Method and apparatus for processing search request

Abstract

The application discloses a search request processing method and device, and relates to the technical field of artificial intelligence identification and classification. A specific embodiment of the method comprises the following steps: receiving a search request input by a user, wherein the search request comprises a search keyword; determining whether the search keyword hits a pre-trained model, wherein the pre-trained model is obtained based on a search rule, and the search rule comprises an association relationship between a target keyword and a click result; in the case that the search keyword hits the pre-trained model, obtaining an identification result of the pre-trained model for the search keyword; and determining a return result corresponding to the search request according to the identification result, wherein the identification result comprises search behavior information. The embodiment can accurately determine the return result corresponding to the search request.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence recognition and classification technology, and in particular to a method and apparatus for processing search requests. Background Technology

[0002] With the development of the internet industry, online resources are becoming increasingly abundant, and information data is constantly growing. Users often use search services to find online resources of interest from a vast pool of information. Searching has gradually become a skill for people to learn about new things. Using search services to search for information on the internet has become an important way of acquiring information. Search methods have also become one of the key research focuses for researchers in this field.

[0003] However, most search engines use rule-based parsing for their retrieval logic, which often results in inaccurate results. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method and apparatus for processing search requests, which can accurately determine the return results of search requests.

[0005] In a first aspect, embodiments of the present invention provide a method for processing search requests, comprising:

[0006] Receive a search request input by a user, the search request including search keywords;

[0007] Determine whether the search keyword matches the pre-trained model, which is trained based on search rules, including the association between the target keyword and the click result;

[0008] If the search keyword matches the pre-trained model, the recognition result of the pre-trained model for the search keyword is obtained; based on the recognition result, the return result corresponding to the search request is determined, and the recognition result includes: search behavior information.

[0009] Optionally, before determining whether the search keyword matches the pre-trained model, the method further includes:

[0010] Obtain search click event information, which includes: historical keywords and click results;

[0011] Multiple target keywords are identified from the multiple historical keywords, and the click results corresponding to each target keyword are determined;

[0012] The search rules are generated based on the target keywords and the click results corresponding to the target keywords.

[0013] Optionally, determining multiple target keywords from the multiple historical keywords includes:

[0014] Determine the frequency of occurrence of each of the aforementioned historical keywords;

[0015] Based on the frequency of occurrence, multiple target keywords are determined from the multiple historical keywords.

[0016] Optionally, determining whether the search keyword matches the pre-trained model includes:

[0017] Obtain multiple target keywords from the search rules;

[0018] Calculate the distance value between each target keyword and the search keyword;

[0019] Determine if there is a matching keyword whose distance to the search keyword is less than a preset distance;

[0020] If it exists, then it is determined that the search keyword matches the pre-trained model;

[0021] If it does not exist, then it is determined that the search keyword did not match the pre-trained model.

[0022] Optionally, the recognition results of the pre-trained model for the search keywords include: multiple recommendation results and recommendation values ​​of the recommendation results;

[0023] The step of determining the return result corresponding to the search request based on the recognition result includes:

[0024] Obtain the search results corresponding to the search request;

[0025] The search results are sorted based on the multiple recommendation results and their recommendation values ​​to obtain the returned results.

[0026] Optionally, the recognition result of the pre-trained model for the search keywords includes: modification information;

[0027] After obtaining the recognition results of the pre-trained model for the search keywords, the method further includes:

[0028] Receive the modified text input by the user regarding the modified information;

[0029] Identify the modification keywords included in the modified text;

[0030] Determine whether the modified keywords match the pre-trained model.

[0031] Optionally, after determining whether the search keyword matches the pre-trained model, the method further includes:

[0032] If the search keyword does not match the pre-trained model, obtain the relevant recommendations from the pre-trained model for the search keyword;

[0033] Based on the relevant recommendations, determine the return results corresponding to the search request.

[0034] Optionally, obtaining the relevant recommendations of the pre-trained model for the search keywords includes:

[0035] Obtain multiple target keywords from the search rules;

[0036] Calculate the distance value between each target keyword and the search keyword;

[0037] Based on the distance value, relevant keywords corresponding to the search keyword are determined from multiple target keywords;

[0038] Based on the search rules, the relevant click results corresponding to the relevant keywords are determined, and the relevant click results are identified as relevant recommendations for the search keywords.

[0039] Optionally, determining the return results corresponding to the search request based on the relevant recommendations includes:

[0040] Obtain the search results corresponding to the search request;

[0041] Determine the recommendation value for each of the relevant recommendations;

[0042] The search results are sorted based on the relevant recommendations and their recommendation values ​​to obtain the returned results.

[0043] Secondly, embodiments of the present invention provide a search request processing apparatus, comprising:

[0044] The request receiving module is used to receive search requests input by the user, wherein the search requests include search keywords;

[0045] The model hit module is used to determine whether the search keyword hits the pre-trained model, which is trained based on search rules, including the association between the target keyword and the click result.

[0046] The result determination module is used to obtain the recognition result of the pre-trained model for the search keyword when the search keyword matches the pre-trained model; and to determine the return result corresponding to the search request based on the recognition result, wherein the recognition result includes: search behavior information.

[0047] Optionally, the device further includes:

[0048] The rule generation module is used to obtain search click event information, which includes: historical keywords and click results;

[0049] Multiple target keywords are identified from the multiple historical keywords, and the click results corresponding to each target keyword are determined;

[0050] The search rules are generated based on the target keywords and the click results corresponding to the target keywords.

[0051] Optionally, the rule generation module is specifically used for:

[0052] Determine the frequency of occurrence of each of the aforementioned historical keywords;

[0053] Based on the frequency of occurrence, multiple target keywords are determined from the multiple historical keywords.

[0054] Optionally, the model hit module is specifically used for:

[0055] Obtain multiple target keywords from the search rules;

[0056] Calculate the distance value between each target keyword and the search keyword;

[0057] Determine if there is a matching keyword whose distance to the search keyword is less than a preset distance;

[0058] If it exists, then it is determined that the search keyword matches the pre-trained model;

[0059] If it does not exist, then it is determined that the search keyword did not match the pre-trained model.

[0060] Optionally, the recognition results of the pre-trained model for the search keywords include: multiple recommendation results and recommendation values ​​of the recommendation results;

[0061] The result determination module is specifically used for:

[0062] Obtain the search results corresponding to the search request;

[0063] The search results are sorted based on the multiple recommendation results and their recommendation values ​​to obtain the returned results.

[0064] Thirdly, embodiments of the present invention provide an electronic device, including:

[0065] One or more processors;

[0066] Storage device for storing one or more programs.

[0067] When the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any of the above embodiments.

[0068] Fourthly, embodiments of the present invention provide a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the methods described in any of the above embodiments.

[0069] Fifthly, embodiments of the present invention provide a computer program product, including a computer program that, when executed by a processor, implements the methods described in any of the above embodiments.

[0070] One embodiment of the above invention has the following advantages or beneficial effects: it determines whether the search keywords in a search request match a pre-trained model, which is trained based on search rules. Search rules can reflect the characteristics of a user's historical search behavior, and can be determined through the user's historical search behavior. If the search keywords match the pre-trained model, it indicates the existence of historical search behaviors similar to the current search behavior. By referencing the pre-trained model's recognition results for search keywords, the returned results corresponding to the current search request can be accurately determined.

[0071] Furthermore, the solution in this embodiment of the invention determines the search behavior information corresponding to the search request, and then determines the return results corresponding to the search request based on the search behavior information, which makes user operation more convenient and improves the user experience. For example, users do not need to click to enter the product details page and scroll to the bottom of the page to perform operations such as purchasing or consulting. Users can directly perform operations such as purchasing or consulting based on the return page of the search request.

[0072] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0073] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein:

[0074] Figure 1 This is a flowchart illustrating a search request processing method provided in the first embodiment of the present invention;

[0075] Figure 2 This is a flowchart illustrating a search request processing method provided in the second embodiment of the present invention;

[0076] Figure 3 This is a flowchart illustrating a search request processing method provided in the third embodiment of the present invention;

[0077] Figure 4 This is a schematic diagram of a training method for a pre-trained model provided in the fourth embodiment of the present invention;

[0078] Figure 5 This is a flowchart illustrating a search request processing method provided in the fourth embodiment of the present invention;

[0079] Figure 6 This is a schematic diagram of the structure of a search request processing device provided in an embodiment of the present invention;

[0080] Figure 7 This is a schematic diagram of the structure of a computer system suitable for implementing terminal devices or servers of the present invention. Detailed Implementation

[0081] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0082] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0083] Figure 1 This is a flowchart illustrating a search request processing method provided in the first embodiment of the present invention, as shown below. Figure 1 As shown, the method includes:

[0084] Step 101: Receive the user's search request, which includes search keywords.

[0085] Search keywords are words used to characterize a user's search intent and search characteristics. Search text can be obtained from the search request, segmented into words, and search keywords can be determined based on the part-of-speech or weight of each segment. Alternatively, search keywords can be determined by comparing the search text with a pre-stored keyword database.

[0086] Step 102: Determine whether the search keywords match the pre-trained model. The pre-trained model is trained based on search rules, which include the correlation between the target keywords and the click results.

[0087] Search rules include the correlation between target keywords and click results; that is, by using search rules, click results with a high degree of relevance to target keywords can be identified. Search rules can reflect the characteristics of users' historical search behavior, and can be determined by analyzing users' historical search behavior. Users' historical search behavior includes: search text, search keywords, click results, favorites, adding to cart, following, purchasing, etc.

[0088] Click results may include information such as the item clicked by the user and the user's action. Different systems providing different services correspond to different click results. For e-commerce platforms, click results may include products sold on the platform, product favorites, adding products to the cart, and purchasing products. For search systems, click results may include web page links, opening links in new windows, adding to favorites, and inspecting elements. For banking systems, click results may include various financial products, financial product favorites, adding financial products to the cart, and purchasing financial products.

[0089] The pre-trained model is trained based on search rules. The model used in the pre-training can be set according to the actual situation and specific needs. The pre-trained model can use a language representation model or an attention model. Examples of language representation models may include, but are not limited to, BERT (Bidirectional Encoder Representation from Transformers) models, CNN models, DNN models, and RNN models.

[0090] Step 103: If the search keyword matches the pre-trained model, obtain the recognition result of the pre-trained model for the search keyword; based on the recognition result, determine the return result corresponding to the search request.

[0091] The identification results include search behavior information. These results are the output returned by the trained model for the search keywords. Search behavior information may include: favorites information, shopping cart information, purchase information, etc. There are many ways to determine the return results corresponding to a search request. For example, the search item information generated by the pre-trained model for the search keywords can be directly used as the return results. Alternatively, the search item information generated by the pre-trained model for the search keywords can be combined with the engine results determined by the search engine to obtain the return results. Another option is to remove the results that the user has clicked from the search item information generated by the pre-trained model for the search keywords and use that as the return results.

[0092] When the recognition results include search behavior information, the returned results corresponding to the search request can be determined based on the search behavior information. The returned results can correspond to multiple returned items, or they can correspond to an operation bar or floating bar, which is used to perform operations on the returned items. Specifically, the returned page includes a list of multiple returned items, and each returned item corresponds to an operation bar or floating bar. The content of the operation bar or floating bar can be set according to the search behavior information. For example, if the search behavior information includes "favorite" and "add to cart" behavior information, then buttons or icons corresponding to the "favorite" and "add to cart" behaviors can be set on the operation bar or floating bar to help users quickly operate on the target items.

[0093] In this embodiment of the invention, the search rules can reflect the characteristics of a user's historical search behavior, and the search rules can be determined through the user's historical search behavior. The pre-trained model is trained based on the search rules. It is determined whether the search keywords in the search request match the pre-trained model. If the search keywords match the pre-trained model, it indicates that there are historical search behaviors similar to the current search behavior. By referring to the recognition results of the search keywords by the pre-trained model, the return results corresponding to the current search request can be accurately determined.

[0094] Furthermore, the solution in this embodiment of the invention determines the search behavior information corresponding to the search request, and then determines the return results corresponding to the search request based on the search behavior information, which makes user operation more convenient and improves the user experience. For example, users do not need to click to enter the product details page and scroll to the bottom of the page to perform operations such as purchasing or consulting. Users can directly perform operations such as purchasing or consulting based on the return page of the search request.

[0095] Figure 2 This is a flowchart illustrating a search request processing method provided in the second embodiment of the present invention, as shown below. Figure 2 As shown, the method includes:

[0096] Step 201: Obtain search click event information, which includes historical keywords and click results.

[0097] Step 202: Identify multiple target keywords from a range of historical keywords and determine the click results for each target keyword.

[0098] This involves identifying multiple target keywords from a pool of historical keywords using various methods. For example, determining the frequency of each historical keyword; then identifying target keywords based on this frequency. Historical keywords with a frequency exceeding a preset frequency or the highest preset number of occurrences are selected as target keywords. Another method is to match multiple historical keywords with a preset keyword table to identify the historical keywords stored in the preset keyword table, which are then used as target keywords.

[0099] Step 203: Generate search rules based on the target keywords and the corresponding click results.

[0100] Search rules may include: target keywords, item information clicked by users corresponding to the target keywords, and user behavior information corresponding to the target keywords.

[0101] Step 204: Receive the user's search request, which includes search keywords.

[0102] Step 205: Determine whether the search keywords match the pre-trained model, which is trained based on search rules.

[0103] Step 206: If the search keyword matches the pre-trained model, obtain the recognition result of the pre-trained model for the search keyword; based on the recognition result, determine the return result corresponding to the search request.

[0104] The recognition results can include search behavior information, search item information, etc. Target keywords are determined from multiple historical keywords. Then, based on the target keywords and their corresponding click results, search rules are generated. These search rules are used to train a pre-trained model, enabling the model to more accurately determine the keyword recognition results.

[0105] In one embodiment of the present invention, determining whether a search keyword hits a pre-trained model includes: obtaining multiple target keywords in the search rules; calculating the distance value between each target keyword and the search keyword; determining whether there is a hit keyword whose distance value between it and the search keyword is less than a preset distance; if so, determining that the search keyword hits the pre-trained model; if not, determining that the search keyword does not hit the pre-trained model.

[0106] The cosine similarity, Minkowski distance, or Euclidean distance between the target keyword and the search keyword can be used as the distance value between them. If the distance value between the target keyword and the search keyword is less than a first preset distance, then the similarity between the target keyword and the search keyword is relatively high. In this case, the target keyword is taken as the hit keyword for the search keyword, and the search keyword is determined to hit the pre-trained model.

[0107] In one embodiment of the present invention, the pre-trained model's identification results for search keywords include: multiple recommended results and their recommendation values; based on the identification results, determining the return results corresponding to the search request includes: obtaining the search results corresponding to the search request; and sorting the search results according to the multiple recommended results and their recommendation values ​​to obtain the return results. The recommendation value characterizes the degree of matching between the recommended results and the search keywords. The higher the recommendation value, the higher the degree of matching between the recommended results and the search keywords. The lower the recommendation value, the lower the degree of matching between the recommended results and the search keywords. Based on the recommendation value, the search results are sorted so that search results with higher matching degrees are placed at the top of the return results for user convenience.

[0108] Figure 3 This is a flowchart illustrating a search request processing method provided in the third embodiment of the present invention, as shown below. Figure 3 As shown, the method includes:

[0109] Step 301: Receive the user's search request, which includes search keywords.

[0110] Step 302: Determine whether the search keywords match the pre-trained model. The pre-trained model is trained based on search rules, which include the relationship between the target keywords and the click results.

[0111] Step 303: Obtain the recognition results of the pre-trained model for the search keywords. The recognition results include: modification information.

[0112] If the identification result includes modification information, it means that the user's entered search keywords are incorrect or inaccurate, and the user needs to modify and supplement them.

[0113] Step 304: Receive the modified text input by the user regarding the modified information.

[0114] Step 305: Identify the keywords to be modified in the text; determine whether the keywords to be modified match the pre-trained model.

[0115] The modified keywords can be used to represent the user's search intent and search characteristics. The modified text can be segmented into words, and the modified keywords can be determined based on the part-of-speech or weight of each segment. Alternatively, the modified keywords can be determined by comparing the modified text with a pre-stored keyword database.

[0116] Step 306: If the modified keyword hits the pre-trained model, obtain the recognition result of the pre-trained model for the modified keyword; based on the recognition result, determine the return result corresponding to the search request.

[0117] In this embodiment of the invention, if the search keywords entered by the user are incorrect or inaccurate, the recognition result output by the pre-trained model will include modification information. The system receives the modified text input by the user regarding the modification information and determines the return result corresponding to the search request based on the modified keywords in the modified text.

[0118] In one embodiment of the present invention, after determining whether the search keyword matches the pre-trained model, the method further includes: if the search keyword does not match the pre-trained model, obtaining relevant recommendations from the pre-trained model for the search keyword; and determining the return result corresponding to the search request based on the relevant recommendations.

[0119] The relevant recommendations of the pre-trained model for the search keyword can be obtained in the following way: obtaining multiple target keywords in the search rules; calculating the distance value between each target keyword and the search keyword; determining the relevant keywords corresponding to the search keyword from the multiple target keywords based on the distance value; determining the relevant click results corresponding to the relevant keywords according to the search rules, and determining the relevant click results as the relevant recommendations of the search keyword.

[0120] The distance between the target keyword and the search keyword can be calculated using methods such as cosine similarity, Minkowski distance, or Euclidean distance. If the distance between the target keyword and the search keyword is less than a second preset distance, the similarity between the two keywords is considered high, and the target keyword is considered a related keyword. The distance between related keywords and the target keyword is usually less than the distance between the hit keyword and the target keyword.

[0121] In one embodiment of the present invention, determining the return result corresponding to the search request based on the relevant recommendations includes: obtaining the search results corresponding to the search request; determining the recommendation value of each of the relevant recommendations; and sorting the search results according to the relevant recommendations and their recommendation values ​​to obtain the return result. The recommendation value characterizes the degree of matching between the recommendation result and the search keyword. The higher the recommendation value, the higher the degree of matching between the recommendation result and the search keyword. The lower the recommendation value, the lower the degree of matching between the recommendation result and the search keyword. The search results are sorted according to the recommendation value so that search results with higher matching degrees are placed at the top of the return results. Alternatively, search behaviors with higher matching degrees can be placed at the top of the operation bar or floating bar to facilitate user selection.

[0122] To make the method of this invention easier to understand, a specific search scheme is explained below. The scheme of this invention includes three parts: 1. Extracting rules and generating a pre-trained model by collecting search click event logs; 2. Searches that match the pre-trained model; 3. Recommending related keywords that do not match the pre-trained model.

[0123] Figure 4 This is a schematic diagram of a training method for a pre-trained model provided in the fourth embodiment of the present invention. Figure 4 As shown, the specific steps for extracting rules and generating a pre-trained model from search click event logs collected through data collection are as follows:

[0124] Step S11: Collect click events of customer search results and record search keywords and search click results.

[0125] Step S12: The log analysis platform analyzes and statistically analyzes search keywords and search click results.

[0126] Step S13: View the ranking of search keywords entered by the customer and the corresponding search results through the backend.

[0127] Step S14: Analyze the correlation between search keywords and search click results, and extract rules.

[0128] Step S15: For search keywords, use deep learning to calculate and form a pre-trained model offline, and then deploy it online.

[0129] Figure 5 This is a flowchart illustrating a search request processing method provided in the fourth embodiment of the present invention. Figure 5 As shown, the specific steps for searching the pre-trained model are as follows:

[0130] Step S21: The intelligent search receives the customer's search request and proceeds to step S22.

[0131] Step S22: Intelligent search first uses the search terms to match the pre-trained model. For those with a matching degree of more than 90%, they are considered to be hits, and proceed to step S23.

[0132] Step S23: Intelligent search reads the rules that hit the pre-trained model, and then proceeds to steps S24 and S25.

[0133] Step S24: The rules of the pre-trained model that the intelligent search hits. If the rules modify the search terms, it means that the search terms entered by the customer are incorrect or inaccurate and need to be actively modified and supplemented. Proceed to step S25.

[0134] Step S25: If the rules of the pre-trained model that the intelligent search hit contain recommended results, the intelligent search needs to weight the recommended results and display them first, then proceed to step S26.

[0135] Step S26: The intelligent search sorts the search results according to the recognition results of the pre-trained model, obtains the returned results, and proceeds to step S27.

[0136] Step S27: The intelligent search will return the results to the customer.

[0137] like Figure 5 As shown, the specific steps for making related recommendations for search keywords that do not match the pre-trained model are as follows:

[0138] Step S31: For search terms that are not matched by the pre-trained model, the pre-trained model will provide some similar recommendations.

[0139] Step S32: View relevant recommendations for pre-trained models through the backend.

[0140] Step S33: Utilize deep learning to generate a pre-trained model offline and then deploy it online.

[0141] Figure 6 This is a schematic diagram of the structure of a search request processing device provided in one embodiment of the present invention, as shown below. Figure 6 As shown, the device includes:

[0142] The request receiving module 601 is used to receive a search request input by a user, wherein the search request includes search keywords;

[0143] The model hit module 602 is used to determine whether the search keyword hits the pre-trained model, which is trained based on search rules, including the association between the target keyword and the click result.

[0144] The result determination module 603 is used to obtain the recognition result of the pre-trained model for the search keyword when the search keyword matches the pre-trained model; and to determine the return result corresponding to the search request based on the recognition result.

[0145] Optionally, the device further includes:

[0146] Rule generation module 604 is used to obtain search click event information, which includes: historical keywords and click results;

[0147] Multiple target keywords are identified from the multiple historical keywords, and the click results corresponding to each target keyword are determined;

[0148] The search rules are generated based on the target keywords and the click results corresponding to the target keywords.

[0149] Optionally, the rule generation module 604 is specifically used for:

[0150] Determine the frequency of occurrence of each of the aforementioned historical keywords;

[0151] Based on the frequency of occurrence, multiple target keywords are determined from the multiple historical keywords.

[0152] Optionally, the model hit module 602 is specifically used for:

[0153] Obtain multiple target keywords from the search rules;

[0154] Calculate the distance value between each target keyword and the search keyword;

[0155] Determine if there is a matching keyword whose distance to the search keyword is less than a preset distance;

[0156] If it exists, then it is determined that the search keyword matches the pre-trained model;

[0157] If it does not exist, then it is determined that the search keyword did not match the pre-trained model.

[0158] Optionally, the recognition results of the pre-trained model for the search keywords include: multiple recommendation results and recommendation values ​​of the recommendation results;

[0159] The result determination module 603 is specifically used for:

[0160] Obtain the search results corresponding to the search request;

[0161] The search results are sorted based on the multiple recommendation results and their recommendation values ​​to obtain the returned results.

[0162] Optionally, the recognition result of the pre-trained model for the search keywords includes: modification information;

[0163] The device also includes:

[0164] Modification module 605 is used to receive modified text input by the user regarding the modification information;

[0165] Identify the modification keywords included in the modified text;

[0166] Determine whether the modified keywords match the pre-trained model.

[0167] Optionally, the result determination module 603 is further configured to obtain relevant recommendations from the pre-trained model for the search keyword if the search keyword does not match the pre-trained model.

[0168] Based on the relevant recommendations, determine the return results corresponding to the search request.

[0169] Optionally, the result determination module 603 is specifically used for:

[0170] Obtain multiple target keywords from the search rules;

[0171] Calculate the distance value between each target keyword and the search keyword;

[0172] Based on the distance value, relevant keywords corresponding to the search keyword are determined from multiple target keywords;

[0173] Based on the search rules, the relevant click results corresponding to the relevant keywords are determined, and the relevant click results are identified as relevant recommendations for the search keywords.

[0174] Optionally, the result determination module 603 is specifically used for:

[0175] Obtain the search results corresponding to the search request;

[0176] Determine the recommendation value for each of the relevant recommendations;

[0177] The search results are sorted based on the relevant recommendations and their recommendation values ​​to obtain the returned results.

[0178] This invention provides an electronic device, comprising:

[0179] One or more processors;

[0180] Storage device for storing one or more programs.

[0181] When one or more programs are executed by one or more processors, the one or more processors implement the methods of any of the above embodiments.

[0182] This invention provides a computer program product, including a computer program that, when executed by a processor, implements the search request processing method of this invention.

[0183] The following is for reference. Figure 7 It shows a schematic diagram of the structure of a computer system 700 suitable for implementing a terminal device of the present invention. Figure 7The terminal device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of the present invention.

[0184] like Figure 7 As shown, the computer system 700 includes a central processing unit (CPU) 701, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 702 or programs loaded from storage section 708 into random access memory (RAM) 703. The RAM 703 also stores various programs and data required for the operation of the system 700. The CPU 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0185] The following components are connected to the I / O interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0186] In particular, according to the embodiments disclosed in this invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 709, and / or installed from removable medium 711. When the computer program is executed by central processing unit (CPU) 701, it performs the functions defined above in the system of this invention.

[0187] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0188] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0189] The modules described in the embodiments of the present invention can be implemented in software or hardware. The described modules can also be located in a processor, for example, they can be described as: a request receiving module, a model hitting module, and a result determining module. The names of these modules do not necessarily limit the module itself; for example, the request receiving module can also be described as "a module that receives search requests input by the user, the search requests including search keywords."

[0190] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to include:

[0191] Receive a search request input by a user, the search request including search keywords;

[0192] Determine whether the search keyword matches the pre-trained model, which is trained based on search rules, including the association between the target keyword and the click result;

[0193] If the search keyword matches the pre-trained model, the recognition result of the pre-trained model for the search keyword is obtained; based on the recognition result, the return result corresponding to the search request is determined, and the recognition result includes: search behavior information.

[0194] According to the technical solution of this invention, it is determined whether the search keywords in the search request match a pre-trained model, where the pre-trained model is trained based on search rules. Search rules can reflect the characteristics of a user's historical search behavior, and can be determined through the user's historical search behavior. If the search keywords match the pre-trained model, it indicates the existence of historical search behaviors similar to the current search behavior. By referencing the pre-trained model's recognition results for search keywords, the returned results corresponding to the current search request can be accurately determined.

[0195] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method of processing a search request, characterized by, include: Receive a search request input by a user, the search request including search keywords; Determining whether the search keyword matches the pre-trained model includes: obtaining multiple target keywords in the search rules; calculating the distance value between each target keyword and the search keyword; determining whether there is a matching keyword whose distance value between it and the search keyword is less than a preset distance; if so, determining that the search keyword matches the pre-trained model; if not, determining that the search keyword does not match the pre-trained model. The pre-trained model is trained based on search rules, which include the correlation between target keywords and click results; If the search keyword matches the pre-trained model, obtain the recognition result of the pre-trained model for the search keyword; determine the return result corresponding to the search request based on the recognition result, wherein the recognition result includes: search behavior information; The returned page of the results includes a list of multiple returned items, each of which corresponds to an operation bar or floating bar. The content of the operation bar or floating bar is set according to the search behavior information. It also includes: collecting click events from customer search results, recording search keywords and search click results; analyzing and statistically analyzing search keywords and search click results through a log analysis platform; viewing the ranking of search keywords entered by customers and the corresponding related search results through the backend; analyzing the correlation between search keywords and search click results and extracting rules; and using deep learning to form pre-trained models offline and deploying them online for search keywords.

2. The method according to claim 1, characterized in that, Before determining whether the search keyword matches the pre-trained model, the method further includes: Obtain search click event information, which includes: historical keywords and click results; Multiple target keywords are identified from the multiple historical keywords, and the click results corresponding to each target keyword are determined; The search rules are generated based on the target keywords and the click results corresponding to the target keywords.

3. The method according to claim 2, characterized in that, The determination of multiple target keywords from the multiple historical keywords includes: Determine the frequency of occurrence of each of the aforementioned historical keywords; Based on the frequency of occurrence, multiple target keywords are determined from the multiple historical keywords.

4. The method according to claim 1, characterized in that, The recognition results of the pre-trained model for the search keywords include: multiple recommendation results and the recommendation values ​​of the recommendation results; The step of determining the return result corresponding to the search request based on the recognition result includes: Obtain the search results corresponding to the search request; The search results are sorted based on the multiple recommendation results and their recommendation values ​​to obtain the returned results.

5. The method according to claim 1, characterized in that, The pre-trained model's recognition results for the search keywords include: modification information; After obtaining the recognition results of the pre-trained model for the search keywords, the method further includes: Receive the modified text input by the user regarding the modified information; Identify the modification keywords included in the modified text; Determine whether the modified keywords match the pre-trained model.

6. The method according to claim 1, characterized in that, After determining whether the search keyword matches the pre-trained model, the process further includes: If the search keyword does not match the pre-trained model, obtain the relevant recommendations from the pre-trained model for the search keyword; Based on the relevant recommendations, determine the return results corresponding to the search request.

7. The method according to claim 6, characterized in that, The step of obtaining relevant recommendations from the pre-trained model for the search keywords includes: Obtain multiple target keywords from the search rules; Calculate the distance value between each target keyword and the search keyword; Based on the distance value, relevant keywords corresponding to the search keyword are determined from multiple target keywords; Based on the search rules, the relevant click results corresponding to the relevant keywords are determined, and the relevant click results are identified as relevant recommendations for the search keywords.

8. The method according to claim 6, characterized in that, The step of determining the return result corresponding to the search request based on the relevant recommendations includes: Obtain the search results corresponding to the search request; Determine the recommendation value for each of the relevant recommendations; The search results are sorted based on the relevant recommendations and their recommendation values ​​to obtain the returned results.

9. A search request processing apparatus, characterized in that, include: The request receiving module is used to receive search requests input by the user, wherein the search requests include search keywords; The model hit module is used to determine whether the search keyword hits the pre-trained model, which is trained based on search rules, including the association between the target keyword and the click result. The result determination module is used to obtain the recognition result of the pre-trained model for the search keyword when the search keyword matches the pre-trained model; Based on the identification results, the return results corresponding to the search request are determined. The identification results include: search behavior information. The return page of the return results includes a list of multiple returned items. Each returned item corresponds to an operation bar or floating bar. The content of the operation bar or floating bar is set according to the search behavior information. The model hit module is specifically used for: acquiring multiple target keywords in the search rules; calculating the distance value between each target keyword and the search keyword; determining whether there is a hit keyword whose distance value between it and the search keyword is less than a preset distance; if so, determining that the search keyword hits the pre-trained model; if not, determining that the search keyword does not hit the pre-trained model. The model hit module is specifically used for: collecting click events of customer search results, recording search keywords and search click results; analyzing and statistically analyzing search keywords and search click results through the log analysis platform; viewing the ranking of search keywords entered by customers and the corresponding related search results through the backend; analyzing the correlation between search keywords and search click results and extracting rules; and using deep learning to offline compute and form a pre-trained model for search keywords, and then deploying it online.

10. The apparatus according to claim 9, characterized in that, The device also includes: The rule generation module is used to obtain search click event information, which includes: historical keywords and click results; Multiple target keywords are identified from the multiple historical keywords, and the click results corresponding to each target keyword are determined; The search rules are generated based on the target keywords and the click results corresponding to the target keywords.

11. The apparatus according to claim 10, characterized in that, The rule generation module is specifically used for: Determine the frequency of occurrence of each of the aforementioned historical keywords; Based on the frequency of occurrence, multiple target keywords are determined from the multiple historical keywords.

12. The apparatus according to claim 9, characterized in that, The recognition results of the pre-trained model for the search keywords include: multiple recommendation results and the recommendation values ​​of the recommendation results; The result determination module is specifically used for: Obtain the search results corresponding to the search request; The search results are sorted based on the multiple recommendation results and their recommendation values ​​to obtain the returned results.

13. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-8.

14. A computer-readable medium having a computer program stored thereon, characterized in that... When the program is executed by the processor, it implements the method as described in any one of claims 1-8.

15. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-8.

Images