Search processing method, device, electronic device and storage medium

Abstract

The present disclosure relates to a search processing method, device, electronic device and storage medium, including: obtaining multiple search terms and the search time of each search term; dividing the search operation corresponding to the search term into multiple intent slices according to the search time and feature information of the search term; detecting the operation information in the search result page of the search operation corresponding to each search term, and in the case of detecting the preset operation information, determining the corresponding search operation as the target search operation, and determining the search step length according to the number of target search operations and the search order of the target search operations in the corresponding intent slices. In this way, the intent slice method is adopted to define a continuous behavior with the same user intention, which is closer to the actual understanding of the user behavior, can improve the accuracy of the search processing results, and meet more complex search scenarios.

Description

Technical Field

[0001] The present disclosure relates to the field of data search, and in particular to a search processing method, device, electronic device and storage medium. Background Art

[0002] In the website's search business, the core demand of users is often "to find results as quickly as possible". Therefore, the website needs to count the users' search efficiency, attribute the current inefficient search scenarios, and further break down the inefficiency issues to optimize the website's search efficiency.

[0003] In the prior art, search efficiency is usually determined based on the point ratio of the first search result on the comprehensive page and the first click position on the specific content vertical search page, wherein the point ratio of the first search result on the comprehensive page refers to the probability of a click on the first search result when at least one search result is exposed on the comprehensive page, and the first click position on the specific content vertical search page refers to the position on average where user result clicks occur on the specific content vertical search page.

[0004] However, the above method of determining search efficiency is relatively isolated and only considers the efficiency of a single search. In some cases, if the user does not get satisfactory search results in the first search, until a satisfactory search result is found in the second search term with a similar intention, this is actually a relatively negative experience, but it cannot be reflected in the current measurement system. Therefore, the accuracy of the search processing results obtained by the current search processing method is limited and it is difficult to meet more complex search scenarios. Summary of the invention

[0005] The present disclosure provides a search processing method, device, electronic device and storage medium to at least solve the problem that the accuracy of search processing results in related technologies is limited and it is difficult to meet the needs of more complex search scenarios. The technical solutions of the present disclosure are as follows:

[0006] According to a first aspect of an embodiment of the present disclosure, a search processing method is provided, including:

[0007] Get multiple search terms and the search time of each search term;

[0008] Divide the search operation corresponding to the search term into multiple intent slices according to the search time and feature information of the search term;

[0009] Detect the operation information in the search results page of the search operation corresponding to each search term. When preset operation information is detected, determine the corresponding search operation as the target search operation, and determine the search step size based on the number of the target search operations and the search order of the target search operations in the intent slice to which they belong.

[0010] Optionally, dividing the search operation corresponding to the search term into multiple intent slices according to the search time and feature information of the search term includes:

[0011] For each search term, calculate the difference between the search time of the current search term and the search time of the previous search term;

[0012] According to the feature information of the current search term and the difference, determining whether the current search term and a previous search term of the current search term meet a matching condition;

[0013] If satisfied, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

[0014] Optionally, judging whether the current search term and a previous search term of the current search term satisfy a matching condition according to the feature information of the current search term and the difference value includes:

[0015] The feature information of the current search term and the difference are input into a pre-trained intention slice classification model for processing, and a determination is made based on the classification result as to whether the current search term and the previous search term of the current search term meet a matching condition.

[0016] Optionally, the feature information of the current search term includes at least one of the following:

[0017] The kernel density estimate of the current search term;

[0018] The cosine similarity between the current search term and the previous search term of the current search term;

[0019] The cosine similarity between the current search term and the next search term of the current search term;

[0020] The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the previous search word;

[0021] The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the current search word.

[0022] Optionally, the preset operation information includes at least one of the following:

[0023] Added follow operation;

[0024] Click operations on followed users;

[0025] Browsing information of any user.

[0026] Optionally, determining the search step length according to the number of the target search operations and the search order of the target search operations in the corresponding intent slices includes:

[0027] Obtain the number of searches required to search for the target search operation in each belonging intent slice, and take the sum of the number of searches as the first quantity;

[0028] determining the number of intent slices including the target search operation as a second number;

[0029] The ratio of the first number to the second number is used as the search step length.

[0030] Optionally, after determining the search step length according to the number of the target search operations and the search order of the target search operations in the corresponding intent slice, the method further includes:

[0031] Calculate a search adjustment parameter according to the search results of the search operation included in the intent slice to which the target search operation belongs;

[0032] According to the search adjustment parameters, attribution analysis is performed on the target search operation to obtain an attribution analysis result.

[0033] Optionally, the search adjustment parameters include one or more of the following:

[0034] The probability that the search result page of the target search operation includes search type switching operation information;

[0035] The probability that the target search operation has a different search term from a previous search operation of the target search operation.

[0036] According to a second aspect of an embodiment of the present disclosure, a search processing device is provided, including:

[0037] An acquisition unit is configured to acquire a plurality of search terms and a search time of each search term;

[0038] a grouping unit, configured to divide the search operation corresponding to the search term into a plurality of intent slices according to the search time and feature information of the search term;

[0039] The determination unit is configured to execute detection of operation information in the search results page of the search operation corresponding to each search term, and when preset operation information is detected, the corresponding search operation is determined as a target search operation, and the search step is determined according to the number of the target search operations and the search order of the target search operations in the corresponding intent slice.

[0040] Optionally, the grouping unit is specifically configured to execute:

[0041] For each search term, calculate the difference between the search time of the current search term and the search time of the previous search term;

[0042] According to the feature information of the current search term and the difference, determining whether the current search term and a previous search term of the current search term meet a matching condition;

[0043] If satisfied, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

[0044] Optionally, the grouping unit is specifically configured to execute:

[0045] The feature information of the current search term and the difference are input into a pre-trained intention slice classification model for processing, and a determination is made based on the classification result as to whether the current search term and the previous search term of the current search term meet a matching condition.

[0046] Optionally, the feature information of the current search term includes at least one of the following:

[0047] The kernel density estimate of the current search term;

[0048] The cosine similarity between the current search term and the previous search term of the current search term;

[0049] The cosine similarity between the current search term and the next search term of the current search term;

[0050] The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the previous search word;

[0051] The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the current search word.

[0052] Optionally, the preset operation information includes one or more of the following:

[0053] Added follow operation;

[0054] Click operations on followed users;

[0055] Browsing information of any user.

[0056] Optionally, the determining unit is configured to execute:

[0057] Obtain the number of searches required to search for the target search operation in each belonging intent slice, and take the sum of the number of searches as the first quantity;

[0058] determining the number of intent slices including the target search operation as a second number;

[0059] The ratio of the first number to the second number is used as the search step length.

[0060] Optionally, the determining unit is further configured to execute:

[0061] Calculate a search adjustment parameter according to the search results of the search operation included in the intent slice to which the target search operation belongs;

[0062] According to the search adjustment parameters, attribution analysis is performed on the target search operation to obtain an attribution analysis result.

[0063] Optionally, the search adjustment parameters include one or more of the following:

[0064] The probability that the search result page of the target search operation includes search type switching operation information;

[0065] The probability that the target search operation has a different search term from a previous search operation of the target search operation.

[0066] According to a third aspect of an embodiment of the present disclosure, there is provided a search processing electronic device, including:

[0067] processor;

[0068] a memory for storing instructions executable by the processor;

[0069] Wherein, the processor is configured to execute the instructions to implement the search processing method described in the first item above.

[0070] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of a search processing electronic device, the search processing electronic device is enabled to perform any of the search processing methods described above.

[0071] According to a fifth aspect of an embodiment of the present disclosure, a computer program product is provided, comprising a computer program / instruction, wherein when the computer program / instruction is executed by a processor, the search processing method described in the first item above is implemented.

[0072] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0073] Obtain multiple search terms and the search time of each search term; divide the search operations corresponding to the search terms into multiple intent slices based on the search time and feature information of the search terms; detect the operation information in the search results page of the search operation corresponding to each search term, and when preset operation information is detected, determine the corresponding search operation as the target search operation, and determine the search step size based on the number of target search operations and the search order of the target search operations in the intent slice to which they belong.

[0074] In this way, we can connect the search terms of each search and consider the user intentions represented by similar search terms in multiple search scenarios. That is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if he changes to a search term with similar intention and finds satisfactory search results the second time, then these two searches belong to the same intent slice. Therefore, using intent slicing to define a continuous behavior with the same user intention is closer to reality for understanding user behavior, which can improve the accuracy of search processing results and meet more complex search scenarios.

[0075] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0077] Figure 1 The figure is a flowchart of a search processing method according to an exemplary embodiment.

[0078] Figure 2 is a flowchart showing another search processing method according to an exemplary embodiment.

[0079] Figure 3 It is a logical schematic diagram of a training intent slice classification model according to an exemplary embodiment.

[0080] Figure 4 It is a schematic diagram of intention slice distribution with different step lengths for finding a person according to an exemplary embodiment.

[0081] Figure 5 The figure is a schematic diagram showing several operations when a user is satisfied with the search results according to an exemplary embodiment.

[0082] Figure 6 The figure is a schematic diagram of search efficiency attribution according to an exemplary embodiment.

[0083] Figure 7 It is a block diagram of a search processing device according to an exemplary embodiment.

[0084] Figure 8 It is a block diagram of an electronic device for search processing according to an exemplary embodiment.

[0085] Fig. 9 The invention is a block diagram of a device for search processing according to an exemplary embodiment. DETAILED DESCRIPTION

[0086] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0087] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0088] Figure 1 is a flowchart of a search processing method according to an exemplary embodiment. Figure 1 As shown, the search processing method includes the following steps.

[0089] In step S11, a plurality of search words and the search time of each search word are obtained.

[0090] The commonly performed search process is to obtain the search term input by the user, perform a search operation on the search term in the database, that is, retrieve the data matching the search term as the search result, and generate a corresponding search result page to display it to the user. In the present disclosure, the search term and search time of each search can be recorded to obtain a data source for intent slicing, thereby realizing the subsequent grouping of intent slices for the search operation. In step S12, the search operation corresponding to the search term is divided into multiple intent slices according to the search time and feature information of the search term.

[0091] Among them, intent slice refers to the combination of multiple searches for the same intent within a period of time. In other words, if several consecutive search terms include the same or similar content, then the search operations corresponding to these search terms can be divided into the same intent slice. In this way, when calculating the efficiency of several searches in the future, the intent slices are used as groups for statistics, and the same or similar searches can be comprehensively judged, so as to have a more accurate evaluation of the search efficiency.

[0092] In this step, the search operations corresponding to the search terms are divided into multiple intent slices, which may specifically include the following steps: for each search term, calculate the difference between the search time of the current search term and the search time of the previous search term; based on the feature information and difference of the current search term, determine whether the current search term and the previous search term of the current search term meet the matching conditions; if so, add the search operation corresponding to the current search term to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

[0093] That is to say, whether the current search term and the previous search term with a short time interval meet the matching conditions can be used to determine whether the current search term can be divided into the same intent slice as the previous search term. It can be understood that two search operations with a long time interval may be caused by unexpected interruption of the previous search operation. It is unreasonable to divide it and the operation corresponding to the current search term into the same intent slice for subsequent search efficiency analysis.

[0094] For example, the feature information of the current search term may include one or more of the following: the kernel density estimate of the current search term; the cosine similarity between the current search term and the previous search term; the cosine similarity between the current search term and the next search term; the ratio of the maximum sequence length between the previous search term and the next search term to the sequence length of the previous search term; the ratio of the maximum sequence length between the previous search term and the next search term to the sequence length of the search term; etc. These feature information are helpful in judging the relevance between the current search term and the previous and next search terms, thereby realizing the division of the intent of the search operation and obtaining accurate intent slicing and grouping results.

[0095] Among them, the cosine similarity between the current search term and the previous search term, and the cosine similarity between the current search term and the next search term are both text cosine similarities, that is, the distance between the two search terms. When calculating, the search terms can be first converted into embedding vectors, and then the cosine distance between the embedding vectors of the two search terms can be further calculated.

[0096] In step S13, the operation information in the search results page of the search operation corresponding to each search term is detected. When the preset operation information is detected, the corresponding search operation is determined as the target search operation. The search step is determined according to the number of target search operations and the search order of the target search operations in the corresponding intent slice.

[0097] In this step, if the operation information in the search results page includes preset operation information, then it can be considered that the search results of this search operation include results that meet the user's needs. Then, based on the search order of the target search operation in each intent slice, it can be determined how many times the user has obtained satisfactory search results, and then the search step can be determined. The search step can be used to measure the efficiency of the search.

[0098] For example, the preset operation information may include one or more of the following: a new follow-up operation, a click operation on a user that has been followed, and an information browsing operation on any user, etc. Different preset operation information may correspond to different user search needs. For example, a new follow-up operation may meet the user's need to make friends, a click operation on a user that has been followed may meet the user's need to find a person that the user has previously followed, and an information browsing operation on any user may meet the user's need to consume the works of a specific person, etc. Other preset operation information may also be determined according to user needs, and the specifics are not limited thereto.

[0099] In one implementation, the step of determining the search step may specifically include: obtaining the number of searches required to search for the target search operation in each intent slice, and taking the sum of the number of searches as the first number; determining the number of intent slices including the target search operation as the second number; and taking the ratio of the first number to the second number as the search step. It can be understood that the search step can express how many times it takes to achieve the goal of the search results meeting the user's needs, reflects the cost of the user obtaining the search results that meet the user's needs, and is a negative indicator that needs to be reduced.

[0100] In the present disclosure, the search step length is used to measure the search efficiency. On the one hand, it has better directionality. For example, if the user switches to vertical search or changes the search term to search again, it is considered that the comprehensive page does not meet the user's search needs, which is a negative experience. The search step length will also increase, but the original indicator cannot reflect this negative experience. On the other hand, it has better sensitivity. For example, compared with satisfying the search needs at the end of each page, it is considered that the situation where the search needs are not met on each page can better reflect the user's dissatisfaction. Therefore, the search step length indicator has better sensitivity than the original indicator.

[0101] From the above, it can be seen that in this solution, the search terms of each search can be linked to consider the user intentions represented by similar search terms in multiple search scenarios, that is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if a satisfactory search result is found by changing the search term with similar intention for the second time, then these two searches belong to the same intent slice. Therefore, the intent slice method is used to define a continuous behavior with the same user intention. The understanding of user behavior is closer to reality, which can improve the accuracy of search processing results and meet more complex search scenarios.

[0102] Figure 2 is a flowchart of another search processing method according to an exemplary embodiment. Figure 2 As shown, the search processing method includes the following steps.

[0103] In step S21, a plurality of search words and the search time of each search word are obtained.

[0104] This step is consistent with step S11 in the above embodiment and will not be described again here.

[0105] In step S22, for each search word, the difference between the search time of the current search word and the search time of the previous search word is calculated.

[0106] Intent slicing refers to the combination of multiple searches for the same intent within a period of time. In this step, it is necessary to calculate the difference between the search time of the current search term and the search time of the previous search term, and then determine whether the current search term can be divided into the same intent slice as the previous search term. It is understandable that two search operations with a relatively long time interval may be caused by an unexpected interruption of the previous search operation. It is unreasonable to divide it and the operation corresponding to the current search term into the same intent slice for subsequent search efficiency analysis.

[0107] In step S23, based on the feature information and difference of the current search term, it is determined whether the current search term and the previous search term of the current search term meet the matching condition; if so, step S24 is executed.

[0108] Among them, based on the feature information and difference of the current search term, judging whether the current search term and the previous search term of the current search term meet the matching conditions, specifically can include: inputting the feature information and difference of the current search term into the pre-trained intent slice classification model for processing, and judging whether the current search term and the previous search term of the current search term meet the matching conditions based on the classification result.

[0109] The pre-trained intent slice classification model can be a GBDT (Gradient Boosting Decision Tree) model, an XGBoost (Extreme Gradient Boosting) model, or any technical sequence labeling model commonly used in natural language processing, such as a BILSTM (Bi-directional Long Short-Term Memory) series CRF (conditional random field algorithm) model, etc., without specific limitation.

[0110] By using the pre-trained intent slice classification model, the intent slices are divided according to the feature information of the search term and the difference in search time with the previous search term. Compared with the division method that simply relies on the similarity between search terms, it is more in line with the recognition of search intent and the division results are more accurate.

[0111] For example, the feature information of the current search term may include one or more of the following: a kernel density estimate of the current search term; cosine similarity between the current search term and the previous search term; cosine similarity between the current search term and the next search term; ratio of the maximum sequence length between the previous search term and the next search term to the sequence length of the previous search term; ratio of the maximum sequence length between the previous search term and the next search term to the sequence length of the search term; and the like.

[0112] In step S24, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

[0113] That is to say, if several consecutive search terms include the same or similar content, then the search operations corresponding to these search terms can be grouped into the same intent slice. In this way, when calculating the efficiency of several searches in the future, the intent slices can be used as groups for statistics, and the same or similar searches can be comprehensively judged, thereby having a more accurate evaluation of the search efficiency.

[0114] In step S25, operation information in the search result page of the search operation corresponding to each search word is detected, and when preset operation information is detected, the corresponding search operation is determined as a target search operation.

[0115] If the operation information in the search results page includes preset operation information, then it can be considered that the search results of this search operation include results that meet the user's needs. Then, according to the search order of the target search operation in each intent slice, it can be determined how many times the user has obtained satisfactory search results, and then the search step can be determined. The search step can be used to measure the efficiency of the search. For example, the preset operation information can include one or more of the following: adding a new follow operation, clicking on a user who has been followed, and browsing information for any user, etc.

[0116] In step S26, the number of searches required to search for the target search operation in each intended slice is obtained, and the sum of the number of searches is taken as the first number; the number of intended slices including the target search operation is determined as the second number; and the ratio of the first number to the second number is taken as the search step.

[0117] This step can be expressed as:

[0118] Search step length = sum (number of searches required to find the target search operation in each intent slice) / number of intent slices enclosing the target search operation

[0119] It can be understood that the search step size can express how many times it is necessary to search to achieve the goal of satisfying user needs with search results. It reflects the cost for users to obtain search results that meet their needs and is a negative indicator that needs to be reduced.

[0120] In step S27, the search adjustment parameters are calculated based on the search results of the search operation included in the intent slice to which the target search operation belongs; and the attribution analysis is performed on the target search operation based on the search adjustment parameters to obtain the attribution analysis results.

[0121] That is, the effect of each search can be further attributed according to the search adjustment parameters, and the problems that need to be solved can be quickly determined according to the attribution analysis results, which is helpful for developers to determine the direction of improving the search algorithm and further improve the efficiency of subsequent searches. Among them, the search adjustment parameters may include one or more of the following: the probability that the search result page of the target search operation includes search type switching operation information; the probability that the search term of the target search operation is different from the search term of the previous search operation of the target search operation.

[0122] Different search adjustment parameters can reflect different problems with search results. For example, if the search result page of the target search operation has a high probability of including search type switching operation information, then it means that for the same search term, the search results corresponding to different search types are quite different, and the search results obtained by the switched search type are more in line with user needs, and the search method of the search type before switching needs to be improved, etc. Search adjustment parameters can be set according to different business needs, and there is no specific limitation.

[0123] From the above, it can be seen that in this solution, the search terms of each search can be linked to consider the user intentions represented by similar search terms in multiple search scenarios, that is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if a satisfactory search result is found by changing the search term with similar intention for the second time, then these two searches belong to the same intent slice. Therefore, the intent slice method is used to define a continuous behavior with the same user intention. The understanding of user behavior is closer to reality, which can improve the accuracy of search processing results and meet more complex search scenarios.

[0124] The search processing method provided by the present disclosure is described below through a specific implementation.

[0125] In the first stage, before performing intent slicing on the search operation based on multiple search terms and the search time of each search term, the intent slicing classification model can be trained first.

[0126] like Figure 3 The figure shows a logical diagram of training an intent slice classification model in one implementation, which includes five steps.

[0127] Among them, the first step is data labeling. In this step, based on the understanding of the business, 50,000 search operations can be pre-grouped by intent slices, and the grouping results can be manually labeled. When labeling, each search operation corresponds to the identification information of an intent slice group. For example, if the identification information of the intent slice group corresponding to a search operation is "B", then it can be indicated that the search operation is the first search operation corresponding to the intent slice group to which it belongs. If the identification information of the intent slice group corresponding to a search operation is "I", then it can be indicated that the search operation is the same intent slice group as the previous search operation.

[0128] The data used for intent slicing includes information such as the search term, search time, and search identifier corresponding to each search operation, and each information corresponds to a different field. For example, as shown in Table 1, it is a list of fields included in the data used for intent slicing, where "keyword" represents the search term for each search, and its field format is STRING, "Search_session_id" represents the identifier corresponding to each search, and its field format is STRING, and "time" represents the search time, and its field format is TIMESTAMP.

[0129] Table 1 List of fields included in the data used for intention slicing

[0130] Fields Field meaning Field Format keyword Search terms per search STRING Search_session_id The identifier corresponding to each search STRING time Search time TIMESTAMP

[0131] As shown in Table 2, this is the result of intent slicing based on search terms. When the search terms are "Character B cartoon", "Character C", "Character C fishing", "Character C fishing monitoring", "Character C fishing monitoring" and "Character D", the intent slicing grouping results are shown on the right side of the table. "Character B cartoon" is the starting search of an intent slicing, "Character C", "Character C fishing", "Character C fishing monitoring" and "Character C fishing monitoring" have the same or similar search intents and belong to the same intent slicing grouping, and "Character D" is the starting search of another intent slicing.

[0132] Table 2 Results of slicing based on search term intent

[0133] Search term Intention slicing results Character B Animation is the starting search for an intent slice Character C is a starting search for an intent slice Character C Fishing Same intent slice as the previous search Person C fishing monitoring Same intent slice as the previous search Person C fishing monitoring Same intent slice as the previous search Character D is the starting search for an intent slice

[0134] The second step is to survey the data distribution, that is, to obtain basic data such as the search term distribution and time span of the search operation, so as to prepare for the subsequent feature selection of these data.

[0135] The third step is feature engineering. In this step, the basic data obtained in the second step can be used to obtain some feature information of the search term of the search operation, so as to further use this feature information to determine whether the current search term meets the previous search term matching conditions. Among them, six different feature information are listed, including:

[0136] 1. Calculate the difference between the search time of each search term and the search time of the previous search term. Then, the difference can be normalized to make the calculation result more intuitive and convenient for further analysis. 2. For each search, calculate the probability distribution KDE (Kernel Density Estimation) of the search term in the search, and get the corresponding binary classification label as a feature information. 3. For each search, use the pre-trained XLnet model to calculate the cosine similarity between the search term and the previous search term as a feature information. The XLnet model is an improved BERT (Bidirectional Encoder Representations from Transformers, bidirectional encoder) model; 4. For each search, use the pre-trained XLnet model to calculate the cosine similarity between the search term and the next search term as a feature information; 5. For each search, calculate the ratio of the maximum sequence length between the previous search term and the next search term to the sequence length of the previous search term, and then normalize it as a feature information; 6. For each search, calculate the ratio of the maximum sequence length between the previous search term and the next search term to the sequence length of the search term, and then normalize it as a feature information.

[0137] Among them, the ratio of the sequence lengths between any two search terms is the repeated and continuous parts of the two search terms. For example, if the repeated and continuous parts of "Game A Character" and "Game A Skin" are "Game A", then the maximum continuous sequence length of "Game A Character" and "Game A Skin" is 3, which is the sequence length corresponding to "Game A".

[0138] The fourth step is model training. In this step, the GBDT model or XGBoost model can be used to analyze the feature information obtained in the third step to obtain a binary classification result of whether each search term meets the matching conditions with the previous search term, that is, output the result of "whether it is the same intent slice group as the previous search".

[0139] The fifth step is model selection. The accuracy of the GBDT model is 0.9765, the recall rate is 0.9735, and the F1 value (F1Score) is 0.9748. The importance of each feature information in the third step is ranked as 5>6>1>3>4>2. The accuracy of the XGBoost model is 0.9738, the recall rate is 0.9716, and the F1 value is 0.9723. The importance of each feature information in the third step is ranked as 3>1>4>5>6>2. In comparison, the GBDT model with better effect is selected.

[0140] In the second stage, after the intent slice classification model is trained, the search operation can be divided into multiple intent slices based on the multiple search terms obtained and the search time of each search term.

[0141] Specifically, multiple search terms and the search time of each search term can be obtained, and for each search term, the difference between the search time of the current search term and the search time of the previous search term is calculated. Then, the feature information and difference of the current search term are input into the pre-trained intent slice classification model for processing, and the classification result is used to determine whether the current search term and the previous search term of the current search term meet the matching condition. As can be seen from the above, the training effect of the GBDT model is better than that of the XGBoost model, so the intent slice classification model in this embodiment is the GBDT model. If the current search term and the previous search term of the current search term meet the matching condition, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

[0142] Among them, intent slicing refers to the combination of multiple searches for the same intent within a period of time. In other words, if several consecutive search terms include the same or similar content, then the search operations corresponding to these search terms can be divided into the same intent slice. For example, if the search terms for several consecutive search operations are "Game A", "Beautiful skins for Game A", and "Character skills for Game A", it can be understood that the intentions of these search operations are all search terms related to Game A. Therefore, these search operations can be divided into the same intent slice.

[0143] In the third stage, after the search operation is divided into multiple intent slices, the search effect can be analyzed based on the intent slices.

[0144] Specifically, the operation information in the search results page of the search operation corresponding to each search term can be detected, and when the preset operation information is detected, the corresponding search operation is determined as the target search operation. If the operation information in the search results page includes the preset operation information, then it can be considered that the search results of this search operation include results that meet the user's needs. Then, according to the search order of the target search operation in each intent slice, it can be determined how many times the user has obtained satisfactory search results, and then the search step can be determined. The search step can be used to measure the efficiency of the search.

[0145] For example, as shown in Table 3, it is a list of fields included in the data source that can be used to determine the search step in this step, where "keyword" represents the search term for each search, and its field format is STRING; "Request_api" represents the searched page, which can be a comprehensive search page or a user vertical search page, and its field format is STRING; "Search_session_id" represents the identifier corresponding to each search, and its field format is STRING; "Is_user_show" represents whether the user card is exposed, and its field format is BINARY; "Is_us "er_click" indicates whether the user card is clicked, and its field format is BINARY; "Is_user_profile_valid_play" indicates whether clicking the user card to enter the personal homepage has a long video broadcast, and its field format is BINARY; "Is_user_follow" indicates whether to follow the user, and its field format is BINARY; "Is_user_click_followed" indicates whether the clicked user is previously followed, and its field format is BINARY; "Goal_id" indicates the identifier of each intent slice, and its field format is BINARY. Among them, "Is_user_show", "Is_user_click", "Is_user_profile_valid_play", "Is_user_follow" and "Is_user_click_followed" fields can all be used as fields corresponding to preset operation information.

[0146] Table 3 List of fields included in the data source used to determine the search step length

[0147]

[0148] Then, the operation information in the search results page of the search operation corresponding to each search term can be detected. When the preset operation information is detected, the corresponding search operation is determined as the target search operation, and the search step length is determined according to the number of target search operations and the search order of the target search operations in the corresponding intent slice.

[0149] For example, if a user searches for "live broadcast", then searches for "emotional anchor", switches to user vertical search and finds the anchor of interest in the user vertical search, and then follows the anchor, that is, the preset operation information is detected in the third search, then the search step length is: 3 / 1=3. It can be understood that the search step length can express how many times it is necessary to search to achieve the goal of search results meeting user needs. It reflects the cost of users obtaining search results that meet their needs, and is a negative indicator that needs to be reduced.

[0150] Based on the search step length, the shortcomings of the current search method can be determined, which makes it easier for developers to optimize. Figure 4 As shown in the figure, it is a schematic diagram of the distribution of intent slices with different steps for finding people when the purpose of the search operation is to find people. It can be seen that the vast majority of intent slices that find people successfully find people in the first search, and the proportion of people found after more than 3 searches is extremely low, indicating that the recommendation accuracy of the current search for people has reached a certain level, but it can also be explained that there is still a certain proportion of intent slices that find people whose search needs are met in the second search, so there is still room for business optimization.

[0151] In addition, the effect of the current search method can be further evaluated through the operation information in the search result page of the search operation corresponding to each search word. Figure 5 As shown, it is a schematic diagram of several operations that represent user satisfaction with search results. Among them, when the purpose of the search operation is to find a person, the percentage of user results clicked in the total search is x%. In the case of user result clicks, it can also be divided into the intention slice of actually finding the person and the intention slice of no satisfactory behavior, which account for y% and 1-y% of user result clicks respectively. Then, the percentage of the intention slice of actually finding the person in the total search is x%*y%=z%. The intention slice of actually finding the person can also be divided into four situations: only the behavior of consuming the user's work after clicking the user's result, only the behavior of paying attention, only clicking on the user who has been followed, and at least two of the above behaviors, which account for a%, b%, c% and d% respectively. Then the probability of the user being satisfied with the search results is a%+b%+c%+d%=z%.

[0152] Furthermore, the search adjustment parameters may be calculated based on the search results of the search operations included in the intent slice to which the target search operation belongs; and based on the search adjustment parameters, attribution analysis may be performed on the target search operation to obtain attribution analysis results.

[0153] For example, Figure 6As shown, it is a schematic diagram of search step attribution in one implementation method, which attributes the slices of search intent with search step>1. Among them, x1%, x2%, and x3% represent the probabilities of three scenarios in which the user did not find the person he wanted on the comprehensive page, and found the person after changing the search term and searching again. The sum of the three is the probability that the target search operation is different from the search term of the previous search operation of the target search operation. y1%, y2%, and y3% represent the probabilities of three scenarios in which the user did not find the person he wanted on the comprehensive page, and found the person after switching the user vertical search. The sum of the three is the probability that the search result page of the target search operation includes the search type switching operation information. The proportions of these scenarios can be calculated, and the corresponding user behavior trajectory can be cited to check the data, such as finding problems such as search term correction and search term intent recognition, and drive improvements.

[0154] For example, when searching for "3014225580", after the non-target search operation, the search continues for "301425580", and the user results are followed in the second search, in this case, a search term error correction occurs; when searching for "Did Zhou Zhou break up?", no user results are displayed, and the user clicks on the user results after searching for "Zhou Zhou" for the second time, in this case, a search term intent recognition error occurs, etc. After discovering these areas of improvement, corresponding special projects are driven to be implemented, such as content intent word expansion, user result recall, error correction improvement, and relationship chain introduction.

[0155] From the above, it can be seen that in this solution, the search terms of each search can be linked to consider the user intentions represented by similar search terms in multiple search scenarios, that is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if a satisfactory search result is found by changing the search term with similar intention for the second time, then these two searches belong to the same intent slice. Therefore, the intent slice method is used to define a continuous behavior with the same user intention. The understanding of user behavior is closer to reality, which can improve the accuracy of search processing results and meet more complex search scenarios.

[0156] Figure 6 is a block diagram of a search processing device according to an exemplary embodiment, the device comprising:

[0157] An acquisition unit 301 is configured to acquire a plurality of search terms and a search time of each search term;

[0158] A grouping unit 302 is configured to divide the search operation corresponding to the search word into multiple intent slices according to the search time and feature information of the search word;

[0159] The computing unit 304 is configured to execute detection of operation information in the search results page of the search operation corresponding to each search term, and when preset operation information is detected, determine the corresponding search operation as a target search operation, and determine the search step length according to the number of the target search operations and the search order of the target search operations in the intent slice to which they belong.

[0160] In one implementation, the grouping unit 302 is specifically configured to perform:

[0161] For each search term, calculate the difference between the search time of the current search term and the search time of the previous search term;

[0162] According to the feature information of the current search term and the difference, determining whether the current search term and a previous search term of the current search term meet a matching condition;

[0163] If satisfied, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

[0164] In one implementation, the grouping unit 302 is specifically configured to perform:

[0165] The feature information of the current search term and the difference are input into a pre-trained intention slice classification model for processing, and a determination is made based on the classification result as to whether the current search term and the previous search term of the current search term meet a matching condition.

[0166] In one implementation, the feature information of the current search term includes at least one of the following:

[0167] The kernel density estimate of the current search term;

[0168] The cosine similarity between the current search term and the previous search term of the current search term;

[0169] The cosine similarity between the current search term and the next search term of the current search term;

[0170] The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the previous search word;

[0171] The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the current search word.

[0172] In one implementation, the preset operation information includes one or more of the following:

[0173] Added follow operation;

[0174] Click operations on followed users;

[0175] Browsing operation on any user's information.

[0176] In one implementation, the determining unit 303 is configured to execute:

[0177] Obtain the number of searches required to search for the target search operation in each belonging intent slice, and take the sum of the number of searches as the first quantity;

[0178] determining the number of intent slices including the target search operation as a second number;

[0179] The ratio of the first number to the second number is used as the search step length.

[0180] In one implementation, the determining unit 303 is further configured to execute:

[0181] Calculate a search adjustment parameter according to the search results of the search operation included in the intent slice to which the target search operation belongs;

[0182] According to the search adjustment parameters, attribution analysis is performed on the target search operation to obtain an attribution analysis result.

[0183] In one implementation, the search adjustment parameter includes one or more of the following:

[0184] The probability that the search result page of the target search operation includes search type switching operation information;

[0185] The probability that the target search operation has a different search term from a previous search operation of the target search operation.

[0186] From the above, it can be seen that the technical solution provided by the embodiments of the present disclosure can be linked to the search terms in each search, and consider the user intentions represented by similar search terms in multiple search scenarios, that is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if a satisfactory search result is found by changing to a search term with similar intention for the second time, then these two searches belong to the same intent slice. Therefore, the intent slice method is used to define a continuous behavior with the same user intention. The understanding of user behavior is closer to reality, which can improve the accuracy of the search processing results and meet more complex search scenarios.

[0187] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0188] Figure 7 It is a block diagram of an electronic device for search processing according to an exemplary embodiment.

[0189] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided, such as a memory including instructions, and the above instructions can be executed by a processor of an electronic device to perform the above method. Optionally, the computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0190] In an exemplary embodiment, a computer program product is also provided. When the computer program product is executed on a computer, the computer implements the above-mentioned search processing method.

[0191] From the above, it can be seen that the technical solution provided by the embodiments of the present disclosure can be linked to the search terms in each search, and consider the user intentions represented by similar search terms in multiple search scenarios, that is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if a satisfactory search result is found by changing to a search term with similar intention for the second time, then these two searches belong to the same intent slice. Therefore, the intent slice method is used to define a continuous behavior with the same user intention. The understanding of user behavior is closer to reality, which can improve the accuracy of the search processing results and meet more complex search scenarios.

[0192] Figure 8 is a block diagram of a device 800 for search processing according to an exemplary embodiment.

[0193] For example, apparatus 800 may be a mobile phone, a computer, a digital broadcast electronic device, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.

[0194] Reference Figure 8 , the device 800 may include one or more of the following components: a processing component 802 , a memory 804 , a power component 806 , a multimedia component 808 , an audio component 810 , an input / output (I / O) interface 812 , a sensor component 814 , and a communication component 816 .

[0195] The processing component 802 generally controls the overall operation of the device 800, such as operations associated with display, phone calls, data communications, camera operations, and recording operations. The processing component 802 may include one or more processors 820 to execute instructions to complete all or part of the steps of the above-mentioned method. In addition, the processing component 802 may include one or more modules to facilitate the interaction between the processing component 802 and other components. For example, the processing component 802 may include a multimedia module to facilitate the interaction between the multimedia component 808 and the processing component 802.

[0196] The memory 804 is configured to store various types of data to support operations on the device 800. Examples of such data include instructions for any application or method operating on the device 800, contact data, phone book data, messages, pictures, videos, etc. The memory 804 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0197] The power supply component 807 provides power to the various components of the device 800. The power supply component 807 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device 800.

[0198] The multimedia component 808 includes a screen that provides an output interface between the device 800 and the user. In some embodiments, the screen may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen may be implemented as a touch screen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor may not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation. In some embodiments, the multimedia component 808 includes a front camera and / or a rear camera. When the device 800 is in an operating mode, such as a shooting mode or a video mode, the front camera and / or the rear camera may receive external multimedia data. Each front camera and rear camera may be a fixed optical lens system or have a focal length and optical zoom capability.

[0199] The audio component 810 is configured to output and / or input audio signals. For example, the audio component 810 includes a microphone (MIC), and when the device 800 is in an operating mode, such as a call mode, a recording mode, and a speech recognition mode, the microphone is configured to receive an external audio signal. The received audio signal can be further stored in the memory 804 or sent via the communication component 816. In some embodiments, the audio component 810 also includes a speaker for outputting audio signals.

[0200] I / O interface 812 provides an interface between processing component 802 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: home button, volume button, start button, and lock button.

[0201] The sensor assembly 814 includes one or more sensors for providing various aspects of status assessment for the device 800. For example, the sensor assembly 814 can detect the open / closed state of the device 800, the relative positioning of components, such as the display and keypad of the device 800, and the sensor assembly 814 can also detect the position change of the device 800 or a component of the device 800, the presence or absence of user contact with the device 800, the orientation or acceleration / deceleration of the device 800, and the temperature change of the device 800. The sensor assembly 814 may include a proximity sensor configured to detect the presence of a nearby object without any physical contact. The sensor assembly 814 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications. In some embodiments, the sensor assembly 814 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.

[0202] The communication component 816 is configured to facilitate wired or wireless communication between the device 800 and other devices. The device 800 can access a wireless network based on a communication standard, such as WiFi, a carrier network (such as 2G, 3G, 4G or 5G), or a combination thereof. In an exemplary embodiment, the communication component 816 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 816 also includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0203] In an exemplary embodiment, the apparatus 800 may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to execute the methods described in the first and second aspects.

[0204] In an exemplary embodiment, a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 804 including instructions, and the instructions can be executed by the processor 820 of the device 800 to complete the above method. Optionally, for example, the storage medium can be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0205] In an exemplary embodiment, a computer program product containing instructions is also provided. When the computer program product is run on a computer, the computer is enabled to execute the search processing method described in the first embodiment above.

[0206] From the above, it can be seen that the technical solution provided by the embodiments of the present disclosure can be linked to the search terms in each search, and consider the user intentions represented by similar search terms in multiple search scenarios, that is, user behavior is often continuous, and the experience of the previous search is likely to be the reason for the next search. In other words, if the user does not get satisfactory search results in the first search, if a satisfactory search result is found by changing to a search term with similar intention for the second time, then these two searches belong to the same intent slice. Therefore, the intent slice method is used to define a continuous behavior with the same user intention. The understanding of user behavior is closer to reality, which can improve the accuracy of the search processing results and meet more complex search scenarios.

[0207] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0208] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A search processing method, characterized in that: include: Get multiple search terms and the search time of each search term; Divide the search operation corresponding to the search term into multiple intent slices according to the search time and feature information of the search term; Detecting operation information in the search result page of the search operation corresponding to each search term, determining the corresponding search operation as a target search operation when preset operation information is detected, and determining the search step length according to the number of the target search operations and the search order of the target search operations in the intent slice to which they belong; the preset operation information includes at least one of the following: adding a new follow operation; a click operation on a followed user; Information browsing operations for any user; Performing attribution analysis on the intent slices with a search step size greater than one, determining improvement space for the search method based on the attribution analysis result, and driving corresponding improvements; The improvement space of the determining search method can be obtained by reviewing the user behavior trajectory data; The improvements include: content intent word expansion, user result recall, error correction improvement, and relationship chain introduction.

2. The search processing method according to claim 1, characterized in that: The step of dividing the search operation corresponding to the search term into multiple intention slices according to the search time and feature information of the search term includes: For each search term, calculate the difference between the search time of the current search term and the search time of the previous search term; According to the feature information of the current search term and the difference, determining whether the current search term and a previous search term of the current search term meet a matching condition; If satisfied, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

3. The search processing method according to claim 2, characterized in that: The determining, based on the feature information of the current search term and the difference, whether the current search term and a previous search term of the current search term meet a matching condition includes: The feature information of the current search term and the difference are input into a pre-trained intention slice classification model for processing, and a determination is made based on the classification result as to whether the current search term and the previous search term of the current search term meet a matching condition.

4. The search processing method according to claim 2, characterized in that: The characteristic information of the current search term includes at least one of the following: The kernel density estimate of the current search term; The cosine similarity between the current search term and the previous search term of the current search term; The cosine similarity between the current search term and the next search term of the current search term; The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the previous search word; The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the current search word.

5. The search processing method according to claim 1, characterized in that: Determining the search step length according to the number of the target search operations and the search order of the target search operations in the corresponding intent slices includes: Obtain the number of searches required to search for the target search operation in each belonging intent slice, and take the sum of the number of searches as the first quantity; determining the number of intent slices including the target search operation as a second number; The ratio of the first number to the second number is used as the search step length.

6. The search processing method according to claim 1, characterized in that: After determining the search step length according to the number of the target search operations and the search order of the target search operations in the corresponding intent slice, the method further includes: Calculate a search adjustment parameter based on the search results of the search operation included in the intent slice to which the target search operation belongs; According to the search adjustment parameters, attribution analysis is performed on the target search operation to obtain an attribution analysis result.

7. The search processing method according to claim 6, characterized in that: The search adjustment parameters include one or more of the following: The probability that the search result page of the target search operation includes search type switching operation information; The probability that the target search operation has a different search term from a previous search operation of the target search operation.

8. A search processing device, characterized in that: include: An acquisition unit is configured to acquire a plurality of search terms and a search time of each search term; a grouping unit, configured to divide the search operation corresponding to the search term into a plurality of intent slices according to the search time and feature information of the search term; The determination unit is configured to detect operation information in the search result page of the search operation corresponding to each search word, and when the preset operation information is detected, determine the corresponding search operation as a target search operation, and determine the search step length according to the number of the target search operations and the search order of the target search operations in the intent slice to which they belong; the preset operation information includes one or more of the following: adding a new follow operation; a click operation on a followed user; Information browsing operations for any user; Performing attribution analysis on the intent slices with a search step size greater than one, determining improvement space for the search method based on the attribution analysis result, and driving corresponding improvements; The improvement space of the determining search method can be obtained by reviewing the user behavior trajectory data; The improvements include: content intent word expansion, user result recall, error correction improvement, and relationship chain introduction.

9. The search processing device according to claim 8, characterized in that: The grouping unit is specifically configured to execute: For each search term, calculate the difference between the search time of the current search term and the search time of the previous search term; According to the feature information of the current search term and the difference, determining whether the current search term and a previous search term of the current search term meet a matching condition; If satisfied, the search operation corresponding to the current search term is added to the intent slice to which the search operation corresponding to the previous search term of the current search term belongs.

10. The search processing device according to claim 9, characterized in that: The grouping unit is specifically configured to execute: The feature information of the current search term and the difference are input into a pre-trained intention slice classification model for processing, and a determination is made based on the classification result as to whether the current search term and the previous search term of the current search term meet a matching condition.

11. The search processing device according to claim 9, characterized in that: The characteristic information of the current search term includes at least one of the following: The kernel density estimate of the current search term; The cosine similarity between the current search term and the previous search term of the current search term; The cosine similarity between the current search term and the next search term of the current search term; The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the previous search word; The ratio of the maximum sequence length between the previous search word and the next search word to the sequence length of the current search word.

12. The search processing device according to claim 8, characterized in that: The determining unit is configured to execute: Obtain the number of searches required to search for the target search operation in each belonging intent slice, and take the sum of the number of searches as the first quantity; determining the number of intent slices including the target search operation as a second number; The ratio of the first number to the second number is used as the search step length.

13. The search processing device according to claim 8, characterized in that: The determining unit is further configured to execute: Calculate a search adjustment parameter based on the search results of the search operation included in the intent slice to which the target search operation belongs; According to the search adjustment parameters, attribution analysis is performed on the target search operation to obtain an attribution analysis result.

14. The search processing device according to claim 13, characterized in that: The search adjustment parameters include one or more of the following: The probability that the search result page of the target search operation includes search type switching operation information; The probability that the target search operation has a different search term from a previous search operation of the target search operation.

15. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the search processing method according to any one of claims 1 to 7.

16. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of a search processing electronic device, the search processing electronic device is enabled to perform the search processing method according to any one of claims 1 to 7.

17. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the search processing method according to any one of claims 1 to 7 is implemented.

Images