Research On Sentiment Analysis Of Hotel Reviews Based On Dictionary And Machine Learning

With the development of the current Internet technology,a large number of Internet platforms have been born and brought a huge amount of Internet information.User reviews on the Internet platform are now become an important part of Internet information.The expressions of these network users' comments are different from the expression of general objective documents.These comments hide a lot of subjective personal emotional information.The information hidden in these comments is very high for Internet merchants,media and government departments.However,mining these hidden information only by manual methods requires high costs and a lot of time,so how to mine and classify the unstructured comment text on the Internet has become a popular applied research direction in natural language processing related research one.This article mainly conducts sentiment analysis on the text information of user reviews in the hotel field,and excavates users' attitudes,opinions and suggestions on hotel services,and provides suggestions for hotels to improve service quality.The main work of this article is as follows:1.Based on the basic sentiment dictionary,the template rule method and the point mutual information method are combined to build a sentiment dictionary in the hotel field which enriches the sentiment dictionary in the hotel field.2.Use the combination of the sentiment dictionary in the hotel field proposed by{1}and the dependency syntactic structure to make fine-grained sentiment discrimination for hotel reviews.According to the dependency syntax and analyze the dependency relationship between the words in the sentence,find the use of words such as adverbs and negative words,the weight of the emotional influence factor,establish the text evaluation object and calculate the emotional tendency value for each evaluation object in the text and seek and judge the strength of the hotel's emotional tendency by the sum of the values.3.In sentiment analysis,the semantic connection between words and the importance of different words are easily overlooked so this paper proposes a weighted word vector text representation method.This method first uses Word2 Vec to train the document word vector,then calculates the weight of the vocabulary in the document according to the TFIDF algorithm,and finally weights the feature weight with the Word2 Vec word vector to form a weighted word vector text representation.4.Use machine learning methods to conduct sentiment analysis on hotel reviews.Use the sentiment tendency analysis method proposed by{2}to construct the training set of the classifier,use{3}weighted word vector text representation method to generate the word vector,train the classifier by machine learning method and use the test set to classify its classification effect verification.The experimental results show that the method of combining sentiment dictionary and machine learning has improved the classification effect.