OntoED: Low-resource Event Detection with Ontology Embedding

Abstract

Event Detection (ED) aims to identify event trigger words from a given text and classify it into an event type. Most of current methods to ED rely heavily on training instances, and almost ignore the correlation of event types. Hence, they tend to suffer from data scarcity and fail to handle new unseen event types. To address these problems, we formulate ED as a process of event ontology population: linking event instances to pre-defined event types in event ontology, and propose a novel ED framework entitled OntoED with ontology embedding. We enrich event ontology with linkages among event types, and further induce more event-event correlations. Based on the event ontology, OntoED can leverage and propagate correlation knowledge, particularly from datarich to data-poor event types. Furthermore, OntoED can be applied to new unseen event types, by establishing linkages to existing ones. Experiments indicate that OntoED is more predominant and robust than previous approaches to ED, especially in data-scarce scenarios.

Ontology-enhanced Prompt-tuning for Few-shot Learning

“…To integrate the ontology knowledge, [24] propose to tackle the zero-shot event detection problem by mapping each event mentioned to a specific type in a target event ontology. [10] propose an event detection framework based on ontology embedding with event correlations, which interoperates symbolic rules with popular deep neural networks. [18] propose a novel ZSL framework called OntoZSL which not only enhances the class semantics with an ontological schema but also employs an ontology-based generative model to synthesize training samples for unseen classes.…”

“…These drawbacks lead to the research of an important technique, few-shot learning (FSL), which can significantly improve the learning capabilities of machine intelligence and practical adaptive applications by accessing only a small number of labeled examples. Over the past few years, FSL has been introduced in a wide range of machine learning tasks, such as relation extraction [6,32,62,67,68], event extraction [10,47] and knowledge graph completion [77].…”

AliCG: Fine-grained and Evolvable Conceptual Graph Construction for Semantic Search at Alibaba

“…Specifically, we first collect one month's instances (queries) clicked tag pairs and then aggregate the of queries with the same clicked concept tag? 8 . In other words, we estimate the number of queries for each concept tag in the last month.…”

“…Then we leverage expert rules to define the taxonomy between level1 and level2. We further determine the relation 8 The search logs contain the history of user clicked concept tags. between level2 and level3 via probabilistic inference.…”

LOGEN: Few-shot Logical Knowledge-Conditioned Text Generation with Self-training

“…To address this issue, it is intuitive to leverage external logical knowledge for better generation [35], [36], [37], [38], [39]. [34] firstly propose text generation using logical inferences from a table.…”