Brand image mining method and system

Abstract

The invention discloses a brand image mining method, which comprises the following steps: acquiring user comment data of each brand to obtain a comment set of each brand; processing the comment set of each brand to obtain an image word set and a word and sentence set of each brand; calculating the occurrence frequency and word vectors of the intention words in the intention word set to obtain an intention representation vector of each brand; processing the image representation vector to obtain a brand cognitive map; wherein the word and sentence set at least comprises an adjective set. The brand image information is fully extracted from massive comment texts by means of natural language technologies such as word vectors and the like, and the relative position distribution among the brands is intuitively expressed on a two-dimensional plane, so that the brand insight efficiency and accuracy are improved, and periodic experience monitoring work can be carried out at low cost.

Description

Technical Field

[0001] The present invention relates to the field of natural language processing, and in particular to a brand image mining method and system in the field of the Internet and artificial intelligence. Background Art

[0002] Brand image refers to the mental picture that consumers quickly evoke when they come into contact with the brand name. The mental picture is portrayed by screening and summarizing image words. The easier it is to quickly evoke the mental picture, the higher its priority tends to be. High-priority image words easily remind people of the appearance and internal quality of brand-related products and are easy to extract from memory. Nowadays, social media is becoming increasingly diversified, and the frequency of interaction between consumers and brands is increasing. Consumers can learn about brands, consume brands, evaluate brands, and promote brands through various channels such as online TV, forums, and short videos. Even if a company's potential consumers have not purchased related products, they have directly or indirectly come into contact with relevant information about the company's products during the process of participating in social media interactions, and have even built an initial impression of the company's brand in their minds. Building and optimizing brand image is a key topic for companies to carry out digital marketing. It is particularly important for corporate managers to grasp their own brand image and the brand image of competing products in a timely manner so that they can make scientific decisions. Summary of the invention

[0003] The main purpose of the present invention is to provide a brand image mining method and system to address the shortcomings of the prior art that it is difficult to efficiently and accurately obtain brand images.

[0004] To achieve the above object, the present invention provides a brand image mining method, comprising the following steps:

[0005] Obtain user review data for each brand and obtain a review collection for each brand;

[0006] Process the review collection of each brand to obtain the image word collection and the phrase collection of each brand;

[0007] The image representation vector of each brand is obtained by calculating the frequency of occurrence and word vector of the intention word in the intention word set;

[0008] Process the image representation vector to obtain the brand cognitive map;

[0009] The phrase set at least includes an adjective set.

[0010] In the brand image mining method provided by the present invention, the steps of processing the review set of each brand to obtain the image word set and the word and sentence set of each brand include:

[0011] Divide the review text of each brand in the review collection into sentences, and save the segmented short sentences into the text collection;

[0012] Determine the sentiment tendency of each short sentence in the text collection through the sentiment classification model;

[0013] If the sentiment classification model determines that the sentiment tendency of a short sentence in the text collection is positive, the word segmentation tool is used to segment and tag the short sentence, extract the tagged words and sentences in the short sentence and add them to the word and sentence list, and traverse all brands to obtain the word and sentence list of each brand;

[0014] Merge the word lists of all brands, remove duplicate elements from the merged word lists, and obtain an image word set;

[0015] The words and sentences at least include adjectives that comment on the brand, and the sentiment classification model at least includes a SKEP sentiment pre-training model.

[0016] In the brand image mining method provided by the present invention, the step of obtaining the brand image representation vector by calculating the occurrence frequency and word vector of the intention word in the intention word set includes:

[0017] The occurrence frequency of the intention words in the intention word set is used as the weight, and the image word vectors of each image word are weighted averaged to obtain the image representation vector of the brand.

[0018] In the brand image mining method provided by the present invention, the step of processing the image representation vector to obtain the brand cognitive map includes:

[0019] The word vector of each image word and the image representation vector of each brand are reduced to a two-dimensional vector through a dimensionality reduction model to obtain a two-dimensional vector of the image word and a two-dimensional vector of the image representation;

[0020] According to the relative positions of each brand and image word on the plane coordinate system, a brand recognition map is obtained;

[0021] The dimension reduction model at least includes a UMAP dimension reduction model.

[0022] In addition, to achieve the above purpose, the present invention also provides a brand image mining system, which includes an acquisition module, a processing module and a calculation module.

[0023] The acquisition module acquires user review data of each brand and obtains a review set of each brand;

[0024] The processing module processes the review set of each brand to obtain the image word set and the word and sentence set of each brand;

[0025] The calculation module obtains the image representation vector of each brand through the occurrence frequency and word vector calculation of the intention words in the intention word set;

[0026] The processing module also processes the image representation vector to obtain the brand recognition map.

[0027] The phrase set at least includes an adjective set.

[0028] In the brand image mining system provided by the present invention, the processing module divides the review text of each brand in the review collection into sentences, and saves the segmented short sentences into the text collection;

[0029] Determine the sentiment tendency of each short sentence in the text collection through the sentiment classification model;

[0030] If the sentiment classification model determines that the sentiment tendency of a short sentence in the text collection is positive, the word segmentation tool is used to segment and tag the short sentence, extract the tagged words and sentences in the short sentence and add them to the word and sentence list, and traverse all brands to obtain the word and sentence list of each brand;

[0031] Merge the word lists of all brands, remove duplicate elements from the merged word lists, and obtain an image word set;

[0032] The words and sentences at least include adjectives that comment on the brand, and the sentiment classification model at least includes a SKEP sentiment pre-training model.

[0033] In the brand image mining system provided by the present invention, the calculation module uses the occurrence frequency of the intention words in the intention word set as the weight, performs weighted average calculation on the image word vector of each image word, and obtains the brand image representation vector.

[0034] In the brand image mining system provided by the present invention, the processing module reduces the word vector of each image word and the image representation vector of each brand to a two-dimensional vector through a dimensionality reduction model to obtain the image word two-dimensional vector and the image representation two-dimensional vector;

[0035] According to the relative positions of each brand and image word on the plane coordinate system, a brand recognition map is obtained;

[0036] The dimension reduction model at least includes a UMAP dimension reduction model.

[0037] In addition, to achieve the above object, the present invention further provides a terminal device, including:

[0038] Memory for storing computer programs;

[0039] A processor is used to implement the steps of the above brand image mining method when executing a computer program.

[0040] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the steps of the above-mentioned brand image mining method are implemented.

[0041] The brand image mining method and system provided by the present invention fully extract brand image information from massive review texts with the help of natural language technologies such as word vectors, and intuitively represent the relative position distribution of various brands on a two-dimensional plane, thereby improving the efficiency and accuracy of brand insight and helping to carry out periodic experience monitoring work at a low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings in the following description are only embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on the provided drawings without creative work:

[0043] Figure 1 FIG. 4 is a flow chart of a brand image mining method provided by an embodiment of the present invention.

[0044] Figure 2 FIG. 4 is a schematic diagram of the structure of a brand image mining system provided by an embodiment of the present invention.

[0045] Figure 3 Shown is a schematic diagram of a terminal device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to facilitate understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Typical embodiments of the present invention are given in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive.

[0047] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used in the specification of the present invention herein are only for the purpose of describing specific embodiments and are not intended to limit the present invention.

[0048] In order to better understand the above-mentioned technical scheme, the above-mentioned technical scheme will be described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0049] Reference Figure 1 , Figure 1FIG. 1 is a flow chart of a brand image mining method provided by an embodiment of the present invention. In this embodiment, the brand image mining method includes:

[0050] Step S10, obtaining user review data of each brand to obtain a review set of each brand;

[0051] Step S20, processing the review set of each brand to obtain an image word set and a phrase set of each brand;

[0052] Step S30, obtaining the image representation vector of each brand through the occurrence frequency and word vector calculation of the intention words in the intention word set;

[0053] Step S40, processing the image representation vector to obtain a brand recognition map;

[0054] The phrase set at least includes an adjective set.

[0055] In one embodiment, user review data of each brand is obtained, that is, the user review text of the brand is obtained to obtain a review set of each brand, and the name of the brand is recorded as brand i (i∈[1,n]), there are n brands in total, brand brand i The corresponding comment set is recorded as

[0056] Step S20 processes the review set of each brand to obtain an image word set and a phrase set of each brand, including:

[0057] Step S201, dividing the review text of each brand in the review set into sentences, and saving the segmented short sentences into the text set;

[0058] Step S202, determining the sentiment tendency of each short sentence in the text collection through a sentiment classification model;

[0059] Step S203: if the sentiment classification model determines that the sentiment tendency of the short sentence in the text set is positive, the short sentence is segmented and POS tagged using a word segmentation tool, the tagged words and phrases in the short sentence are extracted and added to the word and phrase list, and all brands are traversed to obtain the word and phrase list of each brand;

[0060] Step S204, merging the word lists of all brands, removing duplicate elements from the merged word list, and obtaining an image word set; wherein the words and sentences at least include adjectives that comment on the brand, and the sentiment classification model at least includes a SKEP sentiment pre-training model.

[0061] The image word set is a set of intention words summarized for each brand. The comment sets of each brand are processed to obtain the adjective set of each brand and the image word set summarized for all brands.

[0062] In one embodiment, taking the adjective list as an example, for the brand i Collection of comments Use punctuation marks as boundaries The text of each comment in Split into short sentences and save the segmented short sentences to the text collection middle.

[0063] For text collections Each short sentence in Use the sentiment classification model to output each short sentence If the sentiment is positive (positive emotion), such as praise, the word segmentation tool is used to segment the sentence. Perform word segmentation and part-of-speech tagging to extract short sentences Words marked as adjectives in the list are added to the adjective list In, traverse all brands brand i (i∈[1,n]) Get the adjective list of each brand Among them, the sentiment classification model can be a SKEP sentiment pre-training model (Sentiment Knowledge Enhanced Pre-training for Sentiment Analysis), and the word segmentation tool can be a Baidu LAC word segmentation tool.

[0064] Merge the adjective lists of all brands, remove duplicate elements from the merged adjective lists, and obtain the image word set Union.

[0065] Step S30 calculates the brand image representation vector by using the occurrence frequency and word vector of the intention word in the intention word set, including:

[0066] Step S301, taking the occurrence frequency of the intention words in the intention word set as the weight, performing weighted average calculation on the image word vectors of each image word, and obtaining the image representation vector of the brand.

[0067] In one embodiment, the number of words contained in the image word set Union is recorded as N Union , image word is recorded as w k (k∈[1,N Union ]), image word w k List of adjectives in various brands The frequency of occurrence in is recorded as Image word w kThe assigned word vector is denoted as Vector(w k )(k∈[1,N Union ]).

[0068] Tencent AI Lab Embedding Corpus for Chinese Words and Phrases can be used as the embedding vector for each image word w in the image word set Union. k Assign a 200-dimensional word vector Vector(w k )(k∈[1,N Union ]), brand i The image representation vector The calculation method is:

[0069]

[0070] That is Take the weighted average of the vectors at the same position.

[0071] Step S40 processes the image representation vector to obtain a brand recognition map, including:

[0072] Step S401, reducing the word vector of each image word and the image representation vector of each brand to a two-dimensional vector through a dimensionality reduction model, thereby obtaining an image word two-dimensional vector and an image representation two-dimensional vector;

[0073] Step S402, obtaining a brand recognition map according to the relative positions of each brand and image word on the plane coordinate system; wherein the dimensionality reduction model at least includes a UMAP dimensionality reduction model.

[0074] In one embodiment, the vector is represented by the imagery of all brands As a data sample, input the UMAP dimensionality reduction model for training, set the UMAP model parameter n_components = 2, which means that the dimension of the vector after dimensionality reduction is 2, and the image representation vector The corresponding image after dimensionality reduction represents a two-dimensional vector denoted as Image word w k -Word vector Vector(w k ) After the UMAP model reduces the dimension of the image word, the two-dimensional vector is recorded as Vector′(w k ).

[0075] Representing the image as a two-dimensional vector and the two-dimensional vector Vector′(w k ) as the direction of the x-axis, and the second dimension as the direction of the y-axis, representing the image as a two-dimensional vector As a corresponding brand i The coordinate point of the image word is the two-dimensional vector Vector′(w k ) is used as the coordinate point of the image word to represent the relative position of all brands and all image words on the plane coordinate system. The image words near each brand are circled respectively. In the brand recognition map, the closer the image word is to a brand, the more the image word can represent the brand.

[0076] Based on the above method, with the help of natural language technologies such as word vectors, brand image information can be fully extracted from massive comment texts, and the relative position distribution of each brand can be intuitively represented on a two-dimensional plane, which improves the efficiency and accuracy of brand insights and helps to carry out periodic experience monitoring work at a low cost.

[0077] Reference Figure 2 Accordingly, the present invention also provides a brand image mining system 200, the system 200 includes an acquisition module 201, a processing module 202 and a calculation module 203.

[0078] The acquisition module 201 acquires user comment data of each brand to obtain a comment set of each brand;

[0079] The processing module 202 processes the review set of each brand to obtain an image word set and a word and sentence set of each brand;

[0080] The calculation module 203 calculates the image representation vector of each brand through the occurrence frequency and word vector of the intention word in the intention word set;

[0081] The processing module 202 also processes the image representation vector to obtain a brand recognition map;

[0082] The phrase set at least includes an adjective set.

[0083] Furthermore, in one embodiment of the present invention, the processing module 202 divides the review text of each brand in the review set into sentences, and saves the divided short sentences into the text set;

[0084] Determine the sentiment tendency of each short sentence in the text collection through the sentiment classification model;

[0085] If the sentiment classification model determines that the sentiment tendency of a short sentence in the text collection is positive, the word segmentation tool is used to segment and tag the short sentence, extract the tagged words and sentences in the short sentence and add them to the word and sentence list, and traverse all brands to obtain the word and sentence list of each brand;

[0086] Merge the word lists of all brands, remove duplicate elements from the merged word lists, and obtain the image word set and the word set of each brand;

[0087] The words and sentences at least include adjectives that comment on the brand, and the sentiment classification model at least includes a SKEP sentiment pre-training model.

[0088] Furthermore, in an embodiment of the present invention, the calculation module 203 uses the occurrence frequency of the intention words in the intention word set as a weight, performs weighted average calculation on the image word vectors of each image word, and obtains the image representation vector of the brand.

[0089] Furthermore, in one embodiment of the present invention, the processing module 202 reduces the word vector of each image word and the image representation vector of each brand to a two-dimensional vector through a dimensionality reduction model to obtain an image word two-dimensional vector and an image representation two-dimensional vector;

[0090] According to the relative positions of each brand and image word on the plane coordinate system, a brand recognition map is obtained;

[0091] The dimension reduction model at least includes a UMAP dimension reduction model.

[0092] Reference Figure 3 The embodiment of the present invention further provides a terminal device 100, which may include:

[0093] A memory 101, used for storing computer programs;

[0094] The processor 102, when used to execute the computer program stored in the memory 101, can implement the following steps:

[0095] Process the review collection of each brand to obtain the image word collection and the phrase collection of each brand;

[0096] The image representation vector of each brand is obtained by calculating the frequency of occurrence and word vector of the intention words in the intention word set;

[0097] Process the image representation vector to obtain the brand cognitive map;

[0098] The phrase set at least includes an adjective set.

[0099] The embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following steps can be implemented;

[0100] Process the review collection of each brand to obtain the image word collection and the phrase collection of each brand;

[0101] The image representation vector of each brand is obtained by calculating the frequency of occurrence and word vector of the intention word in the intention word set;

[0102] Process the image representation vector to obtain the brand cognitive map;

[0103] The phrase set at least includes an adjective set.

[0104] The computer-readable storage medium may include: a USB flash drive, a mobile hard disk, a read-only memory (ROM)> a random access memory (RAM), a magnetic disk or an optical disk, and other media that can store program codes.

[0105] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0106] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, various features of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the intention that the claimed invention requires more features than those explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0107] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0108] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0109] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components in the embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., computer program and computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0110] It should be noted that the above embodiments illustrate rather than limit the invention, and that a person skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of means, several of these means may be embodied by the same item of hardware.

Claims

1. A brand image mining method, characterized in that: The method comprises the following steps: Obtain user review data for each brand to obtain a review set for each brand; Processing the review collections of each brand to obtain an image word collection and a word and sentence collection of each brand; The image representation vector of each brand is obtained by calculating the occurrence frequency and word vector of the intention word in the intention word set; Processing the image representation vector to obtain a brand cognition map; The phrase set at least includes an adjective set.

2. The brand image mining method according to claim 1, characterized in that: The step of processing the review sets of each brand to obtain an image word set and a word and sentence set of each brand includes: Divide the review text of each brand in the review collection into sentences, and save the segmented short sentences into the text collection; Determine the sentiment tendency of each short sentence in the text collection by using a sentiment classification model; If the sentiment classification model determines that the sentiment tendency of the short sentence in the text collection is positive, the short sentence is segmented and tagged with parts of speech using a word segmentation tool, the tagged words and sentences in the short sentence are extracted and added to the word and sentence list, and all brands are traversed to obtain the word and sentence list of each brand; Merge the word lists of all brands, remove duplicate elements from the merged word lists, and obtain an image word set; The word and sentence list includes at least an adjective list, and the sentiment classification model includes at least a SKEP sentiment pre-training model.

3. The brand image mining method according to claim 2, characterized in that: The step of calculating the image representation vector of the brand by using the occurrence frequency and word vector of the intention words in the intention word set includes: The occurrence frequency of the intention words in the intention word set is used as the weight, and the image word vectors of the image words are weighted averaged to obtain the image representation vector of the brand.

4. The brand image mining method according to claim 3, characterized in that: The step of processing the image representation vector to obtain a brand awareness map includes: The word vector of each image word and the image representation vector of each brand are reduced to a two-dimensional vector through a dimensionality reduction model to obtain a two-dimensional vector of the image word and a two-dimensional vector of the image representation; According to the relative positions of each brand and image word on the plane coordinate system, a brand recognition map is obtained; Wherein, the dimension reduction model at least includes a UMAP dimension reduction model.

5. A brand image mining system, characterized in that: The system includes an acquisition module, a processing module and a calculation module. The acquisition module acquires user comment data of each brand to obtain a comment set of each brand; The processing module processes the review set of each brand to obtain an image word set and a word and sentence set of each brand; The calculation module calculates the image representation vector of each brand through the occurrence frequency and word vector of the intention word in the intention word set; The processing module further processes the image representation vector to obtain a brand recognition map; Wherein, the phrase set at least includes an adjective set.

6. The brand image mining system according to claim 5, characterized in that: The processing module divides the review text of each brand in the review set into sentences, and saves the segmented short sentences into the text set; Determine the sentiment tendency of each short sentence in the text collection by using a sentiment classification model; If the sentiment classification model determines that the sentiment tendency of the short sentence in the text collection is positive, the short sentence is segmented and tagged with parts of speech using a word segmentation tool, the tagged words and sentences in the short sentence are extracted and added to the word and sentence list, and all brands are traversed to obtain the word and sentence list of each brand; Merge the word lists of all brands, remove duplicate elements from the merged word lists, and obtain an image word set; The words and sentences at least include adjectives that comment on the brand, and the sentiment classification model at least includes a SKEP sentiment pre-training model.

7. The brand image mining system according to claim 6, characterized in that: The calculation module uses the occurrence frequency of the intention words in the intention word set as a weight, performs weighted average calculation on the image word vectors of each image word, and obtains the image representation vector of the brand.

8. The brand image mining system according to claim 7, characterized in that: The processing module reduces the word vector of each image word and the image representation vector of each brand to a two-dimensional vector through a dimensionality reduction model to obtain an image word two-dimensional vector and an image representation two-dimensional vector; According to the relative positions of each brand and image word on the plane coordinate system, a brand recognition map is obtained; Wherein, the dimension reduction model at least includes a UMAP dimension reduction model.

9. A terminal device, characterized in that: The terminal device comprises: a memory, a processor, and a data processing program stored in the memory and executable on the processor, wherein the data processing program implements the method according to any one of claims 1 to 4 when executed by the processor.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a data processing program, and when the data processing program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images