ArgGen: Prompting Text Generation Models for Document-Level Event-Argument Aggregation

Most of the existing discourse-level Information Extraction tasks have been modeled to beextractive in nature. However, we argue thatextracting information from larger bodies ofdiscourse-like documents requires more naturallanguage understanding and reasoning capabilities. In our work, we propose the novel taskof document-level event argument aggregationwhich generates consolidated event-argumentsat a document-level with minimal loss of information. More specifically, we focus on generating precise document-level information framesin a multilingual setting using prompt-basedmethods. In this paper, we show the effectiveness of prompt-based text generation approachto generate document-level argument spans ina low-resource and zero-shot setting. We alsorelease the first of its kind multilingual eventargument aggregation dataset that can be leveraged in other related multilingual text generation tasks as well: https://github.com/DebanjanaKar/ArgGen.