Brand identification method and device for materials

By conducting brand relationship analysis and level screening of creative text information, the industry analysis model is used to improve brand recognition accuracy and efficiency, the problem of low brand recognition accuracy and efficiency in the existing technology is solved, and adapting to complex and changeable creatives.

CN113420146BActive Publication Date: 2024-11-29GUANGZHOU YOUMI INFORMATION TECH
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202110644897.1
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-06-09
Publication Date
2024-11-29
Estimated Expiration
2041-06-09

AI Technical Summary

Technical Problem

The prior art has low brand identification accuracy and efficiency in creative text information, making it particularly difficult to deal with complex and changeable creative text.

Method used

By performing brand identification operations on text information of the identification material, judging the relationship between brands, analyzing brand collections, determining levels or filtering brands with the longest string length, and using industry analysis models and brand recognition models to improve brand identification accuracy and efficiency .

Benefits of technology

No large deep network resources are required to improve the accuracy and efficiency of brand identification in creative text information, optimize the recall rate of brand identification, and adapt to complex and changeable creatives.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113420146B_ABST
    Figure CN113420146B_ABST
Patent Text Reader

Abstract

The present invention discloses a material brand recognition method and device. After the brand of the text information of the material to be recognized is recognized, the brand result corresponding to the material is determined by automatically analyzing whether there are brands with a preset brand relationship among multiple brands in the recognized text information, such as whether there are brands with a parent-child brand relationship. No large-scale deep network resources (such as Bert, etc.) are required, nor are powerful equipment resources and training resources. Even when dealing with complex and changeable advertising materials, the recognition accuracy and efficiency of the brand in the text information of the material (such as the advertising material) can be improved by using only limited resources, thereby improving the determination accuracy and efficiency of the brand result corresponding to the material, and optimizing the recall rate of brand recognition.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a material brand recognition method and device. Background Art

[0002] With the rapid development of Internet technology, e-commerce is also developing rapidly. Based on the characteristics of fast transmission efficiency and wide coverage, Internet advertising (such as advertising on TV, websites and other media) has become an important means of promoting products. Through Internet advertising, the information to be conveyed can be transmitted to the consumer end to attract the attention of consumers.

[0003] In practice, when a large amount of advertising creative text needs to be sorted out, it is necessary to manually annotate and identify the brands in the text information of the advertising creatives. After obtaining the brands of the advertising creative texts, different dimensions of brands can be compared; or the brands of the advertising creative texts can be identified through general named entity recognition (NER). However, practice has found that manual annotation has the problems of low accuracy and low efficiency. General NER is only applicable to the text information of regular advertising creatives and still cannot identify the accurate brand. In addition, in order to attract the attention of the audience, the text style of the advertising creatives is often adjusted according to hot spots and is complex and changeable. Therefore, it is particularly important to propose a solution to improve the recognition accuracy of the brand of advertising creatives. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a material brand recognition method and device, which can improve the recognition accuracy and efficiency of the brand in the text information of the material.

[0005] In order to solve the above technical problems, the present invention discloses a material brand recognition method in a first aspect, the method comprising:

[0006] Performing a brand recognition operation on the text information of the identification material to obtain a brand set of the text information, wherein the brand set includes at least one brand;

[0007] Determine whether there is a brand with a preset brand relationship among all the brands in the brand set, and obtain a determination result;

[0008] According to the judgment result, all the brands in the brand set are analyzed to obtain brand results corresponding to the material to be identified.

[0009] As an optional implementation manner, in the first aspect of the present invention, analyzing all the brands in the brand set according to the judgment result to obtain the brand result corresponding to the material to be identified includes:

[0010] When the judgment result is used to indicate that there is a brand with the preset brand relationship among all the brands in the brand set, a brand with a next-lower level is determined from all the brands in the brand set as the brand result corresponding to the material to be identified.

[0011] As an optional implementation manner, in the first aspect of the present invention, analyzing all the brands in the brand set according to the judgment result to obtain the brand result corresponding to the material to be identified includes:

[0012] When the judgment result indicates that there is no brand having the preset brand relationship among all the brands in the brand set, determining the character string length of each brand among all the brands in the brand set;

[0013] According to the character string length, a brand with the longest character string length is selected from all the brands in the brand set as a brand result corresponding to the material to be identified.

[0014] As an optional implementation, in the first aspect of the present invention, before determining the character string length of each of the brands in the brand set, the method further includes:

[0015] Based on the determined industry brand set, judging whether there is a non-ambiguous brand among all the brands in the brand set, wherein all the brands in the industry brand set include ambiguous brands and non-ambiguous brands;

[0016] When it is determined that there is a non-ambiguous brand among all the brands in the brand set, all the non-ambiguous brands are screened from all the brands in the brand set;

[0017] The step of determining the character string length of each brand among all the brands in the brand set includes:

[0018] The string length of each of the brands among all the brands that are unambiguous is determined.

[0019] As an optional embodiment, in the first aspect of the present invention, the method further comprises:

[0020] When it is determined that there is no unambiguous brand among all the brands in the brand set, the text information is input into the determined industry analysis model for analysis, and an analysis result output by the industry analysis model is obtained as the industry to which the material to be identified belongs;

[0021] Determine whether there is a brand among all the brands in the brand set that does not match the industry to which the material to be identified belongs;

[0022] When it is determined that there is no brand that does not match the industry to which the to-be-identified material belongs, the operation of determining the character string length of each of the brands in the brand set is performed.

[0023] As an optional implementation manner, in the first aspect of the present invention, when it is determined that there is a brand that does not match the industry to which the material to be identified belongs, the brand that does not match the industry to which the material to be identified belongs is removed from all the brands in the brand set to obtain a first brand set;

[0024] The step of determining the character string length of each of the brands in the brand set includes:

[0025] The character string length of each of the brands in all the brands in the first set of brands is determined.

[0026] As an optional implementation, in the first aspect of the present invention, the source type of the text information includes a slogan source type and / or a description source type;

[0027] Furthermore, after removing brands that do not match the industry to which the text information belongs from all the brands in the brand set to obtain the first brand set, the method further includes:

[0028] Determine whether the number of brands in the first brand set is greater than or equal to 1. When it is determined that the number of brands in the first brand set is greater than or equal to 1, perform the operation of determining the string length of each of the brands in all the brands in the first brand set.

[0029] As an optional embodiment, in the first aspect of the present invention, the method further comprises:

[0030] When it is determined that the number of brands in the first brand set is less than 1, determining image text information related to the material to be identified, and performing a brand identification operation on the image text information to obtain a brand set of the image text information;

[0031] It is determined whether the number of brands in the brand set of the image text information is equal to 1. When the determination result is yes, the brand of the brand set of the image text information is determined to be the brand result corresponding to the material to be identified.

[0032] As an optional embodiment, in the first aspect of the present invention, the method further comprises:

[0033] Collecting training text information of training materials of multiple industries, the number of training materials of each industry is greater than or equal to the determined material number threshold, and the number of training text information of each training material is greater than or equal to the determined text number threshold;

[0034] Establishing an association relationship between each of the training materials and the industry to which the training material belongs, and performing a training operation on the determined classification model based on the training text information of all the training materials of each industry and the association relationship between each of the training materials and the training material, to obtain a trained classification model;

[0035] The trained classification model is determined to be the determined industry analysis model.

[0036] As an optional implementation manner, in the first aspect of the present invention, performing a brand recognition operation on the text information of the to-be-recognized material to obtain a brand set of the text information includes:

[0037] Inputting the text information of the material to be identified into the determined brand identification model for analysis, and obtaining the analysis result output by the brand identification model as the brand set of the text information;

[0038] And, the method further comprises:

[0039] Collecting a training brand set, wherein the training brand set includes brands from multiple industries;

[0040] Based on the initial recognition model determined by all the brands in the training brand set, a trained initial recognition model is obtained, and the trained initial recognition model is determined as the determined brand recognition model.

[0041] As an optional implementation manner, in the first aspect of the present invention, after performing a brand recognition operation on the text information of the to-be-recognized material to obtain a brand set of the text information, the method further comprises:

[0042] Determining whether there is a brand of the determined brand type among all the brands in the brand set;

[0043] When it is determined that no brand exists, the operation of determining whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a determination result is performed;

[0044] And, the method further comprises:

[0045] When it is determined that the brand exists, the brand of the brand type is deleted from all the brands in the brand set to obtain a second brand set, and the operation of determining whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a determination result is performed, and the second brand set is the brand set.

[0046] As an optional implementation manner, in the first aspect of the present invention, after performing a brand recognition operation on the text information of the to-be-recognized material to obtain a brand set of the text information, the method further comprises:

[0047] Determine whether all the brands in the brand set are main brands;

[0048] When the judgment result is yes, the operation of judging whether there is a brand with a preset brand relationship among all the brands in the brand set is performed to obtain a judgment result;

[0049] And, the method further comprises:

[0050] When the judgment result is no, the brands that are not main brands in all the brands in the brand set are converted into main names to obtain a third brand set of the text information, and the third brand set includes at least one main brand.

[0051] A second aspect of the present invention discloses a material brand identification device, the device comprising:

[0052] An identification module, configured to perform a brand identification operation on the text information of the identification material to obtain a brand set of the text information, wherein the brand set includes at least one brand;

[0053] A judgment module, used for judging whether there is a brand with a preset brand relationship among all the brands in the brand set, and obtaining a judgment result;

[0054] The analysis module is used to analyze all the brands in the brand set according to the judgment result to obtain the brand result corresponding to the material to be identified.

[0055] As an optional implementation, in the second aspect of the present invention, the analysis module analyzes all the brands in the brand set according to the judgment result to obtain the brand result corresponding to the material to be identified in the following manner:

[0056] When the judgment result is used to indicate that there is a brand with the preset brand relationship among all the brands in the brand set, a brand with a next-lower level is determined from all the brands in the brand set as the brand result corresponding to the material to be identified.

[0057] As an optional implementation, in the second aspect of the present invention, the analysis module includes:

[0058] a determination submodule, configured to determine the character string length of each of the brands in the brand set when the judgment result indicates that there is no brand having the preset brand relationship among all the brands in the brand set;

[0059] The screening submodule is used to screen the brand with the longest string length from all the brands in the brand set according to the string length as the brand result corresponding to the material to be identified.

[0060] As an optional implementation, in the second aspect of the present invention, the analysis module further includes:

[0061] a judging submodule, configured to judge whether there is a non-ambiguous brand among all the brands in the brand set based on the determined industry brand set before the determining submodule determines the character string length of each brand in all the brands in the brand set, wherein all the brands in the industry brand set include ambiguous brands and non-ambiguous brands;

[0062] The screening submodule is further configured to screen all the non-ambiguous brands from all the brands in the brand set when the judging submodule judges that there are non-ambiguous brands among all the brands in the brand set;

[0063] The determination submodule determines the character string length of each brand in all the brands in the brand set in the following manner:

[0064] The string length of each of the brands among all the brands that are unambiguous is determined.

[0065] As an optional implementation, in the second aspect of the present invention, the analysis module further includes:

[0066] an analysis submodule, configured to input the text information into the determined industry analysis model for analysis when the judgment submodule judges that there is no unambiguous brand among all the brands in the brand set;

[0067] An acquisition submodule, used for acquiring the analysis result output by the industry analysis model as the industry to which the material to be identified belongs;

[0068] The judging submodule is further used to judge whether there is a brand that does not match the industry to which the material to be identified belongs among all the brands in the brand set; when it is judged that there is no brand that does not match the industry to which the material to be identified belongs, the determining submodule is triggered to perform the operation of determining the string length of each of the brands in all the brands in the brand set.

[0069] As an optional implementation, in the second aspect of the present invention, the analysis module further includes:

[0070] a removal submodule, configured to remove the brands that do not match the industry to which the to-be-recognized material belongs from all the brands in the brand set when the judgment submodule judges that there are brands that do not match the industry to which the to-be-recognized material belongs, so as to obtain a first brand set;

[0071] The determination submodule determines the character string length of each brand in all the brands in the brand set in the following manner:

[0072] The character string length of each of the brands in all the brands in the first set of brands is determined.

[0073] As an optional implementation, in the second aspect of the present invention, the source type of the text information includes a slogan source type and / or a description source type;

[0074] Furthermore, the judging submodule is further used for eliminating brands that do not match the industry to which the text information belongs from all the brands in the brand set in the eliminating submodule, and after obtaining the first brand set, judging whether the number of brands in the first brand set is greater than or equal to 1; when it is judged that the number of brands in the first brand set is greater than or equal to 1, triggering the determining submodule to perform the operation of determining the string length of each of the brands in all the brands in the first brand set.

[0075] As an optional implementation, in the second aspect of the present invention, the determination submodule is further configured to determine image text information related to the material to be identified when the determination submodule determines that the number of brands in the first brand set is less than 1;

[0076] The recognition module is further used to perform a brand recognition operation on the image text information to obtain a brand set of the image text information;

[0077] The judging submodule is further used to judge whether the number of brands in the brand set of the image text information is equal to 1;

[0078] The determination submodule is further configured to determine, when the determination submodule determines that the result is yes, that the brand in the brand set of the image text information is the brand result corresponding to the material to be identified.

[0079] As an optional implementation, in the second aspect of the present invention, the device further includes:

[0080] A first collection module is used to collect training text information of training materials of multiple industries, the number of training materials of each industry is greater than or equal to a determined material number threshold, and the number of training text information of each training material is greater than or equal to a determined text number threshold;

[0081] An establishing module, used for establishing an association relationship between each of the training materials and the industry to which the training material belongs;

[0082] A first training module is used to perform a training operation on the determined classification model based on the training text information of all the training materials of each industry and the association relationship between each training material and the training material, so as to obtain a trained classification model;

[0083] The determination module is used to determine the trained classification model as the determined industry analysis model.

[0084] As an optional implementation, in the second aspect of the present invention, the recognition module performs a brand recognition operation on the text information of the recognition material, and obtains the brand set of the text information in a specific manner as follows:

[0085] Inputting the text information of the material to be identified into the determined brand identification model for analysis, and obtaining the analysis result output by the brand identification model as the brand set of the text information;

[0086] And, the device further comprises:

[0087] A second collection module is used to collect a training brand set, where the training brand set includes brands from multiple industries;

[0088] A second training module is used to obtain a trained initial recognition model based on the initial recognition model determined by training all the brands in the training brand set;

[0089] The recognition module is further used to determine the trained initial recognition model as the determined brand recognition model.

[0090] As an optional implementation, in the second aspect of the present invention, the judgment module is further used to, after the recognition module performs a brand recognition operation on the text information of the to-be-recognized material and obtains a brand set of the text information, determine whether a brand of the determined brand type exists among all the brands in the brand set; and when it is determined that the brand does not exist, trigger the operation of determining whether a brand with a preset brand relationship exists among all the brands in the brand set to obtain a determination result;

[0091] And, the device further comprises:

[0092] A deletion module is used to delete the brand of the brand type from all the brands in the brand set when it is determined that it exists, to obtain a second brand set, and to trigger the judgment module to perform the operation of judging whether there is a brand with a preset brand relationship among all the brands in the brand set, to obtain a judgment result, and the second brand set is this brand set.

[0093] As an optional implementation, in the second aspect of the present invention, the judgment module is further used for, after the recognition module performs a brand recognition operation on the text information of the material to be recognized and obtains a brand set of the text information, judging whether all the brands in the brand set are all main brands; when the judgment result is yes, performing the operation of judging whether there are brands with a preset brand relationship among all the brands in the brand set to obtain the judgment result;

[0094] And, the device further comprises:

[0095] The conversion module is used for, when the judgment result is no, performing primary name conversion on the brands that are not primary brands in all the brands in the brand set to obtain a third brand set of the text information, wherein the third brand set includes at least one primary brand.

[0096] The third aspect of the present invention discloses another material brand identification device, the device comprising:

[0097] A memory storing executable program code;

[0098] a processor coupled to the memory;

[0099] The processor calls the executable program code stored in the memory to execute part or all of the steps in the material brand identification method disclosed in the first aspect of the present invention.

[0100] The fourth aspect of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute some or all of the steps in the brand identification method of materials disclosed in the first aspect of the present invention.

[0101] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0102] In an embodiment of the present invention, a brand recognition operation is performed on the text information of the material to be recognized to obtain a brand set of the text information, and the brand set includes at least one brand; it is determined whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a judgment result; according to the judgment result, all the brands in the brand set are analyzed to obtain a brand result corresponding to the material. It can be seen that after the brand of the text information of the material to be recognized is recognized, the brand result corresponding to the material is determined by automatically analyzing whether there are brands with a preset brand relationship for multiple brands in the recognized text information, such as whether there are brands with a parent-child brand relationship. There is no need for large-scale deep network resources (such as Bert, etc.), nor for powerful equipment resources and training resources. Even when dealing with complex and changeable advertising materials, only limited resources can be used to improve the recognition accuracy and efficiency of brands in the text information of the material (such as advertising materials), thereby improving the determination accuracy and efficiency of the brand result corresponding to the material, and optimizing the recall rate of brand recognition. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0104] Figure 1 It is a flow chart of a material brand identification method disclosed in an embodiment of the present invention;

[0105] Figure 2 It is a flow chart of another material brand recognition method disclosed in an embodiment of the present invention;

[0106] Figure 3 It is a structural schematic diagram of a material brand identification device disclosed in an embodiment of the present invention;

[0107] Figure 4 It is a structural schematic diagram of another material brand identification device disclosed in an embodiment of the present invention;

[0108] Figure 5It is a structural schematic diagram of another material brand identification device disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0109] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings 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.

[0110] The terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish different objects rather than to describe a specific order. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, device, product or end including a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units that are not listed, or may optionally include other steps or units inherent to these processes, methods, products or ends.

[0111] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0112] The present invention discloses a material brand recognition method and device, which can, after identifying the brand of the text information of the material to be identified, automatically analyze whether there are brands with a preset brand relationship among multiple brands in the identified text information, such as whether there are brands with a parent-child brand relationship, to determine the brand result corresponding to the material, without requiring the resources of a large deep network (such as Bert, etc.), nor requiring powerful equipment resources and training resources. Even in response to complex and changeable advertising materials, only limited resources can be used to improve the recognition accuracy and efficiency of the brand in the text information of the material (such as the advertising material), thereby improving the determination accuracy and efficiency of the brand result corresponding to the material, optimizing the recall rate of brand recognition, and facilitating improving the classification accuracy of the material. The following are detailed descriptions.

[0113] Embodiment 1

[0114] See also Figure 1 , Figure 11 is a flow chart of a material brand identification method disclosed in an embodiment of the present invention. Figure 1 The described method can be applied to a material brand recognition device, such as a server, etc., and the embodiment of the present invention does not limit this. Figure 1 As shown, the brand identification method of the material may include the following operations:

[0115] 101. Perform a brand recognition operation on text information of a to-be-recognized material to obtain a brand set of the text information, where the brand set includes at least one brand.

[0116] In the embodiment of the present invention, the material to be identified includes any advertising material or non-advertising material whose brand needs to be identified. And the source type of the text information of the material to be identified includes one or more combinations of slogan / title source type, descriptive source type and image source type. The text information of the material to be identified can be automatically collected by the brand identification device of the material, or can be input by any user to the brand identification device of the material.

[0117] In the embodiment of the present invention, the brand of the brand set of the text information includes at least one of a Chinese name, an English name, an alias, and a brand ID.

[0118] In the embodiment of the present invention, it should be noted that when the number of brands included in the brand set is 0 and the source type of the text information of the material to be identified includes a slogan / title source type, a brand identification operation is performed on the image text information of the material to be identified.

[0119] 102. Determine whether there is a brand with a preset brand relationship among all brands in the brand set, and obtain a determination result.

[0120] In the embodiment of the present invention, the preset brand relationship is any brand relationship that can represent the relationship between brands, such as a parent-child brand relationship.

[0121] 103. According to the judgment result, all brands in the brand set are analyzed to obtain brand results corresponding to the material to be identified.

[0122] In the embodiment of the present invention, as an optional implementation mode, all brands in the brand set are analyzed according to the judgment result to obtain the brand result corresponding to the material to be identified, including:

[0123] When the judgment result is used to indicate that there is a brand with a preset brand relationship among all brands in the brand set, a brand with a next-lower level is determined from all brands in the brand set as the brand result corresponding to the material to be identified;

[0124] When the judgment result is used to indicate that there is no brand with a preset brand relationship among all brands in the brand set, determining the character string length of each brand in all brands in the brand set;

[0125] According to the string length, the brand with the longest string length is selected from all brands in the brand set as the brand result corresponding to the material to be identified.

[0126] In this optional implementation, the latter level is also referred to as the next level or a smaller level.

[0127] In an embodiment of the present invention, all brands in the brand set can be optionally sorted, and in this case, the brand corresponding to the longest string length is the brand ranked first in the brand set. Specifically: calculate the string length of each brand in the brand set, sort them from large to small according to the string length, and obtain a sorted brand set; or, for each brand in the brand set, use each brand as a keyword to search for a preset number (such as 100) of materials, and obtain the number of materials with real brands corresponding to each brand (such as 40) or the proportion of the preset number (such as 0.4), and sort the number corresponding to each brand or the proportion of the preset number from large to small according to the number, to obtain a sorted brand set. In this way, all brands in the brand set can be sorted by multiple methods, which can improve the sorting flexibility of all brands in the brand set.

[0128] For example, if the text information of the material to be identified is "Anbao sunscreen 100ml under Hou Shengtang", then the brand set of the text information of the material to be identified is [Hou Shengtang, Anbao]. If the relationship between [Hou Shengtang] and [Anbao] is a parent-child brand relationship, and [Hou Shengtang] is the parent brand and [Anbao] is the child brand, then [Anbao] is determined to be the brand result corresponding to "Anbao sunscreen 100ml under Hou Shengtang"; if the relationship between [Hou Shengtang] and [Anbao] is not a parent-child brand relationship, then the string length of [Hou Shengtang] and the string length of [Anbao] are determined, and the string length of [Hou Shengtang] is greater than the string length of [Anbao], then [Hou Shengtang] is determined to be the brand result corresponding to "Anbao sunscreen 100ml under Hou Shengtang".

[0129] It can be seen that this optional implementation method can improve the recognition accuracy of the brand results corresponding to the material, as well as the determination efficiency and accuracy of the brand results corresponding to the material by taking the brand of the next level or the brand corresponding to the longest character string length among all brands in the text information of the material to be identified as the brand result corresponding to the material according to different brand recognition results.

[0130] It can be seen that the implementation Figure 1The described method can, after identifying the brand of the text information of the material to be identified, automatically analyze whether there are brands with a preset brand relationship among multiple brands in the identified text information, such as whether there are brands with a parent-child brand relationship, to determine the brand result corresponding to the material. It does not require the resources of a large deep network (such as Bert, etc.), nor does it require powerful equipment resources and training resources. Even when dealing with complex and changeable advertising materials, it is possible to improve the recognition accuracy and efficiency of the brand in the text information of the material (such as the advertising material) by using only limited resources, thereby improving the determination accuracy and efficiency of the brand result corresponding to the material, and optimizing the recall rate of brand recognition.

[0131] In an optional embodiment, before determining the character string length of each brand in all brands in the brand set, the method further includes the following steps:

[0132] Based on the determined industry brand set, determine whether there is a non-ambiguous brand among all the brands in the brand set;

[0133] When it is determined that there are non-ambiguous brands among all brands in the brand set, all non-ambiguous brands are screened from all brands in the brand set;

[0134] In this optional embodiment, optionally, determining the string length of each brand in all brands in the brand set includes:

[0135] Determine the string length for each brand among all brands that are unambiguous.

[0136] In this optional embodiment, optionally, judging whether there is a non-ambiguous brand among all brands in the brand set based on the determined industry brand set includes:

[0137] All brands in the brand set are compared with all brands in the determined industry brand set to obtain a brand comparison result, which is used to indicate whether there are unambiguous brands or no unambiguous brands among all brands in the brand set.

[0138] In this optional embodiment, all brands in the industry brand set include ambiguous brands and unambiguous brands. Wherein, whether it is an ambiguous brand or an unambiguous brand, each brand includes one or more combinations of a main brand, a parent brand and a sub-brand, wherein the main brand, the parent brand and the sub-brand of each brand include one or more combinations of a Chinese brand, an English brand, an alias brand and a brand ID. Wherein, an ambiguous brand is a brand that belongs to a common word and / or a brand exists in at least one industry, for example: "Qingshan Brand" belongs to a common word, "Lanjian Brand" exists in the clothing industry and the catering industry, then "Qingshan Brand" and "Lanjian Brand" are both ambiguous brands. Further, the industry brand set also includes the industry to which each brand belongs. In this way, the more content included in the industry brand set, the more conducive it is to improve the accuracy, comprehensiveness and efficiency of the brand extracted from the text information of the material to be identified, thereby helping to improve the accuracy and efficiency of the brand results of the material to be identified. Wherein, the main brand is the final official name of the brand. The parent brand is the parent brand of the sub-brand, such as Hou Shengtang Anhushai, then Hou Shengtang is the parent brand, Anhushai is the sub-brand, and the main brand of Anhushai is Anbaofu.

[0139] It can be seen that this optional embodiment first analyzes whether there are unambiguous brands among all the brands in the identified text information before determining the string length of each brand in the text information of the material to be identified, and directly determines the string length of each unambiguous brand when there are unambiguous brands. This can reduce the situation where ambiguous brands in the identified text information are used as brand results corresponding to the material to be identified, thereby further improving the accuracy of determining the brand results corresponding to the material to be identified.

[0140] In another optional embodiment, the method may further include the following steps:

[0141] When it is determined that there is no unambiguous brand among all brands in the brand set (that is, all brands in the identified text information are ambiguous brands), the text information is input into the determined industry analysis model for analysis, and the analysis result output by the industry analysis model is obtained as the industry to which the material to be identified belongs;

[0142] Determine whether there is a brand in the brand set that does not match the industry to which the material to be identified belongs;

[0143] When it is determined that there is no brand that does not match the industry to which the material to be identified belongs, the above operation of determining the character string length of each brand in all brands in the brand set is performed.

[0144] It can be seen that when this optional embodiment determines that all brands in the identified text information are ambiguous brands, it further automatically determines the industry to which the material to be identified belongs based on the text information of the material to be identified, and when it is ensured that there are no brands that do not match the industry to which the material to be identified belongs among all the brands in the identified text information, it directly determines the string length of each identified brand. This can reduce the situation where existing brands that do not match the industry to which the material to be identified belongs are mistakenly identified as brand results corresponding to the material to be identified, thereby further improving the accuracy of determining the brand results corresponding to the material to be identified.

[0145] In yet another optional embodiment, the method may further include the following steps:

[0146] When it is determined that there is a brand that does not match the industry to which the material to be identified belongs, the brand that does not match the industry to which the material to be identified belongs is removed from all brands in the brand set to obtain a first brand set;

[0147] In this optional embodiment, optionally, determining the string length of each brand in all brands in the brand set includes:

[0148] Determine the string length for each brand among all brands in the first brand set.

[0149] It can be seen that when this optional embodiment determines that there are brands that do not match the industry to which the material to be identified belongs among all the brands of the identified text information, the unmatched brands are first eliminated, so that more accurate brands can be obtained, thereby improving the efficiency of determining the brand string length, thereby further improving the accuracy and efficiency of determining the brand results corresponding to the material to be identified.

[0150] In another optional embodiment, the source type of the text information includes a slogan / title source type and / or a description source type. And, after removing brands that do not match the industry to which the text information belongs from all brands in the brand set and obtaining the first brand set, the method may further include the following steps:

[0151] It is determined whether the number of brands in the first brand set is greater than or equal to 1. When it is determined that the number of brands in the first brand set is greater than or equal to 1, the above operation of determining the string length of each brand in all brands in the first brand set is performed.

[0152] It can be seen that, when the source type of the text information of the material to be identified is a slogan / title source type and / or a description source type and there is a brand that does not match the industry to which the material to be identified belongs, this optional embodiment further determines whether there is still a brand. If so, it continues to determine the string length of the brand. This can reduce the occurrence of the step of determining the string length of the brand even though the brand does not exist, thereby increasing the possibility of successfully determining the material to be identified and reducing the power consumption of the device.

[0153] In yet another optional embodiment, the method may further include the following operations:

[0154] When it is determined that the number of brands in the first brand set is less than 1, determining the image text information of the material to be identified, and performing a brand identification operation on the image text information to obtain a brand set of the image text information;

[0155] It is determined whether the number of brands in the brand set of the image text information is equal to 1. When the result of the determination is yes, the brand of the brand set of the image text information is determined to be the brand result corresponding to the material to be identified.

[0156] In this optional embodiment, optionally, when it is determined that the number of brands in the brand set of the image text information is greater than 1 or equal to 0, it is determined that the brand result corresponding to the material to be identified is empty, that is, there is no corresponding brand result for the material to be identified.

[0157] It can be seen that after this optional embodiment determines that the brand result corresponding to the material to be identified is not determined based on the text information of the slogan / title source type and / or the description source type, the text information of the image source type of the material to be identified is further identified and analyzed to obtain the brand result corresponding to the material to be identified. This can increase the probability of successfully determining the brand result corresponding to the material to be identified, and enrich the intelligent function of the brand identification device of the material.

[0158] In yet another optional embodiment, the method may further include the following steps:

[0159] Collect training text information of training materials of multiple industries, the number of training materials of each industry is greater than or equal to a determined material number threshold (such as 100 pieces), and the number of training text information of each training material is greater than or equal to a determined text number threshold (such as 500 pieces);

[0160] Establishing an association relationship between each training material and the industry to which the training material belongs, and performing a training operation on the determined classification model based on the training text information of all training materials in each industry and the association relationship between each training material and the training material, so as to obtain a trained classification model;

[0161] Determine the trained classification model as the determined industry analysis model.

[0162] In this optional embodiment, the multiple industries include but are not limited to at least one of the digital home appliance industry, beauty and personal care industry, clothing, shoes and bags industry, maternal and child industry, automobile industry, financial industry, sports and fitness industry, food and beverage industry and smart home industry.

[0163] In this optional embodiment, the determined classification model may include one or more combinations of SVM, LR, and Fast Text.

[0164] In this optional embodiment, optionally, after obtaining the trained classification model, the method may further include the following steps:

[0165] The trained classification model is continuously verified based on the verification text information of each verification material in each industry, and the accuracy corresponding to the verified classification model is continuously determined until the accuracy of the classification model is greater than or equal to the determined accuracy threshold (such as 95%). Then, the verification operation on the classification model is stopped, and the classification model verified for the last time is determined as the determined industry analysis model.

[0166] It can be seen that this optional embodiment can obtain an industry classification model that can be applied to industry identification of materials by training the existing classification model based on multiple training text information of each training material in each industry, which is conducive to accurately and quickly identifying the industry to which the material to be identified belongs based on the industry classification model, and further helps to improve the accuracy and efficiency of determining the brand results of the material to be identified; and based on the verification data, the obtained industry classification model is verified, and the verification operation on the industry classification model is stopped when the accuracy of the industry classification model reaches the set accuracy threshold, so that a more accurate industry classification model can be obtained, thereby further improving the recognition accuracy and efficiency of the industry to which the material to be identified belongs.

[0167] In yet another optional embodiment, performing a brand recognition operation on the text information of the identification material to obtain a brand set of the text information includes:

[0168] The text information of the material to be identified is input into the determined brand identification model for analysis, and the analysis result output by the brand identification model is obtained as the brand set of the text information.

[0169] And, the method may further include the following steps:

[0170] Collect a training brand collection, which includes brands from multiple industries;

[0171] Based on the initial recognition model determined by training all brands in the training brand set, the trained initial recognition model is obtained, and the trained initial recognition model is determined as the determined brand recognition model. The initial recognition model includes an Aho-Corasick automaton recognition model. Further, the initial recognition model also includes a BFS (Breadth First Search) search model.

[0172] In this alternative embodiment, for the descriptions of the brands and industries in the training brand set, please refer to the detailed description of the relevant content of the present invention, which will not be elaborated here.

[0173] For example, the training brand set includes three brands: [Half Good and Bad], [Graceful], and [Half]. The first root node is constructed as an empty node, the second root nodes [Happy], [Excellent], and [Half] are constructed, the third root nodes [Excellent], [Half], and [Beautiful] are constructed, the fourth root node [Half] is constructed, the fifth root node [Half] is constructed, and pointers are set behind each root node. At this time, the [Happy] of the second root node is linked to the [Excellent] of the third root node, the [Excellent] of the third root node is linked to the [Excellent] of the second root node and the [Half] of the fourth root node, the [Excellent] of the second root node is linked to the [Beautiful] of the third root node, the [Half] of the fourth root node is linked to the [Half] of the fifth root node and the [Half] of the second root node, the [Half] of the second root node is linked to the [Half] of the third root node, and the [Half] of the fifth root node is linked to the [Half] of the third root node. Among them, the pointers behind the [Happy], [Excellent], [Half] of the second root node, the [Beautiful] of the third root node, and the [Half] of the third root node are fail pointers and are all linked to the first root node (i.e., the empty node), indicating that no other characters are linked. It should be noted that other training brands are constructed in sequence to obtain the brand recognition model.

[0174] For example, the text information of the material to be recognized is "Half Good and Bad". First, search whether there is [Happy]. If it exists, then search whether there is [Excellent] after [Happy]. If [Excellent] exists, continue to search after [Excellent]. At this time, [Half] is searched, and the level of [Half] is the same as that of [Happy]. Continue to search after [Half], and [Half] is searched. Then, the brands in the brand set of the text information of the material to be recognized: "Half Good and Bad" include [Half Good and Bad] and [Half].

[0175] It can be seen that in this alternative embodiment, by training the existing recognition model based on multiple training text information of each training material in each industry, a brand recognition model that can be applied to the text information of the material can be obtained, which is conducive to accurately and quickly identifying the brands of the text information of the material to be recognized based on this brand recognition model.

[0176] In yet another optional embodiment, after performing a brand recognition operation on the text information of the to-be-recognized material and obtaining a brand set of the text information, the method may further include the following steps:

[0177] Determine whether there is a brand matching the determined common word among all brands in the brand set, and when it is determined that there is no brand matching the common word, perform the above operation of determining whether there is a brand having a preset brand relationship among all brands in the brand set to obtain a determination result;

[0178] The method may further comprise the following steps:

[0179] When it is determined that there is a brand that matches the common word, the brand that matches the common word is deleted from the brand set to obtain a fourth brand set, and the above-mentioned operation of determining whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a determination result is performed. At this time, the brand set in this step is the fourth brand set.

[0180] In this optional embodiment, different industries have corresponding common words. For example, the common words in the beauty and personal care industry are "after", "perfect" and "finish", and the common words in the digital home appliance industry are "line", "shell", etc.

[0181] It can be seen that this optional embodiment can obtain a more accurate brand set by performing the subsequent operation of determining whether there are brands with a preset brand relationship among all the brands in the brand set when it is determined that there are no brands that match the common words in the brand set of the identified text information, or after it is determined that there are brands that match the common words in the brand set of the identified text information and it is deleted. This is beneficial for improving the accuracy of subsequent operations, such as improving the accuracy of determining the brand results corresponding to the material to be identified.

[0182] In yet another optional embodiment, the method may further include the following steps:

[0183] Perform part-of-speech tagging on the text information of the material to be identified to obtain the part-of-speech tagging result of each word in the text information of the material to be identified, determine the part-of-speech tagging result of the brand result corresponding to the material to be identified based on the part-of-speech tagging result of each word in the text information of the material to be identified, and judge whether the part-of-speech tagging result of the brand result corresponding to the material to be identified is the same as the determined part-of-speech tagging result of the brand result of the material to be identified. When the judgment result is yes, output the brand result corresponding to the material to be identified.

[0184] It can be seen that the embodiment of the present invention determines whether the part-of-speech tagging of the brand result corresponding to the material to be identified in its text information is consistent with the prescribed part-of-speech tagging. If they are consistent, the brand result corresponding to the material to be identified is output, which can reduce the occurrence of brands being used as verbs and common words, further improve the recognition accuracy of the brand results corresponding to the material to be identified, and improve the output accuracy of the brand results corresponding to the material to be identified.

[0185] In yet another optional embodiment, the method may further include the following operations:

[0186] Performing a formatting operation on the text information of the material to be identified to obtain the formatted text information of the material to be identified, and performing the above-mentioned brand identification operation on the text information of the material to be identified to obtain the brand set operation of the text information;

[0187] In this optional embodiment, the formatting operation includes removing irregular characters and / or converting text, wherein the irregular characters include: emoticon string, separators such as |, *, @, ¥, &, etc., and the text conversion operation includes converting all English words to lowercase or converting all English words to uppercase.

[0188] In this optional embodiment, optionally, a formatting operation is performed on the text information of the material to be identified by using a regular expression.

[0189] It can be seen that this optional embodiment can obtain text information that meets the processing requirements by first performing a formatting operation on the text information of the material to be identified, which is beneficial to further improve the accuracy and efficiency of identifying brands in text information, thereby further improving the recognition accuracy and efficiency of brand results corresponding to the material to be identified.

[0190] Embodiment 2

[0191] See also Figure 2 , Figure 2 is a flow chart of another material brand recognition method disclosed in an embodiment of the present invention. Figure 2 The described method can be applied to a material brand recognition device, such as a server, etc., and the embodiment of the present invention does not limit this. Figure 2 As shown, the brand identification method of the material may include the following operations:

[0192] 201. Perform a brand recognition operation on text information of a to-be-recognized material to obtain a brand set of the text information, where the brand set includes at least one brand.

[0193] 202. Determine whether a brand of the determined brand type exists among all brands in the brand set; if it is determined that the brand does not exist, execute step 204; if it is determined that the brand does exist, execute step 203.

[0194] In the embodiment of the present invention, optionally, the determined brand type includes at least one of the types of brands with the same affixes, the same brand type, the equivalent brand type, and the same alternative brand type. In this way, the more content included in the brand type, the more conducive it is to improve the accuracy, efficiency, and comprehensiveness of judging whether there are or not brands that do not meet the requirements in the brand set of the text information, so as to improve the accuracy, efficiency, and comprehensiveness of deleting the brands that do not meet the requirements when it is judged that there are brands that do not meet the requirements, and further help to further improve the recognition accuracy of brands in the text information of the material to be identified, and help to further improve the recognition accuracy of the brand results corresponding to the material to be identified.

[0195] 203. Delete the brands with the existing brand type from all brands in the brand set to obtain a second brand set. At this time, the brand set in step 204 is the second brand set.

[0196] 204. Determine whether there is a brand with a preset brand relationship among all brands in the brand set, and obtain a determination result.

[0197] 205. According to the judgment result, all brands in the brand set are analyzed to obtain brand results corresponding to the material to be identified.

[0198] In the embodiment of the present invention, it should be noted that for the relevant description of step 201, step 204, and step 205, please refer to the detailed description of step 101 to step 103 in the first embodiment, and the embodiment of the present invention will not be repeated.

[0199] It can be seen that the embodiment of the present invention can also first determine whether there are any unmatched brands in the brand set of the identified text information. If not, the subsequent brand analysis steps are directly executed. If so, the brands that do not meet the requirements are first deleted from the brand set. This can reduce the occurrence of brand result recognition errors corresponding to the material to be identified and reduced recognition efficiency of the brand results due to the presence of unmatched brands in the brand set but continuing the brand result analysis, which is conducive to further improving the recognition accuracy and efficiency of the brand results corresponding to the material to be identified.

[0200] It can be seen that the implementation Figure 2The described method can, after identifying the brand of the text information of the material to be identified, automatically analyze whether there are brands with a preset brand relationship among the multiple brands in the identified text information, such as whether there are brands with a parent-child brand relationship, to determine the brand result corresponding to the material. It does not require the resources of a large deep network (such as Bert, etc.), nor does it require powerful equipment resources and training resources. Even when dealing with complex and changeable advertising materials, it can improve the recognition accuracy and efficiency of the brand in the text information of the material (such as the advertising material) by using only limited resources, thereby improving the determination accuracy and efficiency of the brand result corresponding to the material, and optimizing the recall rate of brand recognition. In addition, it can also reduce the occurrence of brand result recognition errors corresponding to the material to be identified and the reduction of the recognition efficiency of the brand result due to the presence of brands that do not meet the requirements in the brand set but still continuing the brand result analysis, which is conducive to further improving the recognition accuracy and efficiency of the brand result corresponding to the material to be identified.

[0201] In an optional embodiment, determining whether a brand of the determined brand type exists among all brands in the brand set includes:

[0202] Determine whether a word in the determined word set exists in the text information of the material to be identified. If it is determined that a word exists, determine that a brand of the determined brand type exists among all brands in the brand set of the identified text information, and the brand preceding the word is a brand of the determined brand type; or,

[0203] It is determined based on the text information of the material to be identified whether there are brands with the same word among all the brands in the brand set of the identified text information. When it is determined that there are brands with the same word, it is determined that there are brands of the determined brand type among all the brands in the brand set of the identified text information, and the brands with the same word are brands of the determined brand type.

[0204] In this optional embodiment, optionally, the word set includes at least one of the same item, a parallel substitute, and a same item replacement.

[0205] In this optional embodiment, when it is determined that a certain word in the determined word set does not exist in the text information of the material to be identified, the following operation of determining whether the brand included in all brands in the brand set of the identified text information exists according to the text information of the material to be identified can be triggered; or, when it is determined that the included brand does not exist, the above operation of determining whether a certain word in the determined word set exists in the text information of the material to be identified can be triggered. It should be noted that a certain word can be understood as at least one word.

[0206] In this optional embodiment, further, the method may also include the following steps:

[0207] When it is determined that there is a brand with the same word, it is determined whether the same word exists in the determined set of prefixes and suffixes. When it is determined that the same word exists, the operation of determining that there is a brand of the determined brand type among all brands in the brand set of the identified text information is performed, and the brand with the same word is a brand of the determined brand type;

[0208] Furthermore, the method may also include the following steps:

[0209] When it is determined that the same word does not exist in the set of prefix and suffixes, it is determined whether the index positions of each brand with the same word are the same. When it is determined that they are different, the above-mentioned operation of determining whether a brand of the determined brand type exists among all brands in the brand set of the identified text information is performed, and the brand with a shorter string length among all brands with the same word is a brand of the determined brand type.

[0210] In this optional embodiment, the prefix and suffix set includes at least one prefix and suffix, such as: big, small, medium.

[0211] For example, the text information of the material to be identified is "sky fairy water affordable alternative to Hanmei lotion". At this time, [sky] is not the correct brand of the material to be identified, and it is necessary to delete the affordable alternative [sky] in front; the text information of the material to be identified is "IELTS Meidai Xiyou Canban". At this time, three brands [IELTS Meidai], [Xiyou Canban] and [Canban] will be identified, among which the index position of [Xiyou Canban] is 2 and the string length is 4, and the index position of [Canban] is 4 and the string length is 2. At this time, if [Xiyou] is a word in the set of prefix and suffixes, then delete [Xiyou Canban] and [Canban]. If [Xiyou] is not a word in the set of prefix and suffixes, then keep [Xiyou Canban] and delete [Canban].

[0212] It can be seen that this optional embodiment can also, when it is identified that there are words such as "substitute" and "same style" in the text information of the material to be identified, and / or when there are included brands, it can determine that there are brands that do not meet the requirements in the brand set of the identified text information, and can enrich the method of determining that there are brands that do not meet the requirements in the brand set of the identified text information, thereby improving the probability and efficiency of determining that there are brands that do not meet the requirements in the brand set of the identified text information, which is beneficial to improving the recognition accuracy and reliability of the brand results corresponding to the material to be identified.

[0213] In an optional embodiment, after performing a brand recognition operation on the text information of the to-be-recognized material and obtaining a brand set of the text information, or obtaining a second brand set, the method may further include the following steps:

[0214] Determine whether all brands in the brand set are main brands;

[0215] When the judgment result is yes, the operation of judging whether there is a brand having a preset brand relationship among all brands in the brand set is performed to obtain the judgment result;

[0216] And, the method may further include the following steps:

[0217] When the judgment result is no, the brands that are not main brands in all brands in the brand set are converted into main names to obtain a third brand set of text information, and the third brand set includes at least one main brand.

[0218] It can be seen that this optional embodiment can first determine whether the brands in the brand set of the text information of the material to be identified are all main brand names. If so, continue to execute the subsequent brand analysis operation. If not, perform the main name conversion first, and deduplicate the brands in the brand set of the text information of the material to be identified, thereby reducing the amount of calculation, which is conducive to further improving the recognition accuracy and efficiency of the brand results of the material to be identified.

[0219] Embodiment 3

[0220] See also Figure 3 , Figure 3 Schematic diagram of a brand identification device for a material disclosed in an embodiment of the present invention. Figure 3 As shown, the brand identification device of the material may include:

[0221] The identification module 301 is used to perform a brand identification operation on the text information of the identification material to obtain a brand set of the text information, and the brand set includes at least one brand.

[0222] The judgment module 302 is used to judge whether there is a brand with a preset brand relationship among all brands in the brand set, and obtain a judgment result.

[0223] The analysis module 303 is used to analyze all brands in the brand set according to the judgment result to obtain brand results corresponding to the material to be identified.

[0224] It can be seen that the implementation Figure 3The described material brand identification device can, after identifying the brand of the text information of the material to be identified, automatically analyze whether there are brands with a preset brand relationship among multiple brands in the identified text information, such as whether there are brands with a parent-child brand relationship, to determine the brand result corresponding to the material. It does not require the resources of a large deep network (such as Bert, etc.), nor does it require powerful equipment resources and training resources. Even when dealing with complex and changeable advertising materials, it can improve the recognition accuracy and efficiency of the brand in the text information of the material (such as advertising material) by using only limited resources, thereby improving the determination accuracy and efficiency of the brand result corresponding to the material, and optimizing the recall rate of brand recognition.

[0225] In an optional embodiment, if Figure 3 As shown, the analysis module 303 analyzes all brands in the brand set according to the judgment result, and obtains the brand result corresponding to the material to be identified in the following manner:

[0226] When the judgment result is used to indicate that there are brands with a preset brand relationship among all brands in the brand set, a brand with a next-lower level is determined from all brands in the brand set as the brand result corresponding to the material to be identified.

[0227] It can be seen that the implementation Figure 3 The described material brand recognition device can also improve the recognition accuracy and efficiency of the brand results corresponding to the material by selecting the brand with the next lower level among all brands in the text information of the material to be recognized as the brand result corresponding to the material.

[0228] In another optional embodiment, Figure 4 As shown, the analysis module 303 includes:

[0229] The determination submodule 3031 is configured to determine the character string length of each brand in the brand set when the judgment result indicates that there is no brand in the brand set whose relationship is the preset brand relationship.

[0230] The screening submodule 3032 is used to screen the brand with the longest string length from all brands in the brand set according to the string length as the brand result corresponding to the material to be identified.

[0231] It can be seen that the implementation Figure 4 The described material brand identification device can improve the recognition accuracy of the brand results corresponding to the material by taking the brand corresponding to the longest character string length among all brands in the text information of the material to be identified as the brand result corresponding to the material, as well as improve the determination efficiency and accuracy of the brand results corresponding to the material.

[0232] In yet another optional embodiment, Figure 4 As shown, the analysis module 303 further includes:

[0233] The judging submodule 3033 is used to judge whether there are unambiguous brands among all brands in the brand set based on the determined industry brand set before the determining submodule 3031 determines the string length of each brand in all brands in the brand set, where all brands in the industry brand set include ambiguous brands and unambiguous brands.

[0234] The screening submodule 3032 is further configured to screen all the non-ambiguous brands from all the brands in the brand set when the judging submodule 3033 judges that there are non-ambiguous brands among all the brands in the brand set.

[0235] The determination submodule 3031 determines the character string length of each brand in the brand set in the following manner:

[0236] Determine the string length for each brand among all brands that are unambiguous.

[0237] It can be seen that the implementation Figure 4 The brand identification device of the described material can also analyze whether there are unambiguous brands among all the brands in the identified text information before determining the string length of each brand in the text information of the material to be identified, and directly determine the string length of each unambiguous brand when there are unambiguous brands. This can reduce the situation where ambiguous brands are present in the identified text information and are not used as brand results corresponding to the material to be identified, thereby further improving the accuracy of determining the brand results corresponding to the material to be identified.

[0238] In yet another optional embodiment, Figure 4 As shown, the analysis module 303 further includes:

[0239] The analysis submodule 3034 is used to input the text information into the determined industry analysis model for analysis when the judgment submodule 3033 judges that there is no unambiguous brand among all the brands in the brand set.

[0240] The acquisition submodule 3035 is used to obtain the analysis result output by the industry analysis model as the industry to which the material to be identified belongs.

[0241] The judgment submodule 3033 is also used to judge whether there is a brand that does not match the industry to which the material to be identified belongs among all the brands in the brand set; when it is judged that there is no brand that does not match the industry to which the material to be identified belongs, the determination submodule 3031 is triggered to perform the above-mentioned operation of determining the string length of each brand in all the brands in the brand set.

[0242] It can be seen that the implementation Figure 4 The brand recognition device of the described material can also automatically determine the industry to which the material to be identified belongs based on the text information of the material to be identified when it is determined that all brands in the identified text information are ambiguous brands, and directly determine the string length of each identified brand when it is ensured that there are no brands that do not match the industry to which the material to be identified belongs among all brands in the identified text information. This can reduce the situation where existing brands that do not match the industry to which the material to be identified belongs are mistakenly identified as brand results corresponding to the material to be identified, thereby further improving the accuracy of determining the brand results corresponding to the material to be identified.

[0243] In yet another optional embodiment, Figure 4 As shown, the analysis module 303 further includes:

[0244] The elimination submodule 3036 is used to eliminate the brands that do not match the industry to which the material to be identified belongs from all brands in the brand set when the judgment submodule 3033 judges that there are brands that do not match the industry to which the material to be identified belongs, so as to obtain a first brand set.

[0245] The determination submodule 3031 determines the character string length of each brand in the brand set in the following manner:

[0246] Determine the string length for each brand among all brands in the first brand set.

[0247] It can be seen that the implementation Figure 4 The brand recognition device for the described material can also eliminate the unmatched brands first when it is determined that there are brands that do not match the industry to which the material to be identified belongs among all the brands in the identified text information, so as to obtain more accurate brands, thereby improving the efficiency of determining the brand string length, thereby further improving the accuracy and efficiency of determining the brand results corresponding to the material to be identified.

[0248] In yet another optional embodiment, the source type of the text information includes a slogan source type and / or a description source type; Figure 4 As shown, the judging submodule 3033 is also used to eliminate brands that do not match the industry to which the text information belongs from all the brands in the brand set in the eliminating submodule 3036, and after obtaining the first brand set, determine whether the number of brands in the first brand set is greater than or equal to 1. When it is determined that the number of brands in the first brand set is greater than or equal to 1, the determining submodule 3031 is triggered to perform the above-mentioned operation of determining the string length of each brand in all the brands in the first brand set.

[0249] It can be seen that the implementation Figure 4The brand identification device of the described material can also further determine whether there is still a brand when the source type of the text information of the material to be identified is a slogan / title source type and / or a description source type and there is a brand that does not match the industry to which the material to be identified belongs. If so, it continues to determine the string length of the brand. This can reduce the occurrence of the step of determining the string length of the brand when the brand does not exist, thereby increasing the possibility of successfully determining the material to be identified and reducing the power consumption of the device.

[0250] In yet another optional embodiment, Figure 4 As shown, the determination submodule 3031 is further used to determine the image text information related to the material to be identified when the determination submodule 3033 determines that the number of brands in the first brand set is less than 1.

[0251] The identification module 301 is further used to perform a brand identification operation on the image text information to obtain a brand set of the image text information.

[0252] The judging submodule 3033 is further used to judge whether the number of brands in the brand set of the image text information is equal to 1.

[0253] The determination submodule 3031 is further configured to determine that the brand of the brand set of the image text information is the brand result corresponding to the material to be identified when the determination submodule 3033 determines that the result is yes.

[0254] It can be seen that the implementation Figure 4 The brand identification device of the described material can also further identify and analyze the text information of the image source type of the material to be identified, after determining that the brand result corresponding to the material to be identified is not determined based on the text information of the slogan / title source type and / or the description source type, to obtain the brand result corresponding to the material to be identified. This can increase the probability of successfully determining the brand result corresponding to the material to be identified and enrich the intelligent function of the brand identification device of the material.

[0255] In yet another optional embodiment, Figure 4 As shown, the device also includes:

[0256] The first collection module 304 is used to collect training text information of training materials for multiple industries, the number of training materials for each industry is greater than or equal to a determined material quantity threshold (such as 100), and the number of training text information for each training material is greater than or equal to a determined text quantity threshold (such as 500).

[0257] The establishing module 305 is used to establish an association relationship between each training material and the industry to which the training material belongs.

[0258] The first training module 306 is used to perform a training operation on the determined classification model based on the training text information of all training materials of each industry and the association relationship between each training material and the training material, so as to obtain a trained classification model.

[0259] The determination module 307 is used to determine the trained classification model, which is the determined industry analysis model.

[0260] It can be seen that the implementation Figure 4 The brand identification device of the described material can also train the existing classification model based on multiple training text information of each training material in each industry, and can obtain an industry classification model that can be applied to the industry identification of the material, so as to accurately and quickly identify the industry to which the material to be identified belongs based on the industry classification model, and then help to improve the accuracy and efficiency of determining the brand results of the material to be identified; and perform a verification operation on the obtained industry classification model based on the verification data, and stop performing the verification operation on the industry classification model when the accuracy of the industry classification model reaches the set accuracy threshold, so as to obtain a more accurate industry classification model, thereby further improving the recognition accuracy and efficiency of the industry to which the material to be identified belongs.

[0261] In yet another optional embodiment, Figure 4 As shown, the identification module 301 performs a brand identification operation on the text information of the identification material, and obtains the brand set of the text information in the following manner:

[0262] Inputting the text information of the material to be identified into the determined brand identification model for analysis, and obtaining the analysis result output by the brand identification model as the brand set of the text information;

[0263] And, if Figure 4 As shown, the device also includes:

[0264] The second collection module 308 is used to collect a training brand set, where the training brand set includes brands from multiple industries.

[0265] The second training module 309 is used to train the determined initial recognition model based on all brands in the training brand set to obtain a trained initial recognition model.

[0266] The recognition module 301 is further used to determine a trained initial recognition model as the determined brand recognition model, and the initial recognition model includes an AC automaton recognition model.

[0267] It can be seen that the implementation Figure 4The brand recognition device for the described material can also train the existing recognition model by using multiple training text information based on each training material in each industry, and can obtain a brand recognition model that can be applied to brand recognition of the text information of the material, thereby facilitating accurate and rapid identification of the brand of the text information of the material to be identified based on the brand recognition model.

[0268] In yet another optional embodiment, Figure 4 As shown, the judgment module 302 is further used to perform a brand recognition operation on the text information of the identification material in the identification module 301, and after obtaining a brand set of the text information, determine whether there is a brand of the determined brand type among all the brands in the brand set; when it is determined that there is no brand, trigger the above-mentioned operation of determining whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a judgment result;

[0269] And, if Figure 4 As shown, the device also includes:

[0270] The deletion module 310 is used to delete the brand of the existing brand type from all brands in the brand set when it is determined that it exists, to obtain a second brand set, and trigger the judgment module 302 to perform the above-mentioned operation of judging whether there is a brand with a preset brand relationship among all brands in the brand set, and obtain the judgment result. The second brand set is this brand set.

[0271] It can be seen that the implementation Figure 4 The brand identification device of the described material can also first determine whether there are any unmatched brands in the brand set of the identified text information. If not, the subsequent brand analysis steps are directly executed. If so, the unqualified brands are first deleted from the brand set. This can reduce the occurrence of brand result identification errors corresponding to the material to be identified and reduced recognition efficiency of the brand results due to the existence of unmatched brands in the brand set but continuing the brand result analysis, which is conducive to further improving the recognition accuracy and efficiency of the brand results corresponding to the material to be identified.

[0272] In yet another optional embodiment, Figure 4 As shown, the judgment module 302 is further used to perform a brand recognition operation on the text information of the identification material in the identification module 301, and after obtaining the brand set of the text information, determine whether all the brands in the brand set are all main brands; when the judgment result is yes, perform the operation of determining whether there are brands with a preset brand relationship among all the brands in the brand set, and obtain the judgment result;

[0273] And, if Figure 4 As shown, the device also includes:

[0274] The conversion module 311 is used to convert the main name of all brands in the brand set that are not main brands when the judgment result is no, to obtain a third brand set of text information, and the third brand set includes at least one main brand.

[0275] It can be seen that the implementation Figure 4 The brand identification device of the described material can also first determine whether the brands in the brand set of the text information of the material to be identified are all main brand names. If so, continue to perform subsequent brand analysis operations. If not, perform main name conversion first, and can deduplicate the brands in the brand set of the text information of the material to be identified, thereby reducing the amount of calculation, which is conducive to further improving the recognition accuracy and efficiency of the brand results of the material to be identified.

[0276] Embodiment 4

[0277] See also Figure 5 , Figure 5 FIG. 1 is a schematic diagram of a structure of another material brand identification device disclosed in an embodiment of the present invention. Figure 5 As shown, the device may include:

[0278] A memory 501 storing executable program codes;

[0279] a processor 502 coupled to the memory 501;

[0280] Furthermore, it may also include an input interface 503 and an output interface 504 coupled to the processor 502;

[0281] The processor 502 calls the executable program code stored in the memory 501 to execute the steps of the material brand identification method disclosed in any one of the first embodiment or the second embodiment of the present invention.

[0282] Embodiment 5

[0283] An embodiment of the present invention discloses a computer storage medium, which stores computer instructions. When the computer instructions are called, they are used to execute the steps in the material brand identification method disclosed in any one of the first embodiment or the second embodiment of the present invention.

[0284] The device embodiments described above are only illustrative, wherein the modules described as separate components may or may not be physically separated, and the components displayed as modules may or may not be physical modules, i.e., they may be located in one place, or they may be distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0285] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the above technical solution can be essentially or partly contributed to the prior art in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0286] Finally, it should be noted that the brand identification method and device for a material disclosed in the embodiment of the present invention discloses only the preferred embodiments of the present invention, which are only used to illustrate the technical solution of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, it should be understood by those skilled in the art that the technical solutions described in the aforementioned embodiments can still be modified, or some of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A material brand identification method, characterized in that: The method comprises: Performing a brand recognition operation on the text information of the identification material to obtain a brand set of the text information, wherein the brand set includes at least one brand; Determine whether there is a brand with a preset brand relationship among all the brands in the brand set, and obtain a determination result; According to the judgment result, all the brands in the brand set are analyzed to obtain brand results corresponding to the material to be identified; The step of analyzing all the brands in the brand set according to the judgment result to obtain a brand result corresponding to the material to be identified includes: When the judgment result is used to indicate that there is a brand with the preset brand relationship among all the brands in the brand set, a brand with a next-lower level is determined from all the brands in the brand set as the brand result corresponding to the material to be identified; When the judgment result indicates that there is no brand having the preset brand relationship among all the brands in the brand set, determining the character string length of each brand among all the brands in the brand set; Selecting the brand with the longest string length from all the brands in the brand set according to the string length as the brand result corresponding to the material to be identified; After performing the brand recognition operation on the text information of the to-be-recognized material to obtain the brand set of the text information, the method further includes: Determine whether there is a brand of the determined brand type among all the brands in the brand set, wherein the brand type includes at least one of a brand type with the same affix, a brand type of the same style, a substitute brand type, and a substitute brand type of the same style; When it is determined that no brand exists, the operation of determining whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a determination result is performed; When it is determined that the brand exists, the brand of the brand type is deleted from all the brands in the brand set to obtain a second brand set, and the operation of determining whether there is a brand with a preset brand relationship among all the brands in the brand set to obtain a determination result is performed, and the second brand set is the brand set.

2. The material brand identification method according to claim 1, characterized in that: Before determining the character string length of each of the brands in the brand set, the method further includes: Based on the determined industry brand set, judging whether there is a non-ambiguous brand among all the brands in the brand set, wherein all the brands in the industry brand set include ambiguous brands and non-ambiguous brands; When it is determined that there is a non-ambiguous brand among all the brands in the brand set, all the non-ambiguous brands are screened from all the brands in the brand set; The step of determining the character string length of each brand among all the brands in the brand set includes: The string length of each of the brands among all the brands that are unambiguous is determined.

3. The material brand identification method according to claim 2, characterized in that: The method further comprises: When it is determined that there is no unambiguous brand among all the brands in the brand set, the text information is input into the determined industry analysis model for analysis, and an analysis result output by the industry analysis model is obtained as the industry to which the material to be identified belongs; Determine whether there is a brand among all the brands in the brand set that does not match the industry to which the material to be identified belongs; When it is determined that there is no brand that does not match the industry to which the to-be-identified material belongs, the operation of determining the character string length of each of the brands in the brand set is performed.

4. The material brand identification method according to claim 3, characterized in that: The method further comprises: When it is determined that there is a brand that does not match the industry to which the to-be-identified material belongs, the brand that does not match the industry to which the to-be-identified material belongs is removed from all the brands in the brand set to obtain a first brand set; The step of determining the character string length of each brand among all the brands in the brand set includes: The character string length of each of the brands in all the brands in the first set of brands is determined.

5. The material brand identification method according to claim 4, characterized in that: The source type of the text information includes a slogan source type and / or a description source type; Furthermore, after removing brands that do not match the industry to which the text information belongs from all the brands in the brand set to obtain the first brand set, the method further includes: Determine whether the number of brands in the first brand set is greater than or equal to 1. When it is determined that the number of brands in the first brand set is greater than or equal to 1, perform the operation of determining the string length of each of the brands in all the brands in the first brand set.

6. The material brand identification method according to claim 5, characterized in that: The method further comprises: When it is determined that the number of brands in the first brand set is less than 1, determining image text information related to the material to be identified, and performing a brand identification operation on the image text information to obtain a brand set of the image text information; It is determined whether the number of brands in the brand set of the image text information is equal to 1. When the determination result is yes, the brand of the brand set of the image text information is determined to be the brand result corresponding to the material to be identified.

7. The material brand identification method according to any one of claims 3 to 6, characterized in that: The method further comprises: Collecting training text information of training materials of multiple industries, the number of training materials of each industry is greater than or equal to the determined material number threshold, and the number of training text information of each training material is greater than or equal to the determined text number threshold; Establishing an association relationship between each of the training materials and the industry to which the training material belongs, and performing a training operation on the determined classification model based on the training text information of all the training materials of each industry and the association relationship between each of the training materials and the training material, to obtain a trained classification model; The trained classification model is determined to be the determined industry analysis model.

8. The material brand identification method according to any one of claims 1 to 6, characterized in that: The performing of a brand recognition operation on the text information of the to-be-recognized material to obtain a brand set of the text information includes: Inputting the text information of the material to be identified into the determined brand identification model for analysis, and obtaining the analysis result output by the brand identification model as the brand set of the text information; And, the method further comprises: Collecting a training brand set, wherein the training brand set includes brands from multiple industries; Based on the initial recognition model determined by all the brands in the training brand set, a trained initial recognition model is obtained, and the trained initial recognition model is determined as the determined brand recognition model.

9. The material brand identification method according to any one of claims 1 to 6, characterized in that: After performing the brand recognition operation on the text information of the to-be-recognized material to obtain the brand set of the text information, the method further includes: Determine whether all the brands in the brand set are main brands; When the judgment result is yes, the operation of judging whether there is a brand with a preset brand relationship among all the brands in the brand set is performed to obtain a judgment result; And, the method further comprises: When the judgment result is no, the brands that are not main brands in all the brands in the brand set are converted into main names to obtain a third brand set of the text information, and the third brand set includes at least one main brand.

10. A material brand identification device, characterized in that: The device is used to execute the material brand recognition method according to any one of claims 1 to 9, and the device comprises: An identification module, configured to perform a brand identification operation on the text information of the identification material to obtain a brand set of the text information, wherein the brand set includes at least one brand; A judgment module, used for judging whether there is a brand with a preset brand relationship among all the brands in the brand set, and obtaining a judgment result; The analysis module is used to analyze all the brands in the brand set according to the judgment result to obtain the brand result corresponding to the material to be identified.

Citation Information

Patent Citations

  • Brand word recognition method and device, equipment and storage medium

    CN110750985A