Product word processing method and apparatus, device, medium, and product

Abstract

The application discloses a product word processing method and device, equipment, medium and product. The method comprises the following steps: performing word segmentation processing on a commodity title to obtain a plurality of ordered segmented words, and forming a segmented word sequence; calculating the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the commodity title, and taking the data distance as the similarity score of each segmented word; quantitatively determining the ranking score of the segmented word according to the ranking information of the segmented word in the segmented word sequence; and outputting the segmented word with the highest comprehensive score as the product word of the commodity title, wherein the comprehensive score is the sum of the similarity score and the ranking score of the segmented word. The application can conveniently, efficiently and accurately determine the corresponding product word from the given commodity title, and provides basic services for downstream tasks such as commodity search, commodity advertisement placement and commodity collection of the independent site of the e-commerce platform, so as to improve the service experience of the e-commerce platform.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of e-commerce information, and in particular to a product word processing method and a corresponding device, computer equipment, computer readable storage medium, and computer program product. BACKGROUND

[0002] To meet the needs of e-commerce platform in product classification, product search, and advertisement placement, it is often necessary to determine the associated products from the product database according to the given individual keywords. In practice, the corresponding product words are determined for each product in advance to facilitate indexing and matching of the products.

[0003] To determine the product words for the products, the product title is usually used as the basis, for example:

[0004] One of the common methods is to match the product title with the pre-set product words. When the product title contains the pre-set product word, the pre-set product word is the product word of the product title. This method has limited effectiveness because the pre-collected product words cannot cover the expressions of a large number of real products.

[0005] Another method is to use a deep learning model to represent the semantic features of the product title and perform classification mapping to determine the corresponding product word for each product title. This method is costly and has limited effectiveness because the classification space is dense, the sample features are over-generalized, and the training is difficult.

[0006] Further, in the cross-border e-commerce scenario, each online store of the e-commerce platform is deployed as an independent site. In actual operation, each independent site organizes the text according to its own language expression habits when processing the expressions of the product-related information such as the product title, which leads to different texts for the same product name, further reducing the effectiveness of identifying the product words of the product title by using the traditional technology.

[0007] In summary, the traditional technology has very limited effectiveness in determining the product words of the product title, and it is difficult to quickly and efficiently determine the corresponding product words for the product title. Therefore, there is still room for improvement in related technology. SUMMARY

[0008] The primary purpose of the present application is to solve at least one of the above problems and provide a product word processing method and a corresponding device, computer equipment, computer readable storage medium, and computer program product.

[0009] To meet the various purposes of the present application, the present application adopts the following technical solutions:

[0010] A product word processing method provided to meet one of the purposes of the present application includes the following steps:

[0011] The title of the commodity is segmented to obtain an ordered plurality of segmented words, constituting a segmented word sequence;

[0012] The data distance between the semantic feature vector of each segmented word and the semantic feature vector of the title of the commodity is calculated, which is correspondingly taken as the similarity score of each segmented word;

[0013] According to the ordering information of the segmented word hitting the word element in the preset product dictionary in the segmented word sequence, the ordering score is quantitatively determined;

[0014] The segmented word with the highest comprehensive score is output as the product word of the title of the commodity, and the comprehensive score is the sum of the similarity score and the ordering score of the corresponding segmented word.

[0015] In a deepened embodiment, the title of the commodity is segmented to obtain a plurality of segmented words, constituting a segmented word sequence, including the following steps:

[0016] Obtain the title of the commodity submitted by the user;

[0017] The title of the commodity is segmented using a preset segmentation algorithm to obtain a plurality of segmented words;

[0018] According to the order of the plurality of segmented words in the title of the commodity, the plurality of segmented words are constructed into a segmented word sequence, and the ordering information is represented by the ordering value of the segmented word in the segmented word sequence.

[0019] In a deepened embodiment, the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the title of the commodity is calculated, which is correspondingly taken as the similarity score of each segmented word, including the following steps:

[0020] Word embedding is performed on each segmented word and the title of the commodity, respectively, to obtain the corresponding embedding vectors of each segmented word and the title of the commodity;

[0021] The text feature extraction model trained to a convergent state is used to perform representation learning on the corresponding embedding vectors of each segmented word and the title of the commodity, respectively, to obtain the corresponding semantic feature vectors;

[0022] The data distance between the semantic feature vector of each segmented word and the semantic feature vector of the title of the commodity is calculated, and the data distance is taken as the similarity score of the segmented word.

[0023] In a deepened embodiment, according to the ordering information of the segmented word hitting the word element in the preset product dictionary in the segmented word sequence, the ordering score is quantitatively determined, including the following steps:

[0024] According to the semantic feature vector of the title of the commodity, the corresponding commodity classification of the title of the commodity is determined;

[0025] For each segmented word, it is detected whether it contains at least one word element in a product dictionary preset for the corresponding commodity category, and when the word element is contained, it is determined that the segmented word is an optional segmented word that hits the word element in the product dictionary.

[0026] The sorting value of the optional segmented word in the sequence of segmented words is determined, and the preset weight setting associated with the sorting value is set as the sorting score value corresponding to the optional segmented word.

[0027] In an embodiment of the deepening part, the segmented word with the highest comprehensive score is output as the product word of the commodity title, including the following steps:

[0028] The sum of the similarity score and the sorting score of each segmented word that hits the word element in the product dictionary is calculated to obtain the comprehensive score of the segmented word;

[0029] According to the comprehensive score, the segmented words that hit the product dictionary are sorted in reverse order, and the first segmented word is determined as the product word of the commodity title;

[0030] The product word is output.

[0031] In an embodiment of the extension part, before the step of obtaining the sorting information of the segmented word that hits the word element in the preset product dictionary in the sequence of segmented words, the following steps are included:

[0032] From the product words corresponding to each commodity category pre-collected, a plurality of word elements are extracted, and the product dictionary of the corresponding commodity category is constructed.

[0033] In an embodiment of the extension part, after the step of outputting the segmented word with the highest comprehensive score as the product word of the commodity title, the following steps are included:

[0034] According to the product word of the commodity title, a target commodity that is consistent or semantically similar to the product word is retrieved from a commodity database;

[0035] The commodity information of the target commodity is pushed to the terminal device that submits the commodity title.

[0036] A product word processing device provided for adapting to one of the purposes of the present application, comprising a word segmentation processing module, a similarity score module, a ranking score module, and a word determination module, wherein: the word segmentation processing module is configured to perform word segmentation processing on a product title to obtain an ordered plurality of segmented words, forming a segmented word sequence; the similarity score module is configured to calculate the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title, which is used as the similarity score of each segmented word; the ranking score module is configured to quantitatively determine the ranking score of a segmented word in the segmented word sequence according to the ranking information of the segmented word that hits a word element in a preset product dictionary; and the word determination module is configured to output the segmented word with the highest comprehensive score as the product word of the product title, wherein the comprehensive score is the sum of the similarity score and the ranking score of the corresponding segmented word.

[0037] In a deepened embodiment, the word segmentation processing module comprises: a title acquisition unit configured to acquire a product title submitted by a user; a word segmentation execution unit configured to perform word segmentation on the product title using a preset word segmentation algorithm to obtain a plurality of segmented words; and a ranking representation unit configured to construct the plurality of segmented words into a segmented word sequence according to the order of the segmented words in the product title, and represent the ranking information of the segmented words in the segmented word sequence by the ranking values of the segmented words in the segmented word sequence.

[0038] In a deepened embodiment, the similarity score module comprises: an encoding processing unit configured to respectively perform word embedding on each segmented word and the product title to obtain the embedding vectors of each segmented word and the product title; a representation learning unit configured to respectively perform representation learning on the embedding vectors of each segmented word and the product title using a text feature extraction model that has been trained to a convergent state to obtain the semantic feature vectors of each segmented word and the product title; and a distance calculation unit configured to calculate the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title, and use the data distance as the similarity score of the segmented word.

[0039] In a deepened embodiment, the ranking score module comprises: a classification mapping unit configured to determine the product category corresponding to the product title according to the semantic feature vector of the product title; a segmented word hitting unit configured to, for each segmented word, detect whether the segmented word contains at least one word element in a preset product dictionary corresponding to the product category, and when the segmented word contains the word element, determine that the segmented word is an optional segmented word that hits the word element in the product dictionary; and a ranking score unit configured to determine the ranking value of the optional segmented word in the segmented word sequence, and set the ranking score corresponding to the optional segmented word by associating the preset weight with the ranking value.

[0040] In the deepened partial embodiment, the word determining module comprises: a score synthesizing unit configured to calculate a sum of the similarity score and the ranking score of each hit word element in the product dictionary to obtain a comprehensive score of the word element; a ranking optimization unit configured to perform reverse ranking on each hit word element in the product dictionary according to the comprehensive score to determine a first word element as the product word of the product title; and a result output unit configured to output the product word.

[0041] In the expanded partial embodiment, the product word processing device further comprises a dictionary constructing module operating before the ranking score module, configured to extract a plurality of word elements from the pre-collected product words corresponding to each product category, and store the word elements to construct a product dictionary of the corresponding product category.

[0042] In the expanded partial embodiment, the product word processing device further comprises the following modules operating after the word determining module: a search executing module configured to search target products from a product database according to the product word of the product title, the target products being consistent with or semantically similar to the product word; and a product pushing module configured to push product information of the target products to a terminal device submitting the product title.

[0043] A computer device is provided to adapt to one of the purposes of the present application, comprising a central processing unit and a memory, the central processing unit being configured to invoke a computer program stored in the memory to execute the steps of the product word processing method.

[0044] A computer readable storage medium is provided to adapt to another purpose of the present application, which stores a computer program implemented according to the product word processing method in the form of computer readable instructions, the computer program being invoked and run by a computer to execute the steps included in the method.

[0045] A computer program product is provided to adapt to another purpose of the present application, comprising computer program / instructions, the computer program / instructions being executed by a processor to implement the steps of the method described in any one of the embodiments of the present application.

[0046] Compared with the prior art, the technical solution of the present application has at least the following technical advantages:

[0047] Firstly, the application obtains a word segmentation sequence by simply word segmentation on a given commodity title, then determines a similarity score of each word segmentation according to a semantic data distance between the word segmentation and the commodity title, and determines a ranking score corresponding to each word segmentation according to position information implied by a natural ranking of the word segmentation in a word segmentation sequence in which a word element hitting a preset product dictionary is located, adds the similarity score of each word segmentation and the ranking score corresponding thereto to obtain a comprehensive score corresponding to the word segmentation, and realizes comprehensive quantification of importance of information in two dimensions of semantics and position of each word segmentation through the comprehensive score, so as to rank each word segmentation and determine a word segmentation with the highest comprehensive score as a product word of the commodity title, which has a simple and fast and efficient implementation process and low implementation cost.

[0048] Secondly, in the process of determining the ranking score of each word segmentation, each word segmentation is matched with each word element in the preset product dictionary to determine whether each word segmentation hits the product dictionary, so as to determine the ranking score corresponding to the word segmentation that hits, and the word segmentation that does not hit cannot obtain the ranking score, wherein the product dictionary realizes compatible recognition of different versions of a product word of the same commodity expressed in different ways by different online stores through word elements, and it is not difficult to understand that with the assistance of the product dictionary, a very high recognition accuracy can be achieved.

[0049] In addition, since the technical solution of the application can efficiently and accurately recognize the product word of the given commodity title with the above advantages, it can be deployed as a basic service of an e-commerce platform, and serve the calling of each independent site of the e-commerce platform, thereby providing effective services for the needs of commodity search, commodity advertisement placement, commodity collection and the like of each online store. BRIEF DESCRIPTION OF DRAWINGS

[0050] The above and / or additional aspects and advantages of the application will become apparent and be readily understood from the following description, taken in conjunction with the drawings, in which:

[0051] Figure 1 a flowchart of a typical embodiment of the product word processing method of the application;

[0052] Figure 2 a flowchart of the process of word segmentation on a commodity title in an embodiment of the application;

[0053] Figure 3 a flowchart of the process of calculating a similarity score in an embodiment of the application;

[0054] Figure 4 a flowchart of the process of calculating a ranking score in an embodiment of the application;

[0055] Figure 5A flowchart of a process for calculating a comprehensive score to determine a product word in an embodiment of the present application;

[0056] Figure 6 A flowchart of an extended embodiment of the product word processing method of the present application;

[0057] Figure 7 A principle block diagram of the product word processing apparatus of the present application;

[0058] Figure 8 A structural diagram of a computer device used in the present application. DETAILED DESCRIPTION

[0059] Embodiments of the present application are described in detail below with reference to the attached drawing figures, wherein the same or like component have the same or similar designations. The embodiments described below are presented by way of example only and are not intended to limit the present application as hereinafter claimed.

[0060] As will be understood by one skilled in the art, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. It should be further understood that the terms "comprise," "comprises," "comprising," "include," "includes," "including," "contain," "contains," "containing," and the like, when used herein, 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 can be present. Furthermore, "connected" or "coupled" as used herein can include wirelessly connected or wirelessly coupled. As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0061] As will be understood by one skilled in the art, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains, unless otherwise defined. 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 relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0062] Those skilled in the art will understand that, as used herein, the terms "client," "terminal," and "terminal device" include both devices that are solely wireless signal receivers and devices that have both receiving and transmitting hardware that can communicate bi-directionally over a bi-directional communication link. Such devices can include cellular or other communication devices with single-line or multiple-line displays, or no display, Personal Communications Service (PCS) devices that can combine a voice and / or data processor, a PDA that can include a radio frequency receiver and a pager, Internet and / or Intranet access, a Web browser, a calendar, and / or a GPS receiver, a conventional laptop and / or palmtop computer and / or other devices that have a radio frequency receiver. As used herein, the terms "client," "terminal," and "terminal device" can be portable, transportable, mounted in a vehicle (aeronautical, maritime, and / or land), or adapted and / or configured for local and / or distributed operation on Earth and / or any other location in space. As used herein, the terms "client," "terminal," and "terminal device" can also be a communication terminal, an Internet terminal, a music / video playing terminal, such as a PDA, a Mobile Internet Device (MID), and / or a mobile phone with music / video playing function, a smart television, a set-top box, and / or the like.

[0063] As used herein, the terms "server," "client," "service node," and the like refer to hardware that has the equivalent capability of a personal computer, i.e., an electronic device having a central processing unit (including an arithmetic unit and a controller), a memory, an input device, and an output device, and the like necessary components disclosed by the Von Neumann principle. A computer program is stored in the memory, the central processing unit loads the program stored in the external memory into the memory and runs it, executes the instructions in the program, and interacts with the input and output devices, thereby completing a specific function.

[0064] It should be noted that the concept of "server" in the present application can also be extended to the case of a server cluster. According to the principle of network deployment understood by those skilled in the art, the servers should be logically divided, and in physical space, these servers can be independent of each other but can be called through an interface, or can be integrated into a physical computer or a computer cluster. Those skilled in the art should understand this variation and should not be restricted by the implementation of the network deployment of the present application.

[0065] One or more technical features of the present application, unless explicitly specified, can be deployed on a server for implementation and accessed by a client remotely calling an online service interface provided by the server, or can be directly deployed and run on a client for implementation and access.

[0066] The neural network model referred to or possibly referred to in the present application, unless explicitly specified, can be deployed on a remote server and remotely called by a client, or can be deployed on a client with sufficient device capability for direct calling. In some embodiments, when it is run on a client, its corresponding intelligence can be obtained through transfer learning, so as to reduce the requirement for client hardware running resources and avoid excessive occupation of client hardware running resources.

[0067] The various data involved in the present application, unless explicitly specified, can be remotely stored on a server or stored on a local terminal device, as long as it is suitable for being called by the technical solutions of the present application.

[0068] Those skilled in the art should know that the various methods of the present application, although based on the same concept and described to present commonality between them, are independently executable unless otherwise specified. Similarly, for each embodiment disclosed in the present application, it is based on the same inventive concept, so the same concept is understood to be equivalent, and although the concept is expressed differently, it is only for the convenience of appropriate transformation of the concept.

[0069] Unless it is explicitly stated that the various embodiments disclosed in the present application are mutually exclusive, the related technical features involved in each embodiment can be cross-combined to flexibly construct new embodiments, as long as such combination does not deviate from the spirit of the present application and can meet the needs of the prior art or solve some aspects of the deficiencies in the prior art. For this variation, those skilled in the art should know.

[0070] The product word processing method of the present application can be programmed as a computer program product, deployed in a client or a server for running to be implemented, for example, in the e-commerce platform application scenario including live e-commerce in the present application, generally deployed in the server for implementation, whereby the interface opened after the running of the computer program product can be accessed, and the process of the computer program product can be interacted with the human-computer interface to execute the method.

[0071] Please refer to Figure 1 The product word processing method of the present application, in its typical embodiment, includes the following steps:

[0072] Step S1100, performing word segmentation processing on the product title to obtain an ordered plurality of segmented words, constituting a segmented word sequence:

[0073] When the downstream task of determining the product word from the product title is required to provide the product title, the word segmentation processing can be started. The product title is generally a title used to describe the goods in the online store of the e-commerce platform. The downstream task can be a product search task, a product advertising task, or a product collection task, etc. The product search task can search for similar products with the same or similar product words according to the product title; the product advertising task is the same, and the product title corresponding to the product title with similar semantics to the product word of the former product is matched according to the product title of the product accessed by the user; the product collection task is mainly to facilitate the aggregation of multiple products according to the same information such as product words in multiple product titles. The downstream tasks such as this can rely on the product words obtained by processing the product title according to the present application, so the product title requiring product word recognition can be submitted by these downstream tasks.

[0074] After obtaining the product title, the product title can be segmented by using a traditional segmentation method. The optional segmentation method includes but is not limited to a mechanical segmentation method based on string matching, a segmentation method based on statistics, a segmentation method based on understanding, etc. Among them, the segmentation method based on statistics can include an algorithm based on word frequency statistics, an algorithm based on sequence probability, an algorithm based on deep learning, etc. More specifically, any one of the popular tool models such as N-Gram, Jieba, HMM, TF-IDF, etc. can be used to implement the segmentation of the product title, so as to obtain a plurality of segmented words, for example, for the product title "winter mature men's colorful sports suit classic", after segmentation, a plurality of segmented words can be obtained, each segmented word is arranged in its inherent order in the product title, constituting a segmented word sequence, which is represented as follows:

[0075] {winter; mature; mature men; men's suit; men's suit; colorful; sports, sports suit, suit; classic; classic}

[0076] Of course, different segmentation methods may have slightly different segmentation sequences for the same product title, and those skilled in the art will naturally understand this, and this example should not limit the scope covered by the creative spirit of the present application.

[0077] Step S1200, calculate the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title, and correspondingly as the similarity score of each segmented word:

[0078] To facilitate the determination of the semantic correlation between each segmented word in the segmented sequence and the product title, the deep semantic information of each segmented word and the product title can be extracted by any feasible feature extraction method to obtain their respective semantic feature vectors, and the semantic information of each segmented word and the product title can be represented. The feature extraction method is usually recommended to be performed by a deep learning model, which is trained to convergence by a preselected deep learning model, so that it has the ability to perform representation learning according to the given text embedding vector to obtain the high-level semantic information of the text. Based on the principles disclosed herein, those skilled in the art can flexibly implement it.

[0079] The same feature extraction method can be used to determine the respective semantic feature vectors of each segmented word and the product title. After obtaining the semantic feature vectors, a preset data distance algorithm can be used to calculate the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title. The data distance algorithm includes but is not limited to cosine similarity algorithm, Euclidean distance algorithm, Pearson correlation coefficient algorithm, Minkowski distance algorithm, Mahalanobis distance algorithm, and Jaccard coefficient algorithm. Those skilled in the art can select any data distance algorithm as long as it can calculate the distance between points.

[0080] After determining the data distance, in an optional embodiment, it can be normalized to a numerical space such as [0, 1] for easy calculation. The larger the value, the closer the data distance, indicating that the segmented word is more likely to be a product word.

[0081] As can be seen, the data distance between each segmented word and the product title represents the degree of semantic association between the segmented word and the product title, so the data distance can be determined as the similarity score of the segmented word, and the association between the segmented word and the product title can be effectively represented from the semantic dimension.

[0082] Step S1300, according to the ordering information of the segmented word that hits the word in the preset product dictionary in the segmented sequence, quantitatively determine the ordering score:

[0083] It should be noted that as a variation of the present embodiment, the present step S1300 and the previous step S1200 can be executed out of order or concurrently, without affecting the implementation of the present application.

[0084] In order to measure the ranking score of each word based on the position information of each word in the title of the product, first, it can be detected by the preset product dictionary whether each word hits the word element in the preset product dictionary.

[0085] The product dictionary is used to support the matching of each word of the present application to determine whether a certain word has the possibility of becoming a product word. In order to support the matching, a large number of word elements are stored in the product dictionary in advance. The word element is usually a higher level of semantic concept used to describe a certain product word or a certain type of product word, which is usually a substring of such product word in the data level. For example, in the Chinese context, "clothing" can be regarded as a word element of "sports clothing", "shirt", "shirt", etc. Similarly, "card" can also be regarded as a word element of "birthday card", "enterprise card", etc. For example, in the English context, "suit" can be regarded as a word element of "Sweetsuit". As such, in general, for a word element stored in the product dictionary, when a word completely contains the word element, the word is matched with the word element, and thus it is considered that the word hits the product dictionary.

[0086] The collection of the word element can be realized by combining the word segmentation statistical technology or manual screening by those skilled in the art, without affecting the embodiment of the spirit of the present application. Generally, those skilled in the art can determine the corresponding product dictionary and the word elements included therein according to the required product categories of the e-commerce platform.

[0087] For the word that hits the word element in the product dictionary, it indicates that it has the possibility of becoming the product word corresponding to the title of the product. For the word that does not hit the word element in the product dictionary, it means that it does not have the possibility of becoming the product word corresponding to the title of the product, and thus it can not be considered, or although it is considered, it cannot obtain the corresponding ranking score in the present step.

[0088] Since the word sequence is ordered according to the order of the appearance of each word in the title of the product, the ranking value of each word in the word sequence represents the position information of the word, and the size of the position information corresponds to the possibility of the word becoming a product word.

[0089] More specifically, according to the expression habit of human language reflected in the commodity title, the attributive is usually placed in front, and the modified word is placed behind, for example, in the exemplary commodity title "Winter mature men's colorful sports suit classic" in the preceding text, the word "sports suit" is relatively placed in a relatively rear position, and thus has a higher possibility of becoming a product word than the word "sports" in front. Similarly, the word "suit" has a higher possibility of becoming a product word than "sports" and "sports suit" because it is further behind. Accordingly, the position dimension provides a certain degree of reference information for the possibility of the word becoming a product word. Thus, the occurrence position of each word in the word sequence, i.e., the order value, can be used to determine the corresponding order score of the word. In an embodiment, the order value can be directly used as the order score of the corresponding word. In a variant embodiment, a preset weight can be associated with the order value to restrict the contribution value of the position information. The order value can be an index of an array representing the word sequence, such as 0, 1, 2, 3, 4, and the like. Accordingly, the quantification of the order score of the word in the hit product dictionary is completed.

[0090] Similarly, for intuitive calculation, the order score can be normalized to a numerical space of [0, 1], so that the larger the value, the higher the possibility of the corresponding word becoming a product word.

[0091] Step S1400, output the word with the highest comprehensive score as the product word of the commodity title, and the comprehensive score is the sum of the similarity score and the order score of the corresponding word:

[0092] After the foregoing steps, each word in the word sequence that hits the product dictionary will obtain its corresponding similarity score and order score. The sum of the two is the comprehensive score of the word. As for the word that does not hit the product dictionary, it can be ignored. The comprehensive score realizes the comprehensive representation of the semantic similarity between the corresponding word and the commodity title and the importance of the position of the corresponding word in the word sequence. It is not difficult to understand that the comprehensive score can measure the possibility of the corresponding word becoming the commodity title. Accordingly, the word with the highest comprehensive score among the words that hit the product dictionary can be selected as the product word of the commodity title, and then output to submit to the downstream task of the commodity title.

[0093] Through the typical embodiments and variant embodiments of the present application, it can be seen that, compared with the prior art, the technical solution of the present application at least contains the following technical advantages:

[0094] Firstly, the application obtains a word segmentation sequence by simply word segmentation on a given commodity title, then determines a similarity score of each word segmentation according to a semantic data distance between the word segmentation and the commodity title, and determines a ranking score of each word segmentation according to position information implied by a natural ranking of a word segmentation sequence in which a word segmentation hitting a preset product dictionary is located, adds the similarity score of each word segmentation and the ranking score corresponding to the word segmentation to obtain a comprehensive score of the word segmentation, and realizes comprehensive quantification of importance of information in two dimensions of semantics and position of each word segmentation through the comprehensive score, so as to rank each word segmentation and determine a word segmentation with the highest comprehensive score as a product word of the commodity title, which has a simple and fast and efficient implementation process and low implementation cost.

[0095] Secondly, in the process of determining the ranking score of each word segmentation, each word segmentation is matched with each preset word in the preset product dictionary to determine whether each word segmentation hits the product dictionary, so as to determine the corresponding ranking score of the word segmentation that hits, and the word segmentation that does not hit cannot obtain the ranking score, wherein the product dictionary realizes compatible recognition of different versions of a product word of the same commodity expressed in different ways by different online stores through word, and it is not difficult to understand that with the assistance of the product dictionary, a very high recognition accuracy can be achieved.

[0096] In addition, since the technical solution of the application can efficiently and accurately recognize the product word of the given commodity title with the above advantages, it can be deployed as a basic service of an e-commerce platform and serve the calling of each independent site of the e-commerce platform, thereby providing effective services for the needs of commodity search, commodity advertisement placement, commodity collection and the like of each online store.

[0097] Please refer to Figure 2 In a deepened embodiment, the step S1100 of performing word segmentation on the commodity title to obtain a plurality of word segmentations and form a word segmentation sequence comprises the following steps:

[0098] The step S1110 of obtaining the commodity title submitted by the user comprises the following steps:

[0099] For the downstream tasks such as commodity search and commodity advertisement placement, a target commodity can be determined by a user of a terminal device, and the commodity title of the target commodity is submitted by the user to the background as the commodity title submitted by the user.

[0100] The step S1120 of performing word segmentation on the commodity title by using a preset word segmentation algorithm to obtain a plurality of word segmentations comprises the following steps:

[0101] As described above, the word segmentation algorithm can adopt various known algorithms in the prior art. In the present embodiment, a statistical N-Gram algorithm is recommended to be used to segment the product title, which takes words from the product title by a preset sliding window. The sliding step of the sliding window can be set to one single word, and the window size can be set to 2 words, 3 words, or 4 words, etc., which can be implemented flexibly by those skilled in the art. For example, the exemplary product title “winter mature men's colorful sports suit classic” can be segmented into the following word set by a four-word sliding window:

[0102] {winter, mature, mature men, mature men's clothing, mature men, mature men's clothing, men's clothing, colorful, sports, sports suit, suit, classic, classic}

[0103] In step S1130, the plurality of segmented words are constructed into a segmented word sequence in the order of the plurality of segmented words in the product title, and the ordering information of the segmented words is represented by the ordering values of the segmented words in the segmented word sequence.

[0104] According to the segmented word set obtained in the previous step, an array can be further used to represent the segmented word set, which is converted and stored as a segmented word sequence. The indices of the array are ordered from small to large, so that the indices can serve as the ordering values of the segmented words, thereby representing the ordering information of the corresponding segmented words.

[0105] In the present embodiment, the segmented word sequence is obtained by segmenting the product title of a target product in response to a product search service. Subsequently, the product words of the product title can be determined according to any one of the other embodiments of the present application, and output to the product search service to perform product search based on the product words. In this process, the product title is segmented by flexibly using a word segmentation algorithm, which has less calculation amount, high segmentation accuracy, and is particularly suitable for application scenarios such as product titles with relatively discrete word meanings.

[0106] Please refer to Figure 3 In some deepened embodiments, step S1200 calculates the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title, which is used as the similarity score of each segmented word. The step includes the following steps.

[0107] In step S1210, word embedding is performed on each segmented word and the product title, respectively, to obtain the embedding vectors of each segmented word and the product title.

[0108] Furthermore, each segmented word and the product title can be vectorized according to a preset word table to construct the corresponding embedding vectors of each segmented word and the product title. In the recommended alternative embodiment, the Word2Vec model can be used after being fine-tuned and trained to a convergent state with sufficient samples. Those skilled in the art can flexibly implement the model according to the principles disclosed herein.

[0109] Step S1220, the text feature extraction model trained to a convergent state is used to perform representation learning on the embedding vectors of the respective segmented words and the product title respectively, to obtain corresponding semantic feature vectors:

[0110] To realize the extraction of high-level semantics of the embedding vectors of the respective segmented words and the product title, a text feature extraction model can be prepared, which can be implemented based on LSTM, Bert, Transformer, TextCNN, etc. basic neural network model, and after being trained to a convergent state by a sufficient amount of corresponding training samples, it can learn the ability to extract deep semantic vectors from the input embedding vectors.

[0111] Accordingly, the text feature extraction model is used to perform representation learning on the embedding vectors of the respective segmented words and the embedding vectors of the product title, to obtain semantic feature vectors of the respective segmented words and semantic feature vectors of the product title.

[0112] Generally, for subsequent processing, the semantic feature vectors are uniformly mapped to high-dimensional vectors of the same dimension.

[0113] Step S1230, calculate the data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title, and take the data distance as the similarity score of the segmented word:

[0114] As mentioned earlier, any known data distance algorithm, such as the Euclidean distance algorithm or the cosine similarity algorithm, is selected to calculate the similarity between the semantic feature vector of each segmented word and the semantic feature vector of the product title, to obtain the corresponding data distance. To make the semantics uniform, it can be normalized to the value space of [0, 1], so that the lower the value, the less similar the two semantic feature vectors are in semantics; and the higher the value, the closer the two semantic vectors are in semantics. Thus, the data distance corresponding to each segmented word can be used as its similarity score to represent the semantic similarity between the segmented word and the product title, and the close degree of the semantics between the two is also quantitatively represented.

[0115] The embodiment exemplarily discloses the determination process of the similarity score of each segmented word, in which the close degree of the semantics between the segmented word and the product title is quantitatively determined based on the vectorization of the segmented word and the product title, and the high-level semantic information is abstracted from both in this process. Therefore, the data distance between them is calculated to determine the similarity score, which provides effective reference information for the determination of the product word.

[0116] Please refer to Figure 4 In a deepened embodiment, the step S1300 of quantitatively determining the ranking score according to the ordering information of the word pieces in the product dictionary that hit the word piece includes the following steps:

[0117] Step S1310, determining the product category corresponding to the product title according to the semantic feature vector of the product title:

[0118] In this embodiment, the product dictionary can be set according to the product category in the classification system of the e-commerce platform. If the classification system contains multiple levels, generally only one corresponding product dictionary is set for each classification of a higher level. It is generally recommended to set the product dictionary corresponding to the highest level of product category, and the word pieces under the product category are pre-collected to construct the product dictionary.

[0119] Accordingly, before determining the ranking score, it is necessary to determine the product category to which the product corresponding to the product title belongs. Therefore, another pre-set product category model can be used to classify and map the semantic feature vector of the product title to obtain the corresponding product category. The product category model can use Bert, TextCNN, etc. as a basic network model, and then connect a multi-classifier. The classification space of the multi-classifier contains all the product categories that need to be identified. A sufficient amount of training samples are used by those skilled in the art to supervise the training of the product category model to a convergent state, so that it learns to represent the input product title embedding vector to obtain the corresponding semantic feature vector, and then classifies and maps the semantic feature vector to obtain the corresponding product category.

[0120] Step S1320, for each word piece, detecting whether it contains at least one word piece in the product dictionary corresponding to the product category, and when it contains the word piece, determining that the word piece is an optional word piece that hits the word piece in the product dictionary:

[0121] After determining the product category, the product dictionary corresponding to the product category can be called. Then, for each word piece in the word piece sequence, it is detected one by one whether it contains at least one word piece in the product dictionary. Generally, when a word piece is detected to contain a word piece in the product dictionary, it is determined that the word piece hits the word piece, thereby hitting the product dictionary. At this time, the word piece becomes an optional word piece in this embodiment, i.e. a word piece that has the possibility of becoming a product word. Thus, a part of the optional word pieces are selected from the word piece sequence.

[0122] Step S1330, determining the ranking value of the optional word piece in the word piece sequence, and setting the ranking value associated with the pre-set weight as the ranking score corresponding to the optional word piece:

[0123] Further, the ranking value of each optional word can be determined according to the ranking value of the optional word in the word sequence, for example, the subscript value of the optional word in the array element of the word sequence.

[0124] To adjust the relationship between the position information of the word and the semantic information of the word in the comprehensive score of the present application, a preset weight can be associated with the ranking value of the optional word, which can be flexibly set by a person skilled in the art according to business needs, and an adjustment mechanism is opened for this purpose. Then, the product of the preset weight and the ranking value is used as the ranking value of the optional word. In some alternative embodiments, the preset weight is also used to associate similar values involved in the embodiments of the present application, which is equivalent to a replacement relationship. In further alternative embodiments, a preset weight is associated between the ranking value and the similar value as a hyperparameter to smooth them when calculating the comprehensive score of a word, so that the final result is normalized to the value space of [0, 1]. Its essence is also the same, which is an equivalent replacement embodiment of the embodiments given in the present application.

[0125] The present embodiment is aimed at the case where there are multiple product categories in an e-commerce platform, and allows setting product dictionaries for each product category respectively to facilitate professional construction of product dictionaries. Accordingly, the corresponding product category is determined according to a given product title, and the corresponding product dictionary is called according to the product category to calculate the ranking value. This can cope with the complex product category of the e-commerce platform, and use professional and accurate product dictionaries to calculate the ranking value of the word segmentation of the product title associated with the corresponding product category, so that the calculation of the ranking value is more accurate and representative.

[0126] Please refer to Figure 5 In some deepened embodiments, the step S1400 of outputting the word with the highest comprehensive score as the product word of the product title comprises the following steps:

[0127] Step S1410, calculating the sum of the similar value and the ranking value of each word segmentation that hits the word element of the product dictionary to obtain the comprehensive score of the word segmentation:

[0128] As mentioned earlier, the word segmentation that hits the word element of the product dictionary has the possibility of becoming a product word, so its comprehensive score can be calculated accordingly. However, the word segmentation that does not hit any word element of the product dictionary can not be considered. Even if it is considered, since the word segmentation that does not hit the product dictionary at least lacks the ranking value, it can be expected that the comprehensive score obtained by it is relatively lower than that of the word segmentation that hits the product dictionary, so it is difficult to interfere with the optimal decision of the product word.

[0129] To select the most representative segmentation from the segments of the product word dictionary, the comprehensive score of each segment is calculated. The comprehensive score of each segment is the sum of the similarity score and the ranking score. The similarity score or the ranking score can be matched with a preset weight, or a hyperparameter can be used as a weight to smooth both scores.

[0130] Step S1420, according to the comprehensive score, the segments of the product word dictionary are sorted in descending order, and the first segment is determined as the product word of the title:

[0131] To select the product word in the title, the segments of the product word dictionary are sorted in descending order according to the comprehensive score. The higher the ranking of the segment, the higher the comprehensive score. Therefore, the first segment can be determined as the unique product word of the title.

[0132] Step S1430, output the product word:

[0133] At this point, the product word can be returned to the downstream task of submitting the title, and the downstream task can continue to perform specific business such as product search and product matching according to the product word.

[0134] The embodiment further discloses the process of determining the unique product word of the title according to the similarity score and the ranking score of the segment. According to this process, the calculation amount is small, efficient and direct, and the occupation of system running resources is low.

[0135] In another embodiment of the present application, the similarity score of the segment hitting the product word dictionary can be calculated according to the method disclosed in the foregoing embodiments of the present application, and it is not necessary to calculate the full amount of each segment. In this way, the system overhead is saved. As for the case where the product word dictionary corresponds to multiple product categories, the steps S1310 and S1320 can be referred to. After the segments of the product word dictionary hitting the corresponding product category are checked out, the similarity score of these segments is calculated, and the segments not hitting the product word dictionary are not considered.

[0136] In another equivalent alternative embodiment of the present application, when calculating the comprehensive score, the corresponding comprehensive score of each word in the word sequence can be considered, at this time, the ranking score of the word that does not hit the product dictionary is 0, but if its similarity score obtains a particularly high value, it is also very likely to obtain the highest comprehensive score, according to which, the word that does not hit the product dictionary can also be determined as the product word of the product title. Since the product word is a new word relative to the product dictionary, it can also be stored in the product dictionary to realize the expansion of the product dictionary, thereby improving the product word recognition service capability of the product dictionary.

[0137] In an extended partial embodiment, the step S1300, before the step of ranking the word sequence according to the information of the word in the product dictionary, includes the following steps: extracting a plurality of word units from the pre-collected product words corresponding to each product category, and storing the product dictionary of the corresponding product category:

[0138] When the product dictionary is constructed in advance, a large number of product words are pre-collected according to the product category corresponding to the product dictionary. These product words can be extracted from the pre-collected product title set by means of other product word extraction technology that has been realized, and then the word units are extracted according to these product words. Specifically, the occurrence frequency of the words and characters of a finer granularity in the large number of product words can be counted first, and then according to a preset threshold, the words and characters with an occurrence frequency higher than the preset threshold are determined as word units and stored in the product dictionary, thereby realizing the construction of the product dictionary.

[0139] As can be understood from this embodiment, the product dictionary can be basically collected without manual collection, but can be intelligently recognized and extracted according to the resources of the e-commerce platform, and then further serve the technical solutions of the present application. Although the initially constructed product dictionary cannot be directly used to match the unique product word of the product title, after the product dictionary is utilized by each embodiment of the present application, effective reasoning and decision of the product word of the product title can be realized, which is low in cost and effective.

[0140] Please refer to Figure 6 In an extended partial embodiment, the step S1400, after the step of outputting the word with the highest comprehensive score as the product word of the product title, further includes the following steps:

[0141] Step S1500, retrieving a target product from the product database according to the product word of the product title, which is consistent or semantically similar to the product word:

[0142] For an independent site, the commodity database corresponding to the online store of the independent site stores the commodity information of each commodity on the shelf, including the commodity title, and any one of the above embodiments of the application can be used to determine the product word corresponding to the commodity title, and the product word is stored as part of the commodity information and is associated with the commodity title.

[0143] Accordingly, according to a given commodity title, the commodity title can be user-submitted information, such as a search string submitted by a terminal consumer user when performing a commodity search, or a commodity title submitted by an advertisement copy editor user or a merchant user, and after determining the corresponding product word for the commodity title through any one of the preceding embodiments of the application, the target commodity corresponding to the product word consistent with the product word or semantically similar to the product word can be determined as a search matching result corresponding to the user task based on rule matching or semantic matching.

[0144] Step S1600, pushing the commodity information of the target commodity to the terminal device that submitted the commodity title:

[0145] Further, the commodity information of the target commodity, including but not limited to the commodity title, the commodity picture, the commodity price, etc., is obtained and pushed to the terminal device of the corresponding user for reference or access by the corresponding user. For example, for a user performing a commodity search, the corresponding target commodity can be directly accessed using the provided commodity information; for an advertisement copy editor user and a merchant user, the provided commodity information can be used to reference the corresponding commodity title for further editing, and the like, which can be flexibly utilized.

[0146] The embodiment further enriches the examples of the product word determined by the application serving various different downstream tasks, and accordingly, through the technical solution of the application, the basic service capability of the e-commerce platform can be improved, and the needs of various different parties such as consumer users, merchant users, and advertisement copy editor users can be met, thereby enriching and improving the service experience of the e-commerce platform.

[0147] In another embodiment of the application, the product word corresponding to the commodity title of each commodity in the commodity database of an independent site can be determined, and then the product word is clustered to realize the collection of commodity information.

[0148] Please refer to Figure 7A product word processing device provided for adapting to one of the purposes of the present application, which is a functional embodiment of the product word processing method of the present application, the device comprising a word segmentation processing module 1100, a similarity score module 1200, a ranking score module 1300, and a word determination module 1400, wherein: the word segmentation processing module 1100 is configured to perform word segmentation processing on a product title to obtain an ordered plurality of segmented words, thereby forming a segmented word sequence; the similarity score module 1200 is configured to calculate a data distance between a semantic feature vector of each segmented word and a semantic feature vector of the product title, which is used as a similarity score of the segmented word; the ranking score module 1300 is configured to quantitatively determine a ranking score of a segmented word in the segmented word sequence according to ranking information of the segmented word in the segmented word sequence, wherein the segmented word hits a word element in a preset product dictionary; and the word determination module 1400 is configured to output a segmented word with the highest comprehensive score as a product word of the product title, wherein the comprehensive score is a sum of the similarity score and the ranking score of the segmented word.

[0149] In an embodiment, the word segmentation processing module 1100 comprises: a title acquisition unit configured to acquire a product title submitted by a user; a word segmentation execution unit configured to perform word segmentation on the product title using a preset word segmentation algorithm to obtain a plurality of segmented words; and a ranking representation unit configured to construct the plurality of segmented words into a segmented word sequence according to an order of the segmented words in the product title, and represent the ranking information of the segmented words in the segmented word sequence by ranking values of the segmented words in the segmented word sequence.

[0150] In an embodiment, the similarity score module 1200 comprises: an encoding processing unit configured to perform word embedding on each segmented word and the product title respectively to obtain an embedding vector corresponding to each segmented word and the product title; a representation learning unit configured to perform representation learning on the embedding vector corresponding to each segmented word and the product title respectively using a text feature extraction model trained to a convergent state to obtain a semantic feature vector corresponding to each segmented word and the product title; and a distance calculation unit configured to calculate a data distance between the semantic feature vector of each segmented word and the semantic feature vector of the product title, and use the data distance as the similarity score of the segmented word.

[0151] In an embodiment, the ranking score module 1300 comprises: a classification mapping unit configured to determine a product category corresponding to the product title according to the semantic feature vector of the product title; a segmented word hitting unit configured to detect, for each segmented word, whether the segmented word contains at least one word element in a preset product dictionary corresponding to the product category, and determine the segmented word as an optional segmented word that hits the word element in the product dictionary when the segmented word contains the word element; and a ranking score unit configured to determine a ranking value of the optional segmented word in the segmented word sequence, and set a ranking score corresponding to the optional segmented word by associating a preset weight with the ranking value.

[0152] In an embodiment, the word determining module 1400 comprises: a score synthesizing unit, configured to calculate a sum value of the similarity score and the ranking score of each hit word element in the product dictionary, to obtain a synthesized score of the word element; a ranking optimization unit, configured to perform reverse ranking on each hit word element in the product dictionary according to the synthesized score, and determine the first word element as the product word of the product title; and a result output unit, configured to output the product word.

[0153] In an embodiment, the product word processing apparatus further comprises a dictionary constructing module, which is operated before the ranking score module 1300, and is configured to extract a plurality of word elements from the pre-collected product words corresponding to each product category, and store the word elements to construct a product dictionary of the corresponding product category.

[0154] In an embodiment, the product word processing apparatus further comprises a module operated after the word determining module 1400, which comprises: a search executing module, configured to search a target product from a product database according to the product word of the product title, the target product being consistent with the product word or semantically similar to the product word; and a product pushing module, configured to push product information of the target product to a terminal device that submits the product title.

[0155] To solve the above technical problems, the embodiment of the present application further provides a computer device. As shown in the Figure 8 internal structure diagram of the computer device. The computer device comprises a processor, a computer readable storage medium, a memory and a network interface connected through a system bus. The computer readable storage medium of the computer device stores an operating system, a database and computer readable instructions. The database can store a control information sequence. When the computer readable instructions are executed by the processor, the processor can implement a product word processing method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device can store computer readable instructions. When the computer readable instructions are executed by the processor, the processor can execute the product word processing method of the present application. The network interface of the computer device is used to connect and communicate with the terminal. Those skilled in the art can understand that Figure 8 the structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can comprise more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0156] In the embodiment, the processor is configured to execute Figure 7The specific functions of each module and its sub-modules in the above embodiment are described above. The memory stores the program codes and various data required for executing the above modules or sub-modules. The network interface is used for data transmission between the user terminal and the server. The memory in the above embodiment stores the program codes and data required for executing all modules / sub-modules in the product word processing device of the present application, and the server can call the program codes and data of the server to execute the functions of all sub-modules.

[0157] The present application also provides a storage medium storing computer readable instructions, which are executed by one or more processors to make the one or more processors perform the steps of the product word processing method of any embodiment of the present application.

[0158] The present application also provides a computer program product, which includes computer programs / instructions, which are executed by one or more processors to implement the steps of the method of any embodiment of the present application.

[0159] A person skilled in the art can understand that all or part of the processes in the above-mentioned embodiments of the present application can be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. The program can include the processes of the above-mentioned embodiments of the method when executed. The storage medium can be a computer readable storage medium such as a magnetic disc, an optical disc, a read-only memory (ROM), or a random access memory (RAM).

[0160] In summary, the present application can conveniently, efficiently and accurately determine the corresponding product word from the given commodity title, and provide basic services for the downstream tasks of the e-commerce platform serving the independent site of the commodity search, commodity advertisement placement, commodity collection, etc. of the commodity, thereby improving the service experience of the e-commerce platform.

[0161] A person skilled in the art can understand that the steps, measures and schemes in the various operations, methods and processes discussed in the present application can be alternated, changed, combined or deleted. Further, other steps, measures and schemes in the various operations, methods and processes discussed in the present application can also be alternated, changed, rearranged, decomposed, combined or deleted. Further, the steps, measures and schemes in the prior art with the various operations, methods and processes disclosed in the present application can also be alternated, changed, rearranged, decomposed, combined or deleted.

[0162] The above merely describes some embodiments of the present application, and it should be pointed out that, for those skilled in the art, some improvements and refinements can be made without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A product word processing method characterized by comprising: The method comprises the following steps: The title of the commodity is subjected to word segmentation processing to obtain an ordered plurality of segmented words, which form a segmented word sequence; The data distance between the semantic feature vector of each segmented word and the semantic feature vector of the title of the commodity is calculated, which is used as the similarity score of each segmented word; A plurality of word units are extracted from the pre-collected product words corresponding to each commodity category, and the word units are stored to form a product dictionary of the corresponding commodity category according to the highest-level commodity category in the classification system of the e-commerce platform, comprising: collecting a large amount of product words in advance according to the commodity category corresponding to the product dictionary; counting the occurrence frequency of words of a relatively fine granularity in the large amount of product words; determining the words with an occurrence frequency higher than a preset threshold as word units and storing them in the product dictionary; The ordering score of the segmented word that hits the word unit in the preset product dictionary is quantitatively determined according to the ordering information of the segmented word in the segmented word sequence, comprising: determining the commodity category corresponding to the title of the commodity by calling a preset commodity classification model according to the semantic feature vector of the title of the commodity; for each segmented word, detecting whether it contains at least one word unit in the preset product dictionary corresponding to the commodity category; when it is detected that the segmented word contains at least one word unit in the product dictionary, determining that the segmented word is a selectable segmented word that hits the word unit in the product dictionary; determining the ordering value of the selectable segmented word in the segmented word sequence, and setting the ordering value associated with the preset weight as the ordering score corresponding to the selectable segmented word, wherein the later the position of the selectable segmented word in the segmented word sequence, the higher the ordering value. The segmented word with the highest comprehensive score is output as the product word of the title of the commodity, and the comprehensive score is the sum of the similarity score and the ordering score of the corresponding segmented word.

2. The product word processing method according to claim 1, characterized by, The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: Obtain the title of the commodity submitted by the user; The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: Obtain the title of the commodity submitted by the user; 3. The product word processing method according to claim 1, wherein The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: Obtain the title of the commodity submitted by the user; The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps:

4. The product word processing method according to claim 1, characterized by, The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to word segmentation processing to obtain a plurality of segmented words, which form a segmented word sequence, comprising the following steps: The title of the commodity is subjected to According to the comprehensive score, the individual word segmentation hitting the product dictionary is inversely sorted, and a word segmentation with the highest ranking is determined as the product word of the product title; Output the product word.

5. The product word processing method according to any one of claims 1 to 4, characterized by, After the step of outputting the word segmentation with the highest comprehensive score as the product word of the product title, the method further comprises the following steps: According to the product word of the product title, a target product consistent with or semantically similar to the product word is retrieved from a product database; The product information of the target product is pushed to a terminal device submitting the product title.

6. A product word processing apparatus characterized by comprising: Comprise: A word segmentation processing module is configured to perform word segmentation processing on a product title to obtain a plurality of word segmentations in order, forming a word segmentation sequence; A similarity score module is configured to calculate a data distance between a semantic feature vector of each word segmentation and a semantic feature vector of the product title, and correspondingly determine a similarity score of each word segmentation; A plurality of word units are extracted from product words pre-collected according to each product category, and the word units are stored to construct a product dictionary of the corresponding product category according to the highest level product category in the classification system of an e-commerce platform, comprising: pre-collecting a large amount of product words according to the product category corresponding to the product dictionary; counting the occurrence frequency of words and characters of a finer granularity in the large amount of product words; determining words and characters with an occurrence frequency higher than a preset threshold as word units and storing them in the product dictionary; A ranking score module is configured to quantitatively determine a ranking score of a word segmentation hitting a preset product dictionary according to the ranking information of the word segmentation in the word segmentation sequence; The ranking score module comprises: a category mapping unit configured to determine the product category corresponding to the product title according to the semantic feature vector of the product title; a word segmentation hitting unit configured to detect whether each word segmentation contains at least one word unit in the preset product dictionary corresponding to the product category; when it is detected that the word segmentation contains at least one word unit in the product dictionary, the word segmentation is determined as an optional word segmentation hitting the word unit in the product dictionary; a ranking score unit configured to determine the ranking value of the optional word segmentation in the word segmentation sequence and set the ranking value associated with a preset weight as the ranking score corresponding to the optional word segmentation, wherein the later the position of the optional word segmentation in the word segmentation sequence, the higher the ranking value; A word determination module is configured to output the word segmentation with the highest comprehensive score as the product word of the product title, wherein the comprehensive score is the sum of the similarity score and the ranking score of the corresponding word segmentation.

7. A computer device comprising a central processing unit and a memory, characterized in that The central processing unit is configured to call and run a computer program stored in the memory to perform the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer program is stored in the form of computer readable instructions and is implemented according to the method of any one of claims 1 to 5, and when the computer program is called and run by a computer, the steps included in the corresponding method are performed.

Images