This paper presents a comprehensive survey of event causality identification (ECI), an essential task in natural language processing (NLP) that automatically detects causal relationships between events in text. We define key concepts, formalize the ECI problem, and present a standard evaluation protocol. We develop a classification framework that categorizes ECI models into two main tasks: sentence-level event causality identification (SECI) and document-level event causality identification (DECI). For SECI, we review models that utilize feature pattern-based matching, machine learning classifiers, deep semantic encoding, prompt-based fine-tuning, and causal knowledge pretraining and data augmentation strategies. For DECI, we focus on approaches that leverage deep semantic encoding, event graph inference, and prompt-based fine-tuning. We pay particular attention to recent advances in multilingual and cross-lingual ECI and zero-shot ECI using large-scale language models (LLMs). We analyze the strengths, limitations, and challenges associated with each approach, and perform extensive quantitative evaluations on four benchmark datasets to rigorously evaluate the performance of various ECI models. Finally, we discuss future research directions and highlight opportunities to further advance this field.
