Method and device for identifying brand of consignment

Abstract

The present application provides a method and device for identifying the brand of consigned goods, and the method for identifying the brand of consigned goods includes: obtaining the consignment introduction text to be identified and the first brand set of the consignment; determining the first brand word vector of each first brand based on the category text of each first brand in the first brand set; determining the second brand word vector based on the consignment introduction text to be identified; and screening the brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain the brand identification result of the consignment. On the one hand, the present application uses multiple categories corresponding to the brand to represent the brand of the consignment, and on the other hand, uses the consignment introduction text to be identified to represent the brand of the consignment, and then screens the brands in the first brand set according to the similarity of the brands expressed in two different ways, so as to obtain a brand identification result with higher accuracy, which can improve the accuracy of consignment brand identification.

Description

Technical Field

[0001] The present application relates to the field of natural language processing and machine learning technology, and in particular to a method and device for identifying a consignment brand. Background Art

[0002] Natural language processing (NLP) is an important direction in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between people and computers using natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field will involve natural language, that is, the language people use in daily life, so it is closely related to the study of linguistics. Natural language processing technology usually includes text processing, semantic understanding, machine translation, robot question answering, knowledge graph and other technologies.

[0003] Machine Learning (ML) is a multi-disciplinary subject that involves probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and other disciplines. It specializes in studying how computers simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications are spread across all areas of artificial intelligence. Machine learning and deep learning usually include artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0004] When using logistics waybill data for user analysis, we usually need to mine users' brand purchasing behavior to gain insights into their purchasing power, consumer category preferences, etc. Therefore, using the consignment description in the logistics waybill data to mine the brand of the current consignment is the only way. The difficulty of this approach is that the text of the consignment description to be identified is very noisy and has a messy format, making it difficult to accurately extract brand information. Existing mining methods usually directly fuzzy match the brand name, and the accuracy of the information obtained is very low, and the accuracy of consignment brand recognition is low.

[0005] That is, the accuracy of brand recognition of consigned items in the prior art is low. Summary of the invention

[0006] The present application aims to provide a method and device for identifying the brand of consigned goods, aiming to solve the problem of low accuracy in identifying the brand of consigned goods in the prior art.

[0007] In one aspect, the present application provides a method for identifying a brand of consigned goods, the method comprising:

[0008] Obtaining a consignment introduction text to be identified and a first brand set of the consignment;

[0009] Determine a first brand word vector for each first brand based on the category text of each first brand in the first brand set;

[0010] Determining a second brand word vector based on the introduction text of the consignment to be identified;

[0011] Brands in the first brand set are screened based on the similarity between the first brand word vector and the second brand word vector to obtain a brand recognition result of the consignment.

[0012] Optionally, the step of obtaining the consignment introduction text to be identified and the first brand set of the consignment includes:

[0013] Acquire a preset corpus, wherein the preset corpus includes a preset brand set and category texts corresponding to each brand in the preset brand set;

[0014] Segmenting the text of the introduction to the consigned object to be identified based on the preset corpus to obtain a segmentation set of the introduction to the consigned object to be identified;

[0015] Calculate the intersection of the word set of the description of the consigned object to be identified and the preset brand set to obtain a second brand set;

[0016] The first brand set is determined based on the second brand set.

[0017] Optionally, the performing brand screening on brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain the brand recognition result of the consignment includes:

[0018] Calculating the similarity between the first brand word vector and the second brand word vector;

[0019] The first brand corresponding to the first brand word vector with the highest similarity is determined as the brand recognition result.

[0020] Optionally, determining the first brand word vector of each first brand in the first brand set based on the category text of each first brand includes:

[0021] vectorizing each category text of the first brand based on the brand word vector conversion model to obtain the category word vector of each category text of the first brand;

[0022] A weighted average of the category word vectors of the first brand is determined as the first brand word vector of the first brand.

[0023] Optionally, the determining a second brand word vector based on the introduction text of the consignment to be identified includes:

[0024] Based on the brand word vector conversion model, each of the to-be-identified consignment introduction participles in the to-be-identified consignment introduction participle set is vectorized to obtain a word vector of each to-be-identified consignment introduction participle;

[0025] The weighted average of the word vectors of the consignment introduction participles to be identified is determined as the second brand word vector.

[0026] Optionally, determining the first brand set based on the second brand set includes:

[0027] Determine whether the number of brands belonging to the same target category in the second brand set exceeds a preset value; if the number of brands belonging to the same target category in the second brand set exceeds the preset value, remove the brands belonging to the same target category in the second brand set to obtain the first brand set;

[0028] And / or, determine whether there is a preset text in the segmentation set of the description of the consignment to be identified; if there is a preset text in the segmentation set of the description of the consignment to be identified, remove multiple brands corresponding to the preset text from the second brand set to obtain the first brand set.

[0029] Optionally, before vectorizing each category text of the first brand based on the brand word vector conversion model to obtain the category word vector of each category text of the first brand, the process includes:

[0030] Obtain multiple consignment introduction texts to be trained;

[0031] Segmenting the plurality of to-be-trained consignment introduction texts respectively to obtain a first to-be-trained consignment introduction segmentation set;

[0032] Eliminate the brands in the first to-be-trained consignment introduction word set to obtain a second to-be-trained consignment introduction word set;

[0033] The second consignment introduction word set to be trained is used as a training set, and an unsupervised learning algorithm is used to train the preset word vector conversion model to obtain the brand word vector conversion model.

[0034] In one aspect, the present application provides a device for identifying a brand of consigned goods, the device comprising:

[0035] An acquiring unit, configured to acquire an introduction text of the consignment to be identified and a first brand set of the consignment;

[0036] A first determining unit, configured to determine a first brand word vector of each first brand based on the category text of each first brand in the first brand set;

[0037] A second determining unit, configured to determine a second brand word vector based on the introduction text of the consignment to be identified;

[0038] A brand screening unit is used to screen brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain a brand recognition result of the consignment.

[0039] Optionally, the acquiring unit is used to:

[0040] Acquire a preset corpus, wherein the preset corpus includes a preset brand set and category texts corresponding to each brand in the preset brand set;

[0041] Segmenting the text of the introduction to the consigned object to be identified based on the preset corpus to obtain a segmentation set of the introduction to the consigned object to be identified;

[0042] Calculate the intersection of the word set of the description of the consigned object to be identified and the preset brand set to obtain a second brand set;

[0043] The first brand set is determined based on the second brand set.

[0044] Optionally, the brand screening unit is used to:

[0045] Calculating the similarity between the first brand word vector and the second brand word vector;

[0046] The first brand corresponding to the first brand word vector with the highest similarity is determined as the brand recognition result.

[0047] Optionally, the first determining unit is configured to:

[0048] vectorizing each category text of the first brand based on the brand word vector conversion model to obtain the category word vector of each category text of the first brand;

[0049] A weighted average of the category word vectors of the first brand is determined as the first brand word vector of the first brand.

[0050] Optionally, the second determining unit is configured to:

[0051] Based on the brand word vector conversion model, each of the to-be-identified consignment introduction participles in the to-be-identified consignment introduction participle set is vectorized to obtain a word vector of each to-be-identified consignment introduction participle;

[0052] The weighted average of the word vectors of the consignment introduction participles to be identified is determined as the second brand word vector.

[0053] Optionally, the acquiring unit is used to:

[0054] Determine whether the number of brands belonging to the same target category in the second brand set exceeds a preset value; if the number of brands belonging to the same target category in the second brand set exceeds the preset value, remove the brands belonging to the same target category in the second brand set to obtain the first brand set;

[0055] And / or, determine whether there is a preset text in the segmentation set of the description of the consignment to be identified; if there is a preset text in the segmentation set of the description of the consignment to be identified, remove multiple brands corresponding to the preset text from the second brand set to obtain the first brand set.

[0056] Optionally, the acquiring unit is used to:

[0057] Obtain multiple consignment introduction texts to be trained;

[0058] Segmenting the plurality of to-be-trained consignment introduction texts respectively to obtain a first to-be-trained consignment introduction segmentation set;

[0059] Eliminate the brands in the first to-be-trained consignment introduction word set to obtain a second to-be-trained consignment introduction word set;

[0060] The second consignment introduction word set to be trained is used as a training set, and an unsupervised learning algorithm is used to train the preset word vector conversion model to obtain the brand word vector conversion model.

[0061] In one aspect, the present application further provides an electronic device, the electronic device comprising:

[0062] one or more processors;

[0063] Memory; and

[0064] One or more applications, wherein the one or more applications are stored in the memory and are configured to be executed by the processor to implement the method for identifying the brand of consigned goods described in any one of the first aspects.

[0065] On the one hand, the present application also provides a computer-readable storage medium on which a computer program is stored, and the computer program is loaded by a processor to execute the steps in the method for identifying the brand of consignment described in any one of the first aspects.

[0066] The present application provides a method for identifying the brand of consigned goods, and the method for identifying the brand of consigned goods includes: obtaining the consignment introduction text to be identified and the first brand set of the consignment; determining the first brand word vector of each first brand based on the category text of each first brand in the first brand set; determining the second brand word vector based on the consignment introduction text to be identified; and screening the brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain the brand identification result of the consignment. On the one hand, the present application uses multiple categories corresponding to the brand to represent the brand of the consignment, and on the other hand, uses the consignment introduction text to be identified to represent the brand of the consignment, and then screens the brands in the first brand set according to the similarity of the brands expressed in two different ways, so as to obtain a brand identification result with higher accuracy, which can improve the accuracy of consignment brand identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0068] Figure 1 A schematic diagram of a scenario of a consignment brand identification system provided in an embodiment of the present application;

[0069] Figure 2 It is a schematic diagram of a flow chart of an embodiment of a method for identifying a brand of consigned goods provided in an embodiment of the present application;

[0070] Figure 3 It is a flow chart of an embodiment of 201 in the method for identifying the brand of consigned goods provided in the embodiment of the present application;

[0071] Figure 4 1 is a schematic diagram of the structure of an embodiment of a device for identifying a brand of consigned goods provided in an embodiment of the present application;

[0072] Figure 5 It is a schematic diagram of the structure of an embodiment of an electronic device provided in the embodiments of the present application. DETAILED DESCRIPTION

[0073] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0074] In the description of the present application, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, which are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0075] In this application, the word "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described in this application as "exemplary" is not necessarily to be construed as being preferred or advantageous over other embodiments. The following description is given to enable any technician in the field to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present application can be implemented without using these specific details. In other instances, well-known structures and processes will not be elaborated in detail to avoid obscuring the description of the present application with unnecessary details. Therefore, the present application is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed in the present application.

[0076] It should be noted that since the method of the embodiment of the present application is executed in an electronic device, the processing objects of each electronic device exist in the form of data or information. For example, time is actually time information. It can be understood that if size, quantity, position, etc. are mentioned in subsequent embodiments, they are all corresponding data for processing by the electronic device. The details will not be elaborated here.

[0077] The embodiments of the present application provide a method and device for identifying the brand of consigned goods, which are described in detail below.

[0078] See also Figure 1 , Figure 1 This is a scenario diagram of a system for identifying a brand of consigned goods provided in an embodiment of the present application. The system for identifying a brand of consigned goods may include an electronic device 100, in which a device for identifying a brand of consigned goods is integrated.

[0079] In the embodiment of the present application, the electronic device 100 may be an independent server, or a server network or server cluster composed of servers. For example, the electronic device 100 described in the embodiment of the present application includes but is not limited to a computer, a network host, a single network server, a plurality of network server sets or a cloud server composed of a plurality of servers. The cloud server is composed of a large number of computers or network servers based on cloud computing.

[0080] Those skilled in the art will understand that Figure 1 The application environment shown in the figure is only one application scenario of the present application solution and does not constitute a limitation on the application scenario of the present application solution. Other application environments may also include Figure 1 More or fewer electronic devices as shown in, e.g. Figure 1 Only one electronic device is shown. It can be understood that the identification system of the consignment brand may also include one or more other servers, which are not limited here.

[0081] In addition, if Figure 1 As shown, the consignment brand identification system may further include a memory 200 for storing data.

[0082] It should be noted that Figure 1 The scenario diagram of the consignment brand identification system shown is merely an example. The consignment brand identification system and scenario described in the embodiment of the present application are intended to more clearly illustrate the technical solution of the embodiment of the present application, and do not constitute a limitation on the technical solution provided in the embodiment of the present application. A person of ordinary skill in the art will appreciate that, with the evolution of the consignment brand identification system and the emergence of new business scenarios, the technical solution provided in the embodiment of the present application is equally applicable to similar technical problems.

[0083] First, a method for identifying a brand of consignment is provided in an embodiment of the present application. The execution subject of the method for identifying a brand of consignment is a device for identifying a brand of consignment. The device for identifying a brand of consignment is applied to an electronic device. The method for identifying a brand of consignment includes:

[0084] Obtaining the consignment introduction text to be identified and the first brand set of the consignment;

[0085] Determine the first brand word vector of each first brand based on the category text of each first brand in the first brand set;

[0086] Determine a second brand word vector based on the introduction text of the consignment to be identified;

[0087] Based on the similarity between the first brand word vector and the second brand word vector, brands in the first brand set are screened to obtain a brand recognition result of the consignment.

[0088] See also Figure 2 , Figure 2 FIG. 1 is a flow chart of an embodiment of a method for identifying a brand of consigned goods provided in an embodiment of the present application. Figure 2 As shown, the identification methods of the consignment brand include:

[0089] 201. Obtain the consignment introduction text to be identified and the first brand set of the consignment.

[0090] In the embodiment of the present application, the introduction text of the consignment to be identified can be obtained by identifying the information on the consignment waybill, or it can be read from the system. For example, the introduction text of the consignment to be identified is "Brand T Men's Spring New Bear Print Korean Breathable Pajamas", "Brand A Charger is suitable for Brand B Brand C Mobile Phones" and "Bear Electric Kettle", etc. The first brand set includes multiple first brands. The first brand in the first brand set is the brand that the introduction text of the consignment to be identified may correspond to. For example, the introduction text of the consignment to be identified is "Brand A Charger is suitable for Brand B Brand C Mobile Phones", and the first brand set includes Brand A, Brand B, Brand C, etc.

[0091] See also Figure 3 In a specific embodiment, obtaining the consignment introduction text to be identified and the first brand set of the consignment may include:

[0092] 301. Obtain a preset corpus, wherein the preset corpus includes a preset brand set and category texts corresponding to each brand in the preset brand set.

[0093] Specifically, brand lexicons of well-known e-commerce platforms on the market are collected through various channels. Brand lexicons can be obtained through manual collection, purchase from third-party platforms, or code crawlers. The brand lexicon contains information including brands and brand categories. Brands on different platforms may have different transliterations or brand duplications. Therefore, brand lexicons on multiple platforms are obtained, and data cleaning is performed on brand lexicons on multiple platforms to obtain cleaned brand lexicons. Data cleaning refers to the last procedure of discovering and correcting identifiable errors in data files, including checking data consistency, handling invalid values ​​and missing values, etc. Specifically, data cleaning of brand lexicons on multiple platforms mainly includes: deduplication, mining similar brands, and manually checking whether they are aliases or sub-brands of each other. Thus, brand lexicons on multiple platforms are merged into a cleaned brand lexicon.

[0094] In a preferred embodiment, the Spark parallel computing framework is used to clean brand word libraries on multiple platforms. Spark is a general parallel framework similar to Hadoop MapReduce, which is open sourced by the AMP Laboratory of the University of California, Berkeley. Spark has the advantages of Hadoop MapReduce; but unlike MapReduce, the intermediate output results of the Job can be stored in memory, so there is no need to read and write HDFS, so Spark is better suitable for MapReduce algorithms that require iteration, such as data mining and machine learning. Spark is an open source cluster computing environment similar to Hadoop, but there are some differences between the two. These useful differences make Spark more superior in certain workloads. In other words, Spark enables memory distributed data sets. In addition to providing interactive queries, it can also optimize iterative workloads. In other embodiments, Python's Pandas framework, Kettle framework, RapidMiner framework, etc. can also be used to clean brand word libraries on multiple platforms.

[0095] Furthermore, the cleaned brand word library is preprocessed to obtain a preset corpus. Specifically, the cleaned brand word library is preprocessed, including converting English brands into lowercase, removing special characters, and removing brands or common brand names whose length is less than a predetermined length.

[0096] The preset corpus includes the preset brand set B and the category texts corresponding to each brand in the preset brand set. For example, brand F has three categories: electric kettle, electric stew pot, and health pot. The category text corresponding to brand F is: "electric kettle, electric stew pot, health pot". Of course, the category text corresponding to brand F can also be the result of word segmentation, such as: "electric kettle", "electric stew pot", "health pot".

[0097] 302. Segment the text of the introduction to the consignment to be identified based on a preset corpus to obtain a segmentation set of the introduction to the consignment to be identified.

[0098] In a specific embodiment, the jieba word segmentation tool is used to segment the text of the consignment introduction to be identified based on a preset corpus to obtain a word segmentation set of the consignment introduction to be identified.

[0099] Specifically, the text of the consignment introduction to be identified is segmented according to formula (1).

[0100] Di_cut_1 = jieba.cut(Di) (1)

[0101] Among them, Di is the introduction text of the consignment to be identified, and Di_cut_1 is the segmentation of the introduction text of the consignment to be identified.

[0102] Specifically, the preset corpus is entered into the jieba word segmentation tool as a custom dictionary, and then the jieba word segmentation tool with the preset corpus entered is used to segment the text of the consignment introduction to be identified, and a word segmentation set of the consignment introduction to be identified is obtained. Entering the preset corpus into the jieba word segmentation tool can make the jieba word segmentation tool have a higher accuracy rate in consignment brand identification.

[0103] The jieba word segmentation tool supports three word segmentation modes: precise mode, which tries to segment the sentence most accurately, suitable for text analysis; full mode, which scans all the words in the sentence that can be formed into words, which is very fast, but cannot resolve ambiguity; search engine mode, based on the precise mode, segments long words again to improve the recall rate, which is suitable for search engine word segmentation. In addition, jieba word segmentation supports custom dictionaries. Of course, the word segmentation tools such as THULAC, SnowNLP, pynlpir, CoreNLP and pyLTP can also be used to segment the text of the consignment introduction to be identified to obtain the word segmentation set of the consignment introduction to be identified, which is not limited in this application.

[0104] 303. The intersection of the word set for the description of the consigned goods to be identified and the preset brand set is obtained to obtain a second brand set.

[0105] Specifically, each brand in the second brand set can be determined according to formula (2).

[0106] Ri_fst_step=Di_cut_1.intersection(B) (2)

[0107] Among them, Di_cut_1 is the participle of the introduction of the consignment to be identified; Ri_fst_step is each brand in the second brand set.

[0108] 304. Determine the first brand set based on the second brand set.

[0109] In a specific embodiment, the second brand set can be determined as the first brand set. However, since brand names may include or be included, or more than one brand name appears in a brand introduction text, the second brand set usually generates multiple results during the initial matching. Therefore, it is necessary to eliminate the brands in the second brand set.

[0110] There will be a variety of noises in the description text of the consignment to be identified, for example, accessories suitable for similar products of multiple brands, such as the description text of the consignment to be identified "Brand A charger is suitable for brand B brand C mobile phone" or "Brand K brand W brand L leather strap", etc. The second brand set that is initially matched will correspond to multiple brands, but because the categories sold by these brands are extremely similar to the categories corresponding to the description text of the consignment to be identified, it may affect the similarity determination in subsequent steps.

[0111] Therefore, in another specific embodiment, determining the first brand set based on the second brand set includes:

[0112] (1) Determine whether the number of brands in the second brand set that belong to the same target category exceeds a preset value.

[0113] Among them, a list of wrong brands of the same category can be preset. list For example, the error brand list W list Including: Brand G, Brand B and Brand C, Brand G, Brand B and Brand C all have the category of mobile phone. The preset value can be set according to the specific situation, for example, the preset value is 1.

[0114] List each wrong brand W list The categories are determined as target categories. For example, "mobile phone" is determined as the target category, and the corresponding brands are Brand G, Brand B, and Brand C.

[0115] (2) If the number of brands in the second brand set that belong to the same target category exceeds a preset value, the brands in the second brand set that belong to the same target category are removed to obtain the first brand set.

[0116] If the number of brands in the second brand set that belong to the target category exceeds the preset value, the wrong brand list W corresponding to the second brand set and the target category is list There are two or more elements in the intersection of , and the description text of the consignment to be identified contains two or more brands in the wrong brand list, indicating that the corresponding consignment may be in the wrong brand list W list For the accessories of the products sold by the brand in the first matching second brand set, the wrong brand list W is removed. list All brands belonging to the same target category in the second brand set are eliminated to obtain the first brand set.

[0117] For example, "Brand A charger is suitable for brand B and brand C mobile phones", there are two brands of the same category, brand B and brand C. All brands corresponding to the category "mobile phone" in the wrong brand list are removed from the second brand set. For example, brand G, brand B and brand C corresponding to the category "mobile phone" are all removed.

[0118] The noise in the description text of the consignment to be identified may also be a joint or series product. For example, if the description text of the consignment to be identified is "Brand D co-branded with Brand E boy's T-shirt", the second brand set initially matched will correspond to two brand results, Brand D and Brand E. This may affect the similarity determination in subsequent steps.

[0119] Therefore, in yet another specific embodiment, determining the first brand set based on the second brand set includes:

[0120] (1) Determine whether there is a preset text in the word set of the consignment description to be identified.

[0121] Specifically, the preset text can be "jointly-branded", "series", etc., which can be set according to the specific situation. The brand list WIP corresponding to common joint names and series can be collected and constructed in advance. For example, the brand list WIP includes: the preset text is "jointly-branded", and the corresponding brands are brand D and brand E.

[0122] (2) If there is a preset text in the word set of the consignment description to be identified, multiple brands corresponding to the preset text are removed from the second brand set to obtain the first brand set.

[0123] If there is a preset text in the segmented word set of the consignment introduction to be identified, it means that the first brand set may have a joint brand or a series brand, and the multiple brands corresponding to the preset text in the second brand set that are initially matched are removed. For example, if "joint brand" appears in the consignment introduction text to be identified, the brand corresponding to "joint brand" is removed from the second brand set to obtain the first brand set.

[0124] Furthermore, it is determined whether the number of brands belonging to the same target category in the second brand set exceeds a preset value; if the number of brands belonging to the same target category in the second brand set exceeds the preset value, the brands belonging to the same target category in the second brand set are eliminated to obtain the first brand set. At the same time, it is determined whether there is a preset text in the segmentation set of the introduction of the consignment to be identified; if there is a preset text in the segmentation set of the introduction of the consignment to be identified, multiple brands corresponding to the preset text in the second brand set are eliminated to obtain the first brand set. The brands are eliminated based on whether the number of brands belonging to the same target category in the second brand set exceeds the preset value, and the brands are eliminated based on whether the preset text exists in the segmentation set of the introduction of the consignment to be identified. The two eliminations are performed together to further reduce the number of brands in the first brand set and improve recognition accuracy.

[0125] Specifically, each brand in the first brand set is determined according to formula (3).

[0126]

[0127] Among them, Ri_scd_step is the various brands in the first brand set, and Ri_fst_step is the various brands in the second brand set.

[0128] S202. Determine a first brand word vector for each first brand based on the category text of each first brand in the first brand set.

[0129] In a specific embodiment, determining the first brand word vector of each first brand based on the category text of each first brand in the first brand set may include:

[0130] (1) Based on the brand word vector conversion model, each category text of the first brand is vectorized to obtain the category word vector of each category text of the first brand.

[0131] Each first brand corresponds to multiple category texts. For example, brand F has three categories: electric kettle, electric stew pot, and health pot. Then the multiple category texts corresponding to brand F are: "electric kettle", "electric stew pot", and "health pot". The correspondence between the first brand and multiple category texts can be obtained and stored in advance, and can be directly read here.

[0132] The brand word vector conversion model can be a trained word vector conversion model. Input each category text of the first brand into the brand word vector conversion model to obtain the category word vector of each category text of the first brand. For example, "electric kettle", "electric stew pot", and "health pot" are input into the brand word vector conversion model to obtain three category word vectors V j .

[0133] Furthermore, since the various category texts of the first brand may be connected as a whole, for example, the presentation of the various category texts of the first brand is "electric kettle, electric stew pot, health pot", it is necessary to first segment the combined category text of the first brand to obtain multiple separated category texts. Specifically, the Jieba word segmentation tool is used to segment the combined category text C of the first brand based on the preset corpus. i (i=1,…,s) to perform word segmentation and obtain multiple category texts. Each category text is: C i_cut =jieba.cut(C i ). For example, if brand A has “electric kettle”, “electric stew pot”, and “health pot”, then s is 3.

[0134] The word vector conversion model can be one or more combinations of the word2vec model, CBOW model, glove model, and kip-gram model. The word vector is a distributed representation of words in deep learning, which represents words as a fixed-length continuous dense vector.

[0135] The word2vec model is a group of related models used to generate word vectors. These models are shallow, two-layer neural networks that are trained to reconstruct linguistic word texts. The network is represented by words and needs to guess the input words in adjacent positions. Under the word bag model assumption in word2vec, the order of words is not important. After training, the word2vec model can be used to map each word to a vector that can be used to represent the relationship between words. The vector is the hidden layer of the neural network. CBOW (Continuous Bag-of-Words Model) is a simplified expression model in natural language processing and information retrieval. Skip-gram is used to predict the context words corresponding to a given central word. It is the opposite of the Continuous Bag of Words (CBOW) algorithm. In Skip-gram, the central word is the input word and the context words are the output words. This process is more difficult because multiple context words need to be predicted.

[0136] (2) The weighted average of the category word vectors of the first brand is determined as the first brand word vector of the first brand.

[0137] Specifically, the average word vector of each category word vector of the first brand is determined as the first brand word vector of the first brand. Of course, a weight can also be set for each category word vector and then the average value is calculated.

[0138] For example, according to formula (4), the first brand word vector of the first brand is determined;

[0139] V Ri =mean(V j for j in C i_cut ) (4)

[0140] Among them, V Ri is the first brand word vector of the first brand.

[0141] Furthermore, before vectorizing each category text of the first brand based on a preset word vector conversion model to obtain the category word vector of each category text of the first brand, the following may be included:

[0142] (1) Obtain multiple consignment introduction texts to be trained.

[0143] In the embodiment of the present application, a plurality of consignment introduction texts to be trained are Di (i=1, ..., n), and different n represents different consignment introduction texts to be identified. The consignment introduction texts to be trained can be collected in various ways.

[0144] (2) Segmenting the multiple consignment introduction texts to be trained respectively to obtain a first consignment introduction segmentation set to be trained.

[0145] For example, the text of the consignment introduction to be trained is "Little Bear Electric Kettle Electric Stew Pot Health Pot". After word segmentation, the first word segmentation set of the consignment introduction to be trained is obtained: "Little Bear", "Electric Kettle", "Electric Stew Pot" and "Health Pot".

[0146] (3) Eliminate the brands in the first set of consignment introduction word segments to be trained, and obtain the second set of consignment introduction word segments to be trained.

[0147] For example, the brand "Little Bear" is removed from the first set of word segments for the introduction of consigned items to be trained, and the second set of word segments for the introduction of consigned items to be trained is obtained: "electric kettle", "electric stew pot" and "health pot".

[0148] (4) The second consignment introduction word segmentation set to be trained is used as a training set, and an unsupervised learning algorithm is used to train the preset word vector conversion model to obtain a brand word vector conversion model.

[0149] In real life, there are often such problems: lack of sufficient prior knowledge, so it is difficult to manually label categories or the cost of manual category labeling is too high. Naturally, we hope that computers can do this for us, or at least provide some help. Solving various problems in pattern recognition based on training samples with unknown categories (not labeled) is called unsupervised learning.

[0150] By training the consignment description segmentation words through unsupervised learning, we can save the cost of manual labeling and accurately predict the consignment brand.

[0151] S203: Determine a second brand word vector based on the introduction text of the consignment to be identified.

[0152] In a specific embodiment, determining the second brand word vector based on the introduction text of the consignment to be identified includes:

[0153] (1) Based on the brand word vector conversion model, each consignment introduction segment to be identified in the consignment introduction segment set is vectorized to obtain the word vector of each consignment introduction segment to be identified.

[0154] Specifically, the brands in the consignment introduction segmentation set to be identified are removed, and each consignment introduction segmentation to be identified after the brands are removed is vectorized to obtain the word vector of each consignment introduction segmentation to be identified.

[0155] For example, according to formula (5), the brands in each of the consignment introduction participles Di_cut_1 to be identified are removed to obtain each of the consignment introduction participles to be identified after the brands are removed.

[0156] D i_cut_2 =D i_cut_1 .difference(B) (5)

[0157] After removing the brands from the segmented words introducing the consignment to be identified, each segmented word introducing the consignment to be identified is vectorized based on the segmented words introducing the consignment after the brands are removed. This can avoid the influence of the existence of the brand on the vectorization of each segmented word introducing the consignment to be identified, improve the accuracy of the second brand word vector representation, and improve the brand recognition accuracy.

[0158] (2) The weighted average of the word vectors of the consignment introduction words to be identified is determined as the second brand word vector.

[0159] For example, the second brand word vector is determined according to formula (6).

[0160] V Di =mean(Vjfor j in D i_cut_2 ). (6)

[0161] Among them, V Di is the second brand word vector, and Vj is the word vector of each consignment introduction segment to be identified.

[0162] S204: Perform brand screening on the brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain a brand recognition result of the consignment.

[0163] In a specific embodiment, brands in the first brand set are screened based on the similarity between the first brand word vector and the second brand word vector to obtain a brand recognition result of the consignment, including:

[0164] (1) Calculate the similarity between the first brand word vector and the second brand word vector.

[0165] Specifically, the cosine similarity between each first brand word vector and each second brand word vector is calculated. Cosine similarity measures the similarity between two vectors by measuring the cosine value of the angle between them. The cosine similarity between the first brand word vector and each second brand word vector is cosine_similarity(V Ri ,V Di ).

[0166] The cosine value of an angle of 0 degrees is 1, while the cosine value of any other angle is not greater than 1; and its minimum value is -1. Therefore, the cosine value of the angle between two vectors determines whether the two vectors point in approximately the same direction. When two vectors have the same direction, the cosine similarity value is 1; when the angle between the two vectors is 90°, the cosine similarity value is 0; when the two vectors point in completely opposite directions, the cosine similarity value is -1. This result is independent of the length of the vector, but only related to the direction in which the vector points. Cosine similarity is usually used in positive space, so the value given is between -1 and 1.

[0167] (2) The first brand corresponding to the first brand word vector with the highest similarity is determined as the brand recognition result.

[0168] Calculate the cosine similarity of each first brand word vector and each second brand word vector, and retain the brand corresponding to the first brand word vector with the highest similarity. Since word vectors can learn the contextual relationship of words, brand matching errors caused by the same name can be avoided. For example, Xiaoxiong is an electrical appliance brand, and the categories sold are "electric kettles, electric stew pots, health pots", etc. However, due to the meaning of the word itself, there will be multiple brands including Xiaoxiong co-occurring in the text of the consignment introduction to be identified, such as "Brand T Men's Spring New Product Xiaoxiong Print Korean Breathable Pajamas". For this type of text, the first brand set will have an incorrect matching result for the brand "Xiaoxiong". After calculating the similarity of word vectors, the similarity score of "Brand T Men's Clothing", which has the same category and the word combination of the consignment introduction to be identified, is higher than that of "Xiaoxiong", so it is correctly predicted.

[0169] For those with a cosine similarity less than 0, it is considered that the first brand word vector and the second brand word vector are completely dissimilar. However, since the description text of the consignment to be identified may only contain the brand name and very few other words, such as "Brand E+Black+22 Code", the description text of the consignment to be identified cannot correctly reflect the brand category, and the cosine similarity is less than 0, directly judging according to the similarity threshold will lead to incorrect judgment.

[0170] Therefore, it is further determined whether the length of the text describing the consignment to be identified is greater than the preset length and whether the cosine similarity is less than the similarity threshold; when the length of the text describing the consignment to be identified is greater than the preset length and the cosine similarity is less than the similarity threshold, the first brand corresponding to the second brand word vector corresponding to the cosine similarity is removed from the first brand set to obtain the first brand set after removal, and the first brand corresponding to the first brand word vector with the highest similarity in the first brand set after removal is determined as the brand recognition result.

[0171] In a specific embodiment, the preset length is 4, the similarity threshold is 0, and the first brand set is eliminated according to formula (7) to obtain the first brand set after elimination;

[0172]

[0173] Among them, Ri_scd_step is the brands in the first brand set, and Ri_lst_step is the brands in the first brand set after elimination.

[0174] Among the brands in the first brand set after elimination, the first brand corresponding to the first brand word vector with the highest similarity is determined as the brand recognition result.

[0175] In order to better implement the method for identifying the brand of consigned goods in the embodiment of the present application, on the basis of the method for identifying the brand of consigned goods, the embodiment of the present application also provides a device for identifying the brand of consigned goods, such as Figure 4 As shown, Figure 4 : is a schematic diagram of the structure of an embodiment of a device for identifying a brand of consigned goods provided in an embodiment of the present application. The device for identifying a brand of consigned goods includes:

[0176] An acquisition unit 401 is used to acquire a consignment introduction text to be identified and a first brand set of the consignment;

[0177] A first determining unit 402, configured to determine a first brand word vector of each first brand based on the category text of each first brand in the first brand set;

[0178] A second determining unit 403, configured to determine a second brand word vector based on the description text of the consignment to be identified;

[0179] The brand screening unit 404 is used to screen the brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain a brand recognition result of the consignment.

[0180] Optionally, the acquiring unit 401 is configured to:

[0181] Obtain a preset corpus, wherein the preset corpus includes a preset brand set and category texts corresponding to each brand in the preset brand set;

[0182] Segment the text of the consignment introduction to be identified based on the preset corpus to obtain a segmentation set of the consignment introduction to be identified;

[0183] The intersection of the word set of the description of the consigned goods to be identified and the preset brand set is obtained to obtain a second brand set;

[0184] The first brand set is determined based on the second brand set.

[0185] Optionally, the brand screening unit 404 is used to:

[0186] Calculate the similarity between the first brand word vector and the second brand word vector;

[0187] The first brand corresponding to the first brand word vector with the highest similarity is determined as the brand recognition result.

[0188] Optionally, the first determining unit 402 is configured to:

[0189] Based on the brand word vector conversion model, each category text of the first brand is vectorized to obtain the category word vector of each category text of the first brand;

[0190] The weighted average of the category word vectors of the first brand is determined as the first brand word vector of the first brand.

[0191] Optionally, the second determining unit 403 is configured to:

[0192] Based on the brand word vector conversion model, each consignment introduction segment word to be identified in the consignment introduction segment word set to be identified is vectorized to obtain the word vector of each consignment introduction segment word to be identified;

[0193] The weighted average of the word vectors of the consignment introduction participles to be identified is determined as the second brand word vector.

[0194] Optionally, the acquiring unit 401 is configured to:

[0195] Determine whether the number of brands belonging to the same target category in the second brand set exceeds a preset value; if the number of brands belonging to the same target category in the second brand set exceeds the preset value, remove the brands belonging to the same target category in the second brand set to obtain the first brand set;

[0196] And / or, determine whether there is a preset text in the word set of the consignment introduction to be identified; if there is a preset text in the word set of the consignment introduction to be identified, remove multiple brands corresponding to the preset text from the second brand set to obtain the first brand set.

[0197] Optionally, the acquiring unit 401 is configured to:

[0198] Obtain multiple consignment introduction texts to be trained;

[0199] Segmenting the multiple consignment introduction texts to be trained respectively to obtain a first consignment introduction segmentation set to be trained;

[0200] Eliminate the brands in the first consignment introduction word set to be trained to obtain the second consignment introduction word set to be trained;

[0201] The second consignment introduction word segmentation set to be trained is used as a training set, and an unsupervised learning algorithm is used to train the preset word vector conversion model to obtain a brand word vector conversion model.

[0202] The present application also provides an electronic device that integrates any one of the consignment brand identification devices provided in the present application. Figure 5 As shown, it shows a schematic diagram of the structure of the electronic device involved in the embodiment of the present application, specifically:

[0203] The electronic device may include components such as a processor 501 with one or more processing cores, a memory 502 with one or more computer-readable storage media, a power supply 503, and an input unit 504. Those skilled in the art will appreciate that the electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. Among them:

[0204] The processor 501 is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 502, and calling data stored in the memory 502, the processor 501 performs various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. Optionally, the processor 501 may include one or more processing cores; preferably, the processor 501 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 501.

[0205] The memory 502 can be used to store software programs and modules. The processor 501 executes various functional applications and data processing by running the software programs and modules stored in the memory 502. The memory 502 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 502 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 502 may also include a memory controller to provide the processor 501 with access to the memory 502.

[0206] The electronic device also includes a power supply 503 for supplying power to each component. Preferably, the power supply 503 can be logically connected to the processor 501 through a power management system, so as to manage charging, discharging, power consumption and other functions through the power management system. The power supply 503 can also include one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, power status indicators and other arbitrary components.

[0207] The electronic device may further include an input unit 504, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0208] Although not shown, the electronic device may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 501 in the electronic device will load the executable files corresponding to the processes of one or more application programs into the memory 502 according to the following instructions, and the processor 501 will run the application programs stored in the memory 502, thereby realizing various functions, as follows:

[0209] Obtaining the consignment introduction text to be identified and the first brand set of the consignment;

[0210] Determine the first brand word vector of each first brand based on the category text of each first brand in the first brand set,

[0211] Determine a second brand word vector based on the introduction text of the consignment to be identified;

[0212] Based on the similarity between the first brand word vector and the second brand word vector, brands in the first brand set are screened to obtain a brand recognition result of the consignment.

[0213] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0214] To this end, the embodiment of the present application provides a computer-readable storage medium, which may include: a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk, etc. A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in any of the consignment brand identification methods provided in the embodiment of the present application. For example, the computer program loaded by the processor can execute the following steps:

[0215] Obtaining the consignment introduction text to be identified and the first brand set of the consignment;

[0216] Determine the first brand word vector of each first brand based on the category text of each first brand in the first brand set;

[0217] Determine a second brand word vector based on the introduction text of the consignment to be identified;

[0218] Based on the similarity between the first brand word vector and the second brand word vector, brands in the first brand set are screened to obtain a brand recognition result of the consignment.

[0219] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the detailed description of other embodiments above, and will not be repeated here.

[0220] In specific implementation, the above units or structures can be implemented as independent entities, or can be arbitrarily combined to be implemented as the same or several entities. The specific implementation of the above units or structures can refer to the previous method embodiments, which will not be repeated here.

[0221] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0222] The above is a detailed introduction to a method and device for identifying a brand of consigned goods provided in an embodiment of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A method for identifying a consignment brand, characterized in that: The identification method comprises: Obtaining a consignment introduction text to be identified and a first brand set of the consignment; Determine a first brand word vector for each first brand based on the category text of each first brand in the first brand set; Determining a second brand word vector based on the introduction text of the consignment to be identified; Performing brand screening on brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain a brand recognition result of the consignment; The step of obtaining the consignment introduction text to be identified and the first brand set of the consignment includes: Acquire a preset corpus, wherein the preset corpus includes a preset brand set and category texts corresponding to each brand in the preset brand set; Segmenting the text of the introduction to the consigned object to be identified based on the preset corpus to obtain a segmentation set of the introduction to the consigned object to be identified; Calculate the intersection of the word set of the description of the consigned object to be identified and the preset brand set to obtain a second brand set; determining the first brand set based on the second brand set; The determining the first brand set based on the second brand set includes: Determine whether the number of brands belonging to the same target category in the second brand set exceeds a preset value; if the number of brands belonging to the same target category in the second brand set exceeds the preset value, remove the brands belonging to the same target category in the second brand set to obtain the first brand set; And / or, determine whether there is a preset text in the set of word segmentations for the description of the consignment to be identified; if there is a preset text in the set of word segmentations for the description of the consignment to be identified, remove the multiple brands corresponding to the preset text from the second brand set to obtain the first brand set, wherein the preset text is pre-constructed with a corresponding brand list, and the brand list includes the preset text and the corresponding multiple brands.

2. The method for identifying the brand of consigned goods according to claim 1, characterized in that: The brand screening of brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector to obtain the brand recognition result of the consignment includes: Calculating the similarity between the first brand word vector and the second brand word vector; The first brand corresponding to the first brand word vector with the highest similarity is determined as the brand recognition result.

3. The method for identifying the brand of consigned goods according to claim 1, characterized in that: The determining the first brand word vector of each first brand based on the category text of each first brand in the first brand set includes: vectorizing each category text of the first brand based on the brand word vector conversion model to obtain the category word vector of each category text of the first brand; A weighted average of the category word vectors of the first brand is determined as the first brand word vector of the first brand.

4. The method for identifying the brand of consigned goods according to claim 1, characterized in that: The determining of the second brand word vector based on the introduction text of the consignment to be identified includes: Based on the brand word vector conversion model, each of the to-be-identified consignment introduction participles in the to-be-identified consignment introduction participle set is vectorized to obtain a word vector of each to-be-identified consignment introduction participle; The weighted average of the word vectors of the consignment introduction participles to be identified is determined as the second brand word vector.

5. The method for identifying the brand of consigned goods according to claim 3, characterized in that: The process of vectorizing each category text of the first brand based on the brand word vector conversion model to obtain the category word vector of each category text of the first brand includes: Obtain multiple consignment introduction texts to be trained; Segmenting the plurality of to-be-trained consignment introduction texts respectively to obtain a first to-be-trained consignment introduction segmentation set; Eliminate the brands in the first to-be-trained consignment introduction word set to obtain a second to-be-trained consignment introduction word set; The second consignment introduction word set to be trained is used as a training set, and an unsupervised learning algorithm is used to train the preset word vector conversion model to obtain the brand word vector conversion model.

6. A device for identifying the brand of consigned goods, characterized in that: The identification device comprises: An acquiring unit, configured to acquire an introduction text of the consignment to be identified and a first brand set of the consignment; A first determining unit, configured to determine a first brand word vector of each first brand based on the category text of each first brand in the first brand set; A second determining unit, configured to determine a second brand word vector based on the introduction text of the consignment to be identified; a brand screening unit, configured to screen brands in the first brand set based on the similarity between the first brand word vector and the second brand word vector, and obtain a brand recognition result of the consignment; The step of obtaining the consignment introduction text to be identified and the first brand set of the consignment includes: Acquire a preset corpus, wherein the preset corpus includes a preset brand set and category texts corresponding to each brand in the preset brand set; Segmenting the text of the introduction to the consigned object to be identified based on the preset corpus to obtain a segmentation set of the introduction to the consigned object to be identified; Calculate the intersection of the word set of the description of the consigned object to be identified and the preset brand set to obtain a second brand set; determining the first brand set based on the second brand set; The determining the first brand set based on the second brand set includes: Determine whether the number of brands belonging to the same target category in the second brand set exceeds a preset value; if the number of brands belonging to the same target category in the second brand set exceeds the preset value, remove the brands belonging to the same target category in the second brand set to obtain the first brand set; And / or, determine whether there is a preset text in the set of word segmentations for the description of the consignment to be identified; if there is a preset text in the set of word segmentations for the description of the consignment to be identified, remove the multiple brands corresponding to the preset text from the second brand set to obtain the first brand set, wherein the preset text is pre-constructed with a corresponding brand list, and the brand list includes the preset text and the corresponding multiple brands.

7. An electronic device, characterized in that: The electronic device comprises: one or more processors; Memory; and One or more applications, wherein the one or more applications are stored in the memory and are configured to be executed by the processor to implement the method for identifying the brand of consignment according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program is loaded by a processor to execute the steps in the method for identifying the brand of consigned goods as described in any one of claims 1 to 5.

Images