Guiding style determination and model training method and device, equipment and medium

Abstract

The invention provides a guide style determination and model training method and device, equipment and a medium, and relates to the technical field of artificial intelligence, in particular to the technical fields of intelligent recommendation, deep learning and the like. The guidance style determination method comprises the following steps: encoding a plurality of preset intervention values and target data to obtain target features; each intervention value is used for identifying one candidate guide style in multiple candidate guide styles, and the target data is used for representing target object information guided by the multiple candidate guide styles; decoding the target feature to obtain a target index value corresponding to each intervention value; and based on the target index value, determining a target guide pattern in the plurality of candidate guide patterns.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to the technical fields of intelligent recommendation, deep learning, etc., and in particular to a method, device, equipment and medium for guiding style determination and model training. Background Art

[0002] In order to stimulate user interest, a plurality of candidate guidance styles may be provided for a certain navigation item, and a final target guidance style may be determined from the plurality of candidate guidance styles for a certain user and provided to the user.

[0003] Therefore, it is necessary to solve the problem of how to determine the target guidance style among multiple candidate guidance styles. Summary of the invention

[0004] The present disclosure provides a guidance style determination and model training method, device, equipment and medium.

[0005] According to one aspect of the present disclosure, a guidance style determination method is provided, comprising: encoding a plurality of preset intervention values ​​and target data to obtain a target feature; each intervention value is used to identify a candidate guidance style among a plurality of candidate guidance styles, and the target data is used to characterize target object information guided by the plurality of candidate guidance styles; decoding the target feature to obtain a target indicator value corresponding to each intervention value; and determining a target guidance style among the plurality of candidate guidance styles based on the target indicator value.

[0006] According to another aspect of the present disclosure, a method for training a guidance style determination model is provided, the model comprising a target decoding layer corresponding to a target intervention value, the target intervention value being used to identify a target candidate guidance style among a plurality of candidate guidance styles, the method comprising: encoding a target data sample and the target intervention value to obtain a target feature; the target data sample being used to characterize target object information guided by the plurality of candidate guidance styles; using the target decoding layer to decode the target feature to obtain a prediction index value corresponding to the target intervention value; constructing a loss function based on the prediction index value and a true index value corresponding to the target guidance style; and adjusting parameters of the target decoding layer based on the loss function.

[0007] According to another aspect of the present disclosure, a guidance style determination device is provided, including: an encoding module, used to encode a plurality of preset intervention values ​​and target data to obtain a target feature; each intervention value is used to identify a candidate guidance style among a plurality of candidate guidance styles, and the target data is used to characterize target object information guided by the plurality of candidate guidance styles; a decoding module, used to decode the target feature to obtain a target indicator value corresponding to each intervention value; and a determination module, used to determine a target guidance style among the plurality of candidate guidance styles based on the target indicator value.

[0008] According to another aspect of the present disclosure, a guidance style determination model training device is provided, the model includes a target decoding layer corresponding to a target intervention value, the target intervention value is used to identify a target candidate guidance style among multiple candidate guidance styles, and the device includes: an encoding module, used to encode a target data sample and the target intervention value to obtain a target feature; the target data sample is used to characterize the target object information guided by the multiple candidate guidance styles; a decoding module, used to decode the target feature using the target decoding layer to obtain a prediction index value corresponding to the target intervention value; a construction module, used to construct a loss function based on the prediction index value and the true index value corresponding to the target guidance style; an adjustment module, used to adjust the parameters of the target decoding layer based on the loss function.

[0009] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any method as described in any of the above aspects.

[0010] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute any one of the methods according to any one of the above aspects.

[0011] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the computer program implements any one of the methods described in any one of the above aspects.

[0012] According to the embodiments of the present disclosure, the accuracy of guiding pattern determination can be improved.

[0013] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0015] Figure 1 is a schematic diagram according to a first embodiment of the present disclosure;

[0016] Figure 2 is a schematic diagram of an application scenario for implementing an embodiment of the present disclosure;

[0017] Figure 3 is a schematic diagram of a guide style provided according to an embodiment of the present disclosure;

[0018] Figure 4 is a schematic diagram of a guidance style determination model provided according to an embodiment of the present disclosure;

[0019] Figure 5 is a schematic diagram according to a second embodiment of the present disclosure;

[0020] Figure 6 is a schematic diagram according to a third embodiment of the present disclosure;

[0021] Figure 7 is a schematic diagram according to a fourth embodiment of the present disclosure;

[0022] Figure 8 is a schematic diagram according to a fifth embodiment of the present disclosure;

[0023] Fig. 9 It is a schematic diagram of an electronic device used to implement the guidance style determination method or guidance style determination model training method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0024] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0025] Some current applications (Application, APP) provide a bottom navigation bar, which includes multiple navigation items, one of which is "Video". For the navigation item "Video", multiple candidate reference styles are provided, which can be respectively referred to as the first guidance style, the second guidance style and the third guidance style. Specifically, the first guidance style is "based on author + avatar" for guidance, which displays the author's avatar in the guidance style, which is more suitable for the situation where the user has a clear preference for a specific author; the second guidance style is "based on author + update" for guidance, which only displays the updates in the guidance style, but does not display the avatar of the author of the update, which is more suitable for the situation where the user pays more attention to the updates; the third guidance style is "based on points of interest" for guidance, which usually displays related content of the content that the user has searched for in the form of bubbles.

[0026] Among them, the first guide style and the second guide style can be collectively referred to as author-based guide, and the third guide style is called other guide.

[0027] In the related art, a hierarchical binary classification model is usually used to make guidance style decisions. Specifically, a binary classification model is first used to determine "author-based guidance" or "other guidance" (the third guidance style) based on the target data. If "author-based guidance" is determined, a second binary classification model is used to determine "author + avatar-based guidance" (first guidance style) or "author + update-based guidance" (second guidance style) based on the target data.

[0028] In this way, there is a problem of information separation between the two binary classification models, which leads to the problem of insufficient accuracy in the determined target guidance style.

[0029] In order to improve the accuracy of guiding pattern determination, the present disclosure provides the following embodiments.

[0030] Figure 1 is a schematic diagram according to the first embodiment of the present disclosure. This embodiment provides a method for determining a guide style, such as Figure 1 As shown, the method includes:

[0031] 101. Encode a plurality of preset intervention values ​​and target data to obtain target features; each intervention value is used to identify a candidate guidance style among a plurality of candidate guidance styles, and the target data is used to characterize target object information guided by the plurality of candidate guidance styles.

[0032] 102. Decode the target feature to obtain a target indicator value corresponding to each intervention value.

[0033] 103. Determine a target guidance style from among the multiple candidate guidance styles based on the target indicator value.

[0034] Among them, multiple candidate guide styles are known. For example, based on the above-mentioned "video" navigation item scenario, there are three candidate guide styles, which are respectively called the first guide style, the second guide style and the third guide style.

[0035] The number of intervention values ​​is the same as the number of candidate guidance patterns, and the specific values ​​of the intervention values ​​are preset.

[0036] Based on the above example, there are three intervention values. Assuming that the intervention value is represented by t, the specific values ​​may be t=0, 1, and 2, respectively identifying a candidate guidance style.

[0037] Target data is used to represent the target object information guided by the candidate guidance style.

[0038] Based on the above example, the candidate guidance style is used to guide the user to watch the video, and for this purpose, the target object information may include video information and / or user information of the candidate video. Video information includes video title, images in the video, etc., and user information includes user gender, age, historical search content, etc.

[0039] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0040] After obtaining the intervention value and target data, these data are encoded to obtain the corresponding vector representation, which is called the target feature.

[0041] After obtaining the target features, they are decoded to obtain the target indicator value corresponding to each intervention value.

[0042] The type of the target indicator value is set. For example, if the target indicator value is click-through rate, the specific value of the click-through rate is obtained. The click-through rate refers to the probability that a user clicks on the "video" navigation item.

[0043] Based on the above example, the target index values ​​corresponding to t=0, 1, and 2 can be obtained respectively.

[0044] After obtaining the target indicator value corresponding to each intervention value, a guidance style decision is made based on the target indicator value. For example, one of the three candidate guidance styles is determined as the target guidance style.

[0045] In this embodiment, the target feature is decoded to obtain the target indicator value corresponding to each intervention value, and the target guidance style is determined based on the target indicator value. The target feature is obtained based on the intervention value and the target data. In this way, information from multiple dimensions can be integrated to make guidance style decisions, thereby improving the accuracy of the determined target guidance style. In addition, by introducing multiple preset intervention values, the intervention values ​​correspond to the candidate guidance styles one by one, and the candidate guidance styles can be identified, thereby further improving the accuracy of the determined target guidance style.

[0046] In order to better understand the present disclosure, the application scenarios involved in the present disclosure are described as follows:

[0047] Figure 2 It is a schematic diagram of an application scenario for implementing the embodiment of the present disclosure.

[0048] like Figure 2 As shown, the application scenario includes: a user terminal 201 and a server 202.

[0049] The user terminal 201 may include: a personal computer (PC), a mobile device (such as a mobile phone), a tablet computer, a laptop computer, a smart wearable device, etc.

[0050] An application (Application, APP) is deployed on the user terminal, such as a search engine APP. A navigation bar is set in the APP, such as a bottom navigation bar, and the navigation bar may include multiple navigation items. For a certain navigation item, multiple candidate guidance styles may be provided. Based on different users, a target guidance style may be determined from the multiple candidate guidance styles, and the target guidance style may be displayed to the user.

[0051] The server 202 interacts with the APP on the user terminal to determine a target guidance style from multiple candidate guidance styles, and sends information of the target guidance style to the APP, which then displays the target guidance style to the user.

[0052] The server may specifically adopt a guidance style determination model to process the input intervention value and target data, so as to determine a target guidance style from a plurality of candidate guidance styles.

[0053] Figure 3 It is a schematic diagram of a guide style provided according to an embodiment of the present disclosure.

[0054] Take search engine apps as an example. Figure 3 As shown, the bottom navigation bar of the APP may include navigation items such as "Video", "Message", and "My".

[0055] For the navigation item "video", a variety of candidate guide styles may be provided. In this embodiment, three candidate guide styles are taken as examples, which are respectively referred to as the first guide style, the second guide style and the third guide style.

[0056] Specifically, Figure 3 As shown, the first guidance style is the "author + avatar based" guidance, in which the avatar of the author of the content update is displayed; the second guidance style is the "author + update based" guidance, in which only the author of the content update is indicated, but the avatar of the author is not displayed; the third guidance style is the "interest point based" guidance, in which the content that the user may be interested in can be displayed in the form of bubbles.

[0057] For a certain user, it is necessary to determine a target guidance style from the three candidate guidance styles mentioned above, and display the target guidance style to the user.

[0058] In order to improve accuracy, in this embodiment, a guidance style determination model is used to determine a target guidance style from a plurality of candidate guidance styles.

[0059] Figure 4 It is a schematic diagram of a guidance style determination model provided according to an embodiment of the present disclosure.

[0060] The input of the guided style determination model includes: target data, and multiple preset intervention values; the output includes: the target indicator value corresponding to each intervention value.

[0061] like Figure 4 As shown, the model includes: an encoding layer and a decoding layer.

[0062] Specifically, the coding layer includes: an intervening coding layer and a non-intervening coding layer.

[0063] The non-interference coding layer is used to encode the target data and obtain non-interference features.

[0064] The intervention encoding layer is used to encode multiple intervention values ​​to obtain intervention features.

[0065] Different intervention values ​​can be encoded using the same intervention coding layer.

[0066] Due to different intervention values, different intervention characteristics can be obtained.

[0067] For example, the intervention value is represented by t, t=0, 1, 2. After encoding t=0 using the intervention coding layer, the first intervention feature is obtained. After encoding t=1 using the intervention coding layer, the second intervention feature is obtained. After encoding t=2 using the intervention coding layer, the third intervention feature is obtained.

[0068] After obtaining the intervention features and non-intervention features, they are concatenated to obtain the target features.

[0069] For example, both the intervention feature and the non-intervention feature are N-dimensional vectors, and after concatenation, a 2N-dimensional vector is obtained.

[0070] In addition, when splicing, the non-intervention feature is spliced ​​with each intervention feature separately. For example, the non-intervention feature is spliced ​​with the first intervention feature to obtain the first target feature, the non-intervention feature is spliced ​​with the second intervention feature to obtain the second target feature, and the non-intervention feature is spliced ​​with the third intervention feature to obtain the third target feature.

[0071] The decoding layers correspond to the intervention values ​​one by one. For example, if the intervention values ​​are three, then there are three decoding layers, and the target indicator values ​​corresponding to each intervention value are obtained respectively.

[0072] Taking three intervention values ​​as an example, three index values ​​can be obtained through three decoding layers (the first decoding layer to the third decoding layer), which are represented as y0, y1 and y2 respectively.

[0073] Specifically, the first decoding layer is used to decode the first target feature to obtain the first indicator value y0; the second decoding layer is used to decode the second target feature to obtain the second indicator value y1; the third decoding layer is used to decode the third target feature to obtain the third indicator value y2.

[0074] The above target indicator value may specifically be a click-through rate.

[0075] Based on this, the candidate guidance style corresponding to the maximum click rate can be used as the target guidance style.

[0076] For example, y0 is the largest. Since y0 corresponds to the first intervention value, and the first intervention value corresponds to the first candidate guidance style, the target guidance style is the first candidate guidance style.

[0077] In this way, a guidance style determination model may be used to determine a target index value corresponding to each intervention value based on the intervention value and target data, and a target guidance style may be determined from a plurality of candidate guidance styles based on the target index value.

[0078] In combination with the above application scenarios, the present disclosure also provides the following embodiments.

[0079] Figure 5 is a schematic diagram according to a second embodiment of the present disclosure. This embodiment provides a method for determining a guide style, the method comprising:

[0080] 501. Using the intervention coding layer in the pre-trained guidance style determination model, encode multiple preset intervention values ​​to obtain intervention features; each intervention value is used to identify a candidate guidance style among multiple candidate guidance styles.

[0081] The intervention value corresponds to the candidate guidance style one by one, and the specific value is set. For example, if there are three candidate guidance styles, the intervention value t=0, 1, 2 can be set.

[0082] After obtaining the intervention value, combined with Figure 4 , multiple intervention values ​​are input into the intervention encoding layer respectively, and the output is the intervention feature corresponding to each intervention value.

[0083] In this embodiment, the intervention value is encoded through the intervention coding layer in the model to obtain the intervention feature, which can utilize the excellent performance of the model to improve the accuracy of the intervention feature and further improve the accuracy of the guided style decision.

[0084] 502. Use a non-interference coding layer in a pre-trained guidance style determination model to encode the target data to obtain the non-interference feature; the target data is used to characterize the target object information guided by the multiple candidate guidance styles.

[0085] The target data is information related to the target object guided by the candidate guiding style. For example, if the guiding style is to guide the user to watch a video, the target data may include video information and / or user information.

[0086] After obtaining the target data, combined with Figure 4 , the target data is input into the non-interference encoding layer, and the output is the non-interference features corresponding to the target data.

[0087] In this embodiment, the target data is encoded through the non-interference coding layer in the model to obtain non-interference features. The excellent performance of the model can be used to improve the accuracy of the non-interference features, thereby improving the accuracy of the guided style decision.

[0088] 503. Concatenate the intervention feature and the non-intervention feature to obtain a target feature.

[0089] Combination Figure 4 , the non-intervention feature can be concatenated with each intervention feature separately to obtain the target feature corresponding to each intervention value.

[0090] In this embodiment, the target data is encoded to obtain non-intervention features, the intervention value is encoded to obtain intervention features, and the non-intervention features and intervention features are spliced ​​to obtain target features. The target data and intervention values ​​can be combined to obtain the target features, thereby improving the accuracy of the target features and thereby improving the accuracy of the guided style decision.

[0091] 504. Determine the decoding layer in the model using the pre-trained guided style to decode the target feature to obtain the target indicator value corresponding to each intervention value.

[0092] In this embodiment, the target features are encoded through the decoding layer in the model to obtain the target index value, which can utilize the excellent performance of the model to improve the accuracy of the target index value, thereby improving the accuracy of the guided style decision.

[0093] Among them, combined Figure 4 There are multiple decoding layers, each decoding layer corresponds to an intervention value. At this time, each decoding layer can be used to decode the target feature respectively to obtain the target indicator value corresponding to each intervention value.

[0094] For example, if there are three intervention values, three decoding layers are used to decode the target features respectively to obtain three target indicator values.

[0095] In this embodiment, by setting decoding layers corresponding to intervention values ​​one by one, the target indicator value corresponding to each intervention value is obtained through each decoding layer. In this way, each decoding layer can be trained independently during the training stage to improve the accuracy of each decoding layer, thereby improving the accuracy of the guided style decision.

[0096] 505. Determine a target guidance style from among the multiple candidate guidance styles based on the target indicator value.

[0097] Among them, taking the target indicator value being the click-through rate as an example, the intervention value corresponding to the maximum click-through rate can be used as the target intervention value; and the candidate guidance style corresponding to the target intervention value can be used as the target guidance style.

[0098] For example, combined with Figure 4 , y0 is the largest. Assuming that y0 corresponds to the first intervention value, the first intervention value is used to identify the first candidate guidance style, then the target guidance style is the first candidate guidance style.

[0099] In this embodiment, the candidate guidance style corresponding to the maximum click rate is used as the target guidance style. In the scenario of guiding the user to watch a video, the user operation can be simplified and the user experience can be improved.

[0100] Figure 6 is a schematic diagram according to a third embodiment of the present disclosure. This embodiment provides a method for training a guidance style determination model, wherein the model includes a target decoding layer corresponding to a target intervention value, wherein the target intervention value is used to identify a target candidate guidance style among multiple candidate guidance styles, and the method includes:

[0101] 601. Encode a target data sample and the target intervention value to obtain a target feature; the target data sample is used to characterize target object information guided by the multiple candidate guidance styles.

[0102] 602. Use the target decoding layer to decode the target feature to obtain a prediction index value corresponding to the target intervention value.

[0103] 603. Construct a loss function based on the predicted index value and the real index value corresponding to the target guidance style.

[0104] 604. Adjust parameters of the target decoding layer based on the loss function.

[0105] In the training phase, each decoding layer is trained independently.

[0106] For example, combined with Figure 4 , there are three decoding layers, then these three decoding layers can be trained separately.

[0107] The currently trained decoding layer is called the target decoding layer, and the corresponding intervention value and candidate guidance style are called the target intervention value and target guidance style.

[0108] For example, if the target intervention value is t=0, the target data sample and the target intervention value t=0 are encoded to obtain the target feature, and the target feature is decoded by the first decoding layer to obtain the prediction index value corresponding to t=0, such as y0'. After that, a loss function is constructed based on y0' and the true index value y0* corresponding to the first guided style (which can be obtained in advance by marking, etc.), and the parameters of the first decoding layer are adjusted based on the loss function until the preset end condition is reached, thereby obtaining the final first decoding layer. Similarly, the second decoding layer and the third decoding layer are obtained respectively.

[0109] In this embodiment, the target feature is decoded to obtain a predicted index value corresponding to the target intervention value, a loss function is constructed based on the predicted index value and its corresponding true index value, and the parameters of the target decoding layer are adjusted based on the loss function. In this way, information from multiple dimensions can be integrated for model training, thereby improving the model accuracy. When the model is then applied to make guidance style decisions, the accuracy of the determined target guidance style can be improved. In addition, by introducing a target intervention value, the target intervention value corresponds to one of a plurality of candidate guidance styles, and the candidate guidance style can be identified, thereby further improving the accuracy of the determined target guidance style.

[0110] In some embodiments, the model also includes: an intervention coding layer and a non-intervention coding layer; the encoding of the target data sample and the target intervention value to obtain the target feature includes: using the non-intervention coding layer to encode the target data to obtain the non-intervention feature; using the intervention coding layer to encode the target intervention value to obtain the intervention feature; splicing the non-intervention feature and the intervention feature to obtain the target feature.

[0111] Among them, combined Figure 4 ,The model also includes a coding layer, which includes an intervention coding layer and a non-intervention coding layer. The intervention coding layer is used to encode the intervention value, and the non-intervention coding layer is used to encode the target data.

[0112] During the training phase, the target data samples are input into the non-intervention coding layer, and the output is the non-intervention features; the target intervention value (such as t=0) is input into the intervention coding layer, and the output is the intervention features. After that, the non-intervention features and the intervention features are concatenated to obtain the target features.

[0113] After obtaining the target feature, it is input into the target decoding layer (such as the first decoding layer), and the output is the predicted index value corresponding to the target intervention value. Afterwards, the loss function is constructed using the predicted index value and its true index value, and the parameters of the target decoding layer are adjusted using the loss function. In addition, the encoding layer (non-intervention encoding layer and intervention encoding layer) can be pre-trained, that is, when the parameters of the target decoding layer are adjusted using the loss function, the parameters of the encoding layer are kept unchanged.

[0114] In this embodiment, the target data is encoded through the non-intervention coding layer, and the target intervention value is encoded through the intervention coding layer. The network layer of the model can be used to extract the corresponding non-intervention features and intervention features to improve the accuracy of the model, thereby improving the accuracy of the reference style decision when the model is applied to guide the style decision.

[0115] Figure 7 It is a schematic diagram according to the fourth embodiment of the present disclosure. This embodiment provides a guide style determination device. The device 700 includes: an encoding module 701, a decoding module 702 and a determination module 703.

[0116] The encoding module 701 is used to encode multiple preset intervention values ​​and target data to obtain target features; each intervention value is used to identify a candidate guidance style among multiple candidate guidance styles, and the target data is used to characterize the target object information guided by the multiple candidate guidance styles; the decoding module 702 is used to decode the target features to obtain the target indicator value corresponding to each intervention value; the determination module 703 is used to determine the target guidance style among the multiple candidate guidance styles based on the target indicator value.

[0117] In this embodiment, the target feature is decoded to obtain the target indicator value corresponding to each intervention value, and the target guidance style is determined based on the target indicator value. The target feature is obtained based on the intervention value and the target data. In this way, information from multiple dimensions can be integrated to make guidance style decisions, thereby improving the accuracy of the determined target guidance style. In addition, by introducing multiple preset intervention values, the intervention values ​​correspond to the candidate guidance styles one by one, and the candidate guidance styles can be identified, thereby further improving the accuracy of the determined target guidance style.

[0118] In some embodiments, the encoding module 701 is further used to:

[0119] encoding the plurality of intervention values ​​to obtain intervention characteristics;

[0120] encoding the target data to obtain non-intervention features;

[0121] The intervention feature and the non-intervention feature are concatenated to obtain the target feature.

[0122] In this embodiment, the target data is encoded to obtain non-intervention features, the intervention value is encoded to obtain intervention features, and the non-intervention features and intervention features are spliced ​​to obtain target features. The target data and intervention values ​​can be combined to obtain the target features, thereby improving the accuracy of the target features and thereby improving the accuracy of the guided style decision.

[0123] In some embodiments, the target indicator value is a click-through rate;

[0124] The determining module 703 is further used for:

[0125] The intervention value corresponding to the maximum click rate is used as the target intervention value;

[0126] The candidate guidance style corresponding to the target intervention value is used as the target guidance style.

[0127] In this embodiment, the candidate guidance style corresponding to the maximum click rate is used as the target guidance style. In the scenario of guiding the user to watch a video, the user operation can be simplified and the user experience can be improved.

[0128] In some embodiments, the encoding module 701 is further used to:

[0129] The plurality of intervention values ​​are encoded using an intervention encoding layer in a pre-trained guided style determination model to obtain the intervention feature.

[0130] In this embodiment, the intervention value is encoded through the intervention coding layer in the model to obtain the intervention feature, which can utilize the excellent performance of the model to improve the accuracy of the intervention feature and further improve the accuracy of the guided style decision.

[0131] In some embodiments, the encoding module 701 is further used to:

[0132] The target data is encoded using a non-interference encoding layer in a pre-trained guided style determination model to obtain the non-interference features.

[0133] In this embodiment, the target data is encoded through the non-interference coding layer in the model to obtain non-interference features. The excellent performance of the model can be used to improve the accuracy of the non-interference features, thereby improving the accuracy of the guided style decision.

[0134] In some embodiments, the decoding module 702 is further configured to:

[0135] The target feature is decoded by using a decoding layer in a pre-trained guided style determination model to obtain a target indicator value corresponding to each intervention value.

[0136] In this embodiment, the target features are encoded through the decoding layer in the model to obtain the target index value, which can utilize the excellent performance of the model to improve the accuracy of the target index value, thereby improving the accuracy of the guided style decision.

[0137] In some embodiments, there are multiple decoding layers, and each decoding layer corresponds to an intervention value;

[0138] The decoding module 702 is further used for:

[0139] Each decoding layer is used to decode the target feature respectively to obtain the target indicator value corresponding to each intervention value.

[0140] In this embodiment, by setting decoding layers corresponding to intervention values ​​one by one, the target indicator value corresponding to each intervention value is obtained through each decoding layer. In this way, each decoding layer can be trained independently during the training stage to improve the accuracy of each decoding layer, thereby improving the accuracy of the guided style decision.

[0141] Figure 8 It is a schematic diagram according to the fifth embodiment of the present disclosure. This embodiment provides a guidance style determination model training device, the model includes a target decoding layer corresponding to a target intervention value, the target intervention value is used to identify a target candidate guidance style among multiple candidate guidance styles, and the device 800 includes: an encoding module 801, a decoding module 802, a construction module 803 and an adjustment module 804.

[0142] The encoding module 801 is used to encode the target data sample and the target intervention value to obtain the target feature; the target data sample is used to characterize the target object information guided by the multiple candidate guidance styles; the decoding module 802 is used to use the target decoding layer to decode the target feature to obtain the prediction index value corresponding to the target intervention value; the construction module 803 is used to construct a loss function based on the prediction index value and the true index value corresponding to the target guidance style; the adjustment module 804 is used to adjust the parameters of the target decoding layer based on the loss function.

[0143] In this embodiment, the target feature is decoded to obtain a predicted index value corresponding to the target intervention value, a loss function is constructed based on the predicted index value and its corresponding true index value, and the parameters of the target decoding layer are adjusted based on the loss function. In this way, information from multiple dimensions can be integrated for model training, thereby improving the model accuracy. When the model is then applied to make guidance style decisions, the accuracy of the determined target guidance style can be improved. In addition, by introducing a target intervention value, the target intervention value corresponds to one of a plurality of candidate guidance styles, and the candidate guidance style can be identified, thereby further improving the accuracy of the determined target guidance style.

[0144] In some embodiments, the model further comprises: an intervening coding layer and a non-intervening coding layer;

[0145] The encoding module 801 is further used to: use the non-intervention encoding layer to encode the target data to obtain non-intervention features; use the intervention encoding layer to encode the target intervention value to obtain intervention features; and splice the non-intervention features and the intervention features to obtain the target features.

[0146] In this embodiment, the target data is encoded through the non-intervention coding layer, and the target intervention value is encoded through the intervention coding layer. The network layer of the model can be used to extract the corresponding non-intervention features and intervention features to improve the accuracy of the model, thereby improving the accuracy of the reference style decision when the model is applied to guide the style decision.

[0147] It can be understood that in the embodiments of the present disclosure, the same or similar contents in different embodiments can be referenced to each other.

[0148] It can be understood that the “first”, “second”, etc. in the embodiments of the present disclosure are only used for distinction and do not indicate the degree of importance, time sequence, etc.

[0149] It is understandable that unless there is any special limitation on the sequence of steps in the process, it means that the timing relationship between these steps is not limited.

[0150] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0151] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0152] Fig. 9 A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device 900 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0153] like Fig. 9 As shown, the electronic device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the electronic device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0154] Multiple components in the electronic device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the electronic device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0155] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as a guide style determination method or a guide style determination model training method. For example, in some embodiments, the guide style determination method or the guide style determination model training method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the guide style determination method or the guide style determination model training method described above may be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the guidance style determination method or the guidance style determination model training method in any other appropriate manner (for example, by means of firmware).

[0156] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0157] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable task processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0158] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0159] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0160] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0161] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship between the client and the server is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0162] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0163] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A method for determining a guide style, comprising: Encoding a plurality of preset intervention values ​​and target data to obtain target features; Each intervention value is used to identify one of the multiple candidate guidance styles, and the target data is used to characterize the target object information guided by the multiple candidate guidance styles; Decoding the target feature to obtain a target indicator value corresponding to each intervention value; Based on the target indicator value, a target guidance style is determined from among the multiple candidate guidance styles.

2. The method according to claim 1, wherein: The encoding of the preset multiple intervention values ​​and target data to obtain the target features includes: encoding the plurality of intervention values ​​to obtain intervention characteristics; encoding the target data to obtain non-intervention features; The intervention feature and the non-intervention feature are concatenated to obtain the target feature.

3. The method according to claim 1, wherein: The target indicator value is click-through rate; The determining a target guidance style from among the multiple candidate guidance styles based on the target indicator value includes: The intervention value corresponding to the maximum click rate is used as the target intervention value; The candidate guidance style corresponding to the target intervention value is used as the target guidance style.

4. The method according to claim 2, wherein: The encoding of the plurality of intervention values ​​to obtain intervention features comprises: The plurality of intervention values ​​are encoded using an intervention encoding layer in a pre-trained guided style determination model to obtain the intervention feature.

5. The method according to claim 2, wherein: The encoding of the target data to obtain non-intervention features comprises: The target data is encoded using a non-interference encoding layer in a pre-trained guided style determination model to obtain the non-interference features.

6. The method according to claim 1, wherein: The decoding of the target feature to obtain the target indicator value corresponding to each intervention value includes: The target feature is decoded by using a decoding layer in a pre-trained guided style determination model to obtain a target indicator value corresponding to each intervention value.

7. The method according to claim 6, wherein: There are multiple decoding layers, and each decoding layer corresponds to an intervention value; The decoding layer in the pre-trained guided style determination model is used to decode the target feature to obtain the target indicator value corresponding to each intervention value, including: Each decoding layer is used to decode the target feature respectively to obtain the target indicator value corresponding to each intervention value.

8. A method for training a guidance style determination model, the model comprising a target decoding layer corresponding to a target intervention value, the target intervention value being used to identify a target candidate guidance style among a plurality of candidate guidance styles, the method comprising: Encoding the target data sample and the target intervention value to obtain a target feature; The target data samples are used to represent target object information guided by the multiple candidate guidance styles; Using the target decoding layer, decoding the target feature to obtain a prediction index value corresponding to the target intervention value; Constructing a loss function based on the predicted index value and the real index value corresponding to the target guidance style; Based on the loss function, parameters of the target decoding layer are adjusted.

9. The method according to claim 8, wherein: The model also includes: an intervention coding layer and a non-intervention coding layer; The encoding of the target data sample and the target intervention value to obtain the target feature includes: Using the non-interference coding layer, encoding the target data to obtain non-interference features; Using the intervention coding layer, encoding the target intervention value to obtain an intervention feature; The non-intervention feature and the intervention feature are concatenated to obtain the target feature.

10. A guidance style determination device, comprising: An encoding module, used for encoding a plurality of preset intervention values ​​and target data to obtain target features; Each intervention value is used to identify one of the multiple candidate guidance styles, and the target data is used to characterize the target object information guided by the multiple candidate guidance styles; A decoding module, used for decoding the target feature to obtain a target indicator value corresponding to each intervention value; A determination module is used to determine a target guidance style from among the multiple candidate guidance styles based on the target indicator value.

11. The device according to claim 10, wherein: The encoding module is further used for: encoding the plurality of intervention values ​​to obtain intervention characteristics; encoding the target data to obtain non-intervention features; The intervention feature and the non-intervention feature are concatenated to obtain the target feature.

12. The device according to claim 10, wherein: The target indicator value is click-through rate; The determination module is further used for: The intervention value corresponding to the maximum click rate is used as the target intervention value; The candidate guidance style corresponding to the target intervention value is used as the target guidance style.

13. The device according to claim 11, wherein: The encoding module is further used for: The plurality of intervention values ​​are encoded using an intervention encoding layer in a pre-trained guided style determination model to obtain the intervention feature.

14. The device according to claim 11, wherein: The encoding module is further used for: The target data is encoded using a non-interference encoding layer in a pre-trained guided style determination model to obtain the non-interference features.

15. The device according to claim 10, wherein: The decoding module is further used for: The target feature is decoded by using a decoding layer in a pre-trained guided style determination model to obtain a target indicator value corresponding to each intervention value.

16. The device according to claim 15, wherein: There are multiple decoding layers, and each decoding layer corresponds to an intervention value; The decoding module is further used for: Each decoding layer is used to decode the target feature respectively to obtain the target indicator value corresponding to each intervention value.

17. A guidance pattern determination model training device, the model comprising a target decoding layer corresponding to a target intervention value, the target intervention value being used to identify a target candidate guidance pattern among a plurality of candidate guidance patterns, the device comprising: An encoding module, used for encoding the target data sample and the target intervention value to obtain a target feature; The target data samples are used to represent target object information guided by the multiple candidate guidance styles; A decoding module, configured to decode the target feature using the target decoding layer to obtain a prediction index value corresponding to the target intervention value; A construction module, configured to construct a loss function based on the predicted index value and the real index value corresponding to the target guidance style; An adjustment module is used to adjust the parameters of the target decoding layer based on the loss function.

18. The device according to claim 17, wherein: The model also includes: an intervention coding layer and a non-intervention coding layer; The encoding module is further used for: Using the non-interference coding layer, encoding the target data to obtain non-interference features; Using the intervention coding layer, encoding the target intervention value to obtain an intervention feature; The non-intervention feature and the intervention feature are concatenated to obtain the target feature.

19. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 9.

20. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-9.

21. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 9.

Images