阅读说明：本技术 一种基于网络数据的在线医疗患者满意度测量方法 (Online medical patient satisfaction measuring method based on network data ) 是由 由丽萍 王世兴 于 2021-08-16 设计创作，主要内容包括：本发明属于医疗服务评价领域,公开了一种基于网络数据的在线医疗患者满意度测量方法,包括以下步骤：获取在线医疗评论文本；S2、对在线医疗评论文本进行情感语义分析,得到情感语义标注文本；S3、识别评论文本对应的患者满意度评价属性；S4、确定评价文本中各个评价属性的评价值；S5、将医疗评论的情感分析结果转化为向量,以医生为评价对象,得到各个评价对象的评价值向量；基于TF-IDF计算各个评价属性的权值；S6、根据权值和评价值向量,计算各个评价对象的满意度值。本发明以语义深度的情感分析提取患者评论文本中的评价属性体系和评估值；以优化的多评价属性决策法对评论数据进行计算,得到综合测量结果,提高了医疗服务评价的效率和准确性。(The invention belongs to the field of medical service evaluation, and discloses an online medical patient satisfaction measuring method based on network data, which comprises the following steps: acquiring an online medical comment text; s2, performing sentiment semantic analysis on the online medical comment text to obtain a sentiment semantic annotation text; s3, identifying the patient satisfaction evaluation attribute corresponding to the comment text; s4, determining the evaluation value of each evaluation attribute in the evaluation text; s5, converting emotion analysis results of the medical comments into vectors, and obtaining evaluation value vectors of all the evaluation objects by taking doctors as the evaluation objects; calculating the weight of each evaluation attribute based on TF-IDF; and S6, calculating the satisfaction value of each evaluation object according to the weight and the evaluation value vector. The evaluation attribute system and the evaluation value in the comment text of the patient are extracted by emotion analysis of semantic depth; the comment data are calculated by an optimized multi-evaluation attribute decision method to obtain a comprehensive measurement result, so that the efficiency and the accuracy of medical service evaluation are improved.)

1. An online medical patient satisfaction measurement method based on network data, characterized by comprising the following steps:

s1, acquiring an online medical comment text;

s2, performing emotion semantic analysis on the online medical comment text to obtain an emotion semantic labeling text, wherein semantic classes, semantic roles and emotion words are labeled in the emotion voice labeling text;

s3, matching a patient satisfaction evaluation attribute database according to the semantic type identifier and the semantic role identifier in the emotion semantic annotation text, and identifying a patient satisfaction evaluation attribute corresponding to the comment text;

s4, determining evaluation values of all evaluation attributes in the evaluation text according to the emotion word identifications in the emotion semantic annotation text;

s5, converting the emotion analysis results of the medical comments into vectors, wherein the vectors comprise evaluation attributes and evaluation values of the evaluation attributes, which are related to each patient comment, and the evaluation value vectors of the evaluation objects are obtained by taking doctors as the evaluation objects; calculating the weight of each evaluation attribute based on TF-IDF;

and S6, calculating the satisfaction value of each evaluation object according to the weight and the evaluation value vector.

2. The method for measuring satisfaction of medical patients on line based on network data as claimed in claim 1, wherein in step S2, when performing sentiment semantic analysis on the text of the online medical comment, a semantic analysis method combining a semantic dictionary and a pattern matching rule is used to identify sentiment information in the medical online comment, and the online comment is converted into structured information consisting of sentiment semantic classes and semantic roles;

the semantic dictionary is constructed according to the following method: firstly, crawling a plurality of patient comment texts from an online medical application platform; then carrying out word segmentation and part-of-speech tagging pretreatment on the initial word list, extracting verbs and adjectives in the initial word list, and storing the verbs and the adjectives as the initial word list; classifying words expressing the similar emotion category in a word list into the same semantic class, adding definitions and descriptions of the semantic class, wherein the description content comprises semantic class names, all words of the class and the emotion value of each word; the annotation of the emotion value is determined according to the emotion polarity and strength of the words, and the value range is (0-1.0).

3. The method for measuring satisfaction of medical patient on line based on network data as claimed in claim 1, wherein said step S3 further comprises the step of constructing a database of patient satisfaction evaluation attributes, said database of patient satisfaction evaluation attributes is constructed by: on-line comment data based on semantic annotation, a semantic class-semantic role pair is extracted, and is clustered by adopting a hierarchical clustering method to form a patient satisfaction evaluation attribute system and is stored in a patient satisfaction evaluation attribute database.

4. The method for measuring satisfaction of medical patients on line based on network data as claimed in claim 1, wherein in the step S3, in the matching process, if a word is not included in the evaluation attribute database, the semantic similarity between the word and the existing word is calculated, and the category to which the word belongs is determined according to the word with the highest similarity; if the database does not record the semantic role or the semantic role is default, only matching the semantic class with the evaluation attribute database; and if the semantic class corresponds to a plurality of evaluation attribute categories in the database, taking the most frequent one as a final evaluation attribute identification result.

5. The method of claim 1, wherein in step S4, the evaluation value of each rating evaluation attribute is determined based on the emotion value of a word in the dictionary according to semantic character and context feature, wherein the context feature mainly includes a level-word modifier and a negative-word modifier, the level-word modifier adjusts the emotion value, and the negative-word is used to negate the emotion value.

6. The method for measuring satisfaction of medical patient on line based on network data as claimed in claim 1, wherein in said step S5, each evaluation object has an evaluation value vector D_i=（f_i1, f_i2,…, f_im) The calculation formula of each element is as follows:

；

wherein D is_iAn evaluation value vector f representing the ith evaluation target_ijA normalized evaluation value of the ith evaluation object on the jth evaluation attribute, A_ijAnd represents the sum of evaluation values of the ith evaluation object on the jth evaluation attribute.

7. The method for measuring satisfaction of medical patient on line based on network data as claimed in claim 1, wherein in said step S5, the weight calculation formula of each evaluation attribute is:

；

wherein the content of the first and second substances,the word frequency representing the jth evaluation attribute,the inverse file frequency representing the jth evaluation attribute.

8. The method of claim 7, wherein the word frequency and inverse document frequency of each evaluation attribute are calculated by the following formula:

，；

wherein n is_j(d) And n_k(d) Respectively representing the occurrence times of the j-th comment evaluation attribute and the k-th comment evaluation attribute in the comment text set D, | D | is the total number of comment pieces,number of comments containing jth evaluation attribute, t_iDenotes the ith comment, d_mAll comments are represented.

9. The method for measuring satisfaction of medical patient on line based on network data as claimed in claim 1, wherein in said step S6, the calculation formula of satisfaction of each evaluated object is:

；

wherein Qi represents the satisfaction of the i-th evaluation object, v represents the second weight, and S represents the group utility value S_iIs a minimum value of, S-represents a population utility value S_iMaximum in (1), R represents the individual regret value R_iMinimum of (d), R-represents the individual regret value R_iI.e.: s^*=minS_i，S^-=maxS_i，R^*=minR_i，R^-=maxR_i；S_iIndividual regret value, R, representing the ith evaluation object_iTo representIndividual regret value of the ith evaluation object;

wherein S is_iAnd R_iThe calculation formula of (2) is as follows:

；

wherein, W_jWeight representing jth evaluation attribute, f_jA and f_j ^—Maximum and minimum values of jth evaluation attribute evaluation values in evaluation value vectors respectively representing evaluation objects, f_ijAnd the evaluation value of the ith evaluation object on the jth evaluation attribute is shown.

Technical Field

The invention belongs to the field of medical service management and evaluation, and particularly relates to an online medical patient satisfaction measuring method based on network data.

Background

Patient Satisfaction (PS) is the extent to which a Patient's actual experience during a course of treatment is met compared to his expectations. International patient satisfaction is considered to be a key component of measuring and reporting the quality of medical services, and various rating scales have been developed, mainly the series of patient satisfaction questionnaires developed by the university of landes, war 1976, the general medical services patient satisfaction questionnaire developed by Baker in the uk 1990, the european general medical services satisfaction questionnaire developed by Grol et al 1999, and the australian victoria inpatient satisfaction monitoring in 2000.

However, the traditional patient satisfaction measurement mainly depends on expert law and questionnaire survey data, and is difficult to adapt to the need of improving patient experience in an online medical environment. Therefore, how to adapt to the technology and data of the new era, fully utilize network information resources, obtain a richer evaluation attribute system, obtain a patient satisfaction measurement result with a good structure and quantification becomes a problem which needs to be solved urgently before managers.

Disclosure of Invention

The invention overcomes the defects of the prior art, and solves the technical problems that: an online medical patient satisfaction measurement method based on network data is provided.

In order to solve the technical problems, the invention adopts the technical scheme that: an online medical patient satisfaction measurement method based on network data, comprising the following steps:

s1, acquiring an online medical comment text;

s2, performing emotion semantic analysis on the online medical comment text to obtain an emotion semantic labeling text, wherein semantic classes, semantic roles and emotion words are labeled in the emotion voice labeling text;

s3, matching a patient satisfaction evaluation attribute database according to the semantic type identifier and the semantic role identifier in the emotion semantic annotation text, and identifying a patient satisfaction evaluation attribute corresponding to the comment text;

s4, determining evaluation values of all evaluation attributes in the evaluation text according to the emotion word identifications in the emotion semantic annotation text;

s5, converting the emotion analysis results of the medical comments into vectors, wherein the vectors comprise evaluation attributes and evaluation values of the evaluation attributes, which are related to each patient comment, and the evaluation value vectors of the evaluation objects are obtained by taking doctors as the evaluation objects; calculating the weight of each evaluation attribute based on TF-IDF;

and S6, calculating the satisfaction value of each evaluation object according to the weight and the evaluation value vector.

In the step S2, when performing sentiment semantic analysis on the online medical comment text, identifying sentiment information in the medical online comment by adopting a semantic analysis method combining a semantic dictionary and a pattern matching rule, and converting the online comment into structured information consisting of sentiment semantic classes and semantic roles;

the semantic dictionary is constructed according to the following method: firstly, crawling a plurality of patient comment texts from an online medical application platform; then carrying out word segmentation and part-of-speech tagging pretreatment on the initial word list, extracting verbs and adjectives in the initial word list, and storing the verbs and the adjectives as the initial word list; classifying words expressing the similar emotion category in a word list into the same semantic class, adding definitions and descriptions of the semantic class, wherein the description content comprises semantic class names, all words of the class and the emotion value of each word; the annotation of the emotion value is determined according to the emotion polarity and strength of the words, and the value range is (0-1.0).

In step S3, the method further includes a step of constructing a patient satisfaction evaluation attribute database, where the method of constructing the patient satisfaction evaluation attribute database includes: on-line comment data based on semantic annotation, a semantic class-semantic role pair is extracted, and is clustered by adopting a hierarchical clustering method to form a patient satisfaction evaluation attribute system and is stored in a patient satisfaction evaluation attribute database.

In the step S3, in the matching process, if a word is not included in the evaluation attribute database, calculating semantic similarity between the word and an existing word, and determining the category of the word according to the word with the highest similarity; if the database does not record the semantic role or the semantic role is default, only matching the semantic class with the evaluation attribute database; and if the semantic class corresponds to a plurality of evaluation attribute categories in the database, taking the most frequent one as a final evaluation attribute identification result.

In step S4, the evaluation value of each rating evaluation attribute is determined based on the emotion value of the word in the dictionary and based on the semantic character and the context feature, where the context feature mainly includes a degree co-word modifier and a negation word modifier, the degree co-word modifier adjusts the emotion value, and the negation word is used to negate the emotion value.

In step S5, the evaluation value vector D for each evaluation target_i＝(f_i1,f_i2,…,f_im) The calculation formula of each element is as follows:

wherein D is_iAn evaluation value vector f representing the ith evaluation target_ijA normalized evaluation value of the ith evaluation object on the jth evaluation attribute, A_ijAnd represents the sum of evaluation values of the ith evaluation object on the jth evaluation attribute.

In step S5, the weight calculation formula of each evaluation attribute is:

W_j＝TF_j×IDF_j；

wherein, TF_jWord frequency, IDF, representing the jth evaluation attribute_jThe inverse file frequency representing the jth evaluation attribute.

The calculation formula of the word frequency and the reverse file frequency of each evaluation attribute is as follows:

wherein n is_j(d) And n_k(d) The number of occurrences of the j-th and k-th comment evaluation attributes in the comment text set D is represented, where | D | is the total number of comment pieces, | { j: t is t_i∈d_mJ is the number of comments containing the jth rating attribute, t_iDenotes the ith comment, d_mAll comments are represented.

In step S6, the calculation formula of each evaluation target satisfaction is:

wherein Qi represents the satisfaction of the i-th evaluation object, v represents the second weight, and S represents the group utility value S_iIs a minimum value of, S-represents a population utility value S_iMaximum in (1), R represents the individual regret value R_iMinimum of (d), R-represents the individual regret value R_iI.e.: s^*＝minS_i，S^-＝maxS_i，R^*＝minR_i，R^-＝maxR_i；S_iIndividual regret value, R, representing the ith evaluation object_iAn individual regret value representing the ith evaluation object;

wherein S is_iAnd R_iThe calculation formula of (2) is as follows:

wherein, W_jWeight representing jth evaluation attribute, f_jA and f_j ^—Maximum and minimum values of jth evaluation attribute evaluation values in evaluation value vectors respectively representing evaluation objects, f_ijAnd the evaluation value of the ith evaluation object on the jth evaluation attribute is shown.

The invention provides an online medical patient satisfaction measuring method based on network data, which takes emotion analysis of semantic depth as a data collecting and processing mode and extracts an evaluation attribute system and an evaluation value in a patient comment text; the comment data are calculated by an optimized multi-evaluation attribute decision method to obtain a comprehensive measurement result, and compared with the prior art, the method has the following beneficial effects:

(1) deep mining of medical online comments is realized by emotion analysis based on semantic depth, and the problems of insufficient semantic understanding, low precision and insufficient field depth of emotion analysis in the past are solved.

(2) The method takes the large-scale medical online comments as data sources, realizes the extraction and classification of the evaluation attributes of the patient satisfaction degree, and breaks through the limitation of the past evaluation attribute construction based on the expert law and questionnaire data.

(3) The patient satisfaction is measured based on the optimized multi-evaluation attribute method, so that the evaluation attributes related to the patient requirements are reflected, the unintended evaluation attributes are considered, and the evaluation result is more consistent with the characteristics of an online environment.

Drawings

Fig. 1 is a schematic flow chart of an online medical patient satisfaction measuring method based on network data according to an embodiment of the present invention;

FIG. 2 is a schematic flow chart of semantic-based online medical comment sentiment analysis in an embodiment of the present invention;

FIG. 3 is a schematic diagram of a result of labeling an online medical review according to an embodiment of the present invention;

FIG. 4 is a diagram illustrating dependency parsing in an embodiment of the present invention;

FIG. 5 is a flow chart of patient satisfaction evaluation attribute identification and classification in an embodiment of the present invention;

fig. 6 is a flowchart illustrating an evaluation value algorithm used in the embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below, and it is obvious that the described embodiments are some embodiments of the present invention, but not all embodiments; all other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The embodiment of the invention provides an online medical patient satisfaction measuring method based on network data, which takes large-scale medical online comments as data sources, realizes the extraction and classification of patient satisfaction evaluation attributes, and breaks through the limitation of the prior expert law and questionnaire-based data; (2) the emotion information of online comments is analyzed in the semantic depth, the problems of insufficient semantic understanding, low precision and insufficient field depth of the previous emotion analysis are solved, and structured data are provided for the satisfaction evaluation of patients; (3) the patient satisfaction is measured based on an optimized multi-evaluation attribute method, so that the evaluation attributes related to the patient requirements are reflected, the unintended evaluation attributes are considered, and the evaluation result is more consistent with the characteristics of an online environment. As shown in fig. 1, the present invention mainly includes the following steps.

And S1, acquiring the online medical comment text.

S2, performing emotion semantic analysis on the online medical comment text to obtain an emotion semantic labeling text, wherein semantic classes, semantic roles and emotion words are labeled in the emotion voice labeling text.

As shown in fig. 2, a schematic flow chart of performing sentiment semantic analysis on the online medical comment text in this embodiment is shown. In the embodiment, firstly, a semantic analysis method combining a semantic dictionary and a pattern matching rule is adopted to perform sentiment semantic analysis on the medical online comments, identify sentiment information in the medical online comments, and convert the online comments into structured information consisting of sentiment semantic classes and semantic roles. The semantic dictionary is constructed according to the following method: firstly, crawling a large amount of patient comment texts from an online medical application platform; then, carrying out word segmentation and part-of-speech tagging pretreatment on the Chinese characters; then, extracting verbs and adjectives in the initial word list and storing the verbs and the adjectives as the initial word list; and classifying the words expressing the similar emotion categories in the word list into the same semantic class, adding definitions and descriptions of the semantic class, wherein the description content comprises semantic class names, all words of the class and the emotion value of each word. The classification of semantic classes refers to resources such as modern Chinese dictionary, newly compiled synonym forest (newly compiled synonym forest [ M ]. Shanghai dictionary Press, King Yong.2015.) and the like. The annotation of the emotion value is determined according to the emotion polarity and strength of the words, and the value range is (0-1.0). Examples of emotion values corresponding to words in each semantic class are shown in table 1.

TABLE 1 medical Domain semantic Classification dictionary example

The semantic analysis adopts a method based on a dictionary and rules, and the process is as follows: firstly, segmenting words and parts of speech of comment texts, then carrying out dependency syntactic analysis on the comment texts, and then carrying out semantic annotation based on an emotion semantic dictionary and a pattern matching rule. The pattern matching rule determines a semantic labeling rule according to the syntactic dependency relationship of sentences and the corresponding relationship between syntactic components and semantic roles. According to the analysis of the large-scale online medical comment text, the semantic annotation rule is generalized as follows:

(1) if the clause only has one emotional word and no other components, namely a single-word sentence, no semantic role is labeled;

(2) if the syntax structure of the clause where the emotion word is located is a major-minor dependency relationship, marking the semantic role of the dominated component (namely, the subject) in the dependency structure as an evaluation object;

(3) if the syntax structure of the clause where the emotion word is located is a fixed dependency relationship, the semantic role of the dominated component (i.e., the fixed language) in the dependency structure is labeled as an "evaluation target".

The core function of the annotation system is developed by adopting python language, the comment text is stored by using MySQL, progress, owlready2, rdflib, xlwt and other related packages and the MySQL server, and the processing result is shown in FIG. 3.

Take a certain medical review as an example:

original text: the doctor feels responsible and has high medical skill, is particularly close to people and is a very good doctor.

Sentence breaking, word segmentation and part of speech tagging:

clause 1: feel/v this/r doctor/n responsible/v and/c medical/n high Ming/a

Clause 2: for/p person/n special/d affinity/a

Clause 3: is/v very good/a/udoctor/n

Where the letter following the "/" indicates the part of speech of the corresponding word, the result of the dependency parsing is shown in fig. 4. In fig. 4, HED represents a predicate center, VOB represents an object, ATT represents a fixed term, SBV represents a subject, ADV represents a state, POB represents a preposition object, and RAD represents a helpword addition in a syntax structure.

The emotion semantic annotation text obtained finally is shown in table 2.

TABLE 2 Emotion semantic analysis results example

Numbering	Emotional words	Semantic class	Semantic roles	Degree word	Negative word
						00101	Is responsible for	Heart of responsibility	The doctor is	Is free of	Is free of
00102	Ko Ming	Can dry	Medical treatment	Is free of	Is free of
						00103	Intimacy of care	Social behavior evaluation	Is free of	In particular	Is free of
00104	Good taste	Desirability	Doctor	Very much	Is free of

S3, matching a patient satisfaction evaluation attribute database according to the semantic type identifier and the semantic role identifier in the emotion semantic annotation text, and identifying a patient satisfaction evaluation attribute corresponding to the comment text;

in step S3, the method further includes a step of constructing a patient satisfaction evaluation attribute database, where the method of constructing the patient satisfaction evaluation attribute database includes: based on the online comment data after semantic annotation, a semantic class-semantic role pair is extracted, and is clustered by adopting a hierarchical clustering method to form a patient satisfaction evaluation attribute system and is stored in a patient satisfaction evaluation attribute database. The evaluation attribute system is shown in table 3.

TABLE 3 patient satisfaction evaluation Attribute System

Semantic classes, semantic roles and frequency information are extracted from the emotion semantic annotation data obtained in step S2, a mapping relation is established with the patient satisfaction evaluation attribute classification system, and the mapping relation is stored in the patient satisfaction evaluation attribute database. As shown in table 4, the semantic class, semantic role and frequency corresponding to part of the medical evaluation attributes are listed.

TABLE 4 semantic class, semantic role and frequency (part) corresponding to medical evaluation attributes

And matching the patient satisfaction evaluation attribute database based on the emotion semantic annotation text according to semantic classes and semantic role identifiers, so as to identify the patient satisfaction evaluation attribute corresponding to the comment text. In the matching process, if the evaluation attribute database does not contain words, the semantic similarity between the words and the existing words is calculated, and the category of the words is determined according to the words with the highest similarity. If the database does not record the semantic role or the semantic role is default, only matching the semantic class with the evaluation attribute database; and if the semantic class corresponds to a plurality of evaluation attribute categories in the database, taking the most frequent one as a final evaluation attribute identification result. Wherein, the semantic similarity takes 0.7 as the lower limit of the effective value.

The specific processing flow is shown in fig. 5.

Example (c): the doctor collaborating with the hospital not only has high public praise and popularity, but also has good medical practice.

Semantic annotation: e1{ F (high): high-low, Obj: public praise };

e2{ F (high): high-low, Obj: popularity };

e3{ F (good): desirability, Obj: medical };

wherein F represents a semantic class, and Obj represents a semantic role "evaluation object". Matching the desirability and the medical skill under the topic of the medical skill according to a knowledge base; the "high-low + awareness" matches under the "whole" topic. In the "high/low + word-of-mouth", the evaluation object "word-of-mouth" is not included, and thus the semantic similarity between the evaluation object and the word under the "high/low" semantic class is calculated.

TABLE 5 semantic similarity of terms under the evaluation object "word-of-mouth" and "high-low" semantic classes

As can be seen from table 5, the semantic similarity between "word of mouth" and "degree of awareness" is 0.875200, which is more effective than 0.7, and therefore, "high/low + word of mouth" belongs to the evaluation attribute "whole".

And S4, determining evaluation values of all evaluation attributes in the evaluation text according to the emotion word identifications in the emotion semantic annotation text.

The evaluation value of the medical online comment is based on the emotion value of the emotion word and is determined according to the semantic role and the context characteristics. For example, if the context has the adverb of "very", the emotion value is +/-0.1, namely the emotion intensity is enhanced by 0.1, the positive evaluation is increased by 0.1, and the negative evaluation is decreased by 0.1; if a negative modifier appears, (1-original value) is used as the final evaluation value. The specific algorithm is shown in fig. 6.

Evaluation value labeling results of the above comment sentences, for example, are shown in table 6.

TABLE 6 evaluation value labeling results

Numbering	Semantic class	Degree word	Negative word	Evaluation value
					00101	Is responsible for	Is free of	Is free of	0.6
00102	Ko Ming	Is free of	Is free of	0.6
					00103	Intimacy of care	In particular	Is free of	0.8
00104	Good taste	Very much	Is free of	0.8

And S5, converting the emotion analysis result of the medical comment into a vector, wherein the vector comprises the evaluation attribute and the evaluation value of each patient comment, and obtaining the evaluation value vector of each evaluation object by taking a doctor as the evaluation object.

(1) Construction of comment vectors

Converting the emotion analysis result of the medical comment into a vector, representing the evaluation attributes contained in one patient comment and the evaluation values of all the evaluation attributes: r_k＝(a_k1，a_k2，...，a_km) (ii) a Wherein R is_kRepresenting the kth patient review; a is_kjIndicating the evaluation value given on the jth evaluation attribute in the comment.

Taking doctors as evaluation objects, measuring the evaluation value of each doctor's evaluation attribute, namely the sum of the evaluation values marked by all comments corresponding to the evaluation attributes,

wherein D represents a patient evaluation vector obtained by a doctor to be evaluated, and R_kVector representing the kth patient review for this doctor, a_kjThe evaluation value on the jth evaluation attribute is reviewed for the kth patient. The sum of the evaluation values of the ith evaluation object on the jth evaluation attribute is set as A_ijThen, the normalization process is performed, and the formula is:

by equation (2), an evaluation value vector of the ith evaluation target can be calculated: d_i＝(f_i1，f_i2，...，f_im)。f_ijA normalized evaluation value of the ith evaluation object on the jth evaluation attribute, A_ijAnd represents the sum of evaluation values of the ith evaluation object on the jth evaluation attribute.

(2) Weight calculation based on TF-IDF

And calculating the weight of the evaluation attribute according to the TF-IDF index of the evaluation attribute entry. Wherein, Term Frequency (TF) refers to the Frequency of the Term corresponding to the evaluation attribute in the comment text:

wherein TF_jWord frequency, n, representing the jth evaluation attribute_jFor the number of occurrences of the jth evaluation attribute in the comment text set d, n_kFor the number of occurrences of the kth rating attribute in the comment text set, the sum of the denominators represents the number of occurrences of all rating attributes in the text set.

Inverse Document Frequency (IDF) is a measure of how important an attribute is generally evaluated. Dividing the total number of the comments (| D |) by the number of the comments (j) containing the evaluation attribute entry, and taking a logarithm with the base of 10:

where | D | is the total number of comments, | { j: t is t_i∈d_mJ is the number of comments containing the jth evaluation attribute, t_iDenotes the ith comment, d_mAll comments are represented.

Finally, the weight is calculated as the product of TF and IDF, i.e. the higher The Frequency (TF) of occurrence of a certain evaluation attribute is, and the lower the frequency of its files in the review set is, the higher its weight is:

W_j＝TF_j×IDF_j；(5)

wherein, W_jRepresenting the weight of the jth evaluation attribute.

And S6, calculating the satisfaction value of each evaluation object according to the weight and the evaluation value vector.

First, the maximum value f in the normalized evaluation value of each evaluation attribute is determined^* _jAnd minimum value f_j ^-That is, the maximum value and the minimum value of the normalized evaluation values corresponding to the evaluation attribute j are determined for all the doctor evaluation targets. Then, a weighted distance between the normalized evaluation value of the evaluation object and the maximum value thereof is calculated

Thereafter, S for each evaluation object is calculated_iValue and R_iValue, S_iThe value represents the population utility value, R_iValues represent an individual regret value. The calculation formula is as follows:

finally, overall patient satisfaction is calculated:

among them, there are:

S^*＝minS_i，S^-＝maxS_i，R^*＝minR_i，R^-＝maxR_i；(8)

in the formula (7), Q_iRepresenting the satisfaction degree of the ith evaluation object, v representing a second weight, representing the maximum utility of the population and the weight of the individual regret, S represents the utility value S of each population_iS-represents the respective population utility value S_iMaximum in (1), R denotes the individual regret value R_iMinimum of (d), R-represents the individual regret value R_iMaximum value of (2). In equation (7), the value of the second weight v may be set to 0.5, i.e., the balance considers the population maximum utility and the individual regret.

10 ten thousand patient reviews are crawled from an online medical website 'good doctor online', and comprise an evaluation text, star quantity, overall scores and the like. And performing word segmentation, part of speech tagging and semantic tagging on the crawled data. By constructing a patient review vector, a patient satisfaction calculation is calculated that scores attributes. Table 7 shows the evaluation values, normalized to values between 0 and 1.

TABLE 7 normalized evaluation values

Note: the superscript "-" after the number indicates the maximum value and the superscript "-" indicates the minimum value.

Overall patient satisfaction is calculated. First, the maximum and minimum values of each evaluation attribute are determined. Then, the distance between each evaluation value and its weight, i.e., the distance between each evaluation value and its weight is calculatedAnd finally, taking the second weight v as 0.5, namely balancing and considering the influence of the maximum utility of the population and the individual regret on the overall satisfaction, and obtaining a ranking result: a. the>D>E>F>B>C。

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.