Method And Application Of Aspect Category Sentiment Analysis Oriented Towards Tourism Image-text Data

With the continuous improvement of people’s living standards,an increasing number of individuals choose to travel to different places during their spare time to relax,record local natural,historical,and cultural sights,and share their experiences on various tourism platforms.When describing their sentiments and feelings,people often attach one or multiple relevant images to convey their messages in both verbal and visual forms.To improve the operational efficiency and service quality of the tourism industry,the sentiment analysis technology of image and text data can serve as an important technical support.In this paper,we focus on the sentiment analysis of image and text data in the tourism field,and conduct research on joint aspect attention interaction,domain knowledge integration,and image-to-text technology.Based on this,we design and implement an emergency warning system for scenic spots that is based on the fusion of image and text data.The main research works are outlined as follows:(1)Image and text aspect-category sentiment analysis of joint attention interaction.Aiming at the lack of data sets for image and text sentiment analysis in tourism field,we construct a Shanxi tourism image-text sentiment analysis dataset,which fills the gap in sentiment analysis datasets in the tourism field.Image-text data has both inconsistency and correlation,but existing image-text sentiment analysis methods only consider the interaction between image and text modalities,ignoring the characterization of these two characteristics.Therefore,this paper proposes a joint aspect attention interaction network model for imagetext sentiment analysis.This method considers both the inconsistency and correlation of image-text data.We design a multi-level fusion strategy for aspect information and imagetext information and a strategy that removes text and images that are irrelevant to a given aspect,enhancing the sentiment representation of the given aspect-classified image-text data.Experimental results show that the proposed model improves the performance of sentiment classification for image-text aspects on both the public dataset Multi-ZOL and the Shanxi tourism image-text sentiment analysis dataset.(2)Aspect-category sentiment analysis based on domain knowledge fusion and image-generated text technologyBased on this,we propose an aspect-category sentiment analysis method based on domain knowledge fusion and image-generated text techniques,which combines domain knowledge fusion and image-generated text techniques.This method matches the scenic spots in the comments with the ones in the knowledge base,and then obtains vector representations of the scenic spot descriptions by using Bert.Meanwhile,the Transformer in the DETR model is used for non-autoregressive text generation,which generates image descriptions and enriches the input text information for the model.Through experiments on the Shanxi tourism image-text sentiment analysis dataset,the experimental results show the feasibility and effectiveness of the proposed DG-ACSA model.(3)Emergency scenic warning system based on image-text data content analysisThis paper employs aspect-category sentiment analysis technology based on domain knowledge fusion and image-generated text techniques,combined with image processing and intelligent question-and-answer technologies,to assess the emergency response and safety risks of scenic areas in real-time.Based on these technologies,we have developed an emergency warning system based on content analysis of image-text data.This system provides real-time monitoring and early warning functions during the emergency management process of scenic areas,which can better ensure the safety and management.