The dynamic nature of real-world information necessitates knowledge editing (KE) in large language models (LLMs). The edited knowledge should propagate and facilitate the deduction of new information based on existing model knowledge. We term the existing related knowledge in LLM serving as the origination of knowledge propagation as ”deduction anchors”. However, current KE approaches, which only operate on (subject, relation, object) triple. We both theoretically and empirically observe that this simplified setting often leads to uncertainty when determining the deduction anchors, causing low confidence in their answers. To mitigate this issue, we propose a novel task of event-based knowledge editing that pairs facts with event descriptions. This task manifests not only a closer simulation of real-world editing scenarios but also a more logically sound setting, implicitly defining the deduction anchor and enabling LLMs to propagate knowledge confidently. We curate a new benchmark dataset Evedit derived from the CounterFact dataset and validate its superiority in improving model confidence. Moreover, while we observe that the event-based setting is significantly challenging for existing approaches, we propose a novel approach Self-Edit that showcases stronger performance, achieving 55.6% consistency improvement while maintaining the naturalness of generation.

Extensive previous research has focused on post-training knowledge editing (KE) for language models (LMs) to ensure that knowledge remains accurate and up-to-date. One desired property and open question in KE is to let edited LMs correctly handle ripple effects, where LM is expected to answer its logically related knowledge accurately. In this paper, we answer the question of why most KE methods still create messy ripple effects. We conduct extensive analysis and identify a salient indicator, GradSim, that effectively reveals when and why updated knowledge ripples in LMs. GradSim is computed by the cosine similarity between gradients of the original fact and its related knowledge. We observe a strong positive correlation between ripple effect performance and GradSim across different LMs, KE methods, and evaluation metrics. Further investigations into three counter-intuitive failure cases (Negation, Over-Ripple, Multi-Lingual) of ripple effects demonstrate that these failures are often associated with very low GradSim. This finding validates that GradSim is an effective indicator of when knowledge ripples in LMs.

Large Language Models (LLMs) struggle with providing current information due to the outdated pre-training data. Existing methods for updating LLMs, such as knowledge editing and continual fine-tuning, have significant drawbacks in generalizability of new information and the requirements on structured updating corpus. We identify the core challenge behind these drawbacks: the LM-logical discrepancy featuring the difference between language modeling probabilities and logical probabilities. To evaluate and address the core challenge, we propose a new task formulation of the information updating task that only requires the provision of an unstructured updating corpus and evaluates the performance of information updating on the generalizability to question-answer pairs pertaining to the updating information.We further propose a novel and effective pipeline approach for the task, highlighting a self-prompting-based question-answer generation process and a associative distillation methods to bridge the LM-logical discrepancy.We develop two datasets for evaluation, one sourced from news articles published in March and April 2023, and the other from the Natural Questions benchmark.Experimental results demonstrate the superiority of our approach, significantly increasing the factual consistency score (on a scale from 0 to 1) by up to 0.16. Furthermore, our method effectively mitigates forgetting utilizing a compact replay buffer with only 2.3% of the training tokens.

Recent research in cross-lingual learning has found that combining large-scale pretrained multilingual language models with machine translation can yield good performance. We explore this idea for cross-lingual event extraction with a new model architecture that jointly encodes a source language input sentence with its translation to the target language during training, and takes a target language sentence with its translation back to the source language as input during evaluation. However, we observe significant representational gap between the native source language texts during training and the texts translated into source language during evaluation, as well as the texts translated into target language during training and the native target language texts during evaluation. This representational gap undermines the effectiveness of cross-lingual transfer learning for event extraction with machine-translated data. In order to mitigate this problem, we propose an adversarial training framework that encourages the language model to produce more similar representations for the translated text and the native text. To be specific, we train the language model such that its hidden representations are able to fool a jointly trained discriminator that distinguishes translated texts’ representations from native texts’ representations. We conduct experiments on cross-lingual for event extraction across three languages. Results demonstrate that our proposed adversarial training can effectively incorporate machine translation to improve event extraction, while simply adding machine-translated data yields unstable performance due to the representational gap.

Recent research has shown that large-scale pretrained language models, specifically transformers, tend to exhibit issues relating to racism, sexism, religion bias, and toxicity in general. Unfortunately, these pretrained language models are used almost universally in downstream tasks, and natural language processing is often applied to make real-world predictions. Thus, debiasing these language models as early in development as possible is increasingly crucial for preventing unintentional harms caused by natural language systems. To this end, we propose a new technique called partitioned contrastive gradient unlearning (PCGU), a gray-box method for debiasing pretrained masked language models. PCGU aims to optimize only the weights that contribute most to a specific domain of bias, doing so by computing a first-order approximation based on the gradients of contrastive sentence pairs. Our experiments show that PCGU is both low-cost and seems particularly effective at pinpointing the sources of implicit social bias in large pretrained transformers. Although we train using PCGU in the gender-profession domain only, we find that doing so can also partially mitigate bias across other domains. All code for our implementation and experiments can be found at https://github.com/CharlesYu2000/PCGU-UnlearningBias.

The recent explosion of performance of large language models (LLMs) has changed the field of Natural Language Processing (NLP) more abruptly and seismically than any other shift in the field’s 80 year history. This has resulted in concerns that the field will become homogenized and resource-intensive. This new status quo has put many academic researchers, especially PhD students, at a disadvantage. This paper aims to define a new NLP playground by proposing 20+ PhD-dissertation-worthy research directions, covering theoretical analysis, new and challenging problems, learning paradigms and interdisciplinary applications.

The distribution of knowledge elements such as entity types and event types is long-tailed in natural language. Hence information extraction datasets naturally conform long-tailed distribution. Although imbalanced datasets can teach the model about the useful real-world bias, deep learning models may learn features not generalizable to rare or unseen expressions of entities or events during evaluation, especially for rare types without sufficient training instances. Existing approaches for the long-tailed learning problem seek to manipulate the training data by re-balancing, augmentation or introducing extra prior knowledge. In comparison, we propose to handle the generalization challenge by making the evaluation instances closer to the frequent training cases. We design a new transformation module that transforms infrequent candidate mention representation during evaluation with the average mention representation in the training dataset. Experimental results on classic benchmarks on three entity or event extraction datasets demonstrates the effectiveness of our framework.

We introduce RESIN-11, a new schema-guided event extraction&prediction framework that can be applied to a large variety of newsworthy scenarios. The framework consists of two parts: (1) an open-domain end-to-end multimedia multilingual information extraction system with weak-supervision and zero-shot learningbased techniques. (2) schema matching and schema-guided event prediction based on our curated schema library. We build a demo website based on our dockerized system and schema library publicly available for installation (https://github.com/RESIN-KAIROS/RESIN-11). We also include a video demonstrating the system.

To tackle the challenge of accurate and timely communication regarding the COVID-19 pandemic, we present a COVID-19 Claim Radar to automatically extract supporting and refuting claims on a daily basis. We provide a comprehensive structured view of claims, including rich claim attributes (such as claimers and claimer affiliations) and associated knowledge elements as claim semantics (such as events, relations and entities), enabling users to explore equivalent, refuting, or supporting claims with structural evidence, such as shared claimers, similar centroid events and arguments. In order to consolidate claim structures at the corpus-level, we leverage Wikidata as the hub to merge coreferential knowledge elements. The system automatically provides users a comprehensive exposure to COVID-19 related claims, their importance, and their interconnections. The system is publicly available at GitHub and DockerHub, with complete documentation.

Supervised event extraction models require a substantial amount of training data to perform well. However, event annotation requires a lot of human effort and costs much time, which limits the application of existing supervised approaches to new event types. In order to reduce manual labor and shorten the time to build an event extraction system for an arbitrary event ontology, we present a new framework to train such systems much more efficiently without large annotations. Our event trigger labeling model uses a weak supervision approach, which only requires a set of keywords, a small number of examples and an unlabeled corpus, on which our approach automatically collects weakly supervised annotations. Our argument role labeling component performs zero-shot learning, which only requires the names of the argument roles of new event types. The source codes of our event trigger detection1 and event argument extraction2 models are publicly available for research purposes. We also release a dockerized system connecting the two models into an unified event extraction pipeline.

Traditional supervised Information Extraction (IE) methods can extract structured knowledge elements from unstructured data, but it is limited to a pre-defined target ontology. In reality, the ontology of interest may change over time, adding emergent new types or more fine-grained subtypes. We propose a new lifelong learning framework to address this challenge. We focus on lifelong event detection as an exemplar case and propose a new problem formulation that is also generalizable to other IE tasks. In event detection and more general IE tasks, rich correlations or semantic relatedness exist among hierarchical knowledge element types. In our proposed framework, knowledge is being transferred between learned old event types and new event types. Specifically, we update old knowledge with new event types’ mentions using a self-training loss. In addition, we aggregate old event types’ representations based on their similarities with new event types to initialize the new event types’ representations. Experimental results show that our framework outperforms competitive baselines with a 5.1% absolute gain in the F1 score. Moreover, our proposed framework can boost the F1 score for over 30% absolute gain on some new long-tail rare event types with few training instances. Our knowledge transfer module improves performance on both learned event types and new event types under the lifelong learning setting, showing that it helps consolidate old knowledge and improve novel knowledge acquisition.

Distantly supervised relation extraction employs existing knowledge graphs to automatically collect training data. While distant supervision is effective to scale relation extraction up to large-scale corpora, it inevitably suffers from the wrong labeling problem. Many efforts have been devoted to identifying valid instances from noisy data. However, most existing methods handle each relation in isolation, regardless of rich semantic correlations located in relation hierarchies. In this paper, we aim to incorporate the hierarchical information of relations for distantly supervised relation extraction and propose a novel hierarchical attention scheme. The multiple layers of our hierarchical attention scheme provide coarse-to-fine granularity to better identify valid instances, which is especially effective for extracting those long-tail relations. The experimental results on a large-scale benchmark dataset demonstrate that our models are capable of modeling the hierarchical information of relations and significantly outperform other baselines. The source code of this paper can be obtained from https://github.com/thunlp/HNRE.

We present a Few-Shot Relation Classification Dataset (dataset), consisting of 70, 000 sentences on 100 relations derived from Wikipedia and annotated by crowdworkers. The relation of each sentence is first recognized by distant supervision methods, and then filtered by crowdworkers. We adapt the most recent state-of-the-art few-shot learning methods for relation classification and conduct thorough evaluation of these methods. Empirical results show that even the most competitive few-shot learning models struggle on this task, especially as compared with humans. We also show that a range of different reasoning skills are needed to solve our task. These results indicate that few-shot relation classification remains an open problem and still requires further research. Our detailed analysis points multiple directions for future research.