AI-based intelligent advertisement promotion service system and method

Abstract

The invention discloses an AI-based intelligent advertisement promotion service system and method, and relates to the technical field of advertisement information management. Picture information and text information of a to-be-promoted product are obtained, feature words corresponding to feature attributes of all phrases of the text information are respectively recognized in the picture information and the text information, and the feature words are stored in a database; the method comprises the following steps: constructing feature word pairs according to the combination of feature words in a statement, marking the difference between the feature words, calculating the occurrence frequency of the feature words in phrases, obtaining a character evaluation coefficient of the feature words in the phrases, calculating the image proportion of feature attributes in a picture, obtaining a visual evaluation coefficient of the feature words, and obtaining the visual evaluation coefficient of the feature words. And comparing differences between the character evaluation coefficient and the visual evaluation coefficient of the same feature word, accumulating the differences to obtain difference evaluation values of the phrases and the pictures, and matching the picture and the phrases with the smallest difference evaluation value with each phrase.

Description

Technical Field

[0001] The present invention relates to the technical field of advertising information management, and specifically to an AI-based intelligent advertising promotion service system and method. Background Art

[0002] As the application scenarios of online platforms continue to enrich, with the help of the convenience and intensive working mode of the network environment, advertising on online platforms has become an important way of advertising. In the prior art, in order to adapt to the management mode of the network platform, staff are usually required to upload the text description and picture introduction of the product separately. When reading advertisements, readers often need to read the text description and picture introduction at the same time to better understand the product information. When the text and pictures of the advertisement are not strongly related, there is a possibility that readers cannot fully grasp the information that the advertiser wants to convey. Therefore, due to the poor reading experience of the advertisement, the advertiser cannot get the expected advertising effect. Summary of the invention

[0003] The purpose of the present invention is to provide an AI-based intelligent advertising promotion service system and method to solve the problems raised in the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solutions: an AI-based intelligent advertising promotion service system and method, the method comprising:

[0005] Step S100: Acquire the image information and text information of the product to be promoted, and respectively identify the text content and image content corresponding to the same characteristic attribute of the product to be promoted in the image information and the text information;

[0006] Step S200: Acquire feature words corresponding to feature attributes of each paragraph of text information, combine the feature words in the sentence to obtain feature word pairs, and mark the feature words that appear and those that do not appear in the feature word pairs;

[0007] Step S300: Calculate the frequency of occurrence of the feature words in the paragraph, obtain the text evaluation coefficient of the feature words in the paragraph, collect the corresponding relationship between the feature words and the text evaluation coefficients, and form a first evaluation list of the feature words;

[0008] Step S400: Calculate the image proportion of the characteristic attribute in the picture, obtain the visual evaluation coefficient of the characteristic word, collect the corresponding relationship between the characteristic word and the visual evaluation coefficient, and form a second evaluation list of the characteristic word;

[0009] Step S500: compare the differences between the text evaluation coefficient and the visual evaluation coefficient of the same feature word, and accumulate the differences to obtain the difference evaluation values ​​between the paragraph and the picture, match the picture with the smallest difference evaluation value with each paragraph, and display the advertisement content after the paragraph and picture are matched to the readers.

[0010] Further, step S100 includes:

[0011] Step S101: Set a product to be promoted as the target product, collect several pictures of the target product, divide each picture into several image slices, extract the feature information of each picture slice, and gather the feature words in the feature information to obtain the feature word reference set U;

[0012] Step S102: Respectively obtain the text information and picture information of the target product from the promotion content of the target product. Among them, the text information includes at least two paragraphs, and the picture information includes at least two pictures of the target product.

[0013] Further, step S200 includes:

[0014] Step S201: Take any one paragraph as the target paragraph, obtain the number of sentences N in the target paragraph, and identify all the feature words in the target paragraph;

[0015] Step S202: Gather the feature words in the i-th sentence of the target paragraph to form the feature word set sen i , calculate the missing feature word set lack i , lack i = U - sen i ;

[0016] In a clause, there are feature words related to product features and feature words not involved in product features. Respectively count the feature words involved and not involved in each clause to obtain the distribution of feature words in the paragraph and get the feature words with a higher degree of involvement in the paragraph;

[0017] Step S203: Obtain any two feature words from the feature word set sen i to form the first feature word pair, obtain any one feature word from the feature word set sen i , obtain any one feature word from the missing feature word set lack i to form the second feature word pair;

[0018] Step S204: Set the first relationship evaluation value a and the second feature evaluation value b, satisfying the condition 0 < a < b, mark the relationship value of the first feature word pair as a, and mark the relationship value of the second feature word pair as b.

[0019] Further, step S300 includes:

[0020] Step S301: Collect all the feature words in the target text segment to obtain a text segment feature word set, and obtain the j-th feature word w in the text segment feature word set j , and collect all the first feature word pairs and second feature word pairs that include the j-th feature word w j in the target text segment;

[0021] Step S302: Calculate the text evaluation coefficient α of the j-th feature word in the text segment feature word set in the target text segment j ,

[0022] α j = (m 1j × a + m 2j × b) ÷ (m 1j + m 2j ), where m 1j is the number of first feature word pairs that include the feature word w j in the target text segment, and m 2j is the number of second feature word pairs that include the feature word w j in the target text segment;

[0023] In the target text segment, when a certain feature word is mentioned more times, there are more first feature word pairs containing this word, and the value of the text evaluation coefficient of this feature word in the target text segment is smaller. When a certain feature word is mentioned less times, there are more second feature word pairs including this word, and the value of the text evaluation coefficient of this feature word in the target text segment is larger;

[0024] Step S303: Calculate the text evaluation coefficients of all feature words in the target text segment respectively, collect the feature words and the corresponding text evaluation coefficients in the target text segment, and record the corresponding relationship between the feature words and the text evaluation coefficients in the first evaluation list of the feature words.

[0025] Further, step S400 includes:

[0026] Step S401: Obtain the k-th picture of the target product, identify the feature information in the k-th picture, and obtain the image slices of each feature information in the k-th picture;

[0027] Step S402: Obtain the image area D k of the k-th picture, the image slice sli p corresponding to the p-th feature information in the k-th picture, obtain the image area D p of the image slice sl i sp , and calculate the visual evaluation coefficient β p of the p-th feature information, β p = D sp / D k ;

[0028] Step S403: Calculate the visual evaluation coefficients of all types of feature information in the k-th picture respectively, obtain the feature words of the feature information, collect the visual evaluation coefficients of all feature words in the k-th picture, and record the corresponding relationship between the feature words and the visual evaluation coefficients in the second evaluation list of the feature words;

[0029] Obtain the proportion of the picture content corresponding to different feature attributes in a certain picture in the image by the proportion of the picture content features in the image.

[0030] Further, step S500 includes:

[0031] Step S501: Obtain the visual evaluation coefficients of each feature word in the second evaluation list, arrange the feature words in descending order according to the visual evaluation coefficients to obtain a feature word reference sequence, and set a unit weight value E, where,

[0032] E>0, calculate the reference weight value γ of the r1-th feature word in the feature word reference sequence r1 , γ r1 =β r1 ×E;

[0033] Step S502: According to the order of the feature word reference sequence, collect the reference weight values corresponding to each feature word in the sequence to obtain a reference weight sequence;

[0034] Step S503: Obtain the literal evaluation coefficients of all feature words in the target text segment, arrange the feature words in ascending order according to the literal evaluation coefficients to obtain a feature word comparison sequence, and assign weights to the feature words in the feature word comparison sequence in order according to the order of the reference weight values in the reference weight sequence, to obtain a feature word comparison sequence of the feature words. In the feature word comparison sequence, the weight value of the feature word is the comparison weight value of the feature word;

[0035] Step S504: Obtain the reference weight value of the same feature word in the feature word reference sequence and the comparison weight value of the feature word comparison sequence, form an evaluation group of the same feature word, collect the evaluation groups of all feature words in the target text segment to obtain a feature word comparison sequence L, where L includes q rating groups,

[0036] where, L: f 1 (g 11 , g 12 ), f 2 (g 21 , g 22 ), f 3 (g 31 , g 32 ), ……f q (g q1 , g q2), where the v-th evaluation group f v (g v1 , g v2 ), in which g v1 represents the reference weight value of the v-th feature word f v of the k-th picture, and g v2 represents the comparison weight value of the v-th feature word f v ;

[0037] Step S505: Calculate the difference evaluation value H between the target text segment and the k-th picture, Obtain the difference evaluation values between the target text segment and each picture, and take the picture with the smallest difference evaluation value as the matching picture of the target text segment;

[0038] Step S506: Obtain the matching pictures of each text segment, and display the pictures corresponding to each text segment to the readers of the advertisement on the display interface of the advertisement.

[0039] Arrange the distribution rates of the features in the text and in the pictures from high to low respectively, obtain the weight relationship of each feature in the picture, get the reference weight sequence, and assign values to the text features in the order of the reference weight sequence; when the positions of the same feature word in the two sequences are the same, the difference in weight is 0. Therefore, in the two sequences, the higher the similarity of the feature word sorting, the lower the value of the difference evaluation value.

[0040] To better implement the above method, an AI-based intelligent advertising promotion service system is also proposed. The system includes:

[0041] An information management module, a feature word difference management module, a text segment evaluation module, a picture evaluation module, and a text-picture matching module. Among them, the information management module is used to manage the picture information and text information of the product to be promoted. The feature word difference management module is used to mark the differences between feature words. The text segment evaluation module is used to calculate the text evaluation coefficient of feature words in the text segment. The picture evaluation module is used to evaluate the visual evaluation coefficient of the feature content in the picture. The text-picture matching module is used to match the text segment with the corresponding picture;

[0042] Furthermore, the information management module includes: a feature management unit, a feature word management unit, and an information storage unit. Among them, the feature management unit is used to manage the text features and picture features of the promotion information. The feature word management unit is used to manage the reference set of feature words. The information storage unit is used to store the text information and picture information of the target product respectively;

[0043] Furthermore, the feature word difference management module includes: a paragraph management unit, a feature word pair management unit and a relationship value marking unit, wherein the paragraph management unit is used to obtain a paragraph and collect the feature words of each sentence in the paragraph respectively, the feature word pair management unit is used to manage the feature value pairs in the sentence, and the relationship value marking unit is used to mark the relationship value of the first feature word pair and the second feature word pair respectively;

[0044] Further, the paragraph evaluation module includes: a classification management unit, a text evaluation coefficient calculation unit and a first evaluation list management unit, wherein the classification management unit is used to respectively manage a first feature word pair and a second feature word pair including the same feature word, the text evaluation coefficient calculation unit is used to calculate the text evaluation coefficient of the feature word in the target paragraph, and the first evaluation list management unit is used to manage a first evaluation list of the feature word in the target paragraph;

[0045] Further, the picture evaluation module includes: a slice management unit, a visual evaluation coefficient calculation unit and a second evaluation list management unit, wherein the slice management unit is used to manage image slices corresponding to feature information in the picture, the visual evaluation coefficient calculation unit is used to calculate the visual evaluation coefficient of the feature information, and the second evaluation list management unit is used to manage the second evaluation list corresponding to the picture;

[0046] Furthermore, the image-text matching module includes: an image feature weight management unit, a weight sequence management unit, a weight assignment unit and a difference evaluation unit, wherein the image feature weight management unit is used to manage the reference weight value corresponding to the feature content in the image, the weight sequence management unit is used to manage the reference weight sequence, the weight assignment unit is used to assign weight values ​​for comparing feature words in the paragraph, and the difference evaluation unit is used to use the picture with the smallest difference evaluation value as the matching picture of the target paragraph based on the difference evaluation values ​​between the paragraph and each picture.

[0047] Compared with the prior art, the beneficial effects of the present invention are: by identifying the text content features and image content features respectively, the feature content is screened and intelligently matched, and a corresponding matching model between text paragraphs and pictures is established. On the advertising content editing side, the existing workflow is retained as much as possible, while reducing the time for relevant personnel to adjust the content and layout; on the advertising content reading side, readers can intuitively understand the advertising content through matching pictures when reading text information. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a structural schematic diagram of an AI-based intelligent advertising promotion service system of the present invention;

[0049] Figure 2 The present invention is a flowchart of an AI-based intelligent advertising promotion service method. DETAILED DESCRIPTION

[0050] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0051] Embodiment: As Figure 1 and Figure 2 shown, the present invention provides a technical solution, an AI-based intelligent advertising promotion service system and method, wherein the method includes:

[0052] Step S100: Obtain the picture information and text information of the product to be promoted, and respectively identify the text content and picture content corresponding to the same characteristic attribute of the product to be promoted in the picture information and text information;

[0053] Among them, step S100 includes:

[0054] Step S101: Set a certain product to be promoted as the target product, collect several pictures of the target product, divide each picture into several image slices, extract the feature information of each picture slice, and collect the feature words in the feature information to obtain the feature word reference set U;

[0055] Step S102: Respectively obtain the text information and picture information of the target product from the promotion content of the target product, wherein the text information includes at least two paragraphs, and the picture information includes at least two pictures of the target product;

[0056] In the embodiment, in the images of different products, the distribution ratios of different features are different. For example, in the overall appearance picture of the product, there are more types of product features, and from the perspective of image presentation, the distribution of features is relatively uniform. In the local design picture of the product, one or several local features of the product will be highlighted, and the proportion of local features in the image will increase;

[0057] In the embodiment, the feature information includes the shape, color and texture of the product.

[0058] Step S200: Obtain the feature words corresponding to the characteristic attributes of each paragraph of the text information, combine the feature words in the sentence to obtain feature word pairs, and mark the feature words that appear and the feature words that do not appear in the feature word pairs;

[0059] Among them, step S200 includes:

[0060] Step S201: Take any paragraph as the target paragraph, obtain the number of sentences N in the target paragraph, and identify all the characteristic words in the target paragraph;

[0061] Step S202: Gather the characteristic words in the i-th sentence of the target paragraph to form the characteristic word set sen of the i-th sentence i , calculate the missing characteristic word set lack i , lack i = U - sen i ;

[0062] Step S203: Obtain any two characteristic words from the characteristic word set sen i to form the first characteristic word pair, obtain any one characteristic word from the characteristic word set sen i , and obtain any one characteristic word from the missing characteristic word set lack i to form the second characteristic word pair;

[0063] Step S204: Set the first relationship evaluation value a and the second characteristic evaluation value b, satisfying the condition 0 < a < b, mark the relationship value of the first characteristic word pair as a, and mark the relationship value of the second characteristic word pair as b;

[0064] For example, if U includes 4 characteristic words: Q1, Q2, Q3, Q4, and a certain sentence involves the characteristic words Q1, Q2, and Q3, then the first characteristic word pairs of this sentence include (Q1, Q2), (Q1, Q3), and (Q2, Q3), and the second characteristic word pairs include (Q1, Q4), (Q2, Q4), and (Q3, Q4).

[0065] Step S300: Calculate the frequency of occurrence of the characteristic words in the paragraph to obtain the text evaluation coefficient of the characteristic words in the paragraph, gather the corresponding relationship between the characteristic words and the text evaluation coefficient, and form the first evaluation list of the characteristic words;

[0066] Among them, Step S300 includes:

[0067] Step S301: Gather all the characteristic words in the target paragraph to obtain the paragraph characteristic word set, and obtain the j-th characteristic word w in the paragraph characteristic word set j , collect all the first characteristic word pairs and second characteristic word pairs that include the j-th characteristic word w j in the target paragraph;

[0068] Step S302: Calculate the text evaluation coefficient α j ,

[0069] α j = (m 1j × a + m2j × b) ÷ (m 1j + m 2j ), where m 1j is the number of the first feature word pairs including the feature word w j in the target text segment, and m 2j is the number of the second feature word pairs including the feature word w j in the target text segment;

[0070] Step S303: Calculate the literal evaluation coefficients of all the feature words in the target text segment respectively, collect the feature words and the corresponding literal evaluation coefficients in the target text segment, and record the corresponding relationship between the feature words and the literal evaluation coefficients in the first evaluation list of the feature words.

[0071] Step S400: Calculate the image proportion of the feature attributes in the picture to obtain the visual evaluation coefficients of the feature words, collect the corresponding relationship between the feature words and the visual evaluation coefficients, and form the second evaluation list of the feature words;

[0072] Among them, Step S400 includes:

[0073] Step S401: Obtain the k-th picture of the target product, identify the feature information in the k-th picture, and obtain the image slice of each feature information in the k-th picture;

[0074] Step S402: Obtain the image area D k of the k-th picture, the image slice sli p corresponding to the p-th feature information in the k-th picture, obtain the image area D p of the image slice sli sp , and calculate the visual evaluation coefficient β p of the p-th feature information, β p = D sp / D k ;

[0075] Step S403: Calculate the visual evaluation coefficients of all types of feature information in the k-th picture respectively, obtain the feature words of the feature information, collect the visual evaluation coefficients of all the feature words in the k-th picture, and record the corresponding relationship between the feature words and the visual evaluation coefficients in the second evaluation list of the feature words.

[0076] Step S500: Compare the differences between the literal evaluation coefficients and the visual evaluation coefficients of the same feature word, and accumulate the differences to obtain the difference evaluation value between the text segment and the picture. Match the picture with the smallest difference evaluation value for each text segment with the text segment, and display the advertising content after matching the text segment with the picture to the readers;

[0077] Among them, Step S500 includes:

[0078] Step S501: Obtain the visual evaluation coefficients of each feature word in the second evaluation list, arrange the feature words in descending order according to the visual evaluation coefficients to obtain a reference sequence of feature words, set a unit weight value E, where E > 0, and calculate the reference weight value γ of the r1-th feature word in the reference sequence of feature words r1 , γ r1 = β r1 ×E;

[0079] Step S502: According to the order of the reference sequence of feature words, pool the reference weight values corresponding to each feature word in the sequence to obtain a reference weight sequence;

[0080] Step S503: Obtain the literal evaluation coefficients of all feature words in the target text segment, arrange the feature words in ascending order according to the literal evaluation coefficients to obtain a comparison sequence of feature words, and assign weights to the feature words in the comparison sequence of feature words in order according to the order of the reference weight values in the reference weight sequence, to obtain a comparison sequence of feature words of the feature words. In the comparison sequence of feature words, the weight value of the feature word is the comparison weight value of the feature word;

[0081] Step S504: Obtain the reference weight value of the same feature word in the reference sequence of feature words and the comparison weight value of the comparison sequence of feature words, form an evaluation group of the same feature word, pool the evaluation groups of all feature words in the target text segment to obtain a comparison sequence of feature words L, where L includes q rating groups,

[0082] where, L: f 1 (g 11 , g 12 ), f 2 (g 21 , g 22 ), f 3 (g 31 , g 32 ), …… f q (g q1 , g q2 ), where in the v-th evaluation group f v (g v1 , g v2 ), g v1 represents the reference weight value of the v-th feature word f v of the k-th picture, and g v2 represents the comparison weight value of the v-th feature word f v ;

[0083] Step S505: Calculate the difference evaluation value H between the target text segment and the k-th picture, Obtain the difference evaluation values between the target text segment and each picture, and take the picture with the smallest difference evaluation value as the matching picture of the target text segment;

[0084] Step S506: Obtain the matching pictures for each text segment, and display the pictures corresponding to each text segment to the readers of the advertisement on the display interface of the advertisement.

[0085] In the embodiment, when the feature word reference sequence and the feature word comparison sequence include the same feature words, the above calculation method is adopted for the cases where the order positions are the same or different. When there is a certain feature word in one sequence but not in the other sequence, a placeholder weight value is set, and the weight of the placeholder weight value is assigned to the weight value of the above feature word to calculate the difference evaluation value. The value of the placeholder weight value is greater than the average value of the feature word weight values in the two sequences.

[0086] The system includes: an information management module, a feature word difference management module, a text segment evaluation module, a picture evaluation module, and a text-picture matching module;

[0087] Among them, the information management module is used to manage the picture information and text information of the product to be promoted. Among them, the information management module includes: a feature management unit, a feature word management unit, and an information storage unit. Among them, the feature management unit is used to manage the text features and picture features of the promotion information, the feature word management unit is used to manage the feature word reference set, and the information storage unit is used to store the text information and picture information of the target product respectively;

[0088] Among them, the feature word difference management module is used to mark the differences between feature words. Among them, the feature word difference management module includes: a text segment management unit, a feature word pair management unit, and a relationship value marking unit. Among them, the text segment management unit is used to obtain text segments and separately collect the feature words of each sentence in the text segment. The feature word pair management unit is used to manage the feature value pairs in the sentence, and the relationship value marking unit is used to separately mark the relationship values of the first feature word pair and the second feature word pair;

[0089] Among them, the text segment evaluation module is used to calculate the text evaluation coefficient of the feature word in the text segment. Among them, the text segment evaluation module includes: a classification management unit, a text evaluation coefficient calculation unit, and a first evaluation list management unit. Among them, the classification management unit is used to separately manage the first feature word pair and the second feature word pair including the same feature word. The text evaluation coefficient calculation unit is used to calculate the text evaluation coefficient of the feature word in the target text segment, and the first evaluation list management unit is used to manage the first evaluation list of the feature word in the target text segment;

[0090] Among them, the picture evaluation module is used for the visual evaluation coefficient of the feature content in the picture. Among them, the picture evaluation module includes: a slice management unit, a visual evaluation coefficient calculation unit, and a second evaluation list management unit. Among them, the slice management unit is used to manage the image slices corresponding to the feature information in the picture, the visual evaluation coefficient calculation unit is used to calculate the visual evaluation coefficient of the feature information, and the second evaluation list management unit is used to manage the second evaluation list corresponding to the picture;

[0091] Among them, the text-picture matching module is used to match the text segment with the corresponding picture. Among them, the text-picture matching module includes: a picture feature weight management unit, a weight sequence management unit, a weight assignment unit, and a difference evaluation unit. Among them, the picture feature weight management unit is used to manage the reference weight values corresponding to the feature content in the picture, the weight sequence management unit is used to manage the reference weight sequence, the weight assignment unit is used to assign comparison weight values to the feature words in the text segment, and the difference evaluation unit is used to take the picture with the smallest difference evaluation value as the matching picture of the target text segment according to the difference evaluation values between the text segment and each picture.

[0092] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and without departing from the spirit or basic characteristics of the present invention, the present invention can be implemented in other specific forms. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be regarded as limiting the claimed rights.

Claims

1. An AI-based intelligent advertising promotion service method, characterized by: The method comprises the steps of: Step S100: Acquire the image information and text information of the product to be promoted, and respectively identify the text content and image content corresponding to the same characteristic attribute of the product to be promoted in the image information and the text information; Step S200: Acquire feature words corresponding to feature attributes of each paragraph of text information, combine the feature words in the sentence to obtain feature word pairs, and mark the feature words that appear and those that do not appear in the feature word pairs; Step S300: Calculate the frequency of occurrence of the feature words in the paragraph, obtain the text evaluation coefficient of the feature words in the paragraph, collect the corresponding relationship between the feature words and the text evaluation coefficients, and form a first evaluation list of the feature words; Step S400: Calculate the image proportion of the characteristic attribute in the picture, obtain the visual evaluation coefficient of the characteristic word, collect the corresponding relationship between the characteristic word and the visual evaluation coefficient, and form a second evaluation list of the characteristic word; Step S500: compare the differences between the text evaluation coefficient and the visual evaluation coefficient of the same feature word, and accumulate the differences to obtain the difference evaluation values ​​between the paragraph and the picture, match the picture with the smallest difference evaluation value with each paragraph, and display the advertisement content after the paragraph and picture are matched to the readers.

2. The AI-based intelligent advertising promotion service method according to claim 1, characterized in that: Step S100 includes: Step S101: a product to be promoted is set as a target product, a number of pictures of the target product are collected, each picture is divided into a number of image slices, feature information of each image slice is extracted, feature words in the feature information are collected, and a feature word reference set U is obtained; Step S102: respectively obtaining text information and picture information of the target product from the promotional content of the target product, wherein the text information includes at least two paragraphs, and the picture information includes at least two pictures of the target product.

3. The AI-based intelligent advertising promotion service method according to claim 2, characterized in that: Step S200 includes: Step S201: taking any segment as a target segment, obtaining the number of sentences N in the target segment, and identifying all characteristic words in the target segment; Step S202: Collect the characteristic words in the i-th sentence in the target segment to form the characteristic word set sen of the i-th sentence i , calculate the missing feature word set lack of the i-th sentence i , lack i =U-sen i ; Step S203: From the feature word set sen i Get any two feature words to form the first feature word pair, and select the feature word set sen i Get any feature word from the missing feature word set lack i Get any feature word from the , and combine the two feature words into a second feature word pair; Step S204: setting a first relationship evaluation value a and a second feature evaluation value b, satisfying the condition 0<a<b, marking the relationship value of the first feature word pair as a, and marking the relationship value of the second feature word pair as b.

4. The AI-based intelligent advertising promotion service method according to claim 3, characterized in that: Step S300 includes: Step S301: Collect all the feature words in the target segment to obtain a segment feature word set, and obtain the jth feature word w in the segment feature word set. j , collect all the target paragraphs containing the j-th feature word w j All first feature word pairs and second feature word pairs of ; Step S302: Calculate the jth feature word w in the segment feature word set j The text evaluation coefficient α in the target segment j , α j =(m 1j ×a+m 2j ×b)÷(m 1j +m 2j ), where m 1j The target segment includes the feature word w j The number of first feature word pairs, m 2j The target segment includes the feature word w j The number of second feature word pairs; Step S303: Calculate the text evaluation coefficients of all feature words in the target segment respectively, collect the feature words and the corresponding text evaluation coefficients in the target segment, and record the correspondence between the feature words and the text evaluation coefficients in the first evaluation list of the feature words.

5. The AI-based intelligent advertising promotion service method according to claim 4, characterized in that: Step S400 includes: Step S401: obtaining the kth picture of the target product, identifying feature information in the kth picture, and obtaining image slices of each feature information in the kth picture; Step S402: Obtain the image area D of the kth image k , the image slice sli corresponding to the p-th feature information in the k-th image p , get the image slice sli p The image area D sp , calculate the visual evaluation coefficient β of the pth feature information p , β p =D sp / D k ; Step S403: Calculate the visual evaluation coefficients of all types of feature information in the kth picture respectively, obtain the feature words of the feature information, collect the visual evaluation coefficients of all feature words in the kth picture, and record the correspondence between the feature words and the visual evaluation coefficients in the second evaluation list of the feature words.

6. The AI-based intelligent advertising promotion service method according to claim 5, characterized in that: Step S500 includes: Step S501: Obtain the visual evaluation coefficient of each feature word in the second evaluation list, arrange the feature words from high to low according to the visual evaluation coefficient, obtain a feature word reference sequence, and set a unit weight value E, where: E>0, calculate the reference weight value γ of the r1th feature word in the feature word reference sequence r1 , γ r1 =β r1 ×E; Step S502: according to the order of the feature word reference sequence, the reference weight values ​​corresponding to each feature word in the sequence are collected to obtain a reference weight sequence; Step S503: obtaining the text evaluation coefficients of all feature words in the target segment, arranging the feature words from low to high according to the text evaluation coefficients to obtain a feature word comparison sequence, assigning weights to the feature words in the feature word comparison sequence in order according to the order of the reference weight values ​​in the reference weight sequence to obtain a feature word comparison sequence of the feature words, wherein the weight value of the feature word in the feature word comparison sequence is the comparison weight value of the feature word; Step S504: Obtain the reference weight value of the same feature word in the feature word reference sequence and the comparison weight value of the feature word comparison sequence, form an evaluation group for the same feature word, collect the evaluation groups of all feature words in the target segment, and obtain a feature word comparison sequence L, wherein L includes q rating groups. Where, L:f1(g 11 , g 12 ), f2(g 21 , g 22 ), f3(g 31 , g 32 ), ...f q (g q1 , g q2 ), in the vth evaluation group f v (g v1 , g v2 ), g v1 Represents the vth feature word f of the kth picture v The reference weight value, g v2 Indicates the vth feature word f v The comparison weight value of Step S505: Calculate the difference evaluation value H between the target segment and the kth picture. Obtain the difference evaluation value between the target paragraph and each picture, and use the picture with the smallest difference evaluation value as the matching picture of the target paragraph; Step S506: Acquire matching images for each paragraph, and display the images corresponding to each paragraph to readers of the advertisement in the advertisement display interface.

7. An AI-based intelligent advertising promotion service system, used to execute an AI-based intelligent advertising promotion service method according to any one of claims 1 to 6, characterized in that: The system includes: An information management module, a feature word difference management module, a paragraph evaluation module, a picture evaluation module and a picture-text matching module, wherein the information management module is used to manage the picture information and text information of the product to be promoted, the feature word difference management module is used to mark the differences between feature words, the paragraph evaluation module is used to calculate the text evaluation coefficient of the feature word in the paragraph, the picture evaluation module is used to evaluate the visual evaluation coefficient of the feature content in the picture, and the picture-text matching module is used to match the paragraph with the corresponding picture.

8. The AI-based intelligent advertising promotion service system according to claim 7, characterized in that: The information management module includes: a feature management unit, a feature word management unit and an information storage unit, wherein the feature management unit is used to manage the text features and picture features of the promotion information, the feature word management unit is used to manage the feature word reference set, and the information storage unit is used to store the text information and picture information of the target product respectively; The feature word difference management module includes: a paragraph management unit, a feature word pair management unit and a relationship value marking unit, wherein the paragraph management unit is used to obtain a paragraph and separately collect the feature words of each sentence in the paragraph, the feature word pair management unit is used to manage the feature value pairs in the sentence, and the relationship value marking unit is used to separately mark the relationship value between the first feature word pair and the second feature word pair.

9. The AI-based intelligent advertising promotion service system according to claim 7, characterized in that: The paragraph evaluation module includes: a classification management unit, a text evaluation coefficient calculation unit and a first evaluation list management unit, wherein the classification management unit is used to respectively manage a first feature word pair and a second feature word pair including the same feature word, the text evaluation coefficient calculation unit is used to calculate the text evaluation coefficient of the feature word in the target paragraph, and the first evaluation list management unit is used to manage a first evaluation list of the feature word in the target paragraph; The picture evaluation module includes: a slice management unit, a visual evaluation coefficient calculation unit and a second evaluation list management unit, wherein the slice management unit is used to manage image slices corresponding to feature information in the picture, the visual evaluation coefficient calculation unit is used to calculate the visual evaluation coefficient of the feature information, and the second evaluation list management unit is used to manage the second evaluation list corresponding to the picture.

10. The AI-based intelligent advertising promotion service system according to claim 7, characterized in that: The image-text matching module includes: an image feature weight management unit, a weight sequence management unit, a weight assignment unit and a difference evaluation unit, wherein the image feature weight management unit is used to manage the reference weight value corresponding to the feature content in the image, the weight sequence management unit is used to manage the reference weight sequence, the weight assignment unit is used to assign weight values ​​for comparing feature words in the paragraph, and the difference evaluation unit is used to use the picture with the smallest difference evaluation value as the matching picture of the target paragraph based on the difference evaluation values ​​between the paragraph and each picture.

Images