多方交谈中的发言者特征建模（英文版）

英文标题：Speaker Profiling in Multiparty Conversations中文摘要：本文通过三个子任务，即人设发现、人设类型识别和人设价值提取，探索了嵌入式对话系统中的说话者画像的任务，并引入了一套名为 SPOT 的神经模型来验证了 SPICE 数据集的可行性。英文摘要：In conversational settings, individuals exhibit unique behaviors, rendering aone-size-fits-all approach insufficient for generating responses by dialogueagents. Although past studies have aimed to create personalized dialogue agentsusing speaker persona information, they have relied on the assumption that thespeaker's persona is already provided. Ho