Recommended search term determination method and apparatus, and electronic device

Abstract

The application is suitable for the field of communication technology, and provides a recommended search word determination method and device and electronic equipment, including: obtaining an input search word input by a user; determining an intermediate category of the input search word according to a preset relationship between a search word and a category, wherein the preset relationship between the search word and the category includes a corresponding relationship between the search word and categories at different levels, and the intermediate category is a category at an intermediate level in a category hierarchical structure; and determining a recommended search word corresponding to the input search word according to the intermediate category. Through the above method, the diversity of the recommended search word is expanded, and the accuracy of the recommended search word is improved.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of communication, and particularly relates to a recommended search word determination method and device, an electronic device, a computer readable storage medium, and a computer program product. BACKGROUND

[0002] With the evolution of Internet technology, the network has become the main channel for obtaining information. In the face of massive content, websites generally integrate search engines to help users quickly locate the required information. Taking an e-commerce platform as an example, a user can input a search word in the search bar of the e-commerce platform, and the search engine of the e-commerce platform searches for corresponding goods according to the search word. In addition, the search engine can also determine a recommended search word according to the search word input by the user, and display the recommended search word through the interface of the e-commerce platform.

[0003] In the existing recommended search word determination method, the recommended search word corresponding to the search word input by the user is usually determined according to the goods corresponding to the search word input by the user, and according to a preset mapping relationship between the goods and the recommended search word. Alternatively, the recommended search word corresponding to the search word input by the user is predicted according to the intermediate word generated in the process of inputting the search word. However, when these methods are used to determine the recommended search word, the diversity of the determined recommended search word is poor. SUMMARY

[0004] The embodiments of the present application provide a recommended search word determination method, device and electronic device, which can solve the problem of poor expansibility of the existing recommended search word.

[0005] In a first aspect, the embodiments of the present application provide a recommended search word determination method, comprising:

[0006] obtaining an input search word input by a user;

[0007] determining an intermediate category of the input search word according to a preset relationship between a search word and a category, wherein the preset relationship between a search word and a category comprises a corresponding relationship between a search word and a category at different levels, and the intermediate category is a category at an intermediate level in a category hierarchical structure;

[0008] determining a recommended search word corresponding to the input search word according to the intermediate category.

[0009] Compared with the prior art, the embodiments of the present application have the beneficial effects that:

[0010] In the embodiments of the present application, after an input search term input by a user is acquired, an intermediate category of the input search term is determined according to a preset relationship between a search term and a category, and a recommended search term corresponding to the input search term is determined according to the intermediate category. Since the preset relationship between a search term and a category includes a corresponding relationship between a search term and a category at different levels, the corresponding relationship between the input search term and the categories at different levels can be determined according to the preset relationship between a search term and a category, and the intermediate category located at an intermediate level in the category hierarchical structure is determined from the categories at different levels, that is, the intermediate category of the input search term is determined. Since the intermediate category includes both the abstract information of part of the input search term and the specific information of part of the input search term, when the recommended search term corresponding to the input search term is determined according to the intermediate category, the recommended search term is not limited by the specific information of the input search term, and part of the abstract information of the input search term is retained, thereby facilitating expansion of the diversity of the recommended search term and improving the accuracy of the recommended search term.

[0011] In a second aspect, the embodiments of the present application provide a recommended search term determination apparatus, including:

[0012] an input search term acquisition module configured to acquire an input search term input by a user;

[0013] an intermediate category determination module configured to determine an intermediate category corresponding to the input search term according to a preset relationship between a search term and a category, wherein the preset relationship between a search term and a category includes a corresponding relationship between a search term and a category at different levels corresponding to the search term, and the intermediate category is a category located at an intermediate level in a category hierarchical structure;

[0014] a recommended search term determination module configured to determine a recommended search term corresponding to the input search term according to the intermediate category.

[0015] In a third aspect, the embodiments of the present application provide an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method of the first aspect when executing the computer program.

[0016] In a fourth aspect, the embodiments of the present application provide a computer readable storage medium, and the computer readable storage medium stores a computer program, and the computer program is executable on a processor to implement the method of the first aspect.

[0017] In a fifth aspect, the embodiments of the present application provide a computer program product, and when the computer program product is executed on an electronic device, the electronic device executes the method of the first aspect.

[0018] It can be understood that the beneficial effects of the second aspect to the fifth aspect described above can be referred to the related description in the first aspect described above, and will not be repeated here. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows.

[0020] Figure 1 is a flowchart of a method for determining a recommended search term provided by the prior art;

[0021] Figure 2 is a structural diagram of a category prediction model provided by an embodiment of the present application;

[0022] Figure 3 is a diagram of a recommended search term displayed on a page provided by an embodiment of the present application;

[0023] Figure 4 is a flowchart of another method for determining a recommended search term provided by an embodiment of the present application;

[0024] Figure 5 is a structural diagram of a device for determining a recommended search term provided by another embodiment of the present application;

[0025] Figure 6 is a structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0026] In the following description, specific details are set forth in order to provide a thorough understanding of the embodiments of the present application. However, persons skilled in the art will understand that the present application can be practiced without these specific details. In other instances, well-known systems, structures, circuits, and methods have not been described in detail in order to avoid obscuring the description of the present application.

[0027] It should be understood that when used in the specification and the appended claims of the present application, the term "comprising" indicates the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0028] It should also be understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations thereof, and includes these combinations.

[0029] In addition, in the description of the present application and the appended claims, the terms "first", "second", and the like are used only to distinguish descriptions, and cannot be understood as indicating or implying relative importance.

[0030] Reference to "one embodiment" or "some embodiments" or the like in the present application description means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present application. Thus, the appearance of the phrases "in one embodiment", "in some embodiments", "in other some embodiments", "in yet some embodiments" or the like in various places in the specification is not necessarily all referring to the same embodiment, but means "one or more but not all embodiments", unless otherwise specifically indicated.

[0031] Currently, when a user searches in a website (such as an e-commerce platform), the e-commerce platform may, in addition to searching according to the search term input by the user, also guess the search intention of the user, and generate a recommended search term according to the guessed search intention, and then display the recommended search term on the interface of the e-commerce platform.

[0032] When generating a recommended search term, if it is generated according to the goods corresponding to the search term input by the user, or if it is generated according to the intermediate term generated in the process of inputting the search term by the user, the recommended search term generated is likely to have a high similarity with the search term, thereby resulting in poor diversity of the recommended search term generated.

[0033] In order to improve the diversity of the recommended search term generated, the present application embodiment provides a recommended search term determination method. In this method, the relationship between the search term and the category is generated in advance, when the input search term input by the user is obtained, the intermediate category corresponding to the input search term is determined according to the relationship between the search term and the category generated in advance, and then the recommended search term corresponding to the input search term is determined according to the intermediate category.

[0034] The recommended search term determination method provided by the present application embodiment will be described below in conjunction with the accompanying drawings.

[0035] Figure 1 A flowchart of a recommended search term determination method provided by the present application embodiment is shown, which can be applied in an electronic device, and the details are as follows:

[0036] S11, obtaining an input search term input by a user.

[0037] Specifically, the electronic device uses the information entered by the user in the search bar (or search box) displayed on the interface as the aforementioned input search term. For example, suppose the user is currently accessing e-commerce platform A, and the electronic device displays the page corresponding to e-commerce platform A, which includes a search bar. If the user enters "milk" in the search bar and clicks the "search" button on the page, the electronic device will use "milk" as the user's input search term.

[0038] In this embodiment of the application, the search terms entered by the user are named "input search terms", while the search terms recommended to the user are named "recommended search terms" to distinguish between the two.

[0039] S12, Based on the preset relationship between search terms and categories, determine the intermediate category of the input search term, wherein the preset relationship between search terms and categories includes: the correspondence between search terms and categories at different levels, and the intermediate category is the category located at the intermediate level in the category hierarchy structure.

[0040] In the preset search term and category relationship, each search term corresponds to a category level of 3 or more. When the number of category levels is even, and there are 2 categories in the middle level of the category hierarchy, either of these 2 categories can be used as the aforementioned intermediate category. When the number of category levels is odd, and there is only 1 category in the middle level of the category hierarchy, this category is the aforementioned intermediate category. For example, assuming the number of category levels is 3, and the names of these levels of categories, from the highest level to the lowest level, are first-level category, second-level category, and third-level category, then the second-level category is the category in the middle level of the category hierarchy, which is the aforementioned intermediate category. As the category level increases (i.e., the closer to the root node), the higher its level of abstraction and the greater the total amount of information it covers or represents. Conversely, the lower the category level, the more local and specific its information. Therefore, the intermediate category of a determined input search term includes both the abstract information of that input search term and the specific information of that input search term.

[0041] In this embodiment of the application, the category corresponding to the search term can be predicted by the category prediction algorithm, and then the relationship between the search term and the predicted category can be determined.

[0042] In the embodiments of the present application, the category prediction algorithm can be implemented through a category prediction model, which can be composed of a Bidirectional Encoder Representations from Transformers (BERT) and a Fully Connected (FC) layer. The BERT is a pre-training language model based on a multi-layer Transformer bidirectional encoder, which learns deep context representations on large-scale unlabeled text in an unsupervised manner. Its bidirectional modeling capability enables the same word to obtain differentiated vectors in different contexts. By connecting the pre-trained BERT to a simple classifier (such as a fully connected layer combined with an activation function), the context features can be mapped to the product category space, realizing efficient multi-category prediction. The structure diagram of the category prediction model is shown in FIG. 1. Figure 2

[0043] In the embodiments of the present application, Figure 2

[0044] [CLS]: This is a special classification mark. When the BERT model is used for a classification task (such as for category prediction of a search word), it is added to the beginning of the input sequence (i.e., a sequence composed of Tok1, Tok2, Tok3, etc. divided words). The BERT model will take the output vector corresponding to this mark as the aggregate representation of the entire input sequence for subsequent classification tasks.

[0045] Tok: represents the input divided words, for example, Tok1, Tok2, Tok3, which respectively represent the various lexical units obtained by dividing the input information of the to-be-predicted category. The BERT model understands the semantics of the input text by processing these divided words, thereby predicting the category to which it belongs. For example, assuming that the classification task is to predict the category of a search word, the above-mentioned information of the to-be-predicted category is the search word; for another example, assuming that the classification task is to predict the category of a product, the above-mentioned information of the to-be-predicted category can be the product title text, such as “RBZST Rose Garden Series Rose Fragrance Body Shower Gel 300mL”.

[0046] E: can represent embedding. For example, , , , etc., refer to the results of embedding operations on input marks (such as [CLS] and each divided word Tok) converted into vector form.

[0047] ​​T: represents the hidden state vector after the transformer processing. For example, T1, T2, etc. The encoder will process the input embedding vector in multiple layers, including self-attention mechanism and feedforward neural network operations, to obtain a hidden state vector containing context information. That is, the encoder is used to capture the deep semantic and dependency relationship of the input text.

[0048] FC: fully connected layer. It is located at the uppermost layer in Figure 2 , used to receive the vector processed by the encoder, and through the linear transformation of the fully connected layer and the activation function operation, the features are mapped to the class label space, and finally the predicted category (i.e. the label in Figure 2 ) is output.

[0049] In the embodiments of the present application, the relationship between the preset search word and the category can be as shown in Table 1.

[0050] Table 1:

[0051]

[0052] In the embodiments of the present application, the relationship between the search word and the category can be determined offline, so as to improve the determination speed of the intermediate category when it is needed to participate in the determination of the intermediate category according to the relationship between the search word and the category. In addition, since the relationship between the search word and the category is determined offline, only the result table needs to be read in the online stage, so the computing power constraint of the model side is transferred from "online delay" to "offline throughput", so that when building the category prediction model, there is no need to accommodate the number of queries that need to be processed per second online, and then the large parameter BERT can be used with confidence to improve the accuracy of the predicted category.

[0053] S13, determining the recommended search word corresponding to the input search word according to the above intermediate category.

[0054] Among them, the recommended search word is the search word recommended by the electronic device to the user.

[0055] In the embodiments of the present application, the intermediate category corresponding to each search word can be determined from the preset relationship between the search word and the category according to the input search word, and the search word other than the input search word in the each search word is determined as the recommended search word corresponding to the input search word (i.e. the recommended search word is determined by recalling the hot search word of the same category). For example, assuming that the input search word is milk, the intermediate category corresponding to milk is dairy products, and from the preset relationship between the search word and the category, it is determined that the search words corresponding to dairy products include "milk" and "yogurt", then "yogurt" is a search word other than "milk", at this time, "yogurt" can be determined as the recommended search word corresponding to "milk".

[0056] In the embodiments of the present application, the clicked commodity corresponding to the input search term can also be determined, and the search term corresponding to the intermediate category of the input search term is determined as the recommended search term of the input search term (i.e. the recommended search term is determined by recalling the related commodity search term) according to the search term corresponding to the clicked commodity and the intermediate category of the input search term. Of course, the recommended search term corresponding to the input search term can also be determined according to other manners, which is not limited here.

[0057] In the embodiments of the present application, after the input search term input by the user is obtained, the intermediate category of the input search term is determined according to the preset relationship between the search term and the category, and the recommended search term corresponding to the input search term is determined according to the intermediate category. Since the preset relationship between the search term and the category includes the corresponding relationship between the search term and the category of different levels, the category of different levels corresponding to the input search term can be determined according to the preset relationship between the search term and the category, and the category at the intermediate level in the category level structure is determined in the categories of different levels, i.e. the intermediate category of the input search term is determined. And since the intermediate category includes the abstract information of part of the input search term and the specific information of part of the input search term, when the recommended search term corresponding to the input search term is determined according to the intermediate category, the recommended search term is not limited by the specific information of the input search term, and part of the abstract information of the input search term is retained, so as to facilitate the expansion of the diversity of the recommended search term and improve the accuracy of the recommended search term.

[0058] In some embodiments, after the recommended search term is determined, the electronic device displays the recommended search term. For example, the recommended search term is displayed in the search bar in which the input search term input by the user is received, or the recommended search term can also be displayed at other positions of the same page in which the search bar is displayed. As shown in Figure 3 It is assumed that the electronic device currently displays a page of an e-commerce platform. The search bar is at the upper end of the page, and "RR milk" in the search bar is the input search term input by the user. The electronic device searches for corresponding commodities in the e-commerce platform according to the input search term, and commodities A, B, C, D and E are displayed as shown in Figure 3 In addition, the recommended search terms corresponding to "RR milk" are displayed at other positions of the page (i.e. positions on the page other than the position of the search bar and the position of displaying the commodities), i.e. yogurt, RR pure milk, pure milk and fresh milk.

[0059] It should be pointed out that in the embodiments of the present application, the recommended search term corresponding to the input search term can also be determined according to other manners, which is not limited here. Figure 3In the embodiment, only four recommended search words are shown, and in actual cases, other numbers of recommended search words can be shown. For example, a larger number of recommended search words can be shown when there is a larger blank space in the page (e.g., when the number of searched commodities is small), and a smaller number of recommended search words can be shown when there is a smaller blank space in the page (e.g., when the number of searched commodities is large). Of course, a fixed number of recommended search words can also be shown, which is not limited herein.

[0060] In the embodiment, when the recommended search word is determined according to the search word corresponding to the intermediate category same as the intermediate category of the input search word, the step S13 of determining the recommended search word corresponding to the input search word according to the intermediate category comprises:

[0061] A1, determining other search words corresponding to the intermediate category of the input search word according to the preset relationship between the search word and the category, to obtain first candidate search words.

[0062] In the embodiment, the other search words corresponding to the intermediate category of the input search word refer to the search words other than the input search word in the search words corresponding to the intermediate category of the input search word.

[0063] In the embodiment, the search words corresponding to the intermediate category same as the intermediate category of the input search word can be found according to the preset relationship between the search word and the category, and the search words other than the input search word are filtered out from the found search words, which are the first candidate search words.

[0064] A2, filtering out the entity word of which the entity word type is "category" from the first candidate search words according to the preset relationship between the search word and the entity word, and the preset relationship between the search word and the entity word comprises the corresponding relationship between the search word and the entity word type and the entity word.

[0065] In the embodiment, the search word (e.g., the input search word) can be identified according to the entity recognition algorithm, to obtain the entity words included in the search word and the entity word type corresponding to each entity word.

[0066] The above entity recognition algorithm can be implemented by an entity recognition model, which can be composed of a pre-trained BERT model and a Conditional Random Field (CRF) model. Among them, BERT is mainly responsible for deep semantic understanding and feature extraction of input text, providing rich semantic information and context association for entity recognition; while CRF focuses on sequence labeling decision based on the features extracted by BERT, considering the dependency between labels, optimizing the labeling results, and the combination of the two realizes efficient and accurate recognition of entity words and their types in the input text (such as search words or product names).

[0067] Among them, the role of the BERT model is:

[0068] (1) Deep semantic understanding: the pre-trained BERT model can learn rich language knowledge and semantic representation. It has mastered the vocabulary, grammar and semantic information in natural language through the learning of a large amount of text data. In the entity recognition task, BERT can convert each word and its context information in the input search word or product name into a vector representation rich in semantics, so as to better understand the meaning of the text and the context in which the entity word is located.

[0069] (2) Provide context information: the bidirectional Transformer structure of BERT model enables it to consider the context information on the left and right of each word at the same time. For example, in "DY whole box of milk", "DY" as a brand word, its meaning and judgment not only depend on itself, but also need to be combined with the following "milk", "whole box" and other words to understand comprehensively, BERT model can capture the association between these contexts, and provide strong support for accurate recognition of entity words.

[0070] (3) Feature extraction: BERT model can extract high-quality features for subsequent entity recognition tasks. These features include semantic, grammatical and positional information of words in the text, which can help the model more accurately identify entity words and their types in the text.

[0071] Among them, the role of CRF is:

[0072] (1) Sequence labeling decision: CRF is a statistical model for sequence labeling, in the entity recognition task, it is mainly used for the final labeling decision of the feature sequence extracted by BERT. CRF can determine the entity word type label (such as brand, packaging form, category, etc.) of each word according to the feature information of the entire feature sequence and the transition probability between labels, for example, in "Hei Kouwei Beifang Laomian Mantou (550g)", CRF can determine which words belong to the brand (i.e. "Hei Kouwei"), which belong to the ingredient (i.e. "Laomian"), which belong to the category (i.e. "Mantou") and the specification (i.e. "550g").

[0073] (2) Consider the dependency between labels: CRF can consider the order and dependency between labels, that is, the rationality of the label sequence. For example, general brand words will not appear in the middle position of ingredient entity words, etc. Through such dependency constraints, the accuracy and rationality of entity recognition can be further improved.

[0074] (3) Optimize the labeling result: Based on the consideration of the features of the entire feature sequence and the label transition probability, CRF can optimize the labeling result, adjust possible labeling errors or unreasonable labeling positions, and thus obtain more accurate, more semantic and logical entity labeling sequences.

[0075] In an embodiment of the present application, the relationship between the preset search word and the entity word can be as shown in Table 2.

[0076] Table 2:

[0077]

[0078] As can be seen from Table 2, when the entity word type is "category", the corresponding entity word contains the most important information amount of the search word, so filtering out the entity word of the type "category" from the first candidate search word is equivalent to filtering out the entity word containing the most important information amount of the first candidate search word from each entity word of the first candidate search word.

[0079] A3, determining the recommended search word corresponding to the input search word according to the filtered entity word.

[0080] Specifically, the filtered entity word can be used as the recommended search word corresponding to the input search word, or the filtered entity word is filtered again, and the recommended search word corresponding to the input search word is determined according to the filtering result. Since the filtered entity word is the entity word of the type "category" in the first candidate search word, and the entity word of the type "category" contains the most important information amount in the first candidate search word, therefore, through the above-mentioned manner, it is conducive to recalling the recommended words of the related category, thereby facilitating to improve the accuracy of the determined recommended search word.

[0081] In some embodiments, the re-screening of the screened entity words can include screening according to the number of clicks. That is, the preset relationship between the search words and the entity words described above further includes a corresponding relationship between the search words and the number of clicks, and at this time, step A3, determining the recommended search words corresponding to the input search words according to the screened entity words, includes:

[0082] A31, determining the number of clicks of the search words corresponding to the screened entity words according to the corresponding relationship between the search words and the number of clicks, to obtain the number of clicks of the screened entity words.

[0083] In the embodiments of the present application, the number of clicks of the search words corresponding to the screened entity words is the number of clicks of the screened entity words.

[0084] Suppose the relationship between the input search words, the first candidate search words, the entity word types, the entity words, and the number of clicks is as shown in Table 3.

[0085] Table 3:

[0086]

[0087] As can be seen from Table 3, the second category of the first candidate search words is the same as the second category of the input search words, and the entity word types corresponding to the entity words in the first candidate search words in Table 3 are all "category".

[0088] In Table 3, the first candidate search words corresponding to the entity word "yogurt" are "yogurt" and "KS yogurt", respectively, wherein the number of clicks corresponding to the first candidate search word "yogurt" is "56539", and the number of clicks corresponding to the first candidate search word "KS yogurt" is "4006", that is, the number of clicks of the entity word "yogurt" is "56539" and "4006", respectively.

[0089] A32, aggregating the number of clicks of the same screened entity words to obtain the aggregated entity words and the number of clicks of the aggregated entity words.

[0090] In the embodiments of the present application, the click times of the same entity words are accumulated when the same entity word corresponds to multiple click times. For example, in Table 3, the entity words are yogurt, pure milk, fresh milk, fresh milk, and yogurt. In Table 3, the same entity word "yogurt" appears, and the click times of "yogurt" are aggregated, and the click times of the yogurt are 56539+4006=60545. The remaining entity words do not have the same entity word, and the click times of the entity words in Table 3 are aggregated, and the aggregated entity words are yogurt, pure milk, fresh milk, and fresh milk, and the click times of these entity words are 60545, 17694, 10035, and 9602.

[0091] A33, the click times of the aggregated entity words are sorted according to the above-mentioned aggregated entity words, and the recommended search words corresponding to the input search words are determined according to the sorting results.

[0092] In the embodiments of the present application, the click times of the aggregated entity words can be sorted in descending order, and the aggregated entity words in the front are selected according to the sorting results, and the aggregated entity words in the front are used as the recommended search words corresponding to the input search words.

[0093] Suppose the sorting results are shown in Table 4.

[0094] Table 4:

[0095]

[0096] In the embodiments of the present application, if the top 3 aggregated entity words are selected as the recommended search words, according to Table 4, the recommended search words corresponding to RR milk are yogurt, pure milk, and fresh milk. Of course, in actual situations, the top 20 (or other values) aggregated entity words can also be selected as the recommended search words corresponding to the input search words, which is not limited here.

[0097] In the embodiments of the present application, the aggregated entity words are sorted according to the click times of the aggregated entity words, and the click times can reflect the matching degree of the search words and the user search intention, so the recommended search words are determined according to the sorting results, which is beneficial to improve the accuracy of the determined recommended search words.

[0098] In the above description, the related content of determining the recommended search words corresponding to the input search words according to the entity words of the search words corresponding to the same intermediate category of the input search words is introduced, and the related content of determining the recommended search words corresponding to the input search words in combination with the clicked goods corresponding to the input search words is introduced.

[0099] That is, in some embodiments, the method for determining recommended search terms provided by the embodiments of the present application further comprises:

[0100] According to the preset relationship between the search term and the commodity triplet, the clicked commodity corresponding to the input search term is determined, and the preset relationship between the search term and the commodity triplet includes the corresponding relationship between the search term and the clicked commodity. According to the preset relationship between the commodity and the category, the intermediate category corresponding to the clicked commodity corresponding to the input search term is determined, and the preset relationship between the commodity and the category includes the corresponding relationship between the commodity and the category of different levels corresponding to the commodity.

[0101] Correspondingly, the step S13 of determining the recommended search term corresponding to the input search term according to the intermediate category includes:

[0102] B1, according to the intermediate category corresponding to the intermediate category corresponding to the input search term, the intermediate category corresponding to the input search term is determined, and the target clicked commodity is obtained.

[0103] B2, according to the preset relationship between the search term and the commodity triplet, the search term corresponding to the target clicked commodity other than the input search term is determined, and the second candidate search term is obtained.

[0104] B3, according to the second candidate search term, the recommended search term corresponding to the input search term is determined.

[0105] In the embodiments of the present application, the relationship between the search term and the commodity triplet can be determined by the way of burying point statistics. In the embodiments of the present application, the relationship between the search term and the commodity triplet includes the corresponding relationship between the search term and the clicked commodity, and can also include one or more information such as the number of search term clicks, the number of commodity clicks and the number of clicks of the search term clicks the commodity.

[0106] In the embodiments of the present application, the corresponding relationship between the search term and the clicked commodity can be as shown in Table 5.

[0107] Table 5:

[0108]

[0109] In Table 5, when the search term (or input search term) is "fruit", the user clicks the commodity "yellow dream crisp melon", and when the input search term is "fruit", the number of clicks of "yellow dream crisp melon" is 179, and the number of commodity clicks of each input search term clicks "yellow dream crisp melon" is 643. When the input search term is "fruit", the number of clicks of the search term corresponding to each commodity is 23429.

[0110] In the embodiments of the present application, after determining the relationship between the search word and the commodity triplets, the input search word input by the user is compared with the "search word" in the relationship to find the "search word" in the relationship that is the same as the input search word, and each commodity corresponding to the found "search word" is taken as the clicked commodity corresponding to the input search word. For example, in Table 5, if the input search word is "fruit", the commodities corresponding to the "fruit" are: Huangmeng crisp honey melon, Yangfeng crisp persimmon, gold passion fruit, and gold pillow durian.

[0111] After finding the clicked commodity corresponding to the input search word, the clicked commodity is compared with the commodities in the preset commodity-category relationship to determine the intermediate category corresponding to the clicked commodity corresponding to the input search word. The preset commodity-category relationship can be determined by using the above-mentioned category prediction model. Specifically, the commodity information (such as commodity description information) is taken as the input of the category prediction model, and the category output by the category prediction model is obtained.

[0112] In the embodiments of the present application, the preset commodity-category relationship can be as shown in Table 6.

[0113] Table 6:

[0114]

[0115] Suppose the clicked commodity is "honeydew melon", the determined intermediate category corresponding to the clicked commodity (i.e. "honeydew melon") corresponding to the input search word is "vegetables".

[0116] After determining the intermediate category corresponding to the clicked commodity, the intermediate categories that are the same as the intermediate category of the input search word are selected from the intermediate categories, and the clicked commodities corresponding to the selected intermediate categories that are the same as the intermediate category of the input search word are determined as the target clicked commodities. Suppose the input search word is "RR milk", the intermediate category of the input search word is "dairy products", if one clicked commodity corresponding to the input search word is "RR concentrated milk", and the intermediate category of the "RR concentrated milk" is also "dairy products", the "RR concentrated milk" is the clicked commodity whose intermediate category is the same as the intermediate category of the "RR milk", and is also the target clicked commodity.

[0117] After the target clicked commodity is determined, according to a preset relationship between the search word and the commodity triple, each search word corresponding to the target clicked commodity is determined, and a search word different from the input search word is filtered out from the each search word, and the filtered search word different from the input search word is the second candidate search word. For example, assuming that the input search word is "RR milk", the target clicked commodity is "RR concentrated milk", and the search words corresponding to the "RR concentrated milk" include "RR milk" and "milk", the "milk" is a search word different from the "RR milk", and at this time, the "milk" is the second candidate search word.

[0118] After the second candidate search word is determined, the second candidate search word can be used as a recommended search word corresponding to the input search word, or the second candidate search word can be screened, and then a recommended search word corresponding to the input search word is determined according to a screening result, which is not limited here.

[0119] In the embodiment of the application, when the recommended search word is determined, the related search word of the clicked commodity corresponding to the input search word is combined, and the clicked commodity reflects that the related search word has a higher matching degree with the commodity, so that when the recommended search word is determined by the above method, the accuracy of the recommended search word is improved.

[0120] In some embodiments, in order to make the determined recommended search word not only concise, but also capable of subdividing the search intention of the user on the basis of the search word of the user, after the second candidate search word is obtained in the above step B2, the method further includes:

[0121] C1, according to a preset relationship between the search word and the entity word, an entity word of a target entity word type is screened out from the second candidate search word, and the target entity word type includes at least one of the following entity word types: series, component, characteristic, brand, category, and class.

[0122] In the embodiment of the application, in the preset relationship between the search word and the entity word, a search word same as the second candidate search word is searched, and an entity word corresponding to a target entity word type of the searched search word is determined. For example, if the target entity word type includes series, component, characteristic, brand, category, and class, the second candidate search word is "Shuhua milk", the entity word type of the "Shuhua" entity word in the "Shuhua milk" is "characteristic", and the entity word type of the "milk" entity word in the "Shuhua milk" is "category", the "Shuhua" and the "milk" are both entity words of the target entity word type screened out from the "Shuhua milk".

[0123] Optionally, considering that the text length of the entity word is 1, the entity word can reflect the limited search intention of the user, therefore, when screening the entity words of the target entity word type, the entity words can also be screened according to the text length of the entity words, for example, only the entity words with the text length greater than 1 and the entity word type being the target entity word type are screened out. When only the entity words with the text length greater than 1 and the entity word type being the target entity word type are screened out, the entity words screened out from the second candidate search word "Shuhua milk" are only "Shuhua", and the entity word "milk" will be screened out.

[0124] C2, judging whether the second candidate search word meets a preset filtering condition according to the screened entity words, wherein the preset filtering condition includes: for the four entity word types of series, component, characteristic and brand, only one entity word type of the four entity word types; for the two entity word types of category and class, only one entity word type of the two entity word types.

[0125] Considering that the user search goods usually contains the words of category or class, such as milk, pure milk, and adding the rhetorical words before the words of category or class will make the information contained in the search words (i.e. the words of category or class added with the rhetorical words) clearer, for example, adding the brand "RR" as the rhetorical word before "milk", the information contained in the new search word "RR milk" is clearer than the information contained in "milk". Similarly, adding the component "oatmeal" as the rhetorical word before "milk", the information contained in the new search word "oatmeal milk" is also clearer than the information contained in "milk". Therefore, in the embodiment of the present application, the filtering condition can be set to further screen the second candidate search word.

[0126] In the embodiment of the present application, the preset filtering condition includes: for the four entity word types of series, component, characteristic and brand, only one entity word type of the four entity word types; for the two entity word types of category and class, only one entity word type of the two entity word types.

[0127] For example, assuming that the entity words of the second candidate search term "TT pure milk" are "TT" and "pure milk" respectively, and the entity word type of "TT" is "series" and the entity word type of "pure milk" is "category". Since the entity word types of "TT" and "pure milk" both belong to the target entity word type, "TT" and "pure milk" are both screened entity words. According to the entity word type of "TT" and the entity word type of "pure milk", it is determined whether "TT pure milk" satisfies the preset filtering condition: since in "TT pure milk", only the entity word type of "series" among the four entity word types of series, composition, characteristic and brand appears, and only the entity word type of "category" among the two entity word types of category and class appears, it is determined that "TT pure milk" satisfies the preset filtering condition.

[0128] For another example, assuming that the second candidate search term is "RR", since the entity word type of "RR" is brand and "RR" only satisfies one of the preset filtering conditions, it is determined that "RR" does not satisfy the preset filtering condition.

[0129] Optionally, in order to avoid the recommendation search term process, the above-mentioned filtering condition can also be set to include that the number of entity words of the second candidate search term is equal to 2. When the filtering condition includes that the number of entity words of the second candidate search term is equal to 2, since the number of entity words of "RR" is 1, according to the filtering condition, it can also be determined that "RR" does not satisfy the preset filtering condition.

[0130] Optionally, when the entity words are also screened according to the text length of the entity words, the above-mentioned step C2 of determining whether the corresponding second candidate search term satisfies the preset filtering condition according to the screened entity words includes:

[0131] If the text length of the screened entity words is greater than 1, it is determined whether the corresponding second candidate search term satisfies the preset filtering condition according to the screened entity words.

[0132] For example, if the second candidate search term is "Shuhua milk", the corresponding entity words are "Shuhua" and "milk", since the text length of the entity word "milk" is equal to 1, the entity word "milk" will be screened out, and only the entity word type of "Shuhua" is used to determine whether "Shuhua milk" satisfies the preset filtering condition.

[0133] Correspondingly, the above-mentioned B3 of determining the recommended search term corresponding to the input search term according to the above-mentioned second candidate search term includes:

[0134] If it is determined that the corresponding second candidate search term satisfies the preset filtering condition according to the screened entity words, the recommended search term corresponding to the input search term is determined according to the second candidate search term that satisfies the preset filtering condition.

[0135] In the embodiments of the present application, the second candidate search term satisfying the preset filtering condition can be taken as the recommended search term corresponding to the input search term. Of course, the second candidate search term satisfying the preset filtering condition can also be filtered, and then the recommended search term corresponding to the input search term is determined according to the filtering result. Since the target entity word type includes at least one of the following entity word types: series, component, feature, brand, category, and class, and these entity word types can more specifically reflect the user's intention, the entity word of the target entity word type filtered from the second candidate search term is equivalent to the entity word that can more specifically reflect the user's intention filtered from the second candidate search term. Since the preset filtering condition includes: for the four entity word types of series, component, feature, and brand, only one of the above four entity word types; for the two entity word types of category and class, only one of the above two entity word types, and the second candidate search term satisfying the filtering condition contains more clear information, therefore, judging whether the corresponding second candidate search term satisfies the preset filtering condition according to the filtered entity word is equivalent to further filtering the search term that can clearly reflect the user's intention from the second candidate search term, so that when the recommended search term is determined according to the second candidate search term satisfying the preset filtering condition, the determined recommended search term can clearly reflect the user's intention.

[0136] In some embodiments, the step B3 of determining the recommended search term corresponding to the input search term according to the second candidate search term includes:

[0137] D1, calculating the relevance score of the second candidate search term and the input search term.

[0138] In the embodiments of the present application, the relevance score can be determined by the number of all clicked goods corresponding to the input search term, the number of all clicked goods corresponding to the second candidate search term, the corresponding number of clicks of the input search term, and the number of clicks corresponding to the second candidate search term. For example, the relevance score can be calculated by the following formula:

[0139] The relevance score of the second candidate search term and the input search term = (the number of all clicked goods corresponding to the input search term + the number of all clicked goods corresponding to the second candidate search term) / (the number of clicks corresponding to the input search term + the number of clicks corresponding to the second candidate search term).

[0140] For example, assuming the input keyword is "RR milk", when the user inputs "RR milk", the clicked commodities are "RRTT pure milk", "RR pure milk in a whole box", and "YLRRTT pure milk", respectively. The number of commodities corresponding to "RR milk" is 3. Assuming "RRTT pure milk" is clicked 1000 times, "RR pure milk in a whole box" is clicked 200 times, and "YLRRTT pure milk" is clicked 2000 times, the number of clicks corresponding to "RR milk" is 1000+200+2000=3200. Similarly, the number of commodities corresponding to the second candidate search keyword and the number of clicks corresponding to the second candidate search keyword are calculated in the same way, which will not be described herein.

[0141] Optionally, to improve the accuracy of the calculated correlation score, the correlation information of the same input keyword is aggregated before the correlation score is calculated, and the correlation information of the same second candidate search keyword is aggregated, and then the corresponding correlation score is calculated according to the aggregation result. The correlation information includes the number of commodities and the number of clicks. For example, assuming there are three input keywords "RR milk", one "RR milk" corresponds to "RRTT pure milk", one "RR milk" corresponds to "RR pure milk in a whole box", and the last "RR milk" corresponds to "YLRRTT pure milk". The number of commodities corresponding to the clicked commodities of the input keyword is aggregated, and the obtained number of commodities is 1+1+1=3. Similarly, the aggregation of the number of clicks is similar to the aggregation of the number of commodities, which will not be described herein.

[0142] D2, according to the above correlation score and the above second candidate search keyword, determining the recommended search keyword corresponding to the input search keyword.

[0143] In the embodiments of the present application, the correlation score and the second candidate search keyword corresponding to the correlation score can be sorted in descending order, and then the recommended search keyword corresponding to the input search keyword is determined according to the sorting result. For example, the second candidate search keyword ranked in the top 5 or top 10 or top 20 (or other numerical value) is taken as the recommended search keyword corresponding to the input search keyword.

[0144] Optionally, considering that the two second candidate search terms have the same meaning although the order of the entity words of the two second candidate search terms is different, before sorting the correlation scores and the second candidate search terms corresponding to the correlation scores in descending order, the second candidate search terms can be filtered first, and then the correlation scores corresponding to the second candidate search terms remaining after the filtering are sorted in descending order. To improve the accuracy of the sorting result, when the second candidate search terms are filtered, the second candidate search terms with higher correlation scores are retained. For example, it is assumed that there are two second candidate search terms, “pure milk RR” and “RR pure milk”, since the two entity words of the two second candidate search terms are the same, only the order is different, it is determined that the two second candidate search terms are repeated search terms, and the second candidate search term with a lower correlation score needs to be filtered out. In the embodiment of the present application, when the electronic device determines the recommended search terms, if the number of the determined recommended search terms is large, the number of the recommended search terms displayed on each page can be set first, and the display order of the recommended search terms is set, and then the corresponding recommended search terms are displayed according to the set result.

[0145] In the embodiment of the present application, since the recommended search terms corresponding to the input search term can be determined in different ways, for example, the recommended search terms can be determined by recalling hot search terms of the same category, and the recommended search terms can also be determined by recalling related commodity search terms, therefore, the electronic device can determine the recommended search terms corresponding to the input search term in turn, for example, first, the recommended search terms are determined by recalling hot search terms of the same category, and the recommended search terms are displayed, then, the recommended search terms are determined by recalling related commodity search terms, and the recommended search terms are displayed. Then, the recommended search terms are determined by recalling hot search terms of the same category again, and the recommended search terms are displayed, and the recommended search terms are determined by recalling related commodity search terms again, and the recommended search terms are displayed. Such a cycle in turn not only can expand the diversity of search terms, but also can ensure that the recommended search terms reflect the more detailed search intention of the current user.

[0146] In order to more clearly describe the method for determining the recommended search terms provided in the embodiments of the present application, the following will be described in combination with Figure 4 .

[0147] Figure 4 A flowchart of another method for determining recommended search terms provided in the embodiments of the present application is shown. In Figure 4 , it is assumed that the number of categories is 3, and therefore the intermediate category is a two-level category.

[0148] S401, obtaining an input search term input by a user.

[0149] S402, determine the secondary category of the input search term.

[0150] S403, determine the clicked commodity corresponding to the input search term.

[0151] S404, determine the secondary category of the clicked commodity.

[0152] S405, determine the clicked commodity corresponding to the secondary category of the input search term, to obtain the target clicked commodity.

[0153] S406, determine the search term of the target clicked commodity corresponding to the non-input search term, to obtain the second candidate search term.

[0154] S407, determine the secondary category of the second candidate search term.

[0155] S408, determine the second candidate search term corresponding to the secondary category identical to the secondary category of the clicked commodity.

[0156] S409, filter the entity term of the target entity term type from the second candidate search term.

[0157] S410, filter the second candidate search term according to the preset filtering condition.

[0158] S411, calculate the relevance score of the second candidate search term and the input search term.

[0159] S412, determine the recommended search term according to the relevance score and the second candidate search term.

[0160] S413, determine other search terms corresponding to the intermediate category of the input search term, to obtain the first candidate search term.

[0161] S414, filter the entity term of the entity term type “category”.

[0162] S415, determine the recommended search term after aggregating the click times of the filtered entity term.

[0163] S416, determine and display the recommended search term in different ways in turn. For example, determine the recommended search term in the way of recalling hot search terms of the same category or in the way of recalling related commodity search terms, and display the determined recommended search term.

[0164] It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution. The execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0165] corresponding to the recommended search term determination method described in the above embodiments, Figure 5A structural block diagram of a recommended search word determination apparatus provided by an embodiment of the present application is shown. For ease of illustration, only parts related to the embodiments of the present application are shown.

[0166] With reference to Figure 5 The recommended search word determination apparatus 5 is applied to an electronic device and includes an input search word acquisition module 51, an intermediate category determination module 52, and a recommended search word determination module 53. The input search word acquisition module 51 is configured to acquire an input search word input by a user. The intermediate category determination module 52 is configured to determine an intermediate category corresponding to the input search word according to a preset relationship between a search word and a category. The preset relationship between a search word and a category includes a corresponding relationship between a search word and different levels of categories corresponding to the search word. The intermediate category is a category at an intermediate level in a category hierarchical structure. The recommended search word determination module 53 is configured to determine a recommended search word corresponding to the input search word according to the intermediate category.

[0167] The input search word acquisition module 51 is configured to acquire an input search word input by a user.

[0168] The intermediate category determination module 52 is configured to determine an intermediate category corresponding to the input search word according to a preset relationship between a search word and a category. The preset relationship between a search word and a category includes a corresponding relationship between a search word and different levels of categories corresponding to the search word. The intermediate category is a category at an intermediate level in a category hierarchical structure.

[0169] The recommended search word determination module 53 is configured to determine a recommended search word corresponding to the input search word according to the intermediate category.

[0170] In the embodiments of the present application, after an input search word input by a user is acquired, an intermediate category of the input search word is determined according to a preset relationship between a search word and a category, and a recommended search word corresponding to the input search word is determined according to the intermediate category. Since the preset relationship between a search word and a category includes a corresponding relationship between a search word and different levels of categories, different levels of categories corresponding to the input search word can be determined according to the preset relationship between a search word and a category, and a category at an intermediate level in a category hierarchical structure is determined from the different levels of categories, i.e., the intermediate category of the input search word is determined. Since the intermediate category includes both abstract information of part of the input search word and specific information of part of the input search word, when a recommended search word corresponding to the input search word is determined according to the intermediate category, the recommended search word is not limited by the specific information of the input search word and retains part of the abstract information of the input search word, thereby facilitating expansion of the diversity of the recommended search word and improvement of the accuracy of the recommended search word.

[0171] Optionally, the recommended search word determination module includes:

[0172] The first candidate search word determination unit is configured to determine other search words corresponding to the intermediate category of the input search word according to the preset relationship between a search word and a category, to obtain a first candidate search word.

[0173] The entity screening unit of the category is configured to screen out an entity word of a category type from the first candidate search word according to a preset search word and entity word relationship. The preset search word and entity word relationship includes a corresponding relationship between a search word and an entity word type and an entity word.

[0174] The hot search word recall unit of the same category is configured to determine the recommended search word corresponding to the input search word according to the screened entity word.

[0175] Optionally, the preset search word and entity word relationship further includes a corresponding relationship between a search word and a click number. The hot search word recall unit of the same category is specifically configured to:

[0176] According to the corresponding relationship between the search word and the click number, the click number of the search word corresponding to the screened entity word is determined to obtain the click number of the screened entity word.

[0177] The click numbers corresponding to the same screened entity words are aggregated to obtain an aggregated entity word and a click number of the aggregated entity word.

[0178] The aggregated entity word is sorted according to the click number of the aggregated entity word, and the recommended search word corresponding to the input search word is determined according to a sorting result.

[0179] Optionally, the recommended search word determination apparatus 5 further includes:

[0180] The clicked commodity determination module is configured to determine the clicked commodity corresponding to the input search word according to a preset search word and commodity triple relationship. The preset search word and commodity triple relationship includes a corresponding relationship between a search word and a clicked commodity.

[0181] The clicked commodity corresponding intermediate category determination module is configured to determine the clicked commodity corresponding intermediate category corresponding to the input search word according to a preset commodity and category relationship. The preset commodity and category relationship includes a corresponding relationship between a commodity and different levels of categories corresponding to the commodity.

[0182] Correspondingly, the recommended search word determination module includes:

[0183] The target clicked commodity determination unit is configured to determine the clicked commodity corresponding to the input search word according to the clicked commodity corresponding intermediate category, to obtain a target clicked commodity.

[0184] The second candidate search word determination unit is configured to determine a search word corresponding to the target clicked commodity and not being the input search word according to a preset relationship between the search word and the commodity triplet, to obtain a second candidate search word.

[0185] The related commodity search word recall unit is configured to determine a recommended search word corresponding to the input search word according to the second candidate search word.

[0186] Optionally, the recommended search word determination apparatus 5 further comprises:

[0187] The target entity word type entity word screening module is configured to screen out an entity word of a target entity word type from the second candidate search word according to a preset relationship between a search word and an entity word after the second candidate search word is obtained, the target entity word type including at least one of the following entity word types: series, ingredient, characteristic, brand, category, and class.

[0188] The second candidate search word filtering module is configured to determine whether the corresponding second candidate search word meets a preset filtering condition according to the screened entity word, wherein the preset filtering condition includes: for the four entity word types of series, ingredient, characteristic, and brand, only one entity word type of the four entity word types; and for the two entity word types of category and class, only one entity word type of the two entity word types.

[0189] Correspondingly, the related commodity search word recall unit is specifically configured to:

[0190] If it is determined according to the screened entity word that the corresponding second candidate search word meets the preset filtering condition, the recommended search word corresponding to the input search word is determined according to the second candidate search word meeting the preset filtering condition.

[0191] Optionally, when determining whether the corresponding second candidate search word meets the preset filtering condition according to the screened entity word, the second candidate search word filtering module is specifically configured to:

[0192] If the text length of the screened entity word is greater than 1, it is determined according to the screened entity word whether the corresponding second candidate search word meets the preset filtering condition.

[0193] Correspondingly, the related commodity search word recall unit is specifically configured to:

[0194] The correlation score between the second candidate search word and the input search word is calculated.

[0195] The recommended search word corresponding to the input search word is determined according to the correlation score and the second candidate search word.

[0196] It should be noted that the information interaction, execution process and the like between the above apparatuses / units are based on the same concept as the method embodiments of the present application, and the specific functions and brought technical effects can be referred to the method embodiments part, which will not be repeated here.

[0197] Figure 6 The structure schematic diagram of an electronic device provided by an embodiment of the present application is shown in the figure. Figure 6 As shown in the figure, the electronic device 6 of the embodiment includes at least one processor 60 (only one processor is shown in the figure), a memory 61, and a computer program 62 stored in the memory 61 and executable on the at least one processor 60, wherein the processor 60 implements the steps in any of the above method embodiments when executing the computer program 62. Figure 6

[0198] The electronic device 6 can be a desktop computer, a notebook computer, a palm computer, a cloud server and the like computing device. The electronic device can include, but is not limited to, the processor 60, the memory 61. Those skilled in the art can understand that, Figure 6 The electronic device 6 is only an example and does not constitute a limitation on the electronic device 6, and can include more or fewer components than shown, or combine certain components, or different components, for example, can also include input / output devices, network access devices and the like.

[0199] The processor 60 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or can also be any conventional processor.

[0200] ​The memory 61 can be an internal storage unit of the electronic device 6, such as a hard disk or a memory of the electronic device 6 in some embodiments. The memory 61 can also be an external storage device of the electronic device 6, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, and the like equipped on the electronic device 6 in other embodiments. Further, the memory 61 can include both an internal storage unit and an external storage device of the electronic device 6. The memory 61 is used to store an operating system, an application program, a BootLoader, data, and other programs, such as program codes of the computer program, and the like. The memory 61 can also be used to temporarily store data that has been output or is to be output.

[0201] It should be clearly understood by those skilled in the art that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or in the form of a software functional unit. In addition, the specific names of the functional units and modules are only for the convenience of mutual distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the system can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0202] The embodiments of the present application also provide a network device, which comprises at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, and the processor implements the steps in any of the method embodiments described above when executing the computer program.

[0203] The embodiments of the present application also provide a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the steps in any of the method embodiments described above.

[0204] The embodiments of the present application provide a computer program product, which, when running on an electronic device, enables the electronic device to execute the steps in any of the method embodiments described above.

[0205] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the present application can implement all or part of the processes in the above-mentioned embodiment methods through a computer program to instruct relevant hardware to complete, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of each method embodiment described above can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium at least includes any entity or device capable of carrying the computer program code to the photographing device / electronic device, recording medium, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium. For example, U disk, mobile hard disk, magnetic disk or optical disk, etc. In some jurisdictions, according to legislation and patent practice, the computer readable medium can not be an electrical carrier signal and a telecommunication signal.

[0206] In the above embodiments, the description of each embodiment has its own focus, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0207] Those skilled in the art can appreciate that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0208] In the embodiments provided in the present application, it should be understood that the disclosed apparatus / network device and method can be implemented in other ways. For example, the apparatus / network device embodiments described above are only schematic. For example, the division of the modules or units is only a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual coupling or direct coupling or communication connection between each of the units can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.

[0209] The units described as separate components may or may not be physically separate, and the components displayed as units may or may not be physical units, that is, may be located in one place, or may also be distributed to multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0210] It should be noted that the information collection process (such as the face image collection process, the fingerprint information collection process, etc.) / feature extraction process involved in the present application is executed with the user's knowledge and with the user's permission, that is, the information collection process / feature extraction process meets the legal and regulatory requirements and does not belong to the act of obstructing public interests.

Claims

1. A method for determining recommended search terms, characterized in that, The method comprises the following steps: obtaining an input search word input by a user; determining an intermediate category of the input search word according to a preset relationship between a search word and a category, wherein the preset relationship between the search word and the category comprises a corresponding relationship between the search word and categories at different levels, and the intermediate category is a category at an intermediate level in a category hierarchical structure; determining a recommended search word corresponding to the input search word according to the intermediate category; determining a clicked commodity corresponding to the input search word according to a preset relationship between a search word and a commodity triple, wherein the preset relationship between the search word and the commodity triple comprises a corresponding relationship between the search word and the clicked commodity; determining an intermediate category corresponding to the clicked commodity corresponding to the input search word according to a preset relationship between a commodity and a category, wherein the preset relationship between the commodity and the category comprises a corresponding relationship between the commodity and categories at different levels corresponding to the commodity; the step of determining the recommended search word corresponding to the input search word according to the intermediate category comprises: determining a clicked commodity corresponding to the intermediate category as the clicked commodity corresponding to the intermediate category of the input search word to obtain a target clicked commodity; determining a search word corresponding to the target clicked commodity but not the input search word according to the preset relationship between the search word and the commodity triple to obtain a second candidate search word; determining the recommended search word corresponding to the input search word according to the second candidate search word.

2. The recommended search term determination method of claim 1, wherein, the step of determining the recommended search word corresponding to the input search word according to the intermediate category comprises: determining other search words corresponding to the intermediate category of the input search word according to the preset relationship between the search word and the category to obtain a first candidate search word; screening an entity word of a product category type from the first candidate search word according to a preset relationship between a search word and an entity word, wherein the preset relationship between the search word and the entity word comprises a corresponding relationship between the search word, the entity word type and the entity word; determining the recommended search word corresponding to the input search word according to the screened entity word.

3. The recommended search term determination method according to claim 2, characterized by, The preset relationship between the search word and the entity word further comprises a corresponding relationship between the search word and a click frequency, and the step of determining the recommended search word corresponding to the input search word according to the screened entity word comprises: determining a click frequency of a search word corresponding to the screened entity word according to the corresponding relationship between the search word and the click frequency to obtain a click frequency of the screened entity word; aggregating click frequencies of the same screened entity words to obtain an aggregated entity word and a click frequency of the aggregated entity word; sorting the aggregated entity word according to the click frequency of the aggregated entity word, and determining the recommended search word corresponding to the input search word according to a sorting result.

4. The recommended search term determination method of claim 1, wherein, After the second candidate search word is obtained, the method further comprises the following steps: screening an entity word of a target entity word type from the second candidate search word according to a preset relationship between a search word and an entity word, wherein the target entity word type comprises at least one of the following entity word types: series, component, feature, brand, product category and category. Based on the selected entity words, it is determined whether the corresponding second candidate search term meets the preset filtering conditions. The preset filtering conditions include: for the four entity word types of series, ingredients, characteristics and brand, only one of the four entity word types is selected; for the two entity word types of category and class, only one of the two entity word types is selected. The step of determining the recommended search term corresponding to the input search term based on the second candidate search term includes: If the selected entity words determine that the corresponding second candidate search term meets the preset filtering conditions, then the recommended search term corresponding to the input search term is determined based on the second candidate search term that meets the preset filtering conditions.

5. The method of claim 4, wherein, The step of determining whether the corresponding second candidate search term meets the preset filtering conditions based on the selected entity words includes: If the text length of the selected entity words is greater than 1, then determine whether the corresponding second candidate search term meets the preset filtering conditions based on the selected entity words.

6. The method of claim 1, wherein, The step of determining the recommended search term corresponding to the input search term based on the second candidate search term includes: Calculate the relevance score between the second candidate search term and the input search term; Based on the relevant score and the second candidate search term, the recommended search term corresponding to the input search term is determined.

7. A recommended search term determination apparatus characterized by comprising: include: The input search term acquisition module is used to acquire the user's input search terms; The intermediate category determination module is used to determine the intermediate category corresponding to the input search term based on the preset relationship between search terms and categories. The preset relationship between search terms and categories includes the correspondence between search terms and their corresponding categories at different levels. The intermediate category is a category located at the intermediate level in the category hierarchy structure. The recommended search term determination module is used to determine the recommended search term corresponding to the input search term based on the intermediate category; The clicked product determination module is used to determine the clicked product corresponding to the input search term based on the preset relationship between the search term and the product triple. The preset relationship between the search term and the product triple includes the correspondence between the search term and the clicked product. The intermediate category determination module for clicked products is used to determine the intermediate category corresponding to the clicked products for the input search term based on a preset product-category relationship. The preset product-category relationship includes the correspondence between the product and its corresponding categories at different levels. The recommended search term determination module includes: The target clicked product determination unit is used to determine the clicked product corresponding to the intermediate category of the input search term based on the intermediate category of the clicked product, and obtain the target clicked product. The second candidate search term determination unit is used to determine the search term that is not the input search term corresponding to the product that the target has been clicked, based on the relationship between the preset search term and the product triple, and to obtain the second candidate search term; The related product search term recall unit is used to determine the recommended search term corresponding to the input search term based on the second candidate search term.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor, when executing the computer program, implements the method according to any one of claims 1 to 6.

9. A computer program product, characterised in that, The computer program is executed so that the method according to any one of claims 1 to 6 is performed.

Images