An evaluation report generation system based on artificial intelligence and big data analysis

Abstract

The present invention relates to the technical field of report generation, and discloses an evaluation report generation system based on artificial intelligence and big data analysis, comprising a data acquisition module, a preprocessing module, a sentiment analysis unit, a commodity analysis unit, and a report generation unit; in the present invention, the data acquisition module acquires multi-dimensional data to lay a foundation for analysis and ensure comprehensiveness and objectivity; the preprocessing module extracts keywords with a specific algorithm and accurately refines key contents, which is conducive to subsequent work; the sentiment analysis unit analyzes sentiment tendencies with the help of sentiment dictionaries, etc., can clearly present consumers' word-of-mouth on commodities, and assist in decision-making adjustments; the commodity analysis unit analyzes commodity attributes in combination with word frequency and information entropy, calculates key attribute word scores and correlations with sentiment tendencies, and provides a basis for optimizing commodities; the report generation unit generates a report covering key contents, which has important reference value for merchants to improve products and formulate strategies, and for consumers to make purchase decisions.

Description

Technical Field

[0001] The present invention relates to the technical field of report generation, and in particular to an evaluation report generation system based on artificial intelligence and big data analysis. Background Art

[0002] In today's booming e-commerce era, the number of products on e-commerce platforms has exploded. When faced with a large number of products, how consumers can accurately understand the true situation of the products has become a key issue. At the same time, it is also crucial for merchants to have in-depth insights into consumers' evaluation of products and the performance of their own products in the market in order to optimize products and marketing strategies. Traditional product evaluation analysis methods have many limitations.

[0003] On the one hand, when consumers browse product reviews, they often need to spend a lot of time reading numerous review texts, and due to the diversity and complexity of the review content, it is difficult to quickly and accurately grasp the core advantages, disadvantages and overall reputation of the product. For example, the reviews of some products may be lengthy and disorganized, containing a lot of irrelevant information, and it is difficult for consumers to extract valuable key information from them, thus affecting the efficiency and accuracy of purchase decisions. On the other hand, when analyzing product review data, merchants can usually only perform simple statistics, such as the number of good reviews, medium reviews, and bad reviews. This analysis method is too superficial and cannot deeply explore the detailed information of consumer emotional tendencies contained in the review text and the inherent connection between the product attributes. For example, merchants cannot know exactly whether consumers' emotional feedback on a specific attribute of a product is due to imperfect functions, poor appearance design, or quality problems, etc., and thus it is difficult to improve and optimize the product in a targeted manner, which puts them at a disadvantage in the fierce market competition.

[0004] In addition, as the amount of data continues to increase, manual processing and analysis of these evaluation data has become increasingly difficult and inefficient. There is an urgent need for an evaluation and analysis system that is intelligent, automated, and able to deeply mine the value of data. Summary of the invention

[0005] The purpose of the present invention is to provide an evaluation report generation system based on artificial intelligence and big data analysis, which solves the technical problems raised in the background technology.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] An evaluation report generation system based on artificial intelligence and big data analysis, comprising:

[0008] The data collection module is used to establish a connection with the database of the e-commerce platform to obtain multi-dimensional data;

[0009] A preprocessing module is used to extract keywords from the evaluation text using a word frequency-inverse document frequency algorithm;

[0010] A sentiment analysis unit is used to analyze the sentiment tendency of the evaluation text based on keywords, and determine the sentiment tendency of the evaluation text according to the analysis results;

[0011] The product analysis unit is used to extract key attribute words from the product details data by combining word frequency statistics with information entropy, and then calculate the importance score of the key attribute words, and at the same time calculate the correlation between each key attribute word and the corresponding emotional tendency of the evaluation text;

[0012] The report generation unit is used to generate an evaluation report according to the results obtained by the sentiment analysis unit and the product analysis unit respectively.

[0013] As a further solution of the present invention: wherein the multi-dimensional data includes:

[0014] User evaluation data: the corresponding evaluation text when consumers evaluate the content of the product;

[0015] Product attribute data: product category, brand, price range, and product detail text corresponding to the product's specifications.

[0016] As a further solution of the present invention: the keyword extraction method is as follows:

[0017] Step 1.1, extract all evaluation texts from the user evaluation data, then perform word segmentation on each evaluation text to obtain multiple evaluation words;

[0018] Step 1.2: Count the number of times each evaluation word appears in each evaluation text and record it as PC i，j ;

[0019] Where i represents the number of evaluation words, i=1, 2, ...n, n represents the number of all evaluation words in the jth evaluation text, j represents the number of evaluation texts, j=1, 2, ...m, m represents the number of evaluation texts;

[0020] Step 1.3, by:

[0021] ;

[0022] Calculate the frequency TF of the i-th evaluation word in the j-th evaluation text i，j ;

[0023] Step 1.4: Select each evaluation word one by one, and count the number of evaluation texts containing each evaluation word in the evaluation text, and record it as D i ;

[0024] Step 1.5: Pass , calculate the inverse document frequency IDF of each evaluation word i ;

[0025] Step 1.6: Pass , calculate the word frequency-inverse document frequency TF-IDF of each evaluation word in each evaluation text i，j ;

[0026] Step 1.7: Then, each TF-IDF i，j Respectively with the preset term frequency-inverse document frequency threshold TF-IDF y For comparison:

[0027] When TF-IDF i，j ≥TF-IDF y , then the evaluation words in the corresponding evaluation text are used as keywords;

[0028] When TF-IDF i，j <TF-IDF y , the evaluation words in the corresponding evaluation text will not be used as keywords.

[0029] As a further solution of the present invention: the sentiment tendency analysis method is as follows:

[0030] Step 2.1, extract the pre-built sentiment dictionary;

[0031] Among them, each emotional word is marked as a category corresponding to positive emotion, negative emotion or neutral emotion, and an initial emotional intensity value is assigned to each emotional word in advance and marked as Q k , k=1, 2, ...v, v represents the number of sentiment words in the sentiment dictionary, Q k Indicates the initial sentiment intensity value of the sentiment word in the sentiment dictionary;

[0032] Among them, the sentiment dictionary is obtained by collecting and organizing sentiment words in advance;

[0033] Step 2.2: traverse all keywords, and then combine the sentiment dictionary to count the initial sentiment intensity values ​​corresponding to each keyword as a sentiment word in each evaluation text, and mark it as QP t，j At the same time, the number of occurrences of each keyword as a sentiment word in the evaluation text is counted and marked as QC t，j ; Where t=1, 2, ...u, u represents the number of keywords corresponding to sentiment words, t∈k, that is, u≤v;

[0034] Then through:

[0035] ;

[0036] Calculate the sentiment intensity score E of each evaluation text j ;

[0037] Step 2.3: Then, score the sentiment intensity E of each evaluation text. j Determine the sentiment tendency of each evaluation text;

[0038] If E j >0, the sentiment tendency of the corresponding evaluation text is judged to be positive;

[0039] If E j <0, the sentiment tendency of the corresponding evaluation text is judged to be negative;

[0040] If E j =0, the sentiment tendency of the corresponding evaluation text is judged to be neutral.

[0041] As a further solution of the present invention: the analysis method of the commodity analysis unit is as follows:

[0042] Step 3.1: Extract all attribute words in all product detail texts from product attribute data;

[0043] Among them, attribute words refer to words used to describe various characteristics and features of products;

[0044] Step 3.2: Count the number of times each attribute word appears in each product detail text and mark it as SC q，p , q represents the attribute word, q=1, 2, ... e, e represents the number of attribute words in the p-th product detail text, p represents the evaluation text, p=1, 2, ... r, r represents the number of product detail texts;

[0045] Step 3.3, by:

[0046] ;

[0047] Calculate the frequency TF1 of the qth attribute word in all product details texts q ;

[0048] Step 3.4, by:

[0049] ;

[0050] Calculate the probability H of the qth attribute word in the pth product details text q，p ;

[0051] Step 3.5, by:

[0052] ;

[0053] Calculate the information entropy X of the qth attribute word q ;

[0054] Step 3.6, by:

[0055] ;

[0056] Calculate the importance score ZY of each attribute word q ;

[0057] In the formula, β is a preset weight coefficient, which is used to adjust the ratio between the attribute word corresponding word frequency and the attribute word corresponding information entropy when calculating the importance score of the attribute word;

[0058] Step 3.7: Score the importance of each attribute word ZY q Respectively with the preset importance score threshold ZY y Make a comparison and determine the key attribute words based on the comparison results:

[0059] When ZY q >ZY y , the corresponding attribute word is determined as the key attribute word;

[0060] When ZY q ≤ZY y , the corresponding attribute word is not determined as a key attribute word;

[0061] Step 3.8: All the determined key attribute words are grouped into a key attribute word set B = {B s}, s represents the key attribute word in the key attribute word set, s=1, 2, ... c, c represents the number of key attribute words; where s∈q, and c≤e;

[0062] At the same time, all evaluation texts are combined into an evaluation text set T = {T j};

[0063] Then count each key attribute word B s In each evaluation text T j The number of occurrences in the s，j ;

[0064] Step 3.9: Extract the sentiment intensity score E of each evaluation text j ;

[0065] Also through:

[0066] ;

[0067] Calculate the correlation M between each key attribute word and the corresponding emotional tendency of the evaluation text s.

[0068] Beneficial effects of the present invention:

[0069] The present invention establishes a connection with the e-commerce platform database through a data acquisition module to obtain multi-dimensional data including user evaluation data and product attribute data, laying a solid data foundation for subsequent comprehensive and in-depth analysis of product-related situations, making the analysis results more comprehensive and objective.

[0070] In the present invention, the preprocessing module adopts a word frequency-inverse document frequency algorithm to extract keywords in the evaluation text. After a series of rigorous steps such as word segmentation, word frequency statistics, and inverse document frequency calculation, the keywords are screened out according to a preset threshold. It can accurately extract valuable and representative content in the evaluation text, avoid interference from irrelevant information, and help to carry out subsequent work such as sentiment tendency analysis more efficiently and accurately.

[0071] In the present invention, the sentiment analysis unit performs sentiment tendency analysis on the evaluation text based on the extracted keywords, calculates the sentiment intensity score by constructing a sentiment dictionary, and determines the sentiment tendency of the evaluation text accordingly, so that merchants or related parties can clearly know the consumers' overall sentiment attitude and reputation towards the products, which is convenient for targeted adjustment of business strategies, improvement of products, etc.

[0072] In the present invention, the product analysis unit extracts key attribute words from the product details data using a method based on combining word frequency statistics with information entropy, and not only calculates the importance score of the key attribute words, but also calculates the correlation between each key attribute word and the corresponding emotional tendency of the evaluation text. This helps to gain a deep understanding of the importance of each attribute of the product in consumer cognition and evaluation, as well as the intrinsic connection between the attributes and the emotional tendency of the consumer, thereby providing a strong basis for optimizing the configuration of product attributes, highlighting advantageous attributes, etc.

[0073] In the present invention, the report generation unit generates an evaluation report based on the results obtained by the sentiment analysis unit and the product analysis unit, covering important contents such as the correlation between key attribute words and the sentiment tendency of the evaluation text, the importance score of the attribute words, and the sentiment tendency of the evaluation text to the product. Whether it is for merchants to formulate product improvements and marketing strategies, or for consumers to make purchasing decisions, it can provide extremely valuable reference basis and assist all parties to make more reasonable and scientific decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The present invention will be further described below in conjunction with the accompanying drawings.

[0075] Figure 1 It is a system block diagram of an evaluation report generation system based on artificial intelligence and big data analysis of the present invention. DETAILED DESCRIPTION

[0076] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Embodiment 1

[0077] See also Figure 1 As shown, the present invention is an evaluation report generation system based on artificial intelligence and big data analysis, comprising:

[0078] The data collection module is used to establish a connection with the database of the e-commerce platform to obtain multi-dimensional data;

[0079] Among them, multi-dimensional data includes:

[0080] User evaluation data: the corresponding evaluation text when consumers evaluate the content of the product;

[0081] A preprocessing module is used to extract keywords from the evaluation text using a word frequency-inverse document frequency algorithm;

[0082] The keyword extraction method is as follows:

[0083] Step 1.1, extract all evaluation texts from the user evaluation data, then perform word segmentation on each evaluation text to obtain multiple evaluation words;

[0084] Among them, the word segmentation process is to divide the continuous text sequence into independent words according to certain rules, which adopts a method based on dictionary matching and maximum matching. In this embodiment, it is a prior art, so it will not be described in detail;

[0085] Step 1.2: Count the number of times each evaluation word appears in each evaluation text and record it as PC i，j ;

[0086] Where i represents the number of evaluation words, i=1, 2, ...n, n represents the number of all evaluation words in the jth evaluation text, j represents the number of evaluation texts, j=1, 2, ...m, m represents the number of evaluation texts;

[0087] Step 1.3, by:

[0088] ;

[0089] Calculate the frequency TF of the i-th evaluation word in the j-th evaluation text i，j ;

[0090] Step 1.4: Select each evaluation word one by one, and count the number of evaluation texts containing each evaluation word in the evaluation text, and record it as D i ;

[0091] Step 1.5: Pass Calculate the inverse document frequency (IDF) of each evaluation word i ;

[0092] Step 1.6: Pass , calculate the word frequency-inverse document frequency TF-IDF of each evaluation word in each evaluation text i，j ;

[0093] Step 1.7: Then, each TF-IDF i，j Respectively with the preset term frequency-inverse document frequency threshold TF-IDF y For comparison:

[0094] When TF-IDF i，j ≥TF-IDF y , then the evaluation words in the corresponding evaluation text are used as keywords;

[0095] When TF-IDF i，j <TF-IDF y , the evaluation words in the corresponding evaluation text are not used as keywords;

[0096] Sentiment analysis unit, used to analyze the sentiment tendency of the evaluation text based on keywords;

[0097] The sentiment analysis method is as follows:

[0098] Step 2.1, extract the pre-built sentiment dictionary;

[0099] Among them, each emotional word is marked as a category corresponding to positive emotion, negative emotion or neutral emotion, and an initial emotional intensity value is assigned to each emotional word in advance and marked as Q k , k=1, 2, ...v, v represents the number of sentiment words in the sentiment dictionary, Q k Indicates the initial sentiment intensity value of the sentiment word in the sentiment dictionary;

[0100] In this embodiment, for example, "good review" is a positive sentiment word, and the initial sentiment intensity value of the good review is: Q 好评 =0.8; "bad review" is a negative sentiment word, and the initial sentiment intensity value of bad review is: Q 差评 =-0.8; "General" is a neutral sentiment word, and the initial sentiment intensity value of "General" is: Q 一般 =0;

[0101] Among them, the sentiment dictionary is obtained by collecting and organizing sentiment words in advance;

[0102] Step 2.2: traverse all keywords, and then combine the sentiment dictionary to count the initial sentiment intensity values ​​corresponding to each keyword as a sentiment word in each evaluation text, and mark it as QP t，j At the same time, the number of occurrences of each keyword as a sentiment word in the evaluation text is counted and marked as QC t，j ; Where t=1, 2, ...u, u represents the number of keywords corresponding to sentiment words, t∈k, that is, u≤v;

[0103] Then through:

[0104] ;

[0105] Calculate the sentiment intensity score E of each evaluation text j ;

[0106] Step 2.3: Then, score the sentiment intensity E of each evaluation text. j Determine the sentiment tendency of each evaluation text;

[0107] If E j >0, the sentiment tendency of the corresponding evaluation text is judged to be positive;

[0108] If E j <0, the sentiment tendency of the corresponding evaluation text is judged to be negative;

[0109] If E j =0, the sentiment tendency of the corresponding evaluation text is judged to be neutral.

[0110] The report generation unit is used to generate an evaluation report according to the results obtained by the sentiment analysis unit.

[0111] This embodiment can obtain user evaluation data from multi-dimensional data on the e-commerce platform, and accurately extract keywords from the evaluation text through the word frequency-inverse document frequency algorithm, and then analyze the evaluation text based on keywords and sentiment dictionaries to obtain the sentiment tendency of each evaluation text. This helps merchants quickly understand consumers' emotional attitudes towards products, provides a basis for merchants to preliminarily judge the reputation of products in the market, and can also help consumers make purchase decisions by referring to the sentiment tendencies of other consumers to a certain extent. Embodiment 2

[0112] As the second embodiment of the present invention, when the present application is specifically implemented, compared with the first embodiment, the technical solution of this embodiment is different from that of the first embodiment only in that in this embodiment, the multi-dimensional data further includes:

[0113] Product attribute data: product category, brand, price range, and product details corresponding to the product’s specifications;

[0114] The system also includes:

[0115] The product analysis unit is used to extract key attribute words from the product details data by combining word frequency statistics with information entropy, and then calculate the importance score of the key attribute words, and at the same time calculate the correlation between each key attribute word and the corresponding emotional tendency of the evaluation text;

[0116] The specific method is as follows:

[0117] Step 3.1: Extract all attribute words in all product detail texts from product attribute data;

[0118] Among them, attribute words refer to words used to describe the various characteristics and features of a product. They are a general description of the specific attributes of a product.

[0119] In this embodiment, the attribute words include:

[0120] Functional attribute words: describe the functions or uses of a product. For example, for a mobile phone, "clear photos", "long battery life", "smooth operation" and so on are functional attribute words; for a washing machine, "frequency conversion energy saving", "multiple washing modes", "large capacity" and so on are also functional attribute words;

[0121] Appearance attribute words: descriptions of the product’s external shape, color, size, etc.; taking clothing as an example, “slim fit”, “simple design”, “black”, “one size fit”, etc. are all common appearance attribute words; for furniture, “solid wood material”, “white coating”, “European style”, etc. also fall into the category of appearance attribute words;

[0122] Quality attribute words: reflect the quality and quality status of the product; such as "superior quality", "fine workmanship", "durable", "environmentally friendly materials", etc. These attribute words can give consumers a preliminary understanding of the quality of the product;

[0123] Specification attribute words: The specific specification parameters of a product are one of the key information that consumers pay attention to when purchasing; for example, a computer’s “Intel Core i7 processor”, “16GB of memory”, “512GB solid-state drive”; a camera’s “24 million pixels”, “50mm focal length”, etc.

[0124] Experience attribute words: describe the feelings or experiences of consumers when using the product. For example, "comfortable to wear", "easy to operate", "good quiet effect", "thoughtful service", etc., can help consumers better imagine the scene and feeling of using the product;

[0125] Step 3.2: Count the number of times each attribute word appears in each product detail text and mark it as SC q，p , q represents the attribute word, q=1, 2, ... e, e represents the number of attribute words in the p-th product detail text, p represents the evaluation text, p=1, 2, ... r, r represents the number of product detail texts;

[0126] Step 3.3, by:

[0127] ;

[0128] Calculate the frequency TF1 of the qth attribute word in all product details texts q ;

[0129] Step 3.4, by:

[0130] ;

[0131] Calculate the probability H of the qth attribute word in the pth product details text q，p ;

[0132] Step 3.5, by:

[0133] ;

[0134] Calculate the information entropy X of the qth attribute word q ;

[0135] Step 3.6, by:

[0136] ;

[0137] Calculate the importance score ZY of each attribute word q ;

[0138] In the formula, β is a preset weight coefficient, which is used to adjust the ratio between the attribute word corresponding word frequency and the attribute word corresponding information entropy when calculating the importance score of the attribute word;

[0139] Step 3.7: Score the importance of each attribute word ZY q Respectively with the preset importance score threshold ZY y Make a comparison and determine the key attribute words based on the comparison results:

[0140] When ZY q >ZY y , the corresponding attribute word is determined as the key attribute word;

[0141] When ZY q ≤ZY y, the corresponding attribute word is not determined as a key attribute word;

[0142] Step 3.8: All the determined key attribute words are grouped into a key attribute word set B = {B s}, s represents the key attribute word in the key attribute word set, s=1, 2, ... c, c represents the number of key attribute words; where s∈q, and c≤e;

[0143] At the same time, all evaluation texts are combined into an evaluation text set T = {T j};

[0144] Then count each key attribute word B s In each evaluation text T j The number of occurrences in the s，j ;

[0145] Step 3.9: Extract the sentiment intensity score E of each evaluation text j ;

[0146] Also through:

[0147] ;

[0148] Calculate the correlation M between each key attribute word and the corresponding emotional tendency of the evaluation text s ;

[0149] The report generating unit is also used to generate an evaluation report according to the results obtained by the commodity analyzing unit.

[0150] Based on the first embodiment, this embodiment incorporates the commodity attribute data, uses a specific method to extract key attribute words and calculate their importance scores and the correlation with the emotional tendency of the evaluation text. This allows merchants to not only know consumers' emotional attitudes towards commodities, but also gain an in-depth understanding of the attention and recognition of each commodity attribute, thereby accurately locating the advantages and disadvantages of commodities and providing strong support for targeted improvement or optimization of commodities; at the same time, it also allows consumers to more comprehensively understand the performance of commodities in different attributes and their relationship with consumers' emotional tendencies, further assisting their purchasing decisions. Embodiment 3

[0151] As the third embodiment of the present invention, when the present application is implemented, compared with the first and second embodiments, the technical solution of this embodiment is to combine the solutions of the first and second embodiments;

[0152] in:

[0153] The evaluation report includes the correlation between the key attribute words and the corresponding sentiment tendency of the evaluation text, the importance score of each attribute word, and the sentiment tendency of each evaluation text to the evaluated product;

[0154] The evaluation reports generated in this implementation have the following advantages:

[0155] Help merchants understand the attention and recognition of each product attribute: By presenting the correlation between key attribute words and the corresponding emotional tendency of the evaluation text and the importance score of each attribute word, merchants can clearly understand the importance of different attributes of the product in the minds of consumers and consumers' emotional tendencies towards these attributes, such as functional attributes, appearance attributes, quality attributes, etc., so as to understand the advantages and disadvantages of the product and make targeted improvements or optimizations to the product;

[0156] Help merchants understand consumers' overall emotional attitudes towards products: With the help of the emotional tendency information of each evaluation text in the report, merchants can intuitively understand whether consumers have a positive, negative or neutral attitude towards the product as a whole, and then analyze the reputation of the product in the market, providing a reference for subsequent marketing strategies, product iterations and other decisions;

[0157] Assist consumers in making purchasing decisions: Consumers can view evaluation reports to understand the performance of products in different attributes and other consumers' overall evaluation sentiments on the products, so as to better judge whether the products meet their needs and expectations and help them make more reasonable purchasing choices.

[0158] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0159] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by any technician familiar with the technical field within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An evaluation report generation system based on artificial intelligence and big data analysis, characterized in that: include: The data collection module is used to establish a connection with the database of the e-commerce platform to obtain multi-dimensional data; A preprocessing module is used to extract keywords from the evaluation text using a word frequency-inverse document frequency algorithm; A sentiment analysis unit is used to analyze the sentiment tendency of the evaluation text based on keywords, and determine the sentiment tendency of the evaluation text according to the analysis results; The product analysis unit is used to extract key attribute words from the product details data by combining word frequency statistics with information entropy, and then calculate the importance score of the key attribute words, and at the same time calculate the correlation between each key attribute word and the corresponding emotional tendency of the evaluation text; The analysis method of the commodity analysis unit is as follows: Step 3.1: Extract all attribute words in all product detail texts from product attribute data; Among them, attribute words refer to words used to describe various characteristics and features of products; Step 3.2: Count the number of times each attribute word appears in each product detail text and mark it as SC q，p , q represents the attribute word, q=1, 2, ... e, e represents the number of attribute words in the p-th product detail text, p represents the evaluation text, p=1, 2, ... r, r represents the number of product detail texts; Step 3.3, by: ； Calculate the frequency TF1 of the qth attribute word in all product details texts q ; Step 3.4, by: ； Calculate the probability H of the qth attribute word in the pth product details text q，p ; Step 3.5, by: ； Calculate the information entropy X of the qth attribute word q ; Step 3.6, by: ； Calculate the importance score ZY of each attribute word q ; In the formula, β is the preset weight coefficient; Step 3.7: Score the importance of each attribute word ZY q Respectively with the preset importance score threshold ZY y Make a comparison and determine the key attribute words based on the comparison results: Step 3.8: All the determined key attribute words are grouped into a key attribute word set B = {B s }, s represents the key attribute word in the key attribute word set, s=1, 2, ... c, c represents the number of key attribute words; where s∈q, and c≤e; At the same time, all evaluation texts are combined into an evaluation text set T = {T j }; Then count each key attribute word B s In each evaluation text T j The number of occurrences in the s，j ; Step 3.9: Extract the sentiment intensity score E of each evaluation text j ; Also through: ； Calculate the correlation M between each key attribute word and the corresponding emotional tendency of the evaluation text s ; The report generation unit is used to generate an evaluation report according to the results obtained by the sentiment analysis unit and the product analysis unit respectively.

2. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 1 is characterized in that: The keyword extraction method is as follows: Step 1.1, extract all evaluation texts from the user evaluation data, then perform word segmentation on each evaluation text to obtain multiple evaluation words; Step 1.2: Count the number of times each evaluation word appears in each evaluation text and record it as PC i，j ; Where i represents the number of evaluation words, i=1, 2, ...n, n represents the number of all evaluation words in the jth evaluation text, j represents the number of evaluation texts, j=1, 2, ...m, m represents the number of evaluation texts; Step 1.3, by: ； Calculate the frequency TF of the i-th evaluation word in the j-th evaluation text i，j ; Step 1.4: Select each evaluation word one by one, and count the number of evaluation texts containing each evaluation word in the evaluation text, and record it as D i ; Step 1.5: Pass , calculate the inverse document frequency IDF of each evaluation word i ; Step 1.6: Pass , calculate the word frequency-inverse document frequency TF-IDF of each evaluation word in each evaluation text i，j ; Step 1.7: Then, each TF-IDF i，j Respectively with the preset term frequency-inverse document frequency threshold TF-IDF y Make a comparison and determine the keywords based on the comparison results.

3. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 2 is characterized in that: In step 1.7, the keywords are determined as follows: When TF-IDF i，j ≥TF-IDF y , then the evaluation words in the corresponding evaluation text are used as keywords; When TF-IDF i，j <TF-IDF y , the evaluation words in the corresponding evaluation text will not be used as keywords.

4. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 2 is characterized in that: The sentiment analysis method is as follows: Step 2.1, extract the pre-built sentiment dictionary; Step 2.2: traverse all keywords, and then combine the sentiment dictionary to count the initial sentiment intensity values ​​corresponding to each keyword as a sentiment word in each evaluation text, and mark it as QP t，j At the same time, the number of occurrences of each keyword as a sentiment word in the evaluation text is counted and marked as QC t，j ; Where t = 1, 2, ... u, u represents the number of keywords corresponding to sentiment words; Then through: ； Calculate the sentiment intensity score E of each evaluation text j ; Step 2.3: Then, score the sentiment intensity E of each evaluation text. j Determine the sentiment tendency of each evaluation text.

5. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 4 is characterized in that: In step 2.3, the emotional tendency of the evaluation text is determined as follows: If E j >0, the sentiment tendency of the corresponding evaluation text is judged to be positive; If E j <0, the sentiment tendency of the corresponding evaluation text is judged to be negative; If E j =0, the sentiment tendency of the corresponding evaluation text is judged to be neutral.

6. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 4 is characterized in that: in, The sentiment dictionary is obtained by collecting and organizing sentiment words in advance.

7. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 1 is characterized in that: In step 3.7, the key attribute words are determined as follows: When ZY q >ZY y , the corresponding attribute word is determined as the key attribute word; When ZY q ≤ZY y , the corresponding attribute word is not determined as a key attribute word.

8. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 1 is characterized in that: in, Multi-dimensional data includes: User evaluation data: the corresponding evaluation text when consumers evaluate the content of the product; Product attribute data: product category, brand, price range, and product detail text corresponding to the product's specifications.

9. The evaluation report generation system based on artificial intelligence and big data analysis according to claim 4 is characterized in that: in, Each sentiment word in the sentiment dictionary is marked as a category corresponding to positive sentiment, negative sentiment or neutral sentiment, and the initial sentiment intensity value corresponding to each sentiment word is a pre-assigned value.

Images