CN113468414A - Commodity searching method and device, computer equipment and storage medium - Google Patents
Commodity searching method and device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN113468414A CN113468414A CN202110631812.6A CN202110631812A CN113468414A CN 113468414 A CN113468414 A CN 113468414A CN 202110631812 A CN202110631812 A CN 202110631812A CN 113468414 A CN113468414 A CN 113468414A
- Authority
- CN
- China
- Prior art keywords
- category
- word vector
- commodity
- categories
- vector matrix
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 52
- 239000013598 vector Substances 0.000 claims abstract description 178
- 239000011159 matrix material Substances 0.000 claims abstract description 149
- 238000011156 evaluation Methods 0.000 claims abstract description 74
- 238000007781 pre-processing Methods 0.000 claims abstract description 25
- 230000006870 function Effects 0.000 claims description 21
- 238000011176 pooling Methods 0.000 claims description 20
- 238000004422 calculation algorithm Methods 0.000 claims description 18
- 238000012545 processing Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 8
- 238000010606 normalization Methods 0.000 claims description 3
- 230000011218 segmentation Effects 0.000 description 15
- 238000004891 communication Methods 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 6
- 238000010586 diagram Methods 0.000 description 6
- 230000004913 activation Effects 0.000 description 5
- 238000004590 computer program Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 230000014509 gene expression Effects 0.000 description 3
- 230000001413 cellular effect Effects 0.000 description 2
- 238000005034 decoration Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9538—Presentation of query results
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a commodity searching method, a commodity searching device, computer equipment and a storage medium, wherein the commodity searching method comprises the following steps: receiving a query request of an e-commerce user, and extracting search keywords in the request; carrying out data preprocessing on the searched keywords and constructing a corresponding word vector matrix; inputting the word vector matrix into a category prediction model trained to be in a convergence state, and executing category prediction operation by the category prediction model to determine evaluation scores of different categories in a word vector matrix hit category structure, wherein the category structure organizes the different categories by at least two levels; and constructing a search result list, and pushing the search result list to the e-commerce users, wherein the search result list comprises a plurality of categories of commodity objects with relatively high evaluation scores. The method combines the category structure and the category prediction model, accurately and efficiently queries matched commodities for the key words from the dimension of the category, displays the commodities in order according to the category hierarchical system, and effectively improves the retrieval relevance.
Description
Technical Field
The embodiment of the invention relates to the technical field of e-commerce, in particular to a commodity searching method, a commodity searching device, computer equipment and a storage medium.
Background
At present, a user of an e-commerce service platform can search for a commodity object required by the user for display by inputting a search keyword, however, when the user inputs the search keyword, it may not be clear what keyword is used for expressing the user's needs, therefore, a search result returned according to the search keyword input by the user may not meet the actual needs of the user, and the number of commodities displayed in a commodity display list may be relatively large, on a terminal device of the user, because the display area of the device is limited, the number of commodities displayed in the current display area is also very limited, therefore, if the platform is inaccurate for the commodities inquired by the user, the user may be required to browse the searched commodities by sliding a screen, and correspondingly, the commodities displayed later may be ignored by the user because of a long user operation path, so that the goods can not get effective exposure opportunity. Therefore, in the process of providing the query result, the accuracy of the query result of the commodity directly affects the ordering condition of the commodity, and the actual transaction result of the commodity is directly affected. Therefore, how to improve the quality of the commodity query result becomes a technical problem to be solved by the technical personnel in the field.
Disclosure of Invention
The application aims to provide a commodity searching method, a commodity searching device, computer equipment and a storage medium.
In order to realize the purpose of the application, the following technical scheme is adopted:
a commodity searching method proposed in accordance with one of the objects of the present application, comprising the steps of:
receiving a query request of an e-commerce user, and extracting search keywords in the request;
carrying out data preprocessing on the search keywords and constructing a corresponding word vector matrix;
inputting the word vector matrix into a category prediction model trained to be in a convergence state, wherein the category prediction model executes category prediction operation to determine evaluation scores of different categories in a word vector matrix hit category structure, and the category structure organizes the different categories by at least two levels;
and constructing a search result list, and pushing the search result list to the e-commerce user, wherein the search result list comprises a plurality of categories of commodity objects with relatively high evaluation scores.
In a further embodiment, the step of performing data preprocessing on the search keyword to construct a corresponding word vector matrix includes:
segmenting the search keywords, and acquiring a plurality of keyword groups of the search keywords based on a grammar rule algorithm;
calculating the weight of each key phrase according to a word co-occurrence graph network algorithm;
and calling a word embedding model, and carrying out vectorization processing on a plurality of key word groups with relatively high weights to construct the word vector matrix.
In a further embodiment, the word embedding model is a Skip-garm model, i.e. a word skipping model, and the category prediction model is a TextCNN model.
In a further embodiment, the step of determining the evaluation scores of the different categories in the word vector matrix hit category structure by the category prediction model by performing a category prediction operation comprises:
converting the word vector matrix into a matrix with a preset size;
performing convolution operation on the word vector matrix conforming to the preset size and convolution kernels of different unit sizes to obtain a vector matrix corresponding to each convolution kernel, wherein the unit size is set in different lengths by taking the width of the preset size as a reference;
performing maximum pooling on the vector matrixes, obtaining pooling results corresponding to the vector matrixes, splicing, constructing pooling vectors, and performing full-connection output;
and calling a normalization index function, predicting the feature matrix output by full connection and the prediction result probabilities of the different categories, and determining the probabilities as evaluation scores of the different categories in the word vector matrix hit category structure.
In a preferred embodiment, the method comprises a pre-step of training the category prediction model, wherein the training process comprises:
acquiring massive commodity data and historical search data, and performing data preprocessing on the data to construct a plurality of corresponding word vector matrixes;
inputting the word vector matrixes into a category prediction model, and determining evaluation scores of different categories in a category structure hit by the word vector matrixes;
and performing iterative training in the same way until a plurality of categories with relatively high evaluation scores corresponding to the word vector matrixes are preset categories corresponding to the commodity data or the historical search data to which the categories belong, and representing that the category prediction model is trained to be in a convergence state.
In a further embodiment, the step of constructing a search result list and pushing the search result list to the e-commerce user includes:
judging whether the evaluation scores of all categories of the word vector matrix exceed target scores or not, and determining the categories corresponding to the evaluation scores exceeding the target scores as target categories;
obtaining commodity objects contained in the target categories respectively, and constructing the search result list according to the commodity objects;
according to the grades and evaluation scores of the categories of the commodity objects stored in the search result list, sorting the commodity objects in a descending order;
and pushing the search result list to the e-commerce user.
In a preferred embodiment, in the step of sorting the commodity objects in descending order according to the level and the evaluation score of the category to which each commodity object belongs stored in the search result list, the commodity objects are sorted within the category range to which the commodity object belongs according to commodity feature information of the commodity object, where the commodity feature information includes price information, sales information, and inventory information.
An article search device proposed in accordance with an object of the present application includes:
the request receiving module is used for receiving a query request of an e-commerce user and extracting search keywords in the request;
the data preprocessing module is used for preprocessing the data of the search keywords and constructing a corresponding word vector matrix;
an evaluation score prediction module for inputting the word vector matrix into a category prediction model trained to a convergence state, the category prediction model performing category prediction operation to determine evaluation scores of different categories in a word vector matrix hit category structure, the category structure organizing the different categories in at least two levels;
and the list construction module is used for constructing a search result list and pushing the search result list to the e-commerce user, wherein the search result list comprises a plurality of categories of commodity objects with relatively high evaluation scores.
In a further embodiment, the data preprocessing module comprises:
the keyword word cutting unit is used for segmenting the search keywords and acquiring a plurality of keyword groups of the search keywords based on a grammar rule algorithm;
the weight calculation unit is used for calculating the weight of each key phrase according to the word co-occurrence graph network algorithm;
and the vectorization processing unit is used for calling the word embedding model, carrying out vectorization processing on a plurality of key word groups with relatively high weights and constructing the word vector matrix.
In further embodiments, the assessment score prediction module comprises:
the size conversion unit is used for converting the word vector matrix into a matrix with a preset size;
the convolution operation unit is used for performing convolution operation on the word vector matrix conforming to the preset size and convolution kernels of different unit sizes to obtain a vector matrix corresponding to each convolution kernel, and the unit sizes are set in different lengths by taking the width of the preset size as a reference;
the maximum pooling unit is used for performing maximum pooling on the vector matrixes, obtaining pooling results corresponding to the vector matrixes for splicing, and constructing pooled vectors for full-connection output;
and the probability prediction unit is used for calling the normalized index function, predicting the probability of the feature matrix output by full connection and the prediction result of the different classes of objects, and determining the probabilities as the evaluation scores of the different classes of the word vector matrix hit in the class structure.
In a further embodiment, the list construction module comprises:
the target category determining unit is used for judging whether the evaluation scores of all categories of the word vector matrix exceed the target scores or not and determining the categories corresponding to the evaluation scores exceeding the target scores as the target categories;
the list construction unit is used for acquiring the commodity objects contained in the target categories and constructing the search result list according to the commodity objects;
the commodity object sorting unit is used for sorting the commodity objects in a descending order according to the grades and the evaluation scores of the categories of the commodity objects stored in the search result list;
and the list pushing unit is used for pushing the search result list to the e-commerce user.
In order to solve the above technical problem, an embodiment of the present invention further provides a computer device, including a memory and a processor, where the memory stores computer-readable instructions, and the computer-readable instructions, when executed by the processor, cause the processor to execute the steps of the product search method.
In order to solve the above technical problem, an embodiment of the present invention further provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the article search method.
The embodiment of the invention has the beneficial effects that:
the system predicts evaluation scores of all categories for search keywords, determines corresponding categories for the keywords according to the evaluation scores, and constructs a search result list for the category commodity objects to be pushed.
The method comprises the steps of firstly receiving a query request of an E-commerce user, extracting search keywords in the request, carrying out data preprocessing on the search keywords, constructing a corresponding word vector matrix, secondly inputting the word vector matrix into a category prediction model trained to be in a convergence state, so that the category prediction model executes category prediction operation to determine evaluation scores of different categories in a category structure hit by the word vector matrix, and finally constructing a plurality of commodity objects of categories with relatively high evaluation scores as a search result list to be pushed to the E-commerce user as a query result.
Therefore, the method and the device have the advantages that data preprocessing is carried out on the search keywords, the search keywords are split to obtain the plurality of keyword groups, the weights of the keyword groups are calculated according to the word co-occurrence graph network configured in advance, the word embedding model is called to construct the keyword groups with higher weights into the word vector matrix, compared with the traditional mode that the word vector matrix is constructed only by word segmentation, the characteristics of the search keywords can be analyzed and determined for the neural network model (category prediction model) in advance, and the accuracy of commodity query is effectively improved.
Secondly, the commodity object is inquired for the search keyword from the dimension of the category level and output, the evaluation scores of different categories hit by the search keyword are predicted by combining the category structure and the category prediction model, the commodity object with the category higher in score is determined to be the commodity object higher in association with the search keyword and output, the commodity inquiry accuracy is greatly improved, and the inquiry experience of a user is effectively improved.
In addition, when the search result list is constructed, the commodity objects with higher category levels, to which the commodity objects belong, in the list are sorted to the earlier positions, so that when the list is output and displayed at a client, the commodity object most consistent with the query commodity represented in the search keyword can be output to the earlier position, a user can receive the needed commodity object for consumption at the fastest speed, and the commodity query experience of the user is improved, and the revenues of the platform are effectively improved.
Drawings
The foregoing and/or additional aspects and advantages of the present application will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
fig. 1 is a schematic diagram of a typical network deployment architecture related to implementing the technical solution of the present application;
FIG. 2 is a schematic flow chart diagram of an exemplary embodiment of a merchandise search method of the present application;
FIG. 3 is a schematic flow chart illustrating a specific step of step S12 in FIG. 2;
FIG. 4 is a schematic flowchart illustrating a specific step of step S13 in FIG. 2;
FIG. 5 is a schematic flowchart illustrating a specific step of step S14 in FIG. 4;
FIG. 6 is a schematic flow chart illustrating the pre-steps added in the method for searching for merchandise according to an embodiment of the present application;
FIG. 7 is a functional block diagram of an exemplary embodiment of an article search device of the present application;
fig. 8 is a block diagram of a basic structure of a computer device according to an embodiment of the present application.
Detailed Description
Reference will now be made in detail to embodiments of the present application, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to the same or similar elements or elements having the same or similar function throughout. The embodiments described below with reference to the drawings are exemplary only for the purpose of explaining the present application and are not to be construed as limiting the present application.
As used herein, the singular forms "a", "an", "the" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and/or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. It will be understood that when an element is referred to as being "connected" or "coupled" to another element, it can be directly connected or coupled to the other element or intervening elements may also be present. Further, "connected" or "coupled" as used herein may include wirelessly connected or wirelessly coupled. As used herein, the term "and/or" includes all or any element and all combinations of one or more of the associated listed items.
It will be understood by those within the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
As will be appreciated by those skilled in the art, "client," "terminal," and "terminal device" as used herein include both devices that are wireless signal receivers, which are devices having only wireless signal receivers without transmit capability, and devices that are receive and transmit hardware, which have receive and transmit hardware capable of two-way communication over a two-way communication link. Such a device may include: cellular or other communication devices such as personal computers, tablets, etc. having single or multi-line displays or cellular or other communication devices without multi-line displays; PCS (Personal Communications Service), which may combine voice, data processing, facsimile and/or data communication capabilities; a PDA (Personal Digital Assistant), which may include a radio frequency receiver, a pager, internet/intranet access, a web browser, a notepad, a calendar and/or a GPS (Global Positioning System) receiver; a conventional laptop and/or palmtop computer or other device having and/or including a radio frequency receiver. As used herein, a "client," "terminal device" can be portable, transportable, installed in a vehicle (aeronautical, maritime, and/or land-based), or situated and/or configured to operate locally and/or in a distributed fashion at any other location(s) on earth and/or in space. The "client", "terminal Device" used herein may also be a communication terminal, a web terminal, a music/video playing terminal, such as a PDA, an MID (Mobile Internet Device) and/or a Mobile phone with music/video playing function, and may also be a smart tv, a set-top box, and the like.
The hardware referred to by the names "server", "client", "service node", etc. is essentially an electronic device with the performance of a personal computer, and is a hardware device having necessary components disclosed by the von neumann principle such as a central processing unit (including an arithmetic unit and a controller), a memory, an input device, an output device, etc., a computer program is stored in the memory, and the central processing unit calls a program stored in an external memory into the internal memory to run, executes instructions in the program, and interacts with the input and output devices, thereby completing a specific function.
It should be noted that the concept of "server" as referred to in this application can be extended to the case of a server cluster. According to the network deployment principle understood by those skilled in the art, the servers should be logically divided, and in physical space, the servers may be independent from each other but can be called through an interface, or may be integrated into one physical computer or a set of computer clusters. Those skilled in the art will appreciate this variation and should not be so limited as to restrict the implementation of the network deployment of the present application.
Referring to fig. 1, the hardware basis required for implementing the related art embodiments of the present application may be deployed according to the architecture shown in the figure. The server 80 is deployed at the cloud end, and serves as a business server, and is responsible for further connecting to a related data server and other servers providing related support, so as to form a logically associated server cluster to provide services for related terminal devices, such as a smart phone 81 and a personal computer 82 shown in the figure, or a third-party server (not shown in the figure). Both the smart phone and the personal computer can access the internet through a known network access mode, and establish a data communication link with the cloud server 80 so as to run a terminal application program related to the service provided by the server.
For the server, the application program is usually constructed as a service process, and a corresponding program interface is opened for remote call of the application program running on various terminal devices.
The application program refers to an application program running on a server or a terminal device, the application program implements the related technical scheme of the application in a programming mode, a program code of the application program can be saved in a nonvolatile storage medium which can be identified by a computer in a form of a computer executable instruction, and is called into a memory by a central processing unit to run, and the related device of the application is constructed by running the application program on the computer.
For the server, the application program is usually constructed as a service process, and a corresponding program interface is opened for remote call of the application program running on various terminal devices.
For various terminal devices which are popular at present, particularly for mobile devices such as tablets and mobile phones, camera devices such as a camera are usually built in, or a personal computer can be externally connected to the camera devices.
The solution in the present application, which is suitable for implementation in a terminal device, may also be programmed to be built into an application providing a merchant transaction, as a part of which the functionality is extended. The e-commerce transaction refers to an online transaction service displayed based on the network deployment structure.
The person skilled in the art will know this: although the various methods of the present application are described based on the same concept so as to be common to each other, they may be independently performed unless otherwise specified. In the same way, for each embodiment disclosed in the present application, it is proposed based on the same inventive concept, and therefore, concepts of the same expression and concepts of which expressions are different but are appropriately changed only for convenience should be equally understood.
Referring to fig. 2, a method for searching for a commodity according to the present application, in an exemplary embodiment, includes the following steps:
step S11, receiving a query request from an e-commerce user, extracting search keywords in the request:
and the server receives the query request pushed by the e-commerce user and extracts the search keyword contained in the query request.
The inquiry request is pushed to the server by a client of the e-commerce user, the e-commerce user inputs the search keyword through a commodity search control in a graphical user interface, the client is triggered to generate the inquiry request according to the search keyword, and the inquiry request is pushed to the server, so that the server extracts the search keyword according to the inquiry request.
The E-commerce user is a user registered in an E-commerce service platform, the E-commerce service platform establishes a data communication link with the server, and pushes the query request to the server through a corresponding interface, so that the server receives the query request, executes corresponding category prediction processing, and constructs a corresponding search result list for feedback. For the implementation of the category prediction processing and the construction of the search result list, please refer to the following steps, which are not repeated herein.
Step S12, data preprocessing is carried out on the search keywords, and a corresponding word vector matrix is constructed:
and the server carries out data preprocessing on the search keyword to construct the word vector matrix corresponding to the search keyword.
The server calls a specific word segmentation device to perform word segmentation processing on the search keyword so as to obtain a plurality of keyword groups from the search keyword, calculates the weight of each keyword group according to a word co-occurrence graph network algorithm, performs vectorization processing on the plurality of keyword groups with relatively high weights in a calculation result by calling a word embedding model, and constructs a word vector matrix so as to complete data preprocessing on the search keyword and facilitate category prediction on the search keyword.
Regarding the selection of the word segmentation device, when the search keyword is a chinese text, selecting an LTP word segmentation device, a THULAC word segmentation device, a jieba word segmentation device, or a KCWS word segmentation device facing the chinese field to segment the search keyword so as to preliminarily obtain all keyword groups contained in the search keyword, and when the search keyword is an english text, obtaining the contained keyword groups by removing spaces and non-english special characters, or selecting a corresponding word segmentation device facing the english field, for example, a large model of space as the word segmentation device. The technical personnel in the field can select the existing word segmentation device for word segmentation according to the actual service scene, which is not repeated.
The word vector matrix is a matrix of text vectors for representing a plurality of keywords with higher weights in the search keywords, the highest keyword in the weights of the keywords after a word co-occurrence graph network algorithm is determined as a central keyword, the keywords are fed into the word embedding model, the word vector matrix is constructed so as to be fed into the category prediction model in the following process, category prediction operation is executed to determine evaluation scores of different categories in a category structure hit by the vector matrix, and as for a specific implementation mode of the category prediction operation, reference is made to the following steps, which are not repeated.
Referring to fig. 3, a specific implementation of the server performing data preprocessing on the search keyword to construct the word vector matrix includes the following specific implementation steps:
step S121, segmenting the search keyword, and acquiring a plurality of keyword groups of the search keyword based on a grammar rule algorithm:
and the server calls a specific word segmentation device to segment the search keywords, and acquires a plurality of keyword groups in the search keywords based on a grammar rule algorithm of the word segmentation device.
Specifically, when the search keyword is "Apple phone charger", the server calls the specific word segmenter, and after executing word segmentation processing based on a grammar rule algorithm of the word segmenter, the obtained keyword group is: [ 'Apple', 'phone', 'charger' ].
Step S122, calculating the weight of each key phrase according to the word co-occurrence graph network algorithm:
and after the server acquires the key phrases, calculating the weights corresponding to the key phrases according to the word co-occurrence graph network algorithm.
The word co-occurrence graph network algorithm is configured to calculate a weight of each keyword group, where the word co-occurrence graph network calculates a word frequency of a co-occurrence of each keyword pair or keyword phrase related to e-commerce shopping in advance, and the word co-occurrence graph network algorithm calculates the weight of each keyword group according to the word frequency calculated in the word co-occurrence graph network, for example, when each keyword group is: and if the co-occurrence frequency of the key phrases in the network is the highest, the weight of the key phrase of the 'Apple' is higher than that of other key phrases, and by analogy, the weight of each keyword is calculated by using the co-occurrence frequency of the key phrases characterized in the word co-occurrence network.
In one embodiment, if there is only one keyword group, the server queries another keyword group that appears together with the keyword group in the word co-occurrence graph network, determines that the frequency of the co-occurrence of the keyword group and the keyword group is higher as the keyword group of the keyword group, and commonly calculates the weights of the keyword group and the original keyword group, through the implementation of the embodiment, the corresponding category can be predicted more accurately by using one keyword group as the search keyword to which the keyword group belongs, for example, when the keyword group is 'Apple', the keyword group can be characterized by a mobile phone brand or a clothing pattern, etc., when the frequency of the co-occurrence graph network for querying the keyword group is higher and determined as the keyword group that is mostly related to the mobile phone, the meaning characterized by the keyword group as 'Apple' can be determined as the mobile phone brand, and calculating the weights of the key phrases obtained by the queries, and predicting corresponding categories for the key phrases.
Step S123, calling a word embedding model, carrying out vectorization processing on a plurality of key word groups with relatively high weights, and constructing the word vector matrix:
and after the weights of the key word groups are calculated through the word co-occurrence network algorithm, calling the word embedding model, feeding the key word group with higher weight in the key word groups into the word embedding model for vectorization, and constructing the key word groups subjected to vectorization into the word vector matrix.
In one embodiment, the server determines whether the weight of each keyword group is greater than or equal to a preset target weight value, and feeds the keyword groups greater than or equal to the target weight value into the word embedding model for vectorization, so as to construct the word vector matrix.
The word embedding model performs vectorization processing on the keyword groups to obtain vector texts corresponding to the keyword groups, for example, when the keyword groups are [ 'Apple', 'phone', 'charge', ] respectively, the vector text of 'Apple' is [1,0,0], 'phone' is [0,1,0], 'charge' is [0,0,1], after the vectorization processing is completed, the word embedding model maps the text vectors of the keyword groups into a training vector matrix respectively to perform matrix operation to obtain the text vectors corresponding to the text vectors and having the same dimension as the training vector matrix, and constructs the text vectors as the word vector matrix, for example, when the number of the keyword groups is N and the dimension of the training vector matrix is M, the dimension of the text vectors obtained by the matrix operation on the keyword groups is M, the matrix size of the word vector matrix formed by the text vectors is N x M.
The word embedding model is a model trained to a convergence state, and is trained by feeding in massive phrases related to E-commerce shopping, vectorizing the phrases and calculating the similarity of the phrases, constructing a training vector matrix with a certain dimension, so that the training vector matrix carries out matrix operation on text vectors of the key phrases according to the training vector matrix, and forming the word vector matrix with dense vectors, so that word vectors in the word vector matrix are more accurate.
In one embodiment, the word embedding model is a Skip-garm model, that is, a word skipping model, and in the training process of the model, a word group with the largest weight can be used as a central word group according to the weight of a fed-in word group, and multiple times of peripheral word prediction is performed through the central word group, so that the word group coverage rate represented by a text vector in the training vector matrix of the training structure is higher, and the word vector expression constructed by the word embedding model for the search keyword can be effectively improved to be more accurate.
Step S13, inputting the word vector matrix into a category prediction model trained to a convergent state, the category prediction model performing category prediction operations to determine evaluation scores of different categories in a word vector matrix hit category structure, the category structure organizing the different categories in at least two levels:
and the server inputs the word vector matrix into a category prediction model trained to be in a convergence state so that the category prediction model executes the category prediction operation to determine the evaluation scores of different categories in the category structure hit by the word vector matrix.
The category structure comprises categories used for representing categories of commodity objects in the corresponding e-commerce service platform, the categories contained in the category structure are organized in at least two levels, the categories of each level have a contained relation, for example, the categories are first-level categories of electronic products, the contained second-level categories can be generally divided into mobile phones, mobile phone accessories, computer complete machines, computer components, peripheral products and the like, the categories are first-level categories of clothes, and the contained second-level categories can be generally divided into coats, lower coats, undergarments and the like. Each hierarchical category constructs the category structure according to the respective inclusion and contained relationships.
The category prediction model is used for predicting the evaluation scores of different categories in the category structure hit by the input word vector matrix, converting the word vector matrix into a matrix with a preset size by the category prediction model, carrying out convolution operation on the word vector matrix with the preset size and a convolution kernel without unit size to obtain vector matrixes corresponding to the convolution kernels, carrying out maximum pooling on the vector matrixes, obtaining pooled results corresponding to the vector matrixes, splicing, constructing pooled vectors, carrying out full-connection output, calling a normalized exponential function, predicting the probability of a feature matrix of the full-connection output and the prediction results of the different categories, wherein the sum of the probabilities is 1, and determining the probabilities as the evaluation scores of the different categories in the category structure hit by the word vector matrix.
In one embodiment, the category prediction model is a TextCNN model, that is, a text convolutional neural network, which applies the convolutional neural network CNN to a text classification task, extracts key information in a text vector by using a plurality of convolutional kernels with different sizes, and accurately captures local correlation of a text.
Referring to fig. 4, an embodiment of performing a category prediction operation on the category prediction model to determine the evaluation scores of different categories in the category structure hit by the word vector matrix includes the following steps:
step S131, converting the word vector matrix into a matrix with a preset size:
the server converts the word vector matrix into a matrix with a preset size, so that when the matrix is input into the category prediction model to perform category prediction operation, the word vector matrix can accord with the matrix size of convolution kernels with different unit sizes to perform convolution operation.
The server judges whether the word vector matrix conforms to the preset size, the prediction size is generally used for limiting the length of the word vector matrix so as to limit the number of key phrases represented in the word vector matrix, if the size of the word vector matrix is larger than the preset size, the server cuts the length of the word vector matrix to the size conforming to the length represented by the prediction size, and if the size of the word vector matrix is smaller than the preset size, the server fills the length of the word vector matrix with 0 until the length of the word vector matrix conforms to the length represented by the prediction size.
Step S132, performing convolution operation on the word vector matrix conforming to the preset size and convolution kernels of different unit sizes to obtain a vector matrix corresponding to each convolution kernel, wherein the unit size is set in different lengths by taking the width of the preset size as a reference:
and after the server finishes the size conversion of the word vector matrix, inputting the word vector matrix into the category prediction model for category prediction operation, and performing convolution operation on the word vector matrix and the convolution kernels with different unit sizes by the category prediction model to obtain the vector matrix corresponding to each convolution kernel.
The unit size of each convolution kernel performing convolution operation in the category prediction model is the same as the width of the preset size, but the length is different from the preset size, and the length is generally smaller than the word vector matrix, so that the category prediction model performs convolution operation on the word vector matrix locally, for example, when the length of the word vector matrix is L, and when a convolution kernel having the length of 1 of the unit size performs convolution operation on the word vector matrix, the matrix size of the vector matrix obtained by performing convolution operation through the convolution kernel is L1, and when a convolution kernel having the length of 2 of the unit size performs convolution operation on the word vector matrix, the matrix size of the vector matrix obtained by performing convolution operation through the convolution kernel is (L-1) × 1. Those skilled in the art can design the length of the convolution kernel according to actual service requirements under the condition that the width conforms to the preset size, which is not repeated.
The category prediction model performs convolution operation on the word vector matrix and a plurality of convolution kernels with unit sizes, the convolution kernels with the same unit sizes generally have a plurality of convolution kernels, and because the category prediction model is trained to be in a convergence state, vectors in the convolution kernels are trained to have the capability of inquiring the characteristics represented by the word vector matrix, and when the category prediction model performs the convolution operation, the step number is generally set to 1, so that the obtained vector matrices can be used for representing the characteristics of the word vector matrix.
Step S133, performing maximal pooling on the vector matrices, obtaining pooling results corresponding to the vector matrices for splicing, and constructing pooled vectors for full-connection output:
the category prediction model obtains the vector matrixes corresponding to the convolution kernels through convolution operation, then performs maximum pooling on the vector matrixes to obtain pooled results corresponding to the vector matrixes, the matrix size of each vector matrix in the pooled results is generally 1 x 1, the category prediction model splices the vector matrixes which complete the maximum pooling, for example, when the convolution kernels owned by the category prediction model are N, the number of the vector matrixes in the pooled results is N, and after the vector matrixes are spliced, the matrix size of the pooled vector is N1.
After the category prediction model obtains the pooling vector, it calls an activation function to perform full-connection output on the pooling vector, and outputs a feature matrix corresponding to the pooling vector, where the feature matrix is suitable for subsequent calculation of normalized index functions, the activation function generally uses a ReLU activation function, and a person skilled in the art can select other activation functions such as Leaky ReLU, Sigmoid, Tanh, etc. according to actual application scenarios, which will not be repeated.
Step S134, calling a normalization index function, predicting the probabilities of the feature matrix output by full connection and the prediction results of the different classes of objects, and determining the probabilities as evaluation scores of the different classes of the word vector matrix hit in the class structure:
and the category prediction model calls the normalized index function to predict the probability of the feature matrix output by the full connection and the prediction results of different categories, the sum of the probabilities of the prediction results is 1, and the category prediction model determines the probabilities of the prediction results as the evaluation scores of all categories hit in the category structure by the word vector matrix.
Specifically, the normalized index function is generally referred to as Softmax activation function, the category prediction model inputs each element in the feature matrix to the normalized index function for calculation, the number of elements in the feature matrix is equal to the number of all categories included in the category structure, and the positions of the elements in the feature matrix correspond to the positions of the categories in the category structure one by one, so that after the elements are calculated by the normalized index function, the category prediction model uses the prediction result probabilities as the evaluation scores of the categories corresponding to the positions in the category structure according to the positions of the prediction result probabilities in the feature matrix.
Step S14, constructing a search result list, and pushing the search result list to the e-commerce user, where the search result list includes commodity objects of several categories with relatively high evaluation scores:
and the server constructs the search result list according to the evaluation scores of different categories in the word vector matrix hit category structure, and pushes the search result list to the E-commerce user to which the query request belongs, so that the E-commerce user outputs and displays the commodity objects to the commodity display control corresponding to the page in sequence for display according to the sequence of the commodity objects in the search result list.
And the server determines a category with a relatively higher evaluation score as a category contained in the search result list according to the evaluation score, acquires the commodity objects to which the categories corresponding to the evaluation scores belong, and stores the commodity objects in the search result list.
The commodity object is a commodity object under the E-commerce service platform flag of the E-commerce user, and after the client of the E-commerce user acquires the search result list, the commodity feature information can be visually output according to commodity feature information of each commodity object included in the search result list, such as a commodity picture, a commodity name, a commodity price, a commodity sales volume and the like, so that the commodity feature information can be displayed in a graphical user interface.
Referring to fig. 5, when the server constructs the search result list, the server sorts the commodity objects included in the search result list, and the specific implementation steps are as follows:
step S141, judging whether the evaluation score of each category of the word vector matrix exceeds a target score, and determining the category corresponding to the evaluation score exceeding the target score as a target category:
the server judges whether the evaluation score of the word vector matrix exceeds a target score or not, and determines the category corresponding to the evaluation score exceeding the target score as a target category.
The target score is generally set to be in the range of 0.6-0.8, so that the server is ensured to determine that the relevance between the commodity object contained in the target category and the search keyword is high, and the commodity object with low relevance to the search keyword in the search result list is prevented from appearing.
Step S142, obtaining the commodity objects included in the target categories, and constructing the search result list according to the commodity objects:
and the server acquires the commodity objects contained in the target categories through a data communication link with the e-commerce service platform, and constructs the search result list according to the commodity objects.
Step S143, according to the grades and evaluation scores of the categories to which the commodity objects belong stored in the search result list, sorting the commodity objects in descending order:
the server sorts the commodity objects with higher levels and grades to the front positions in the search result list according to the levels and evaluation scores of the categories to which the commodity objects respectively belong, wherein the categories are stored in the search result list, so that when the search result list is output to the graphic user interface of the E-commerce user client, the search result list has a synonymous relationship with search keywords in a query instruction pushed by the E-commerce user client on the semantic level, and the commodity query experience of the E-commerce user in the E-commerce service platform is optimized.
In one embodiment, the server sorts the commodity objects within the category range to which the commodity objects belong according to commodity feature information of the commodity objects, specifically, the commodity feature information includes price information, sales information, inventory information, and the like, and the server may sort the commodity objects in an ascending order according to the price information of the commodity objects for a certain category of commodity objects, or sort the commodity objects in a descending order according to commodity feature information such as sales information and inventory information, or execute a comprehensive sorting strategy in combination with the commodity feature information to sort the commodity objects.
Step S144, pushing the search result list to the e-commerce user:
and the server pushes the search result list with the commodity object sequencing completed to the client of the E-commerce user, so that the client can output the commodity objects to a graphical user interface for display in sequence according to the sequencing of the commodity objects in the search result list.
The above exemplary embodiments and variations thereof fully disclose the embodiments of the merchandise search method of the present application, but various variations thereof can be deduced by transforming and amplifying some technical means, and other embodiments are briefly described as follows:
in one embodiment, referring to fig. 6, the method further includes a pre-step of training the category prediction model, wherein the training process is as follows:
step S08, acquiring massive commodity data and historical search data, preprocessing the data, and constructing a plurality of corresponding word vector matrixes:
and the server acquires massive commodity data and historical search data from the e-commerce service platform establishing a data communication link with the e-commerce service platform, and carries out data preprocessing on the commodity data and the historical search data in sequence to construct word vector matrixes corresponding to the data.
Regarding the implementation of the data preprocessing, please refer to the related embodiment in step S12, which is not repeated herein.
Step S09, inputting the word vector matrices into a category prediction model, and determining evaluation scores of different categories in the word vector matrix hit category structure:
and sequentially inputting a plurality of word vector matrixes into the category prediction model by the server so that the category prediction model executes the category prediction operation to determine the evaluation scores corresponding to the category prediction models respectively.
Regarding the implementation manner of the category prediction model for performing the category prediction operation, please refer to the related embodiment in step S13, which is not repeated herein.
Step S10, performing iterative training in this way until a plurality of categories with relatively high evaluation scores corresponding to each word vector matrix are preset categories corresponding to the commodity data or the historical search data to which the categories belong, and then characterizing that the category prediction model has been trained to a convergence state:
the server obtains the evaluation score corresponding to each word vector matrix through the category prediction model, judges whether the category corresponding to the higher evaluation score in the plurality of evaluation scores corresponding to each word vector matrix is the corresponding preset category according to the databases of the plurality of preset categories corresponding to the stored commodity data and the historical search data, and continuously inputs the word vector matrix which is not the preset category into the category prediction model for prediction until the categories corresponding to the higher evaluation scores of all the word vector matrices predicted by the category prediction model are the plurality of preset categories, so that the category prediction model is characterized to be trained to a convergence state and can be put into the commodity search service of the e-commerce service platform.
Further, a product search device of the present application can be constructed by functionalizing the steps in the methods disclosed in the above embodiments, and according to this idea, please refer to fig. 8, in an exemplary embodiment of the product search method, the device includes: the system comprises a request receiving module 11, a data preprocessing module 12, an evaluation score predicting module 13 and a list constructing module 14, wherein the request receiving module 11 is used for receiving a query request of an e-commerce user and extracting search keywords in the request; the data preprocessing module 12 is configured to perform data preprocessing on the search keywords and construct corresponding word vector matrices; an evaluation score prediction module 13, configured to input the word vector matrix into a category prediction model trained to a convergence state, where the category prediction model performs category prediction operations to determine evaluation scores of different categories in a category structure hit by the word vector matrix, and the category structure organizes the different categories by at least two levels; and the list construction module 14 is configured to construct a search result list, and push the search result list to the e-commerce user, where the search result list includes a plurality of categories of commodity objects with relatively high evaluation scores.
In one embodiment, the data pre-processing module 12 comprises: the keyword word cutting unit is used for segmenting the search keywords and acquiring a plurality of keyword groups of the search keywords based on a grammar rule algorithm; the weight calculation unit is used for calculating the weight of each key phrase according to the word co-occurrence graph network algorithm; and the vectorization processing unit is used for calling the word embedding model, carrying out vectorization processing on a plurality of key word groups with relatively high weights and constructing the word vector matrix.
In one embodiment, the assessment score prediction module 13 comprises: the size conversion unit is used for converting the word vector matrix into a matrix with a preset size; the convolution operation unit is used for performing convolution operation on the word vector matrix conforming to the preset size and convolution kernels of different unit sizes to obtain a vector matrix corresponding to each convolution kernel, and the unit sizes are set in different lengths by taking the width of the preset size as a reference; the maximum pooling unit is used for performing maximum pooling on the vector matrixes, obtaining pooling results corresponding to the vector matrixes for splicing, and constructing pooled vectors for full-connection output; and the probability prediction unit is used for calling the normalized index function, predicting the probability of the feature matrix output by full connection and the prediction result of the different classes of objects, and determining the probabilities as the evaluation scores of the different classes of the word vector matrix hit in the class structure.
In one embodiment, the list construction module 14 includes: the target category determining unit is used for judging whether the evaluation scores of all categories of the word vector matrix exceed the target scores or not and determining the categories corresponding to the evaluation scores exceeding the target scores as the target categories; the list construction unit is used for acquiring the commodity objects contained in the target categories and constructing the search result list according to the commodity objects; the commodity object sorting unit is used for sorting the commodity objects in a descending order according to the grades and the evaluation scores of the categories of the commodity objects stored in the search result list; and the list pushing unit is used for pushing the search result list to the e-commerce user.
In order to solve the above technical problem, an embodiment of the present application further provides a computer device, configured to run a computer program implemented according to the product search method. Referring to fig. 8, fig. 8 is a block diagram of a basic structure of a computer device according to the present embodiment.
As shown in fig. 8, the internal structure of the computer device is schematically illustrated. The computer device includes a processor, a non-volatile storage medium, a memory, and a network interface connected by a system bus. The non-volatile storage medium of the computer device stores an operating system, a database and computer readable instructions, the database can store control information sequences, and the computer readable instructions can enable the processor to realize a commodity searching method when being executed by the processor. The processor of the computer device is used for providing calculation and control capability and supporting the operation of the whole computer device. The memory of the computer device may have stored therein computer readable instructions that, when executed by the processor, may cause the processor to perform a method of article searching. The network interface of the computer device is used for connecting and communicating with the terminal. Those skilled in the art will appreciate that the architecture shown in fig. 8 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In this embodiment, the processor is configured to execute specific functions of each module/sub-module in the product search device of the present invention, and the memory stores program codes and various types of data required for executing the modules. The network interface is used for data transmission to and from a user terminal or a server. The memory in this embodiment stores program codes and data necessary for executing all modules/submodules in the product search/output device, and the server can call the program codes and data of the server to execute the functions of all the submodules.
The present application also provides a non-volatile storage medium, wherein the product search method is written as a computer program and stored in the storage medium in the form of computer readable instructions, which when executed by one or more processors, means the execution of the program in a computer, thereby causing the one or more processors to execute the steps of the product search method according to any of the above embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and can include the processes of the embodiments of the methods described above when the computer program is executed. The storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a Read-Only Memory (ROM), or a Random Access Memory (RAM).
In summary, the present application provides a novel commodity search system through technical improvement, the system predicts evaluation scores of all categories for search keywords, determines corresponding categories for the keywords according to the levels of the evaluation scores, constructs a search result list for the category commodity objects, and pushes the search result list.
It should be understood that, although the steps in the flowcharts of the figures are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and may be performed in other orders unless explicitly stated herein. Moreover, at least a portion of the steps in the flow chart of the figure may include multiple sub-steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, which are not necessarily performed in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
Those of skill in the art will appreciate that the various operations, methods, steps in the processes, acts, or solutions discussed in this application can be interchanged, modified, combined, or eliminated. Further, other steps, measures, or schemes in various operations, methods, or flows that have been discussed in this application can be alternated, altered, rearranged, broken down, combined, or deleted. Further, steps, measures, schemes in the prior art having various operations, methods, procedures disclosed in the present application may also be alternated, modified, rearranged, decomposed, combined, or deleted.
The foregoing is only a partial embodiment of the present application, and it should be noted that, for those skilled in the art, several modifications and decorations can be made without departing from the principle of the present application, and these modifications and decorations should also be regarded as the protection scope of the present application.
Claims (10)
1. A commodity searching method is characterized by comprising the following steps:
receiving a query request of an e-commerce user, and extracting search keywords in the request;
carrying out data preprocessing on the search keywords and constructing a corresponding word vector matrix;
inputting the word vector matrix into a category prediction model trained to be in a convergence state, wherein the category prediction model executes category prediction operation to determine evaluation scores of different categories in a word vector matrix hit category structure, and the category structure organizes the different categories by at least two levels;
and constructing a search result list, and pushing the search result list to the e-commerce user, wherein the search result list comprises a plurality of categories of commodity objects with relatively high evaluation scores.
2. The method of claim 1, wherein the step of performing data preprocessing on the search keywords to construct a corresponding word vector matrix comprises:
segmenting the search keywords, and acquiring a plurality of keyword groups of the search keywords based on a grammar rule algorithm;
calculating the weight of each key phrase according to a word co-occurrence graph network algorithm;
and calling a word embedding model, and carrying out vectorization processing on a plurality of key word groups with relatively high weights to construct the word vector matrix.
3. The method of claim 2, wherein the word embedding model is a Skip-garm model, i.e., a word skipping model, and the category prediction model is a TextCNN model.
4. The method of claim 1, wherein the step of the category prediction model performing a category prediction operation to determine the evaluation scores of the word vector matrix for different categories in the category structure comprises:
converting the word vector matrix into a matrix with a preset size;
performing convolution operation on the word vector matrix conforming to the preset size and convolution kernels of different unit sizes to obtain a vector matrix corresponding to each convolution kernel, wherein the unit size is set in different lengths by taking the width of the preset size as a reference;
performing maximum pooling on the vector matrixes, obtaining pooling results corresponding to the vector matrixes, splicing, constructing pooling vectors, and performing full-connection output;
and calling a normalization index function, predicting the feature matrix output by full connection and the prediction result probabilities of the different categories, and determining the probabilities as evaluation scores of the different categories in the word vector matrix hit category structure.
5. The method according to any one of claims 1 to 5, comprising a preceding step of training the category prediction model, wherein the training process comprises:
acquiring massive commodity data and historical search data, and performing data preprocessing on the data to construct a plurality of corresponding word vector matrixes;
inputting the word vector matrixes into a category prediction model, and determining evaluation scores of different categories in a category structure hit by the word vector matrixes;
and performing iterative training in the same way until a plurality of categories with relatively high evaluation scores corresponding to the word vector matrixes are preset categories corresponding to the commodity data or the historical search data to which the categories belong, and representing that the category prediction model is trained to be in a convergence state.
6. The method of any one of claims 1 to 5, wherein the step of constructing a search result list and pushing the search result list to the e-commerce user comprises:
judging whether the evaluation scores of all categories of the word vector matrix exceed target scores or not, and determining the categories corresponding to the evaluation scores exceeding the target scores as target categories;
obtaining commodity objects contained in the target categories respectively, and constructing the search result list according to the commodity objects;
according to the grades and evaluation scores of the categories of the commodity objects stored in the search result list, sorting the commodity objects in a descending order;
and pushing the search result list to the e-commerce user.
7. The method according to claim 6, wherein in the step of sorting the commodity objects in descending order according to the rank of the category to which each commodity object belongs and the evaluation score stored in the search result list, the commodity objects are sorted within the category to which the commodity objects belong according to commodity feature information of the commodity objects, the commodity feature information including price information, sales information, and inventory information.
8. An article search device, comprising:
the request receiving module is used for receiving a query request of an e-commerce user and extracting search keywords in the request;
the data preprocessing module is used for preprocessing the data of the search keywords and constructing a corresponding word vector matrix;
an evaluation score prediction module for inputting the word vector matrix into a category prediction model trained to a convergence state, the category prediction model performing category prediction operation to determine evaluation scores of different categories in a word vector matrix hit category structure, the category structure organizing the different categories in at least two levels;
and the list construction module is used for constructing a search result list and pushing the search result list to the e-commerce user, wherein the search result list comprises a plurality of categories of commodity objects with relatively high evaluation scores.
9. A computer device comprising a memory and a processor, the memory having stored therein computer-readable instructions which, when executed by the processor, cause the processor to carry out the steps of the item searching method according to any one of claims 1 to 7.
10. A storage medium having computer-readable instructions stored thereon, which, when executed by one or more processors, cause the one or more processors to perform the steps of the item search method of any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110631812.6A CN113468414A (en) | 2021-06-07 | 2021-06-07 | Commodity searching method and device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110631812.6A CN113468414A (en) | 2021-06-07 | 2021-06-07 | Commodity searching method and device, computer equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113468414A true CN113468414A (en) | 2021-10-01 |
Family
ID=77868669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110631812.6A Pending CN113468414A (en) | 2021-06-07 | 2021-06-07 | Commodity searching method and device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113468414A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113947456A (en) * | 2021-10-19 | 2022-01-18 | 广州华多网络科技有限公司 | Online shop matching method and device, equipment, medium and product thereof |
| CN114429384A (en) * | 2021-12-30 | 2022-05-03 | 杭州盟码科技有限公司 | Intelligent product recommendation method and system based on e-commerce platform |
| CN114741589A (en) * | 2022-03-24 | 2022-07-12 | 深圳买个宝科技有限公司 | Commodity transaction system based on big data |
| CN114861057A (en) * | 2022-05-17 | 2022-08-05 | 北京百度网讯科技有限公司 | Resource sending method, training of recommendation model and device |
| CN115470322A (en) * | 2022-10-21 | 2022-12-13 | 深圳市快云科技有限公司 | Keyword generation system and method based on artificial intelligence |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102609458A (en) * | 2012-01-12 | 2012-07-25 | 北京搜狗信息服务有限公司 | Method and device for picture recommendation |
| US20150310010A1 (en) * | 2014-03-13 | 2015-10-29 | Shutterstock, Inc. | Systems and methods for multimedia image clustering |
| JP2016018286A (en) * | 2014-07-07 | 2016-02-01 | 日本電信電話株式会社 | Action type discrimination system, action type discrimination method, and action type discrimination program |
| US20180113938A1 (en) * | 2016-10-24 | 2018-04-26 | Ebay Inc. | Word embedding with generalized context for internet search queries |
| CN110232122A (en) * | 2019-05-15 | 2019-09-13 | 上海海事大学 | A kind of Chinese Question Classification method based on text error correction and neural network |
| CN110516070A (en) * | 2019-08-28 | 2019-11-29 | 上海海事大学 | A kind of Chinese Question Classification method based on text error correction and neural network |
| CN110619067A (en) * | 2019-08-27 | 2019-12-27 | 深圳证券交易所 | Industry classification-based retrieval method and retrieval device and readable storage medium |
| CN112241493A (en) * | 2020-10-28 | 2021-01-19 | 浙江集享电子商务有限公司 | Commodity retrieval method and device, computer equipment and storage medium |
-
2021
- 2021-06-07 CN CN202110631812.6A patent/CN113468414A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102609458A (en) * | 2012-01-12 | 2012-07-25 | 北京搜狗信息服务有限公司 | Method and device for picture recommendation |
| US20150310010A1 (en) * | 2014-03-13 | 2015-10-29 | Shutterstock, Inc. | Systems and methods for multimedia image clustering |
| JP2016018286A (en) * | 2014-07-07 | 2016-02-01 | 日本電信電話株式会社 | Action type discrimination system, action type discrimination method, and action type discrimination program |
| US20180113938A1 (en) * | 2016-10-24 | 2018-04-26 | Ebay Inc. | Word embedding with generalized context for internet search queries |
| CN110232122A (en) * | 2019-05-15 | 2019-09-13 | 上海海事大学 | A kind of Chinese Question Classification method based on text error correction and neural network |
| CN110619067A (en) * | 2019-08-27 | 2019-12-27 | 深圳证券交易所 | Industry classification-based retrieval method and retrieval device and readable storage medium |
| CN110516070A (en) * | 2019-08-28 | 2019-11-29 | 上海海事大学 | A kind of Chinese Question Classification method based on text error correction and neural network |
| CN112241493A (en) * | 2020-10-28 | 2021-01-19 | 浙江集享电子商务有限公司 | Commodity retrieval method and device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 张素智 等: "基于语义的KNN 短文本分类算法研究", 郑州轻工业学院学报(自然科学版), vol. 27, no. 6, pages 1 - 4 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113947456A (en) * | 2021-10-19 | 2022-01-18 | 广州华多网络科技有限公司 | Online shop matching method and device, equipment, medium and product thereof |
| CN114429384A (en) * | 2021-12-30 | 2022-05-03 | 杭州盟码科技有限公司 | Intelligent product recommendation method and system based on e-commerce platform |
| CN114741589A (en) * | 2022-03-24 | 2022-07-12 | 深圳买个宝科技有限公司 | Commodity transaction system based on big data |
| CN114741589B (en) * | 2022-03-24 | 2024-05-07 | 深圳买个宝科技有限公司 | Commodity transaction system based on big data |
| CN114861057A (en) * | 2022-05-17 | 2022-08-05 | 北京百度网讯科技有限公司 | Resource sending method, training of recommendation model and device |
| CN114861057B (en) * | 2022-05-17 | 2023-05-30 | 北京百度网讯科技有限公司 | Resource sending method, training of recommendation model and device |
| CN115470322A (en) * | 2022-10-21 | 2022-12-13 | 深圳市快云科技有限公司 | Keyword generation system and method based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11983269B2 (en) | Deep neural network system for similarity-based graph representations | |
| CN113468414A (en) | Commodity searching method and device, computer equipment and storage medium | |
| KR101778679B1 (en) | Method and system for classifying data consisting of multiple attribues represented by sequences of text words or symbols using deep learning | |
| US20230088171A1 (en) | Method and apparatus for training search recommendation model, and method and apparatus for sorting search results | |
| CN110909182B (en) | Multimedia resource searching method, device, computer equipment and storage medium | |
| CN110795527B (en) | Candidate entity ordering method, training method and related device | |
| WO2024041483A1 (en) | Recommendation method and related device | |
| CN111078835A (en) | Resume evaluation method and device, computer equipment and storage medium | |
| CN116521906B (en) | Meta description generation method, device, equipment and medium thereof | |
| CN113157876A (en) | Information feedback method, device, terminal and storage medium | |
| CN114663155A (en) | Advertisement putting and selecting method and device, equipment, medium and product thereof | |
| CN115545832A (en) | Commodity search recommendation method and device, equipment and medium thereof | |
| CN114862480A (en) | Advertisement putting orientation method and its device, equipment, medium and product | |
| CN114626926A (en) | Commodity search category identification method and device, equipment, medium and product thereof | |
| CN111310462A (en) | User attribute determination method, device, equipment and storage medium | |
| CN115099857A (en) | Advertisement commodity combined publishing method and device, equipment, medium and product thereof | |
| CN114580402A (en) | Enterprise-oriented product information acquisition method and device, server and storage medium | |
| CN113821588A (en) | Text processing method and device, electronic equipment and storage medium | |
| CN116823404A (en) | Commodity combination recommendation method, device, equipment and medium thereof | |
| CN116485503A (en) | Commodity combination recommendation method, device, equipment and medium thereof | |
| CN116484105A (en) | Service processing method, device, computer equipment, storage medium and program product | |
| CN116029793A (en) | Commodity recommendation method, device, equipment and medium thereof | |
| CN115907928A (en) | Commodity recommendation method, commodity recommendation device, commodity recommendation equipment and commodity recommendation medium | |
| CN113343684A (en) | Core product word recognition method and device, computer equipment and storage medium | |
| CN115203514A (en) | Commodity query redirection method and device, equipment, medium and product thereof |
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 |