Study On The Influence Of Text Features Of Medical Crowdsourcing Schemes On The Selection Of Optimal Scheme

People’s demand for health resources is rising with the improvement of living standards.However,the relatively backward economy in China restricts people’s demand for medical resources.How to allocate medical resources effectively and reasonably in the current situation is a difficult problem that needs to be solved urgently.The development of computer technology,especially in online medical community,medical crowdsourcing and other new technologies,provides new ideas and methods to solve this problem.Medical crowdsourcing brings together the power of Internet users worldwide to solve problems that are difficult to solve in traditional ways in a convenient way.It enhances people’s ability to access medical information and customize medical programs.With more and more people participating,the platform is also full of a large number of crowd-sourcing information,requiring patients to select a number of options.Most of the schemes in medical crowdsourcing are given in the form of words.It is very important for crowdsourcing publishers to study the characteristics of these texts to choose the best scheme.In this study,we selected five text features,namely,the similarity between the scheme and the question text,the attitude of the scheme text,the readability of the scheme text,the professionalism of the scheme text and the length of the scheme text,to explore the factors affecting the patients’ choice of the best scheme from the perspective of the text characteristics.Firstly,through literature collection,this study combs and summarizes the textual features that influence the choice of alternatives in existing studies.Then the research hypothesis is put forward according to the characteristics of medical field.In this study,"micro-medicine" is selected as the research platform,and the readability of the scheme text is calculated by CRITIC method;the cosine similarity between the problem text and the scheme text is calculated by text analysis method;and the emotional orientation value of the scheme text is calculated based on the emotional dictionary.Then the logistic regression model is established.In order to verify the validity of the model,this study tested the robustness of the computational methods of text sentiment orientation value and text length.Supervised machine learning method is used to learn the labeled corpus data and build a classification model.Random Forest model with the best classification effect is selected by calculating F-Score and drawing ROC curve.The model is used to recalculate the text sentiment orientation.At the same time,the number of punctuation marks is used as the text length of the scheme,and the logistic regression model is re-established.The results of robustness test prove that changing the way variables are calculated does not affect the results of the model.The results show that the similarity between the scheme and the problem,the readability of the scheme text and the professionalism of the scheme text have a positive impact on patients’ choice of the best scheme.Attitude and length of protocol text had no significant positive effect on patient’s choice behavior.One possible explanation is that patients are well aware that "good medicine is good for the mouth",and they pay more attention to the professionalism of doctors than to the attitude of doctors.Long text may increase the reading and understanding barriers of patients,thus affecting patients’ choice of programs.The results of this study will provide guidance for doctors’ behavior.The results can help doctors to provide the answers needed by patients,increase the probability of their being selected by patients,and promote the continuous participation of doctors on the platform.