Business Scene Dialogue (BSD) corpus

Business Scene Dialogue (BSD) 数据集是由东京大学创建的一个日英双语商业对话平行语料库，包含955个场景和30,000个平行句子。该数据集通过精心挑选商业场景，编写单语对话脚本，并由专业翻译人员翻译成另一种语言，以确保对话的自然流畅。BSD数据集特别注重视场景信息的标注，以帮助解决对话翻译中的上下文依赖问题。该数据集主要用于机器翻译系统的训练和评估，特别是在对话和会话翻译领域，旨在提高机器翻译的质量和准确性。Business Scene Dialogue (BSD) Dataset is a Japanese-English bilingual parallel business dialogue corpus created by The University of Tokyo, which contains 955 scenarios and 30,000 parallel sentence pairs. This dataset is constructed by carefully selecting business scenarios, drafting monolingual dialogue scripts, and then having professional translators translate the scripts into the other language to ensure the naturalness and fluency of the dialogues. The BSD dataset places special emphasis on scenario information annotation to address the context dependency problem in dialogue translation. It is mainly used for training and evaluating machine translation systems, especially in the field of dialogue and conversational translation, aiming to improve the quality and accuracy of machine translation.

搜集汇总

背景与挑战

背景概述

Business Scene Dialogue (BSD) corpus是一个精心设计的日英双语商务对话数据集，包含多种商务场景的对话文本，数据以结构化json格式存储，并附带丰富的元数据信息。该数据集适用于机器翻译等自然语言处理任务研究，采用CC BY-NC-SA非商业共享许可协议。