurally associate with each other. To achieve this goal, our work addresses the problems of temporal event relation extraction (TRE) and subevent relation extraction (SRE). The latest methods for such problems have commonly built document-level event graphs for global reasoning across sentences. However, the edges between events are usually derived from external tools heuristically, which are not always reliable and may introduce noise. Moreover, they are not capable of preserving logical constraints among event relations, e.g., coreference constraint, symmetry constraint and conjunction constraint. These constraints guarantee coherence between different relation types, enabling the generation of a unified event evolution graph. In this work, we propose a novel method named LogicERE, which performs high-order event relation reasoning through modeling logic constraints. Specifically, different from conventional event graphs, we design a logic constraint induced graph (LCG) without any external tools. LCG involves event nodes where the interactions among them can model the coreference constraint, and event pairs nodes where the interactions among them can retain the symmetry constraint and conjunction constraint. Then we perform high-order reasoning on LCG with relational graph transformer to obtain enhanced event and event pair embeddings. Finally, we further incorporate logic constraint information via a joint logic learning module. Extensive experiments demonstrate the effectiveness of the proposed method with state-of-the-art performance on benchmark datasets.

Script Event Prediction (SEP) aims to forecast the next event in a sequence from a list of candidates. Traditional methods often use pre-trained language models to model event associations but struggle with semantic ambiguity and embedding bias. Semantic ambiguity arises from the multiple meanings of identical words and insufficient consideration of event arguments, while embedding bias results from assigning similar word embeddings to event pairs with similar lexical features, despite their different meanings. To address above issues, we propose a the Semantic and Sentiment Dual-enhanced Generative Model (SSD-GM). SSD-GM leverages two types of script event information to enhance the generative model. Specifically, it employs a GNN-based semantic structure aggregator to integrate the event-centric structure information, thereby mitigating the impact of semantic ambiguity. Furthermore, we find that local sentiment variability effectively reduces biases in event embeddings, while maintaining global sentiment consistency enhances predictive accuracy. As a result, SSD-GM adeptly captures both global and local sentiment of events through its sentiment information awareness mechanism. Extensive experiments on the Multi-Choice Narrative Cloze (MCNC) task demonstrate that our approach achieves better results than other state-of-the-art baselines.

Event Causality Identification (ECI) aims to detect causal relations between events in unstructured texts. This task is challenged by the lack of data and explicit causal clues. Some methods incorporate explicit knowledge from external knowledge graphs (KGs) into Pre-trained Language Models (PLMs) to tackle these issues, achieving certain accomplishments. However, they ignore that existing KGs usually contain trivial knowledge which may prejudice the performance. Moreover, they simply integrate the concept triplets, underutilizing the deep interaction between the text and external graph. In this paper, we propose an effective pipeline DFP, i.e., Distill, Fuse and Pre-train, to build a commonsense-aware pre-trained model which integrates reliable task-specific knowledge from commonsense graphs. This pipeline works as follows: (1) To leverage the reliable knowledge, commonsense graph distillation is proposed to distill commonsense graphs and obtain the meta-graph which contain credible task-oriented knowledge. (2) To model the deep interaction between the text and external graph, heterogeneous information fusion is proposed to fuse them through a commonsense-aware memory network. (3) Continual pre-training designs three continual pre-training tasks to further align and fuse the text and the commonsense meta-graph. Through extensive experiments on two benchmarks, we demonstrate the validity of our pipeline.

Abstract In recent years, Coronavirus disease 2019 (COVID-19) has become a global epidemic, and some efforts have been devoted to tracking and controlling its spread. Extracting structured knowledge from involved epidemic case reports can inform the surveillance system, which is important for controlling the spread of outbreaks. Therefore, in this paper, we focus on the task of Chinese epidemic event extraction (EE), which is defined as the detection of epidemic-related events and corresponding arguments in the texts of epidemic case reports. To facilitate the research of this task, we first define the epidemic-related event types and argument roles. Then we manually annotate a Chinese COVID-19 epidemic dataset, named COVID-19 Case Report (CCR). We also propose a novel hierarchical EE architecture, named multi-model fusion-based hierarchical event extraction (MFHEE). In MFHEE, we introduce a multi-model fusion strategy to tackle the issue of recognition bias of previous EE models. The experimental results on CCR dataset show that our method can effectively extract epidemic events and outperforms other baselines on this dataset. The comparative experiments results on other generic datasets show that our method has good scalability and portability. The ablation studies also show that the proposed hierarchical structure and multi-model fusion strategy contribute to the precision of our model.

Abstract Named Entity Recognition (NER) has so far evolved from the traditional flat NER to overlapped and discontinuous NER. They have mostly been solved separately, with only several exceptions that concurrently tackle three tasks with a single model. The current best-performing method formalizes the unified NER as word-word relation classification, which barely focuses on mention content learning and fails to detect entity mentions comprising a single word. In this paper, we propose a two-stage span-based framework with templates, namely, T2-NER, to resolve the unified NER task. The first stage is to extract entity spans, where flat and overlapped entities can be recognized. The second stage is to classify over all entity span pairs, where discontinuous entities can be recognized. Finally, multi-task learning is used to jointly train two stages. To improve the efficiency of span-based model, we design grouped templates and typed templates for two stages to realize batch computations. We also apply an adjacent packing strategy and a latter packing strategy to model discriminative boundary information and learn better span (pair) representation. Moreover, we introduce the syntax information to enhance our span representation. We perform extensive experiments on eight benchmark datasets for flat, overlapped, and discontinuous NER, where our model beats all the current competitive baselines, obtaining the best performance of unified NER.

Content representation is a fundamental task in information retrieval. Representation learning is aimed at capturing features of an information object in a low-dimensional space. Most research on representation learning for heterogeneous information networks (HINs) focuses on static HINs. In practice, however, networks are dynamic and subject to constant change. In this article, we propose a novel and scalable representation learning model, M-DHIN , to explore the evolution of a dynamic HIN. We regard a dynamic HIN as a series of snapshots with different time stamps. We first use a static embedding method to learn the initial embeddings of a dynamic HIN at the first time stamp. We describe the features of the initial HIN via metagraphs, which retains more structural and semantic information than traditional path-oriented static models. We also adopt a complex embedding scheme to better distinguish between symmetric and asymmetric metagraphs. Unlike traditional models that process an entire network at each time stamp, we build a so-called change dataset that only includes nodes involved in a triadic closure or opening process, as well as newly added or deleted nodes. Then, we utilize the above metagraph-based mechanism to train on the change dataset. As a result of this setup, M-DHIN is scalable to large dynamic HINs since it only needs to model the entire HIN once while only the changed parts need to be processed over time. Existing dynamic embedding models only express the existing snapshots and cannot predict the future network structure. To equip M-DHIN with this ability, we introduce an LSTM-based deep autoencoder model that processes the evolution of the graph via an LSTM encoder and outputs the predicted graph. Finally, we evaluate the proposed model, M-DHIN , on real-life datasets and demonstrate that it significantly and consistently outperforms state-of-the-art models.

Most existing work on event extraction (EE) either follows a pipelined manner or uses a joint structure but is pipelined in essence.As a result, these efforts fail to utilize information interactions among event triggers, event arguments, and argument roles, which causes information redundancy.In view of this, we propose to exploit the role information of the arguments in an event and devise a Hierarchical Policy Network (HPNet) to perform joint EE.The whole EE process is fulfilled through a two-level hierarchical structure consisting of two policy networks for event detection and argument detection.The deep information interactions among the subtasks are realized, and it is more natural to deal with multiple events issue.Extensive experiments on ACE2005 and TAC2015 demonstrate the superiority of HPNet, leading to state-of-the-art performance and is more powerful for sentences with multiple events.

To represent a complex network, paths are often employed for capturing relationships among node: random walks for (homogeneous) networks and metapaths for heterogeneous information networks (HINs). However, there is structural (and possibly semantic) information loss when using paths to represent the subgraph between two nodes, since a path is a linear structure and a subgraph often is not. Can we find a better alternative for network embeddings? We offer a novel mechanism to capture the features of HIN nodes via metagraphs, which retains more structural and semantic information than path-oriented models. Inspired by developments in knowledge graph embedding, we propose to construct HIN triplets using nodes and metagraphs between them. Metagraphs are generated by harnessing the GRAMI algorithm, which enumerates frequent subgraph patterns in a HIN. Subsequently, the Hadamard function is applied to encode relationships between nodes and metagraphs, and the probability whether a HIN triplet can be evaluated. Further, to better distinguish between symmetric and asymmetric cases of metagraphs, we introduce a complex embedding scheme that is able to precisely express fine-grained features of HIN nodes. We evaluate the proposed model, M-HIN, on real-life datasets and demonstrate that it significantly and consistently outperforms state-of-the-art models.