摘要
旅游平台上的旅游在线评论是消费者出行旅游的重要参考,然而评分不一致与无效评论问题会导致对旅游目的地的评价失真,干扰消费者出行决策。针对旅游目的地综合评价的问题,首先通过基于负样本生成的深度学习BERT二分类模型等方式清理无效评论数据,然后基于改进的BERT粗粒度情感得分与细粒度多维特征匹配的评分计算方法构建旅游目的地的综合评价模型。结果表明模型在实验数据集上取得了较低的误差,并且在考虑数据有效性和指标多样性的前提下能较好地拟合平台得分数据。因此,在保证训练数据客观的准确的前提下,该方法具有一定实用性和泛化性。
Online reviews on tourism platforms are important references for consumers’ decision-making in travelling. However, the inconsistent ratings and invalid reviews may lead to distortion of the evaluation on travel destinations and interfere with consumers’ travel decisions. For the comprehensive evaluation of tourist destinations, invalid comments are firstly cleaned up by using a BERT binary classification model based on negative samples generation. And then a comprehensive evaluation model of tourist destinations is constructed based on the score calculated by the BERT coarse-grained sentiment level and the fine-grained multi-dimensional feature matching score. The results show that the model has achieved extremely low errors on the experimental data set, and it can fit well the platform score data under the premise of the data validity and indicator diversity. Therefore, when the training data are sufficiently objective and accurate, this method has significant practicability and generalization.
出处
《运筹与模糊学》
2022年第2期157-168,共12页
Operations Research and Fuzziology
