CN114661895A - Commodity classification method and device, storage medium and electronic equipment - Google Patents

Commodity classification method and device, storage medium and electronic equipment Download PDF

Info

Publication number
CN114661895A
CN114661895A CN202011540233.2A CN202011540233A CN114661895A CN 114661895 A CN114661895 A CN 114661895A CN 202011540233 A CN202011540233 A CN 202011540233A CN 114661895 A CN114661895 A CN 114661895A
Authority
CN
China
Prior art keywords
commodity
information
sub
type
feature
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011540233.2A
Other languages
Chinese (zh)
Inventor
唐杰
罗干
刘德兵
郭同
杨林
胡懋地
刘怀军
李滔
王栋
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tsinghua University
Beijing Sankuai Online Technology Co Ltd
Original Assignee
Tsinghua University
Beijing Sankuai Online Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tsinghua University, Beijing Sankuai Online Technology Co Ltd filed Critical Tsinghua University
Priority to CN202011540233.2A priority Critical patent/CN114661895A/en
Publication of CN114661895A publication Critical patent/CN114661895A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367—Ontology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/047—Probabilistic or stochastic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0623—Item investigation
    • G06Q30/0625—Directed, with specific intent or strategy
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0631—Item recommendations

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Computational Linguistics (AREA)
  • Software Systems (AREA)
  • Finance (AREA)
  • Molecular Biology (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computing Systems (AREA)
  • Accounting & Taxation (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Databases & Information Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • General Business, Economics & Management (AREA)
  • Animal Behavior & Ethology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The disclosure relates to a commodity classification method, a commodity classification device, a storage medium and an electronic device. The method comprises the following steps: acquiring commodity information of a target commodity, wherein the commodity information comprises information acquired from a plurality of data sources; determining commodity sub-information corresponding to each data type from the commodity information according to a preset data type; inputting all commodity sub-information into a commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to a target commodity; the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for inputting the commodity sub-information into the feature representation module corresponding to the data type of the commodity sub-information to obtain the feature vector of the commodity sub-information aiming at each commodity sub-information; splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of a target commodity; and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain a classification label set output by the classification module.

Description

Commodity classification method and device, storage medium and electronic equipment
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a method and an apparatus for classifying commodities, a storage medium, and an electronic device.
Background
In the commodity knowledge graph of the e-commerce platform, each commodity has an association relation with some tag entities. In the related technology, the method for mining the association between the goods and the tag entities is to extract the association between the goods and the tag entities from the goods name information, the goods picture information or the merchant label information in the related service database of the e-commerce platform.
However, the number of tag entities having an association relationship with a commodity acquired by this method is very limited, so that the commodity has tags that are not complete enough.
Disclosure of Invention
The present disclosure provides a method, an apparatus, a storage medium and an electronic device for classifying commodities, so that the commodities have more accurate and complete classification labels.
In order to achieve the above object, a first part of the embodiments of the present disclosure provides a method of classifying a commodity, the method including:
the method comprises the steps of obtaining commodity information of a target commodity, wherein the commodity information comprises information obtained from a plurality of data sources;
determining commodity sub-information corresponding to each data type from the commodity information according to a preset data type, wherein the preset data type comprises at least one of a commodity name type, a description text type, a picture type and a structural attribute type;
inputting all the commodity sub-information into a commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to the target commodity;
the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for:
aiming at each commodity sub-information, inputting the commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information to obtain a feature vector of the commodity sub-information;
splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of the target commodity;
and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain the classification label set which is output by the classification module and corresponds to the target commodity.
Optionally, the method is applied to an e-commerce platform, and the method further comprises:
determining a target label entity set which is missing in the label entity set according to the classification label set of the target commodity and a label entity set which has an incidence relation with the target commodity in a commodity knowledge graph of the e-commerce platform;
and establishing an incidence relation between the target commodity and each tag entity in the target tag entity set in the commodity knowledge graph.
Optionally, the method is applied to an e-commerce platform, and the method further comprises:
determining an interest tag set of a user based on the classification tag set of each historical commodity searched by the user;
and determining the commodity to be recommended corresponding to the classified label set with the similarity greater than a preset threshold value with the interested label set, and pushing the information of the commodity to be recommended to the user.
Optionally, the feature representation module corresponding to the commodity name type is configured to determine a feature vector of each character in the commodity sub-information of the commodity name type; and are
Determining a feature vector of the commodity sub-information of the commodity name type through the following calculation formula:
Figure BDA0002854680230000021
wherein e isnameA feature vector, h, of the commodity sub-information characterizing the commodity name typeiAnd k is the character number of the commodity sub-information of the commodity name type.
Optionally, the feature representation module corresponding to the description text type is configured to:
determining a feature vector of each character in the commodity sub-information describing the text type;
for each text segment, determining a feature matrix of the text segment characters corresponding to the text segment according to the feature vector of each character in the text segment;
and aiming at the feature matrix of the text segment character corresponding to each text segment, determining a weighted text segment feature vector corresponding to the feature matrix of the text segment character through the following calculation formula:
hpara＝αtokenHtoken，
wherein，αtoken＝softmax(qtokentanh(HtokenWtoken)T)，
Wherein h isparaCharacterizing the feature vectors of weighted text segments, HtokenFeature matrix, alpha, characterizing characters of a text segmenttokenCharacterizing a vector consisting of the weights of the characters, Wtoken、qtokenIs a training parameter of the feature representation module;
according to the weight text segment feature vector corresponding to each text segment and the weight corresponding to each text segment, determining the feature vector of the commodity sub-information describing the text type through the following calculation formula:
eparas＝αpara Hpara，
wherein alpha ispara＝softmax(qparatanh(Hpara Wpara)T)，
Wherein e isparasFeature vectors, H, characterizing the sub-information of the goods of the description text typeparaCharacterizing a matrix composed of weighted text segment feature vectors of each text segment, alphaparaCharacterizing a vector consisting of the weights of the texts, qpara、WparaAre the training parameters of the feature representation module.
Optionally, the feature representation module corresponding to the picture type is configured to:
determining a picture characteristic vector corresponding to each picture in the sub information of the picture type; and are combined
Determining a feature vector of the commodity sub-information of the picture type by the following calculation formula:
eimgs＝αimgHimg；
wherein alpha isimg＝softmax(qimg tanh(HimgWimg)T)；
Wherein e isimgsCharacteristic vector, H, of the sub-information of the goods characterizing the picture typeimgCharacterizing a matrix composed of feature vectors of pictures, alphaimgCharacterizing a vector consisting of the weights of the pictures, Wimg、qimgAre the feature representation module training parameters.
Optionally, the feature representation module corresponding to the structured attribute type is configured to determine a feature vector of the commodity sub-information of the structured attribute type through the following calculation formula:
estr＝ReLU(efeatureWstr)；
wherein e isstrFeature vectors characterizing the sub-information of the goods of said structured attribute type, efeatureA structured attribute feature vector, W, characterizing the commodity sub-information of said structured attribute typestrThe training parameters of the module are represented for the features.
A second part of the disclosed embodiments provides a goods sorting device, the device comprising:
the system comprises an acquisition module, a display module and a display module, wherein the acquisition module is configured to be used for acquiring commodity information of a target commodity, and the commodity information comprises information acquired from a plurality of data sources;
the first determining module is configured to determine commodity sub-information corresponding to each data type from the commodity information according to preset data types, wherein the preset data types include at least one of a commodity name type, a description text type, a picture type and a structured attribute type;
the input module is configured to input all the commodity sub-information into a commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to the target commodity;
the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for:
inputting the commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information aiming at each commodity sub-information to obtain a feature vector of the commodity sub-information;
splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of the target commodity;
and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain the classification label set which is output by the classification module and corresponds to the target commodity.
Optionally, the device is applied to an e-commerce platform, and the device further comprises:
a second determination module, configured to determine, according to the classification tag set of the target commodity and a tag class entity set in a commodity knowledge graph of the e-commerce platform, which has an association relationship with the target commodity, a missing target tag class entity set in the tag class entity set;
the establishing module is configured to establish an association relationship between the target product and each tag class entity in the target tag class entity set in the product knowledge graph.
Optionally, the apparatus is applied to an e-commerce platform, and the apparatus further includes:
a third determination module configured to determine a tag set of interest of a user based on the category tag set of each historical item searched by the user;
the recommending module is configured to determine a to-be-recommended commodity corresponding to the classified label set with the similarity greater than a preset threshold with the interested label set, and push information of the to-be-recommended commodity to the user.
Optionally, the feature representation module corresponding to the commodity name type is configured to determine a feature vector of each character in the commodity sub-information of the commodity name type; and are combined
Determining a feature vector of the commodity sub-information of the commodity name type through the following calculation formula:
Figure BDA0002854680230000041
wherein e isnameA feature vector, h, of the commodity sub-information characterizing the commodity name typeiIs the feature vector of the ith character, k is the type of the commodity nameThe number of characters of the commodity sub-information.
Optionally, the feature representation module corresponding to the description text type is configured to:
determining a feature vector of each character in the commodity sub-information describing the text type;
for each text segment, determining a feature matrix of the text segment characters corresponding to the text segment according to the feature vector of each character in the text segment;
and aiming at the feature matrix of the text segment character corresponding to each text segment, determining a weighted text segment feature vector corresponding to the feature matrix of the text segment character through the following calculation formula:
hpara＝αtokenHtoken，
wherein alpha istoken＝softmax(qtoken tanh(HtokenWtoken)T)，
Wherein h isparaCharacterizing the feature vectors of weighted text segments, HtokenFeature matrix, alpha, characterizing characters of a text segmenttokenCharacterizing a vector consisting of the weights of the characters, Wtoken、qtokenIs a training parameter of the feature representation module;
according to the weight text segment feature vector corresponding to each text segment and the weight corresponding to each text segment, determining the feature vector of the commodity sub-information describing the text type through the following calculation formula:
eparas＝αpara Hpara，
wherein alpha ispara＝softmax(qpara tanh(Hpara Wpara)T)，
Wherein e isparasFeature vectors, H, characterizing the sub-information of the goods of the description text typeparaCharacterizing a matrix composed of weighted text segment feature vectors of text segments, alphaparaCharacterizing a vector consisting of the weights of the texts, qpara、WparaAre the training parameters of the feature representation module.
Optionally, the feature representation module corresponding to the picture type is configured to:
determining a picture characteristic vector corresponding to each picture in the sub information of the picture type; and are combined
Determining a feature vector of the commodity sub-information of the picture type by the following calculation formula:
eimgs＝αimgHimg；
wherein alpha isimg＝softmax(qimg tanh(HimgWimg)T)；
Wherein e isimgsCharacteristic vector, H, of the sub-information of the goods characterizing the picture typeimgCharacterizing a matrix of feature vectors, alpha, of each pictureimgCharacterizing a vector consisting of the weights of the pictures, Wimg、qimgAre the feature representation module training parameters.
Optionally, the feature representation module corresponding to the structured attribute type is configured to determine a feature vector of the commodity sub-information of the structured attribute type through the following calculation formula:
estr＝ReLU(efeatureWstr)；
wherein e isstrFeature vectors characterizing the sub-information of the goods of the structured attribute type, efeatureA structured attribute feature vector, W, characterizing the commodity sub-information of said structured attribute typestrThe training parameters of the module are represented for the features.
A third part of the embodiments of the present disclosure provides a computer-readable storage medium, on which a computer program is stored, which when executed by a processor, performs the steps of the method of any one of the above-mentioned first aspects.
A fourth aspect of the embodiments of the present disclosure provides an electronic apparatus, including:
a memory having a computer program stored thereon;
a processor for executing the computer program in the memory to implement the steps of the method of any of the first aspects above.
By adopting the technical scheme, the following technical effects can be at least achieved:
the commodity information of the target commodity is obtained from a plurality of different data sources, so that the obtained commodity information is richer. And more commodity classification labels can be extracted from the rich commodity information. And determining commodity sub-information respectively corresponding to each data type from the commodity information according to a preset data type, inputting each commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information in the commodity multi-classification model, and obtaining a feature vector of the corresponding commodity sub-information. And splicing the characteristic vectors of the commodity sub-information to obtain a commodity vector of the target commodity. The commodity vectors obtained based on splicing are subjected to multi-classification processing, and the commodity vectors comprise commodity sub-information corresponding to each data type, so that the multi-classification results can comprehensively consider the commodity sub-information of each data type and the relation among the commodity sub-information, and on the basis of obtaining richer commodity classification labels, the obtained commodity classification labels can be more accurate.
Additional features and advantages of the present disclosure will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this specification, illustrate embodiments of the disclosure and together with the description serve to explain the disclosure, but do not constitute a limitation of the disclosure. In the drawings:
fig. 1 is a flowchart illustrating a method of classifying items according to an exemplary embodiment of the present disclosure.
Fig. 2 is a schematic diagram illustrating merchandise information according to an exemplary embodiment of the present disclosure.
FIG. 3 is a block diagram illustrating a feature representation module according to an exemplary embodiment of the present disclosure.
Fig. 4 is a block diagram illustrating an article sorting apparatus according to an exemplary embodiment of the present disclosure.
FIG. 5 is a block diagram illustrating an electronic device in accordance with an example embodiment.
Detailed Description
The following detailed description of specific embodiments of the present disclosure is provided in connection with the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present disclosure, are given by way of illustration and explanation only, not limitation.
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. When the following description refers to the accompanying drawings, like numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the disclosure, as detailed in the appended claims.
In the related technology, a method for mining the association relationship between the goods and the label entities is to respectively extract the association relationship between the goods and the label entities from the goods name information, the goods picture information or the merchant label information in the related service database of the e-commerce platform. Because the information in the related service database of the e-commerce platform is limited, the number of the label entities which are acquired by the method and have the association relation with the commodities is very limited.
In addition, the hidden relationship among the commodity name information, the commodity picture information and the merchant labeling information is not considered in this way, so that the situation that the label entities which are extracted in this way and have the association relationship with the commodity may be contradictory exists, for example, a label entity with spicy taste of the commodity may be extracted from the commodity picture information, and a label entity with light taste of the commodity is extracted from the merchant labeling information, and obviously, the label entity with spicy taste and the label entity with light taste are two label entities which are contradictory. Therefore, the label entities with the association relation with the commodities obtained by the method are inaccurate.
In view of this, the present disclosure provides a method, an apparatus, a storage medium, and an electronic device for classifying a product, so as to obtain a rich and accurate set of product classification tags.
Fig. 1 is a flowchart illustrating a method of classifying items according to an exemplary embodiment of the present disclosure, as shown in fig. 1, the method including:
and S11, acquiring commodity information of the target commodity, wherein the commodity information comprises information acquired from a plurality of data sources.
The data sources comprise Internet platforms such as an e-commerce platform, encyclopedia entries, a public knowledge base, a commenting platform, a teaching website and the like. In one possible approach, the data source further includes manual research data. The present disclosure is not particularly limited thereto.
The commodity information may be a category, a name, a regional attribute, a commodity attribute, ingredients, raw materials, components, an efficacy, a manufacturing method, an appearance image, and the like of any target commodity.
And S12, determining commodity sub-information corresponding to each data type from the commodity information according to preset data types, wherein the preset data types comprise at least one of commodity name types, description text types, picture types and structured attribute types.
The preset data types include a commodity name type, a description text type, a picture type, a structured attribute type, a voice description type, a video type and the like. The present disclosure is not particularly limited thereto.
For example, assuming that the target product is the shredded pork with a fish-flavor, the product information of the shredded pork with a fish-flavor is shown in fig. 2. Then, the commodity sub-information of the commodity name type refers to the commodity text name, i.e., "fish-flavoured shredded pork". The commodity sub-information describing the text type refers to a menu, an encyclopedia introduction, a user comment text and the like of the fish-flavored shredded pork. The commodity sub-information of the picture type refers to the picture shown in the upper right corner of fig. 2. The commodity sub-information of the structured attribute type refers to a structured attribute feature vector extracted from the commodity sub-information of the description text type.
It should be noted that, in an implementation manner, a product attribute set of a product category to which a target product belongs and a vector representation manner of each product attribute may be predefined, for example, 0, 1 and a combination of 0 and 1 are used to represent each predefined product attribute. Furthermore, based on the product attribute set of the product category to which the target product belongs, the product attributes can be extracted from the product information of the target product, and the extracted product attributes can be converted into corresponding vector representations, so that the product sub-information of the structured attribute type, which is correspondingly represented by 0 and 1, of the target product can be obtained.
It is easily understood that, in another implementation, each predefined commodity attribute may not be represented by a combination of 0 and 1, but may be directly represented by text information of each commodity attribute.
S13, inputting all the commodity sub-information into a commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to the target commodity;
the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for: inputting the commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information aiming at each commodity sub-information to obtain a feature vector of the commodity sub-information; splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of the target commodity; and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain the classification label set which is output by the classification module and corresponds to the target commodity.
The classification module comprises an MLP (multi-layer perceptron) and a sigmoid function layer of the multi-classification model of the commodity.
Illustratively, it is assumed that the preset data types include a commodity name type, a description text type, a picture type, a structured attribute type, a voice description type, and a video type. Correspondingly, the commodity multi-classification model comprises feature representation modules respectively corresponding to a commodity name type, a description text type, a picture type, a structured attribute type, a voice description type and a video type.
Specifically, the multi-classification model of the commodity is used for inputting commodity sub-information of the commodity name type into the feature representation module corresponding to the commodity name type to obtain a feature vector corresponding to the commodity name type. And inputting the commodity sub-information of the description text type into a feature representation module corresponding to the description text type to obtain a feature vector corresponding to the description text type. And inputting the commodity sub-information of the picture type into a feature representation module corresponding to the picture type to obtain a feature vector corresponding to the picture type. And inputting the commodity sub-information of the structured attribute type into a feature representation module corresponding to the structured attribute type to obtain a feature vector corresponding to the structured attribute type. And inputting the commodity sub-information of the voice description type into a feature representation module corresponding to the voice description type to obtain a feature vector corresponding to the voice description type. And inputting commodity sub-information of the video type into a feature representation module corresponding to the video type to obtain a feature vector corresponding to the video type.
Further, the commodity multi-classification model splices the feature vector corresponding to the commodity name type, the feature vector corresponding to the description text type, the feature vector corresponding to the picture type, the feature vector corresponding to the structured attribute type, the feature vector corresponding to the voice description type and the feature vector corresponding to the video type to obtain the commodity vector. And inputting the commodity vector into an MLP (multi-layer perceptron) and a sigmoid function layer in the commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to the target commodity.
In one possible embodiment, a weight may be assigned to the feature vector corresponding to each data type of the sub-information of the product, so that when performing multi-classification processing according to the product vector, a plurality of contradictory classification labels may be avoided based on the weight of the feature vector corresponding to each data type of the sub-information of the product.
By adopting the method, the commodity information of the target commodity is obtained from a plurality of different data sources, so that the obtained commodity information is richer. And more commodity classification labels can be extracted from the rich commodity information. And determining commodity sub-information corresponding to each data type from the commodity information according to a preset data type, inputting each commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information in the commodity multi-classification model, and obtaining a feature vector of the corresponding commodity sub-information. And splicing the characteristic vectors of the commodity sub-information to obtain a commodity vector of the target commodity. The commodity vectors obtained based on splicing are subjected to multi-classification processing, and the commodity vectors comprise commodity sub-information corresponding to each data type, so that the multi-classification results can comprehensively consider the commodity sub-information of each data type and the relation among the commodity sub-information, and on the basis of obtaining richer commodity classification labels, the obtained commodity classification labels can be more accurate.
Optionally, the above-mentioned goods classification method may be applied to an e-commerce platform, and in particular, may be applied to a server of the e-commerce platform, and the method may further include the following steps:
determining a target label entity set which is missing in the label entity set according to the classification label set of the target commodity and a label entity set which has an incidence relation with the target commodity in a commodity knowledge graph of the e-commerce platform; and establishing an incidence relation between the target commodity and each tag entity in the target tag entity set in the commodity knowledge graph.
It is easy to understand that, in the related art, the association relationship between the goods and the tag entities in the goods knowledge graph of the e-commerce platform is extracted from the goods name information, the goods picture information or the merchant label information in the related service database of the e-commerce platform respectively. The number of the label entities which are acquired by the method and have the association relation with the commodities is very limited, so that the commodity knowledge graph is not complete.
In order to supplement the commodity knowledge graph of the e-commerce platform, after the step S13, a target tag entity set missing from the tag entity set may be determined according to the classification tag set of the target commodity and a tag entity set having an association relationship with the target commodity in the commodity knowledge graph of the e-commerce platform. Further, in the commodity knowledge graph, an association relationship between the target commodity and each tag class entity in the missing target tag class entity set is established. Therefore, the knowledge map of the commodity can be further enriched and improved.
Optionally, the above method for classifying commodities can be applied to an e-commerce platform, and the method can further include the following steps:
determining an interest tag set of a user based on the classification tag set of each historical commodity searched by the user; and determining the commodity to be recommended corresponding to the classified label set with the similarity greater than a preset threshold value with the interested label set, and pushing the information of the commodity to be recommended to the user.
Specifically, based on the classification label set of each historical commodity (or part of historical commodities) searched by the user, the interest label set of the user is determined. For example, each classification label with the occurrence frequency greater than a preset frequency in the classification label set of each historical commodity searched by the user is used as the interest label of the user, so as to obtain the interest label set of the user. Determining the goods to be recommended corresponding to the classified label set with the similarity greater than the preset threshold with the interested label set, and pushing the information of the goods to be recommended to the user. The quantity of the tags in the interest tag set and the quantity of the tags in the classification tag set of the to-be-recommended goods may be the same or different. The present disclosure is not particularly limited thereto.
In this way, the user can directly and quickly find the required commodities from the pushed commodity information without actively searching and screening the required commodities from the vast commodities, so that the efficiency of screening the commodities by the user is improved.
Optionally, the feature representation module corresponding to the commodity name type is configured to determine a feature vector of each character in the commodity sub-information of the commodity name type; and determining the characteristic vector of the commodity sub-information of the commodity name type through the following calculation formula:
Figure BDA0002854680230000101
wherein e isnameCharacterizing the goods of the goods name typeFeature vector of information, hiAnd k is the feature vector of the ith character, and the number of the characters of the commodity sub-information of the commodity name type.
In an implementation, the product sub-information (i.e. text information) for the product name type or the description text type can be encoded by using the BERT model. In detail, the trained BERT model is obtained by performing fine-tune on the BERT pre-trained model to adapt to the commodity text corpus, and each character in the commodity name type and the commodity sub-information describing the text type can be represented as a corresponding vector by using the trained BERT model.
Among them, the bert (bidirectional Encoder retrieval for transforms) model is a transform-based deep bidirectional coding text representation model proposed by Google, and exhibits excellent coding performance in each NLP (Natural Language processing) basic task through model pre-training and task fine adjustment.
Optionally, the feature representation module corresponding to the description text type is configured to:
determining a feature vector of each character in the commodity sub-information describing the text type; for each text segment, determining a feature matrix of the text segment characters corresponding to the text segment according to the feature vector of each character in the text segment; and aiming at the feature matrix of the text segment character corresponding to each text segment, determining a weighted text segment feature vector corresponding to the feature matrix of the text segment character through the following calculation formula:
hpara＝αtokenHtoken，
wherein alpha istoken＝softmax(qtoken tanh(HtokenWtoken)T)，
Wherein h isparaCharacterizing the feature vectors of weighted text segments, HtokenFeature matrix, alpha, characterizing characters of a text segmenttokenCharacterizing a vector consisting of the weights of the characters, Wtoken、qtokenIs a training parameter of a feature representation module, in particular WtokenIs the firstTrained matrix in the layer attention mechanism, qtokenIs a well-trained vector in the first layer attention mechanism;
according to the weight text segment feature vector corresponding to each text segment and the weight corresponding to each text segment, determining the feature vector of the commodity sub-information describing the text type through the following calculation formula:
eparas＝αpara Hpara，
wherein alpha ispara＝softmax(qpara tanh(Hpara Wpara)T)，
Wherein e isparasFeature vectors, H, characterizing the sub-information of the goods of the description text typeparaCharacterizing a matrix composed of weighted text segment feature vectors of text segments, alphaparaCharacterizing a vector consisting of the weights of the texts, qpara、WparaIs a training parameter of a feature representation module, in particular WparaIs a trained matrix parameter in the second layer attention mechanism, qparaAre well-trained vector parameters in the second tier attention mechanism.
The text segment refers to the text information contained from the ending mark of one text segment to the ending mark of the next text segment. The end mark may be any punctuation mark. In brief, the text paragraphs are divided by the preset punctuation marks.
Optionally, the feature representation module corresponding to the picture type is configured to:
determining a picture characteristic vector corresponding to each picture in the sub information of the picture type; and determining the feature vector of the commodity sub-information of the picture type through the following calculation formula:
eimgs＝αimmg Himg；
wherein alpha isimg＝softmax(qimg tanh(HimgWimg)T)；
Wherein e isimgsCharacteristic vector, H, of the sub-information of the goods characterizing the picture typeimgCharacterization ofMatrix of feature vectors of pictures, alphaimgCharacterizing a vector consisting of the weights of the pictures, wimg、qimgAre the feature representation module training parameters.
Wherein, WimgIs a well-trained matrix of attention training layer, qimgIs a well-trained vector for the attention training layer.
In an implementation manner, the picture feature vector corresponding to each picture in the sub information of the picture type may be determined through a trained Resnet model.
Optionally, the feature representation module corresponding to the structured attribute type is configured to determine a feature vector of the commodity sub-information of the structured attribute type through the following calculation formula:
estr＝ReLU(efeatureWstr)；
wherein e isstrFeature vectors characterizing the sub-information of the goods of the structured attribute type, efeatureA structured attribute feature vector, W, characterizing the commodity sub-information of said structured attribute typestrThe training parameters of the module are represented for the features.
An implementation manner can acquire attribute tag data made by a plurality of merchants or users for a target commodity, vote for each attribute information in commodity sub-information corresponding to a structured attribute type of the target commodity according to the attribute tag data made by the plurality of merchants or users for the target commodity, and take an attribute with a vote number larger than a preset vote number threshold value as a target attribute of the target commodity, thereby obtaining a target attribute set. Since the commodity sub-information of the structured attribute type may be the commodity sub-information of the structured attribute type characterized by 0, 1 and a combination of 0 and 1, the resulting target attribute set may also be the structured attribute feature vector represented by 0 and 1.
Alternatively, a full-connected layer may be used to reduce the dimension of the sparse structured attribute feature vector to a dense low-dimensional structured attribute feature vector, and then the dense low-dimensional structured attribute feature vector is used as e in the above calculation formulafeatureCorresponding value, utilizingThe above calculation formula calculates to obtain the feature vector of the commodity sub-information of the structured attribute type.
By adopting the attribute voting mode, partial invalid attribute information can be screened out, so that the interference of the invalid attribute information on multi-classification results is avoided, and the mode is favorable for further improving the accuracy of the multi-classification results.
Optionally, the feature vectors corresponding to the data types are spliced by the following calculation formula to obtain a commodity vector e:
Figure BDA0002854680230000131
further, the commodity vector e is input into a classification module in a multi-classification model, such as an MLP (multi-layer perceptron) and a sigmoid layer, so as to obtain a corresponding classification label set. The functions of the MLP (multilayer perceptron) and the sigmoid layer are as follows:
p(t|e)＝sigmoid(MLP(e))。
alternatively, the commodity information may be preprocessed after the above step S12 and before S13. For example, a picture in the commodity information may be converted into numerical data. For example, the text information in the commodity information can be converted into numerical data that can be recognized by the multi-classification model.
In order to make the principle of the above-mentioned feature representation modules more easily understood by those skilled in the art, the present disclosure shows a specific architecture diagram of the feature representation modules corresponding to the above-mentioned commodity name type, description text type, picture type, and structured attribute type, respectively, in fig. 3.
Fig. 4 is a block diagram illustrating an article sorting apparatus according to an exemplary embodiment of the present disclosure, and as shown in fig. 4, the apparatus 400 includes:
an obtaining module 410 configured to obtain commodity information of a target commodity, the commodity information including information obtained from a plurality of data sources;
a first determining module 420, configured to determine, according to preset data types, commodity sub-information corresponding to each data type from the commodity information, where the preset data types include at least one of a commodity name type, a description text type, a picture type, and a structured attribute type;
an input module 430, configured to input all the commodity sub-information into a commodity multi-classification model, so as to obtain a classification label set output by the commodity multi-classification model and corresponding to the target commodity;
the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for:
inputting the commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information aiming at each commodity sub-information to obtain a feature vector of the commodity sub-information;
splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of the target commodity;
and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain the classification label set which is output by the classification module and corresponds to the target commodity.
With this apparatus 400, the product information of the target product is acquired from a plurality of different data sources, and the acquired product information can be made richer. And more commodity classification labels can be extracted from the rich commodity information. And determining commodity sub-information corresponding to each data type from the commodity information according to a preset data type, inputting each commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information in the commodity multi-classification model, and obtaining a feature vector of the corresponding commodity sub-information. And splicing the characteristic vectors of the commodity sub-information to obtain a commodity vector of the target commodity. The commodity vectors obtained based on splicing are subjected to multi-classification processing, and the commodity vectors comprise commodity sub-information corresponding to each data type, so that the multi-classification results can comprehensively consider the commodity sub-information of each data type and the relation among the commodity sub-information, and on the basis of obtaining richer commodity classification labels, the obtained commodity classification labels can be more accurate.
Optionally, the apparatus 400 is applied to an e-commerce platform, and the apparatus 400 further includes:
a second determining module 440, configured to determine a missing target tag class entity set in the tag class entity set according to the classification tag set of the target product and a tag class entity set having an association relationship with the target product in a product knowledge graph of the e-commerce platform;
an establishing module 450 configured to establish an association relationship between the target product and each tag class entity in the target tag class entity set in the product knowledge graph.
Optionally, the apparatus 400 is applied to an e-commerce platform, and the apparatus 400 further includes:
a third determining module 460, configured to determine a tag set of interest of the user based on the category tag set of each historical item searched by the user;
the recommending module 470 is configured to determine a to-be-recommended commodity corresponding to the classification tag set with the similarity greater than a preset threshold with the interest tag set, and push information of the to-be-recommended commodity to the user.
Optionally, the feature representation module corresponding to the commodity name type is configured to determine a feature vector of each character in the commodity sub-information of the commodity name type; and are
Determining a feature vector of the commodity sub-information of the commodity name type through the following calculation formula:
Figure BDA0002854680230000141
wherein e isnameA feature vector, h, characterizing the commodity sub-information of the commodity name typeiAnd k is the character number of the commodity sub-information of the commodity name type.
Optionally, the feature representation module corresponding to the description text type is configured to:
determining a feature vector of each character in the commodity sub-information describing the text type;
for each text segment, determining a feature matrix of the text segment characters corresponding to the text segment according to the feature vector of each character in the text segment;
and aiming at the feature matrix of the text segment character corresponding to each text segment, determining a weighted text segment feature vector corresponding to the feature matrix of the text segment character through the following calculation formula:
hpara＝αtoken Htoken，
wherein alpha istoken＝softmax(qtoken tanh(HtokenWtoken)T)，
Wherein h isparaCharacterizing the feature vectors of weighted text segments, HtokenFeature matrix, alpha, characterizing characters of a text segmenttokenCharacterizing a vector consisting of the weights of the characters, Wtoken、qtokenIs a training parameter of the feature representation module;
according to the weight text segment feature vector corresponding to each text segment and the weight corresponding to each text segment, determining the feature vector of the commodity sub-information describing the text type through the following calculation formula:
eparas＝αpara Hpara，
wherein alpha ispara＝softmax(qpara tanh(Hpara Wpara)T)，
Wherein e isparasFeature vectors, H, characterizing the sub-information of the goods of the description text typeparaCharacterizing a matrix composed of weighted text segment feature vectors of each text segment, alphaparaCharacterizing a vector consisting of the weights of the texts, qpara、WparaAre the training parameters of the feature representation module.
Optionally, the feature representation module corresponding to the picture type is configured to:
determining a picture characteristic vector corresponding to each picture in the sub information of the picture type; and are
Determining a feature vector of the commodity sub-information of the picture type through the following calculation formula:
eimgs＝αimgHimg；
wherein alpha isimg＝softmax(qimg tanh(Himg Wimg)T)；
Wherein e isimgsCharacteristic vector, H, of the sub-information of the goods characterizing the picture typeimgCharacterizing a matrix composed of feature vectors of pictures, alphaimgCharacterizing a vector consisting of the weights of the pictures, Wimg、qimgAre the feature representation module training parameters.
Optionally, the feature representation module corresponding to the structured attribute type is configured to determine a feature vector of the commodity sub-information of the structured attribute type through the following calculation formula:
estr＝ReLU(efeature Wstr)；
wherein e isstrFeature vectors characterizing the sub-information of the goods of said structured attribute type, efeatureA structured attribute feature vector, W, characterizing the commodity sub-information of said structured attribute typestrThe training parameters of the module are represented for the features.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
In another exemplary embodiment, a computer program product is also provided, which comprises a computer program executable by a programmable apparatus, the computer program having code portions for performing the above-mentioned article classification method when executed by the programmable apparatus.
Fig. 5 is a block diagram illustrating an electronic device 1900 according to an example embodiment. For example, the electronic device 1900 may be provided as a server. Referring to fig. 5, an electronic device 1900 includes a processor 1922, which may be one or more in number, and a memory 1932 for storing computer programs executable by the processor 1922. The computer program stored in memory 1932 may include one or more modules that each correspond to a set of instructions. Further, the processor 1922 may be configured to execute the computer program to perform the article classification method described above.
Additionally, the electronic device 1900 may also include a power component 1926 and a communication component 1950, the power component 1926 may be configured to perform power management for the electronic device 1900, and the communication component 1950 may be configured to enable communication for the electronic device 1900, e.g., wired or wireless communication. In addition, the electronic device 1900 may also include input/output (I/O) interfaces 1958. The electronic device 1900 may operate based on an operating system, such as Windows Server, stored in memory 1932TM，Mac OS XTM，UnixTM，LinuxTMAnd so on.
In another exemplary embodiment, a computer readable storage medium comprising program instructions which, when executed by a processor, implement the steps of the above-described article sorting method is also provided. For example, the computer readable storage medium may be the memory 1932 described above that includes program instructions executable by the processor 1922 of the electronic device 1900 to perform the article sorting method described above.
The preferred embodiments of the present disclosure are described in detail with reference to the accompanying drawings, however, the present disclosure is not limited to the specific details of the above embodiments, and various simple modifications may be made to the technical solution of the present disclosure within the technical idea of the present disclosure, and these simple modifications all belong to the protection scope of the present disclosure.
It should be noted that the various features described in the above embodiments may be combined in any suitable manner without departing from the scope of the invention. To avoid unnecessary repetition, the disclosure does not separately describe various possible combinations.
In addition, any combination of various embodiments of the present disclosure may be made, and the same should be considered as the disclosure of the present disclosure, as long as it does not depart from the spirit of the present disclosure.

Claims (10)

1. A method of classifying an article, the method comprising:
the method comprises the steps of obtaining commodity information of a target commodity, wherein the commodity information comprises information obtained from a plurality of data sources;
determining commodity sub-information corresponding to each data type from the commodity information according to a preset data type, wherein the preset data type comprises at least one of a commodity name type, a description text type, a picture type and a structural attribute type;
inputting all the commodity sub-information into a commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to the target commodity;
the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for:
inputting the commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information aiming at each commodity sub-information to obtain a feature vector of the commodity sub-information;
splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of the target commodity;
and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain the classification label set which is output by the classification module and corresponds to the target commodity.
2. The method of claim 1, applied to an e-commerce platform, the method further comprising:
determining a target label entity set missing in the label entity set according to the classification label set of the target commodity and a label entity set which has an incidence relation with the target commodity in a commodity knowledge graph of the e-commerce platform;
and establishing an incidence relation between the target commodity and each tag entity in the target tag entity set in the commodity knowledge graph.
3. The method of claim 1, applied to an e-commerce platform, the method further comprising:
determining an interest tag set of a user based on the classification tag set of each historical commodity searched by the user;
and determining the commodity to be recommended corresponding to the classified label set with the similarity greater than a preset threshold value with the interested label set, and pushing the information of the commodity to be recommended to the user.
4. The method according to any one of claims 1 to 3, wherein a feature representation module corresponding to the commodity name type is used for determining a feature vector of each character in the commodity sub-information of the commodity name type; and are
Determining a feature vector of the commodity sub-information of the commodity name type through the following calculation formula:
Figure FDA0002854680220000021
wherein e isnameA feature vector, h, of the commodity sub-information characterizing the commodity name typeiAnd k is the character number of the commodity sub-information of the commodity name type.
5. The method according to any of claims 1-3, wherein the feature representation module corresponding to the descriptive text type is configured to:
determining a feature vector of each character in the commodity sub-information describing the text type;
for each text segment, determining a feature matrix of the text segment characters corresponding to the text segment according to the feature vector of each character in the text segment;
and aiming at the feature matrix of the text segment character corresponding to each text segment, determining a weighted text segment feature vector corresponding to the feature matrix of the text segment character through the following calculation formula:
hpara＝αtokenHtoken，
wherein alpha istoken＝softmax(qtokentanh(HtokenWtoken)T)，
Wherein h isparaCharacterizing the feature vectors of weighted text segments, HtokenFeature matrix, alpha, characterizing characters of a text segmenttokenCharacterizing a vector consisting of the weights of the characters, Wtoken、qtokenIs a training parameter of the feature representation module;
according to the weight text segment feature vector corresponding to each text segment and the weight corresponding to each text segment, determining the feature vector of the commodity sub-information describing the text type through the following calculation formula:
eparas＝αparaHpara，
wherein alpha ispara＝softmax(qparatanh(HparaWpara)T)，
Wherein e isparasFeature vectors, H, characterizing the sub-information of the goods of the description text typeparaCharacterizing a matrix composed of weighted text segment feature vectors of text segments, alphaparaCharacterizing a vector consisting of the weights of the texts, qpara、WparaAre the training parameters of the feature representation module.
6. The method according to any of claims 1-3, wherein the feature representation module corresponding to the picture type is configured to:
determining a picture characteristic vector corresponding to each picture in the sub information of the picture type; and are
Determining a feature vector of the commodity sub-information of the picture type by the following calculation formula:
eimgs＝αimgHimg；
wherein alpha isimg＝softmax(qimgtanh(HimgWimg)T)；
Wherein e isimgsCharacteristic vector, H, of the sub-information of the goods characterizing the picture typeimgCharacterizing a matrix composed of feature vectors of pictures, alphaimgCharacterizing a vector consisting of the weights of the pictures, wimg、qimgAre the feature representation module training parameters.
7. The method according to any one of claims 1 to 3, wherein the feature representation module corresponding to the structured attribute type is configured to determine a feature vector of the commodity sub-information of the structured attribute type by the following calculation formula:
estr＝ReLU(efeatureWstr)；
wherein e isstrFeature vectors characterizing the sub-information of the goods of said structured attribute type, efeatureA structured attribute feature vector, W, characterizing the commodity sub-information of said structured attribute typestrThe training parameters of the module are represented for the features.
8. An article sorting device, said device comprising:
the system comprises an acquisition module, a display module and a display module, wherein the acquisition module is configured to be used for acquiring commodity information of a target commodity, and the commodity information comprises information acquired from a plurality of data sources;
the first determining module is configured to determine commodity sub-information corresponding to each data type from the commodity information according to preset data types, wherein the preset data types include at least one of a commodity name type, a description text type, a picture type and a structured attribute type;
the input module is configured to input all the commodity sub-information into a commodity multi-classification model to obtain a classification label set which is output by the commodity multi-classification model and corresponds to the target commodity;
the commodity multi-classification model comprises feature representation modules corresponding to different data types, and is used for:
inputting the commodity sub-information into a feature representation module corresponding to the data type of the commodity sub-information aiming at each commodity sub-information to obtain a feature vector of the commodity sub-information;
splicing the characteristic vectors of the commodity sub-information of each data type to obtain a commodity vector of the target commodity;
and inputting the commodity vector into a classification module in the commodity multi-classification model to obtain the classification label set which is output by the classification module and corresponds to the target commodity.
9. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 7.
10. An electronic device, comprising:
a memory having a computer program stored thereon;
a processor for executing the computer program in the memory to carry out the steps of the method of any one of claims 1 to 7.
CN202011540233.2A 2020-12-23 2020-12-23 Commodity classification method and device, storage medium and electronic equipment Pending CN114661895A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011540233.2A CN114661895A (en) 2020-12-23 2020-12-23 Commodity classification method and device, storage medium and electronic equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011540233.2A CN114661895A (en) 2020-12-23 2020-12-23 Commodity classification method and device, storage medium and electronic equipment

Publications (1)

Publication Number Publication Date
CN114661895A true CN114661895A (en) 2022-06-24

Family

ID=82024192

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011540233.2A Pending CN114661895A (en) 2020-12-23 2020-12-23 Commodity classification method and device, storage medium and electronic equipment

Country Status (1)

Country Link
CN (1) CN114661895A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115909358A (en) * 2022-07-27 2023-04-04 广州市玄武无线科技股份有限公司 Commodity specification identification method and device, terminal equipment and computer storage medium

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115909358A (en) * 2022-07-27 2023-04-04 广州市玄武无线科技股份有限公司 Commodity specification identification method and device, terminal equipment and computer storage medium
CN115909358B (en) * 2022-07-27 2024-02-13 广州市玄武无线科技股份有限公司 Commodity specification identification method, commodity specification identification device, terminal equipment and computer storage medium

Similar Documents

Publication Publication Date Title
US11409791B2 (en) Joint heterogeneous language-vision embeddings for video tagging and search
CN108763362B (en) Local model weighted fusion Top-N movie recommendation method based on random anchor point pair selection
CN110199276B (en) Method, system and medium for searching
CN109716327B (en) Video capturing framework for visual search platform
US20180336193A1 (en) Artificial Intelligence Based Method and Apparatus for Generating Article
CN109492229B (en) Cross-domain emotion classification method and related device
CN107833082B (en) Commodity picture recommendation method and device
CN107145536B (en) User portrait construction method and device and recommendation method and device
CN113011186B (en) Named entity recognition method, named entity recognition device, named entity recognition equipment and computer readable storage medium
CN110083729B (en) Image searching method and system
US20200210442A1 (en) Identifying and extracting addresses within content
CN102855480A (en) Method and device for recognizing characters in image
CN115659008B (en) Information pushing system, method, electronic equipment and medium for big data information feedback
CN113392236A (en) Data classification method, computer equipment and readable storage medium
CN112288042A (en) Updating method and device of behavior prediction system, storage medium and computing equipment
CN113094549A (en) Video classification method and device, electronic equipment and storage medium
CN107636646B (en) Facility anchoring with geo-located imaging
CN114418032A (en) Five-modal commodity pre-training method and retrieval system based on self-coordination contrast learning
CN112184284A (en) Picture classification method and device and electronic equipment
CN114661895A (en) Commodity classification method and device, storage medium and electronic equipment
US11403339B2 (en) Techniques for identifying color profiles for textual queries
US10402449B2 (en) Information processing system, information processing method, and information processing program
CN117635275A (en) Intelligent electronic commerce operation commodity management platform and method based on big data
CN116958512A (en) Target detection method, target detection device, computer readable medium and electronic equipment
CN112241490A (en) Classification processing and data searching method and device, electronic equipment and intelligent sound box

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination