深度学习驱动的翻译作品接受效果研究——以刘殿爵《论语》英译本读者评论情感分析为例
A Deep-Learning-Driven Study on the Reception of Translated Works:A Case Study of Sentiment Analysis for Readers'Reviews on The Analects Translated by D.C.Lau
陶友兰 1胡伊伊1
作者信息
- 1. 复旦大学外国语言文学学院,上海 200433
- 折叠
摘要
本研究改进深度金字塔卷积神经网络,选用3.178亿词书籍评论数据集(Amazon Review Polarity)训练深度学习模型,将读者评论情感分析分类准确率提高到94.69％,为译著海外接受效果提供更为科学的研究设计与路径.以Amazon网站上D.C.Lau译本The Analects的读者评论为研究对象,发现基于新模型计算出的读者评论情感得分比评星等级更能精准反映读者的真实态度;该译著整体接受程度具体化为积极评论达64.22％,中立评论19.72％,消极评论16.06％;考察译著接受效果不能只看评星等级,应该包括译著思想内容、翻译整体质量、副文本以及出版质量等多个维度.
Abstract
This study improves the deep pyramid convolutional neural network by using a dataset of 317.8 million words from book reviews(Amazon Review Polarity)to train the deep learning model.The enhancement increased the classification accuracy of sentiment analysis for reader reviews to 94.69％,providing a more scientific research design and approach for assessing the overseas reception of translated works.Using the reader reviews of D.C.Lau's translation of The Analects on Amazon as a case study,we find that sentiment scores calculated by the new model can more accurately reflect readers'true attitudes compared to star ratings.The overall reception of this translation was specified as 64.22％positive reviews,19.72％neutral reviews,and 16.06％negative reviews.Evaluating the reception of translated works should not rely solely on star ratings but should include multiple dimensions such as the content of the translation,overall translation quality,para-texts,and publication quality.
关键词
数智人文/深度学习/《论语》英译本/网络读者情感分析/主题建模Key words
Digital-Intelligent Humanities(DlH)/deep learning/The Analects/sentiment analysis for online reviews/topic modeling引用本文复制引用
出版年
2024
NETL
NSTL
万方数据