Evaluating service quality of express logistics service based on online reviews using LDA-LSTM

Logistics services quality affects the performance and competitiveness of express logistics companies. Conventional evaluation methods of service quality are based on constructing a rating index system and then comprehensively evaluating the associated questionnaires and interviews. Such methods are often subjective, time-consuming, and include a limited sample size (data). On the other hand, customers' opinions concerning logistics services can now be deciphered using online customer reviews. Therefore, this paper proposes a method combining the models of latent Dirichlet allocation and long-short term memory (LDA-LSTM) to overcome the limitations of conventional evaluation methods. The main contributions include four aspects. First, the LDA-LSTM model can extract comprehensive aspects and opinions to develop evaluation indexes and scores for logistics service quality, and the reviews of a logistics firm can be examined to verify the effectiveness of this method. Second, the LDA-LSTM model can handle a situation in which several aspects and opinions express one topic or sentiment, and it outperforms the joint sentiment topic model (JST) and naive Bayes classification (NB). Third, positive and negative ratings can reflect a firm's overall service quality, with an excellent rating highlighting the best service quality, which can provide a multi-dimensional evaluation. Fourth, we also identify the indicators of logistics service quality on which customers focus, and we compare the service quality among express enterprises in China.