Sorting method, device, equipment and computer readable storage medium

Abstract

The application discloses a ranking method and device, equipment and a computer readable storage medium, and belongs to the technical field of computers. The method comprises the following steps: obtaining a positive feedback behavior sequence, a negative feedback behavior sequence and a target feature vector; determining a first positive feedback vector and a first negative feedback vector based on the positive feedback behavior sequence, the feedback behavior sequence and the feature vector; determining a second positive feedback vector and a second negative feedback vector based on the first positive feedback vector, the first negative feedback vector and the negative feedback behavior sequence; and determining a ranking result based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector. The application obtains the positive feedback behavior sequence, the negative feedback behavior sequence and the target feature vector, comprehensively considers the influence of the first positive feedback vector existing in the positive feedback behavior sequence and the first negative feedback vector, the second negative feedback vector and the second positive feedback vector existing in the negative feedback behavior sequence on ranking, and improves the reliability of the ranking result.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of computer, in particular, to a ranking method and device, equipment and computer readable storage medium. BACKGROUND

[0002] With the continuous growth of information, the requirements for search engines and APP (Application) searches are increasing. QAC (Query Auto-Completion) is an important module in current search engines and APP searches. QAC retrieves relevant search results based on information input in a search box, recommends search results that meet the current intention, and ranks the search results according to the possibility of meeting the current intention. The search results can include guide words or points of interest. The ranking can mine effective negative feedback behaviors in the input process through a negative feedback method, and adjust the ranking order in a timely manner based on the negative feedback behaviors, so as to achieve a ranking effect that better meets the current intention.

[0003] In related technologies, the negative feedback method can be realized by Exploitation & Exploration (development and exploration) when the user click probability is lower than a certain threshold, such as reducing the exposure opportunity of samples.

[0004] The above-mentioned negative feedback method may disturb the original ranking when balancing Exploitation and Exploration, resulting in a decrease in CTR (Click Through Rate) and reducing the reliability of the ranking result. SUMMARY

[0005] Embodiments of the present application provide a ranking method, device, equipment and computer readable storage medium, which can be used to solve the problems in related technologies. The technical solution is as follows:

[0006] In one aspect, the present application provides a ranking method, the method comprising:

[0007] obtaining a positive feedback behavior sequence, a negative feedback behavior sequence and a target feature vector, the positive feedback behavior sequence comprising a feature vector corresponding to a search result that has been clicked at least once in at least one ranking before the present ranking, the negative feedback behavior sequence comprising a feature vector corresponding to at least one search result that has not been clicked in the present search result, and the target feature vector being a feature vector corresponding to a search result that has appeared in the present input process with the same search target;

[0008] determine a first positive feedback vector based on the positive feedback behavior sequence and the target feature vector; determine a first negative feedback vector based on the negative feedback behavior sequence and the target feature vector;

[0009] determine a second positive feedback vector and a second negative feedback vector based on the first positive feedback vector, the first negative feedback vector and the negative feedback behavior sequence;

[0010] determine a ranking result of a search result corresponding to the input content based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector.

[0011] In a possible implementation, the determining of the first positive feedback vector based on the positive feedback behavior sequence and the target feature vector comprises: determining a first attention matrix based on the positive feedback behavior sequence and the target feature vector; and determining the first positive feedback vector based on the first attention matrix.

[0012] The determining of the first negative feedback vector based on the negative feedback behavior sequence and the target feature vector comprises: determining a second attention matrix based on the negative feedback behavior sequence and the target feature vector; and determining the first negative feedback vector based on the second attention matrix.

[0013] In a possible implementation, the determining of the second positive feedback vector and the second negative feedback vector based on the first positive feedback vector, the first negative feedback vector and the negative feedback behavior sequence comprises:

[0014] determining a first weight based on the first positive feedback vector and the first negative feedback vector;

[0015] determining a second positive feedback vector based on the first weight and the negative feedback behavior sequence; and determining a second negative feedback vector based on the first weight and the negative feedback behavior sequence.

[0016] In a possible implementation, the determining of the second positive feedback vector based on the first weight and the negative feedback behavior sequence comprises:

[0017] weighting at least one feature vector included in the negative feedback behavior sequence by the first weight to obtain the second positive feedback vector.

[0018] In a possible implementation, the determining of the second negative feedback vector based on the first weight and the negative feedback behavior sequence comprises:

[0019] determining a second weight based on the first weight;

[0020] The at least one feature vector included in the negative feedback behavior sequence is weighted by the second weight to obtain the second negative feedback vector.

[0021] In a possible implementation, the ranking result of the guide word or the interest point corresponding to the current input information is determined based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector, and the ranking result includes:

[0022] The first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector are spliced to obtain a positive-negative feedback vector.

[0023] The ranking result is determined based on the positive-negative feedback vector.

[0024] In a possible implementation, the feature vector of any search result in the positive feedback behavior sequence or the negative feedback behavior sequence is used to represent at least one of the following information: title information of the search result, total exposure time information of the search result, maximum exposure time information of the search result, the number of times that a word input in a search box in an input process is at a word boundary of the search result, and the number of times that the search result is exposed at different positions.

[0025] The target feature vector is used to represent at least one of the following information of the search result that appears in the current input process and is the same as the search target: title information of the search result, total exposure time information of the search result, maximum exposure time information of the search result, the number of times that a word input in a search box in an input process is at a word boundary of the search result, and the number of times that the search result is exposed at different positions.

[0026] In another aspect, a ranking device is provided, and the device includes:

[0027] The acquisition module is configured to acquire a positive feedback behavior sequence, a negative feedback behavior sequence, and a target feature vector, the positive feedback behavior sequence includes a feature vector corresponding to a search result that is clicked at least once in at least one ranking before the current ranking, the negative feedback behavior sequence includes a feature vector corresponding to at least one search result that is not clicked in the current search result, and the target feature vector is a feature vector corresponding to a search result that appears in a current input process and is the same as a search target.

[0028] The first determination module is configured to determine a first positive feedback vector based on the positive feedback behavior sequence and the target feature vector, and determine a first negative feedback vector based on the negative feedback behavior sequence and the target feature vector.

[0029] a second determining module configured to determine a second positive feedback vector and a second negative feedback vector based on the first positive feedback vector, the first negative feedback vector and the negative feedback behavior sequence;

[0030] a third determining module configured to determine a ranking result of a search result corresponding to the input content based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector.

[0031] In a possible implementation, the first determining module is configured to determine a first attention matrix based on the positive feedback behavior sequence and the target feature vector; and determine the first positive feedback vector based on the first attention matrix.

[0032] The first determining module is configured to determine a second attention matrix based on the negative feedback behavior sequence and the target feature vector; and determine the first negative feedback vector based on the second attention matrix.

[0033] In a possible implementation, the second determining module is configured to determine a first weight based on the first positive feedback vector and the first negative feedback vector; determine a second positive feedback vector based on the first weight and the negative feedback behavior sequence; and determine a second negative feedback vector based on the first weight and the negative feedback behavior sequence.

[0034] In a possible implementation, the second determining module is configured to weight at least one feature vector included in the negative feedback behavior sequence by the first weight to obtain the second positive feedback vector.

[0035] In a possible implementation, the second determining module is configured to determine a second weight based on the first weight; and weight at least one feature vector included in the negative feedback behavior sequence by the second weight to obtain the second negative feedback vector.

[0036] In a possible implementation, the third determining module is configured to concatenate the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector to obtain a positive-negative feedback vector.

[0037] The ranking result is determined based on the positive-negative feedback vector.

[0038] In a possible implementation, a feature vector corresponding to any search result in the positive feedback behavior sequence or the negative feedback behavior sequence is used to represent at least one of title information of the search result, total exposure time information of the search result, maximum exposure time information of the search result, times information of a word input in a search box in an input process being in a word boundary of the search result, and times information of the search result being exposed at different positions.

[0039] The target feature vector is used to represent at least one of title information of a search result that has occurred in the current input process with the search target, total exposure time information of the search result, maximum exposure time information of the search result, the number of times that a word input in the search box in the input process is at a word boundary of the search result, and the number of times that the search result is exposed at different positions.

[0040] In another aspect, a computer device is provided, which includes a processor and a memory, and the memory stores at least one computer program, which is loaded and executed by the processor to enable the computer device to implement any of the above ranking methods.

[0041] In another aspect, a computer readable storage medium is also provided, which stores at least one computer program, which is loaded and executed by a processor to enable a computer to implement any of the above ranking methods.

[0042] In another aspect, a computer program product or computer program is also provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to implement any of the above ranking methods.

[0043] The technical solutions provided by the embodiments of the present application at least bring the following beneficial effects:

[0044] In the embodiments of the present application, the positive feedback behavior sequence, the negative feedback behavior sequence and the target feature vector are obtained, and the influence of the first positive feedback vector in the positive feedback behavior sequence and the first negative feedback vector, the second negative feedback vector and the second positive feedback vector in the negative feedback behavior sequence on ranking are comprehensively considered, thereby improving the reliability of the ranking result. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0046] Figure 1 is a schematic diagram of an implementation environment provided by an embodiment of the present application;

[0047] Figure 2 is a ranking method flowchart provided by an embodiment of the present application;

[0048] Figure 3 is a kind of sequencing process schematic diagram provided by the embodiment of the application;

[0049] Figure 4 is a kind of method flow chart for determining the interaction score w p of the embodiment of the application;

[0050] Figure 5 is a kind of method flow chart for determining the positive and negative feedback vector of the embodiment of the application;

[0051] Figure 6 is a kind of sequencing device schematic diagram provided by the embodiment of the application;

[0052] Figure 7 is a kind of structure schematic diagram of computer equipment provided by the embodiment of the application;

[0053] Figure 8 is another kind of structure schematic diagram of computer equipment provided by the embodiment of the application. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the present application clearer, the embodiment of the present application will be described in further detail below with reference to the drawings.

[0055] It should be noted that the terms "first", "second" and the like (if any) in the specification and claims of the present application are used to distinguish similar objects, and do not necessarily have to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the application 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 application. Rather, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0056] The embodiment of the present application provides a kind of sequencing method, please refer to Figure 1 , which shows the schematic diagram of the method implementation environment provided by the embodiment of the application. The implementation environment can include: terminal 11 and server 12.

[0057] The terminal 11 can apply the sorting method provided in the embodiments of the present application to perform sorting, and then send the sorting result to the server 12. The server 12 can receive the sorting result sent by the terminal 11 and store the sorting result. Alternatively, the terminal 11 can apply the sorting method provided in the embodiments of the present application to perform sorting and store the sorting result. Alternatively, the server 12 can apply the sorting method provided in the embodiments of the present application to perform sorting, and then send the sorting result to the terminal 11. The terminal 11 can receive the sorting result sent by the server 12 and store the sorting result. Alternatively, the server 12 can apply the sorting method provided in the embodiments of the present application to perform sorting and store the sorting result.

[0058] Alternatively, the terminal 11 can be any electronic product that can interact with a user through one or more of a keyboard, a touchpad, a touch screen, a remote controller, voice interaction, a handwriting device, etc., such as a PC (Personal Computer), a mobile phone, a smart phone, a PDA (Personal Digital Assistant), a wearable device, a PPC (Pocket PC), a tablet computer, a smart car machine, a smart television, a smart speaker, etc. The server 12 can be a server or a server cluster composed of multiple servers, or a cloud computing service center. The terminal 11 and the server 12 establish a communication connection through a wired or wireless network.

[0059] Those skilled in the art should understand that the terminal 11 and the server 12 described above are only examples, and other existing or future terminals or servers, such as those applicable to the present application, should also be included in the protection scope of the present application and are hereby incorporated by reference.

[0060] Based on the above Figure 1 Based on the above Figure 2 The method provided in the embodiments of the present application can include the following steps:

[0061] In step 201, the positive feedback behavior sequence, the negative feedback behavior sequence, and the target feature vector are obtained. The positive feedback behavior sequence includes the feature vector corresponding to the search result that has been clicked at least once in at least one sorting before the present sorting. The negative feedback behavior sequence includes the feature vector corresponding to at least one search result that has not been clicked in the present search result. The target feature vector is the feature vector corresponding to the search result that appears in the present input process and is the same as the search target.

[0062] The embodiments of the present application do not limit the source and type of the positive feedback behavior sequence, the negative feedback behavior sequence and the target feature vector. For example, in the search of any user of an e-commerce platform, the search auto-completion module of the e-commerce platform can use the method provided by the embodiments of the present application to sort the content input by the user in real time. Alternatively, the positive feedback behavior sequence can be the feature vector corresponding to the search result that the user has clicked at least once in at least one previous sorting, the negative feedback behavior sequence can be the feature vector corresponding to at least one search result that the user has not clicked in this search result, and the target feature vector can be the feature vector corresponding to the search result that appears in the current input process with the same search target as the user.

[0063] Alternatively, the feature vector corresponding to any search result in the positive feedback behavior sequence and the negative feedback behavior sequence is used to represent at least one of the title information of the search result, the total exposure time information of the search result, the maximum exposure time information of the search result, the number of times that the text input in the search box in the input process is at the word boundary of the search result, and the number of times that the search result is exposed at different positions.

[0064] The target feature vector is used to represent at least one of the title information of the search result that appears in the current input process with the same search target, the total exposure time information of the search result, the maximum exposure time information of the search result, the number of times that the text input in the search box in the input process is at the word boundary of the search result, and the number of times that the search result is exposed at different positions.

[0065] For example, when the search auto-completion module of the e-commerce platform uses the method provided by the embodiments of the present application for sorting, Figure 3The input process of the search auto-completion of any user of the e-commerce platform as a search target of "couple photo" is shown in the figure. When the user inputs "q", the search begins, and ten search results including "couple dating paradise" in the ninth position appear in the search auto-completion list, and the search target "couple photo" does not appear. Then the user continues to input, and after t1 time, "qing" appears in the search box, at this time "couple dating paradise" is in the third position, and the search target "couple photo" does not appear. Then after t4 time, "qingl" appears in the search box, at this time "couple dating paradise" is in the first position, and the search target "couple photo" does not appear. Then after t5 time, "qinglv" appears in the search box, at this time "couple dating paradise" is in the first position, and the search target "couple photo" is in the tenth position. Then after t6 time, "couple" appears in the search box, at this time "couple dating paradise" no longer appears, and the search target "couple photo" is in the second position, at this time the user clicks "couple photo" in the second position, and the search is completed.

[0066] In the above process, from the user inputting "q" to inputting "qinglv" is the input state, and the user inputs "couple" and clicks "couple photo" appearing at this time is the input completion state. In this example, "couple dating paradise" is a search result that is not clicked in this search result, so the feature vector corresponding to the search result belongs to the negative feedback behavior sequence, the title information corresponding to the search result can be "qinglvyuehuishengdi", the total exposure time information of the search result can be "t1+t4+t5", the maximum exposure time information of the search result can be the maximum value of t1, t4 and t5, the number of times that the text input in the search box is in the word boundary of the search result, that is, the number of times that "qing" or "qinglv" appears in the search box, which can be 2 at this time, and the number of times that the search result is exposed in different positions can be "exposed twice in the first position".

[0067] In this example, "couple photo" is the search target, so in the input process, when "qinglv" and "couple" appear in the search box, the same search result as the search target appears, the last time "couple photo" appears can be obtained, the feature vector corresponding to "couple photo" is the target feature vector, and the determination method of the feature vector can be the same as the method of determining the feature vector corresponding to the negative feedback behavior sequence described above. The determination method of the feature vector corresponding to the positive feedback behavior sequence can also be the same as the method of determining the feature vector corresponding to the negative feedback behavior sequence described above.

[0068] The embodiment of the present application can consider total exposure time information, maximum exposure time information, the number of times that the character input in the search box during input is at the word boundary, and the number of times of exposure at different positions. Among them, the longer the total exposure time is, the greater the probability that the search result is seen is, the longer the maximum exposure time is, the greater the probability that the search result is seen is, the more the number of times that the character input in the search box during input is at the word boundary is, the greater the probability that the search result is seen is, and the more the number of times of exposure at the front position is, the greater the probability that the search result is seen is. The greater the probability that the search result is seen is, the greater the probability that the search result is not interested in when the search result is not clicked is. The embodiment of the present application can better recommend the search result meeting the current intention through the above information, thereby improving the reliability of the sorting result.

[0069] In step 202, a first positive feedback vector is determined based on the positive feedback behavior sequence and the target feature vector; and a first negative feedback vector is determined based on the negative feedback behavior sequence and the target feature vector.

[0070] Optionally, determining the first positive feedback vector based on the positive feedback behavior sequence and the target feature vector comprises: determining a first attention matrix based on the positive feedback behavior sequence and the target feature vector; and determining the first positive feedback vector based on the first attention matrix; and determining the first negative feedback vector based on the negative feedback behavior sequence and the target feature vector comprises: determining a second attention matrix based on the negative feedback behavior sequence and the target feature vector; and determining the first negative feedback vector based on the second attention matrix.

[0071] Exemplarily, the positive feedback behavior sequence and the target feature vector can be embedded and spliced to obtain a positive feedback input matrix B u , and then Q, K and V are obtained based on B u respectively.

[0072] Q=W Q B u , K=W K B u , and V=W V B u

[0073] Wherein, Q can be a query matrix in the Transformer model, K can be a keyword matrix in the Transformer model, V can be a value matrix in the Transformer model, W Q (for example ) can be a mapping matrix corresponding to Q, W K (for example ) can be a mapping matrix corresponding to K, and W V (for example ) can be a mapping matrix corresponding to V, n h (For example, n h is a positive integer) is the feature dimension of the positive feedback input vector, and h (for example, h is a positive integer) is the number of attention heads. Embodiments of the present application do not limit the size of n h and h, which can be set according to the situation.

[0074] Then, a plurality of single-head attention matrices can be calculated:

[0075]

[0076] wherein the subscript i (for example, a positive integer from 1 to h) is the number of attention heads, Q i , K i , and V i may be Q, K, and V corresponding to the i-th attention head, respectively.

[0077] The plurality of single-head attention matrices are then spliced to obtain a first attention matrix, and based on the first attention matrix, a first output matrix F u can be obtained.

[0078] F u = concat (head1,..., head h ) W O

[0079] wherein the first attention matrix can be obtained based on the concat (splicing) function, and W O (for example ) is a mapping matrix corresponding to the first attention matrix.

[0080] After that, the vectors in F u are averaged-pooled to obtain a first positive feedback vector fc.

[0081] The method for determining the second attention matrix can be the same as the method for determining the first attention matrix described above, which will not be described here. The method for determining the first negative feedback vector can be the same as the method for determining the first positive feedback vector described above, which will not be described here.

[0082] In step 203, based on the first positive feedback vector, the first negative feedback vector, and the negative feedback behavior sequence, the second positive feedback vector and the second negative feedback vector are determined.

[0083] Exemplarily, the first positive feedback vector can represent a positive preference corresponding to any search result included in the positive feedback behavior sequence, and the first negative feedback vector can represent a negative preference corresponding to any search result included in the negative feedback behavior sequence. However, there can still be some positive preferences in the negative feedback behavior sequence, for example, in an e-commerce scenario, the reason why the user does not click on a search result can be that the user is not interested in the search result, or the user does not click on the search result because the user does not see the search result. Therefore, it is also necessary to distinguish the real positive preference or real negative preference corresponding to each search result included in the negative feedback behavior sequence.

[0084] Optionally, determining the second positive feedback vector and the second negative feedback vector based on the first positive feedback vector, the first negative feedback vector and the negative feedback behavior sequence comprises: determining a first weight based on the first positive feedback vector and the first negative feedback vector; determining the second positive feedback vector based on the first weight and the negative feedback behavior sequence; and determining the second negative feedback vector based on the first weight and the negative feedback behavior sequence.

[0085] Exemplarily, the first positive feedback vector and the negative feedback behavior sequence {u1, u2, …, un} can be subjected to an outer product and a splicing operation and then input into an MLP (Multilayer Perceptron) to obtain an interaction score w (for example, w is a positive integer), and w is normalized, and finally the first weight a can be obtained based on w. n1 p p p p p

[0086]

[0087] wherein the subscript p is a positive integer from 1 to n1. The embodiments of the present application do not limit the MLP, as long as the normalized w can be output at last. For example, the MLP can include two fully connected layers, and the method for determining w is shown in FIG. 8. First, the first positive feedback vector fc and the negative feedback behavior sequence {u1, u2, …, un} are subjected to an outer product and a splicing operation, and then the result is input into two fully connected layers to obtain the normalized w, and finally the first weight a is obtained based on w. p p Figure 4 n1 p p

[0088] Optionally, determining the second positive feedback vector based on the first weight and the negative feedback behavior sequence comprises: weighting at least one feature vector included in the negative feedback behavior sequence by the first weight to obtain the second positive feedback vector. ​​​​​​​​​​​​

[0089] Exemplarily, the first weight α p Afterwards, each feature vector in the negative feedback behavior sequence {u1, u2, …, u n1} can be weighted and added to obtain a second positive feedback vector f uc+ :

[0090]

[0091] The second positive feedback vector f uc+ may represent the real positive preference corresponding to the search result included in the negative feedback behavior sequence.

[0092] Determining the second negative feedback vector based on the first weight and the negative feedback behavior sequence includes: determining a second weight based on the first weight; weighting at least one feature vector included in the negative feedback behavior sequence by the second weight to obtain the second negative feedback vector.

[0093] Exemplarily, the first weight α p may be reversed to obtain the second weight β p :

[0094]

[0095] The second weight β p is further weighted with each feature vector in the negative feedback behavior sequence {u1, u2, …, u n1} and added to obtain a second negative feedback vector f uc- :

[0096]

[0097] The second negative feedback vector f uc- may represent the real negative preference corresponding to the search result included in the negative feedback behavior sequence.

[0098] In the embodiments of the present application, the influence of the first positive feedback vector existing in the positive feedback behavior sequence and the first negative feedback vector, the second negative feedback vector and the second positive feedback vector existing in the negative feedback behavior sequence on the ranking is comprehensively considered, so as to improve the reliability of the ranking result.

[0099] In step 204, the ranking result of the search result corresponding to the input content is determined based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector.

[0100] Optionally, determining the ranking result of the search result corresponding to the input content based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector comprises: splicing the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector to obtain a positive-negative feedback vector; and determining the ranking result based on the positive-negative feedback vector.

[0101] Exemplarily, the first positive feedback vector, the first negative feedback vector, the second positive feedback vector and the second negative feedback vector can be spliced to obtain the positive-negative feedback vector, and then the positive-negative feedback vector and other feature vectors are input into the neural network model for joint training, and then the trained neural network model can be used to rank the search result corresponding to the input content to obtain the ranking result. The present embodiment does not limit the other feature vectors, which can be set by the user according to the situation. The present embodiment also does not limit the type of neural network model, for example, the neural network model can be DeepFM (Deep Factorization Machines) or Wide&Deep (linear and deep model).

[0102] In the present embodiment, the positive feedback behavior sequence, the negative feedback behavior sequence and the target feature vector are obtained, and the influence of the first positive feedback vector in the positive feedback behavior sequence and the first negative feedback vector, the second negative feedback vector and the second positive feedback vector in the negative feedback behavior sequence on ranking is comprehensively considered, thereby improving the reliability of the ranking result.

[0103] The present embodiment can combine the business characteristics of the QAC level skip bias, propose to determine the first negative feedback vector based on the negative feedback behavior sequence, improve the confidence of the ranking result, and facilitate maintenance.

[0104] The present embodiment can determine the first positive feedback vector and the first negative feedback vector through the target feature vector, the negative feedback behavior sequence and the positive feedback behavior sequence, and distinguish the real positive preference or real negative preference corresponding to each search result included in the negative feedback behavior sequence.

[0105] The present embodiment can be deployed on a terminal to re-rank according to the continuously input content, and can achieve the effect of timely adjusting the ranking result. When the performance allows, it can also be deployed on a server, such as being reused in other search recommendation fields.

[0106] As shown in Figure 5 The present embodiment provides a method for determining a positive-negative feedback vector, which comprises:

[0107] 501, obtain the positive feedback behavior sequence, the negative feedback behavior sequence, and the target feature vector. The implementation of this step can refer to step 201 described above, and details are not repeated here.

[0108] 502, perform multi-head attention calculation and average pooling based on the positive feedback behavior sequence and the target feature vector to determine a first positive feedback vector f c ; perform multi-head attention calculation and average pooling based on the negative feedback behavior sequence and the target feature vector to determine a first negative feedback vector f u The implementation of this step can refer to step 202 described above, and details are not repeated here.

[0109] 503, determine the first weight. The implementation of this step can refer to step 203 described above, and details are not repeated here.

[0110] 504, weight the at least one feature vector included in the negative feedback behavior sequence by the first weight to obtain f uc+ The implementation of this step can refer to step 203 described above, and details are not repeated here.

[0111] 505, determine the second weight based on the first weight. The implementation of this step can refer to step 203 described above, and details are not repeated here.

[0112] 506, weight the at least one feature vector included in the negative feedback behavior sequence by the second weight to obtain f uc- The implementation of this step can refer to step 203 described above, and details are not repeated here.

[0113] 507, determine the positive and negative feedback vector based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector. The implementation of this step can refer to step 203 described above, and details are not repeated here.

[0114] Referring to Figure 6 The embodiment of the present application provides a sorting device, which comprises:

[0115] The acquisition module 601 is configured to acquire a positive feedback behavior sequence, a negative feedback behavior sequence, and a target feature vector. The positive feedback behavior sequence comprises a feature vector corresponding to a search result that has been clicked at least once in at least one previous sorting before the present sorting. The negative feedback behavior sequence comprises a feature vector corresponding to at least one search result that has not been clicked in the present search result. The target feature vector is a feature vector corresponding to a search result that has appeared in the present input process with the same search target.

[0116] The first determination module 602 is configured to determine a first positive feedback vector based on the positive feedback behavior sequence and the target feature vector; and determine a first negative feedback vector based on the negative feedback behavior sequence and the target feature vector.

[0117] The second determination module 603 is configured to determine a second positive feedback vector and a second negative feedback vector based on the first positive feedback vector, the first negative feedback vector, and the negative feedback behavior sequence.

[0118] The third determination module 604 is configured to determine a ranking result of the search result corresponding to the input content based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector.

[0119] Optionally, the first determination module 602 is configured to determine a first attention matrix based on the positive feedback behavior sequence and the target feature vector, and determine the first positive feedback vector based on the first attention matrix.

[0120] The first determination module 602 is configured to determine a second attention matrix based on the negative feedback behavior sequence and the target feature vector, and determine the first negative feedback vector based on the second attention matrix.

[0121] Optionally, the second determination module 603 is configured to determine a first weight based on the first positive feedback vector and the first negative feedback vector, determine the second positive feedback vector based on the first weight and the negative feedback behavior sequence, and determine the second negative feedback vector based on the first weight and the negative feedback behavior sequence.

[0122] Optionally, the second determination module 603 is configured to weight at least one feature vector included in the negative feedback behavior sequence by the first weight to obtain the second positive feedback vector.

[0123] Optionally, the second determination module 603 is configured to determine a second weight based on the first weight, and weight at least one feature vector included in the negative feedback behavior sequence by the second weight to obtain the second negative feedback vector.

[0124] Optionally, the third determination module 604 is configured to concatenate the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector to obtain a positive-negative feedback vector.

[0125] The ranking result is determined based on the positive-negative feedback vector.

[0126] Optionally, the feature vector corresponding to any search result in the positive feedback behavior sequence or the negative feedback behavior sequence is used to represent at least one of the following information: title information of the search result, total exposure time information of the search result, maximum exposure time information of the search result, information about the number of times that the text input in the search box during the input process is at the word boundary of the search result, and information about the number of times that the search result is exposed at different positions.

[0127] The target feature vector is used to represent at least one of title information of a search result that has appeared in the current input process with the search target, total exposure time information of the search result, maximum exposure time information of the search result, the number of times that a word input in the search box in the input process is at the word boundary of the search result, and the number of times that the search result is exposed at different positions.

[0128] In the embodiment of the present application, the positive feedback behavior sequence, the negative feedback behavior sequence and the target feature vector are obtained, and the influence of the first positive feedback vector in the positive feedback behavior sequence and the first negative feedback vector, the second negative feedback vector and the second positive feedback vector in the negative feedback behavior sequence on ranking is comprehensively considered, thereby improving the reliability of the ranking result.

[0129] It should be noted that the device provided in the above embodiments is only used as an example to divide the above functional modules when realizing the functions thereof, and in actual applications, the above functions can be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the above described functions. In addition, the device and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process is detailed in the method embodiments, which will not be described here.

[0130] Figure 7 It is a computer device structure schematic diagram provided by an embodiment of the present application. The server can be quite different due to different configurations or performances, and can include one or more processors (Central Processing Units, CPU) 701 and one or more memories 702. At least one computer program is stored in the one or more memories 702, which is loaded and executed by the one or more processors 701, so that the server implements the ranking method provided by each method embodiment described above. Of course, the server can also have a wired or wireless network interface, a keyboard, an input and output interface and other components for realizing the functions of the device, so as to input and output. The server can also include other components for realizing the functions of the device, which will not be described here.

[0131] Figure 8is a schematic diagram of a computer device structure provided by an embodiment of the present application. The device can be a terminal, for example, can be: a smart phone, a tablet computer, an MP3 (Moving Picture Experts Group Audio Layer III) player, an MP4 (Moving Picture Experts Group Audio Layer IV) player, a notebook computer or a desktop computer. The terminal can also be referred to as a user equipment, a portable terminal, a laptop terminal, a desktop terminal, and other names.

[0132] Generally, the terminal includes a processor 801 and a memory 802.

[0133] The processor 801 can include one or more processing cores, such as a 4-core processor, an 8-core processor, etc. The processor 801 can be implemented in at least one of a hardware form of a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), a PLA (Programmable Logic Array). The processor 801 can also include a main processor and a coprocessor, the main processor is a processor for processing data in an awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor for processing data in a standby state. In some embodiments, the processor 801 can be integrated with a GPU (Graphics Processing Unit) for rendering and drawing the content required to be displayed by the display screen. In some embodiments, the processor 801 can also include an AI (Artificial Intelligence) processor for processing machine learning related computing operations.

[0134] The memory 802 can include one or more computer-readable storage media, which can be non-transitory. The memory 802 can also include a high-speed random access memory, and a non-volatile memory such as one or more disk storage devices, flash storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory 802 is used to store at least one instruction for being executed by the processor 801 to enable the terminal to implement the sorting method provided by the method embodiment of the present application.

[0135] In some embodiments, the terminal can further optionally include a peripheral device interface 803 and at least one peripheral device. The processor 801, the memory 802 and the peripheral device interface 803 can be connected through a bus or a signal line. Each peripheral device can be connected to the peripheral device interface 803 through a bus, a signal line or a circuit board. Specifically, the peripheral device includes at least one of a radio frequency circuit 804, a display screen 805, a camera component 806, an audio circuit 807, a positioning component 808 and a power supply 809.

[0136] The peripheral device interface 803 can be used to connect at least one peripheral device related to I / O (Input / Output) to the processor 801 and the memory 802. In some embodiments, the processor 801, the memory 802 and the peripheral device interface 803 are integrated on the same chip or circuit board; in some other embodiments, any one or two of the processor 801, the memory 802 and the peripheral device interface 803 can be implemented on a separate chip or circuit board, and the present embodiments are not limited in this regard.

[0137] The radio frequency circuit 804 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The radio frequency circuit 804 communicates with a communication network and other communication devices through electromagnetic signals. The radio frequency circuit 804 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals into electrical signals. Optionally, the radio frequency circuit 804 includes an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a subscriber identity module card, and the like. The radio frequency circuit 804 can communicate with other terminals through at least one wireless communication protocol. The wireless communication protocol includes but is not limited to metropolitan area networks, various generations of mobile communication networks (2G, 3G, 4G and 5G), wireless local area networks and / or WiFi (Wireless Fidelity) networks. In some embodiments, the radio frequency circuit 804 can also include NFC (Near Field Communication) related circuitry, which is not limited in the present application.

[0138] The display screen 805 is configured to display a UI (User Interface). The UI can include graphics, text, icons, video, and any combination thereof. When the display screen 805 is a touch display screen, the display screen 805 is further configured to capture touch signals on or above the surface of the display screen 805. The touch signals can be input to the processor 801 as control signals for processing. In this case, the display screen 805 can also be configured to provide virtual buttons and / or virtual keyboard, also known as soft buttons and / or soft keyboard. In some embodiments, the display screen 805 can be one, disposed on the front panel of the terminal; in other embodiments, the display screen 805 can be at least two, respectively disposed on different surfaces of the terminal or in a folding design; in other embodiments, the display screen 805 can be a flexible display screen, disposed on a curved surface or a folding surface of the terminal. Even, the display screen 805 can also be disposed in an irregular shape, i.e., a special-shaped screen. The display screen 805 can be made of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), etc.

[0139] The camera assembly 806 is configured to capture images or videos. Optionally, the camera assembly 806 includes a front camera and a rear camera. Typically, the front camera is disposed on the front panel of the terminal, and the rear camera is disposed on the back of the terminal. In some embodiments, the rear camera is at least two, which are any one of a main camera, a depth-of-field camera, a wide-angle camera, and a telephoto camera, to realize the background blur function by fusing the main camera and the depth-of-field camera, the panoramic shooting and VR (Virtual Reality) shooting function by fusing the main camera and the wide-angle camera, or other fusion shooting functions. In some embodiments, the camera assembly 806 can further include a flash. The flash can be a single-color temperature flash or a dual-color temperature flash. The dual-color temperature flash refers to the combination of a warm light flash and a cold light flash, which can be used for light compensation under different color temperatures.

[0140] The audio circuit 807 can include a microphone and a speaker. The microphone is used to collect sound waves of a user and an environment, and convert the sound waves into an electrical signal input to the processor 801 for processing, or input to the radio frequency circuit 804 to realize voice communication. For the purpose of stereo sound collection or noise reduction, the microphone can be multiple, and arranged at different parts of the terminal. The microphone can also be an array microphone or an omnidirectional collection type microphone. The speaker is used to convert an electrical signal from the processor 801 or the radio frequency circuit 804 into sound waves. The speaker can be a conventional diaphragm speaker, or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, not only can it convert an electrical signal into a sound wave audible to humans, but it can also convert an electrical signal into an inaudible sound wave to humans for ranging purposes, etc. In some embodiments, the audio circuit 807 can also include a headphone jack.

[0141] The positioning component 808 is used to position the current geographic location of the terminal to realize navigation or LBS (Location Based Service, Location Based Service). The positioning component 808 can be a positioning component based on the GPS (Global Positioning System, Global Positioning System) of the United States, the Beidou system of China, the Glonass system of Russia or the Galileo system of the European Union.

[0142] The power supply 809 is used to supply power to each component in the terminal. The power supply 809 can be alternating current, direct current, disposable battery or rechargeable battery. When the power supply 809 includes a rechargeable battery, the rechargeable battery can support wired charging or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0143] In some embodiments, the terminal also includes one or more sensors 810. The one or more sensors 810 include but are not limited to: an acceleration sensor 811, a gyroscope sensor 812, a pressure sensor 813, a fingerprint sensor 814, an optical sensor 815 and a proximity sensor 816.

[0144] The acceleration sensor 811 can detect the acceleration in three coordinate axes of the coordinate system established by the terminal. For example, the acceleration sensor 811 can be used to detect the components of gravitational acceleration in three coordinate axes. The processor 801 can control the display screen 805 to display the user interface in a landscape view or a portrait view according to the gravitational acceleration signal collected by the acceleration sensor 811. The acceleration sensor 811 can also be used for game or user motion data collection.

[0145] The gyro sensor 812 can detect the body direction and rotation angle of the terminal, and the gyro sensor 812 can collect 3D motions of the user on the terminal in cooperation with the acceleration sensor 811. The processor 801 can implement the following functions according to the data collected by the gyro sensor 812: motion sensing (e.g., changing a UI according to a tilt operation of the user), image stabilization when shooting, game control, and inertial navigation.

[0146] The pressure sensor 813 can be disposed at the side frame of the terminal and / or the lower layer of the display 805. When the pressure sensor 813 is disposed at the side frame of the terminal, the grip signal of the user on the terminal can be detected, and the left / right hand recognition or shortcut operation can be performed by the processor 801 according to the grip signal collected by the pressure sensor 813. When the pressure sensor 813 is disposed at the lower layer of the display 805, the operable control on the UI interface can be controlled by the processor 801 according to the pressure operation of the user on the display 805. The operable control includes at least one of a button control, a scroll bar control, an icon control, and a menu control.

[0147] The fingerprint sensor 814 is used to collect the fingerprint of the user, and the identity of the user can be recognized by the processor 801 according to the fingerprint collected by the fingerprint sensor 814, or by the fingerprint sensor 814 according to the collected fingerprint. When the identity of the user is recognized as a trusted identity, the processor 801 authorizes the user to perform a related sensitive operation, which includes unlocking the screen, viewing encrypted information, downloading software, payment, and changing settings, etc. The fingerprint sensor 814 can be disposed at the front, back, or side of the terminal. When the terminal is provided with a physical button or a manufacturer logo (trademark), the fingerprint sensor 814 can be integrated with the physical button or the manufacturer logo.

[0148] The optical sensor 815 is used to collect the ambient light intensity. In one embodiment, the processor 801 can control the display brightness of the display 805 according to the ambient light intensity collected by the optical sensor 815. Specifically, when the ambient light intensity is high, the display brightness of the display 805 is increased, and when the ambient light intensity is low, the display brightness of the display 805 is decreased. In another embodiment, the processor 801 can also dynamically adjust the shooting parameters of the camera assembly 806 according to the ambient light intensity collected by the optical sensor 815.

[0149] The proximity sensor 816, also referred to as a distance sensor, is usually arranged on the front panel of the terminal. The proximity sensor 816 is used to collect the distance between the user and the front of the terminal. In one embodiment, when the proximity sensor 816 detects that the distance between the user and the front of the terminal gradually decreases, the display screen 805 is switched from the bright screen state to the screen-off state under the control of the processor 801; when the proximity sensor 816 detects that the distance between the user and the front of the terminal gradually increases, the display screen 805 is switched from the screen-off state to the bright screen state under the control of the processor 801.

[0150] Those skilled in the art can understand that the structure shown in the foregoing embodiments does not constitute a limitation on the terminal, and the terminal can include more or fewer components than those shown in the drawings, or combine certain components, or adopt a different component arrangement. Figure 8

[0151] In an example embodiment, a computer device is also provided, which includes a processor and a memory having at least one computer program stored therein. The at least one computer program is loaded and executed by one or more processors to enable the computer device to implement any of the above sorting methods.

[0152] In an example embodiment, a computer readable storage medium is also provided, which stores at least one computer program. The at least one computer program is loaded and executed by the processor of the computer device to enable the computer to implement any of the above sorting methods.

[0153] In a possible implementation manner, the above computer readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0154] In an example embodiment, a computer program product or computer program is also provided, which includes computer instructions stored in a computer readable storage medium. The processor of the computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to implement any of the above sorting methods.

[0155] ​It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to searched data, search results, data for analysis, stored data, displayed data, etc.) and signals involved in the present application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data need to comply with relevant laws, regulations and standards of relevant countries and regions. For example, the search results and the like clicked by the user in the embodiments of the present application are obtained under full authorization.

[0156] It should be understood that "multiple" mentioned herein refers to two or more. The "and / or" describes the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally represents that the associated objects before and after it are in an "or" relationship.

[0157] The above only describes exemplary embodiments of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A sorting method, characterized in that, The method includes: Obtain a positive feedback behavior sequence, a negative feedback behavior sequence, and a target feature vector. The positive feedback behavior sequence includes the feature vectors corresponding to the search results that have been clicked at least once in at least one previous sorting. The negative feedback behavior sequence includes the feature vectors corresponding to at least one search result that has not been clicked in the current search results. The target feature vector is the feature vector corresponding to the search results that have appeared in the current input process and are the same as the search target. A first positive feedback vector is determined based on the positive feedback behavior sequence and the target feature vector; a first negative feedback vector is determined based on the negative feedback behavior sequence and the target feature vector. The second positive feedback vector and the second negative feedback vector are determined based on the first positive feedback vector, the first negative feedback vector, and the negative feedback behavior sequence. The sorting result of the search results corresponding to the input content is determined based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector. Determining the first positive feedback vector based on the positive feedback behavior sequence and the target feature vector includes: determining a first attention matrix based on the positive feedback behavior sequence and the target feature vector; and determining the first positive feedback vector based on the first attention matrix. Determining the first negative feedback vector based on the negative feedback behavior sequence and the target feature vector includes: determining a second attention matrix based on the negative feedback behavior sequence and the target feature vector; and determining the first negative feedback vector based on the second attention matrix. The step of determining the second positive feedback vector and the second negative feedback vector based on the first positive feedback vector, the first negative feedback vector, and the negative feedback behavior sequence includes: A first weight is determined based on the first positive feedback vector and the first negative feedback vector; a second positive feedback vector is determined based on the first weight and the negative feedback behavior sequence; a second negative feedback vector is determined based on the first weight and the negative feedback behavior sequence. The step of determining the second positive feedback vector based on the first weight and the negative feedback behavior sequence includes: The first weight is used to weight at least one feature vector of the negative feedback behavior sequence to obtain the second positive feedback vector; The step of determining the second negative feedback vector based on the first weight and the negative feedback behavior sequence includes: A second weight is determined based on the first weight; at least one feature vector included in the negative feedback behavior sequence is weighted by the second weight to obtain the second negative feedback vector.

2. The method according to claim 1, characterized in that, The process of determining the sorting results of the search results corresponding to the input content based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector includes: The first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector are concatenated to obtain the positive and negative feedback vectors. The ranking of the search results corresponding to the input content is determined based on the positive and negative feedback vectors.

3. The method according to claim 1 or 2, characterized in that, The feature vector corresponding to any retrieval result in the positive feedback behavior sequence and the negative feedback behavior sequence is used to represent at least one of the following: the title information of the retrieval result, the total exposure time information of the retrieval result, the maximum exposure time information of the retrieval result, the number of times the text entered in the search box during the input process is at the word boundary of the retrieval result, and the number of times the retrieval result is exposed in different positions. The target feature vector is used to represent at least one of the following: title information of the search results that appeared during the current input process with the same search target; total exposure time information of the search results; maximum exposure time information of the search results; number of times the text entered in the search box during the input process was at the word boundary of the search results; and number of times the search results were exposed in different positions.

4. A sorting device, characterized in that, The device includes: The acquisition module is used to acquire a positive feedback behavior sequence, a negative feedback behavior sequence, and a target feature vector. The positive feedback behavior sequence includes the feature vectors corresponding to the search results that have been clicked at least once in at least one sorting before this sorting. The negative feedback behavior sequence includes the feature vectors corresponding to at least one search result that has not been clicked in this search result. The target feature vector is the feature vector corresponding to the search result that appears in this input process and is the same as the search target. The first determining module is used to determine a first positive feedback vector based on the positive feedback behavior sequence and the target feature vector; and to determine a first negative feedback vector based on the negative feedback behavior sequence and the target feature vector. The second determining module is used to determine the second positive feedback vector and the second negative feedback vector based on the first positive feedback vector, the first negative feedback vector and the negative feedback behavior sequence; The third determining module is used to determine the sorting result of the search results corresponding to the input content based on the first positive feedback vector, the first negative feedback vector, the second positive feedback vector, and the second negative feedback vector; Determining the first positive feedback vector based on the positive feedback behavior sequence and the target feature vector includes: determining a first attention matrix based on the positive feedback behavior sequence and the target feature vector; and determining the first positive feedback vector based on the first attention matrix. Determining the first negative feedback vector based on the negative feedback behavior sequence and the target feature vector includes: determining a second attention matrix based on the negative feedback behavior sequence and the target feature vector; and determining the first negative feedback vector based on the second attention matrix. The step of determining the second positive feedback vector and the second negative feedback vector based on the first positive feedback vector, the first negative feedback vector, and the negative feedback behavior sequence includes: A first weight is determined based on the first positive feedback vector and the first negative feedback vector; a second positive feedback vector is determined based on the first weight and the negative feedback behavior sequence; a second negative feedback vector is determined based on the first weight and the negative feedback behavior sequence. The step of determining the second positive feedback vector based on the first weight and the negative feedback behavior sequence includes: The first weight is used to weight at least one feature vector of the negative feedback behavior sequence to obtain the second positive feedback vector; The step of determining the second negative feedback vector based on the first weight and the negative feedback behavior sequence includes: A second weight is determined based on the first weight; at least one feature vector included in the negative feedback behavior sequence is weighted by the second weight to obtain the second negative feedback vector.

5. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to enable the computer device to implement the sorting method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to enable the computer to implement the sorting method as described in any one of claims 1 to 3.

7. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium, a processor of a computer device reading the computer instructions from the computer-readable storage medium, and the processor executing the computer instructions to cause the computer device to perform the sorting method as described in any one of claims 1 to 3.

Images