CN116703454A - Target recommendation method and device - Google Patents
Target recommendation method and device Download PDFInfo
- Publication number
- CN116703454A CN116703454A CN202310884582.3A CN202310884582A CN116703454A CN 116703454 A CN116703454 A CN 116703454A CN 202310884582 A CN202310884582 A CN 202310884582A CN 116703454 A CN116703454 A CN 116703454A
- Authority
- CN
- China
- Prior art keywords
- recommendation
- recommended
- target
- model
- sequence
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 239000013598 vector Substances 0.000 claims abstract description 88
- 230000007246 mechanism Effects 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 40
- 238000012545 processing Methods 0.000 claims abstract description 15
- 238000012216 screening Methods 0.000 claims abstract description 14
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 12
- 230000003993 interaction Effects 0.000 claims description 26
- 230000006399 behavior Effects 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 17
- 230000008569 process Effects 0.000 claims description 10
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000002708 enhancing effect Effects 0.000 claims description 3
- 230000005012 migration Effects 0.000 claims description 3
- 238000013508 migration Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 230000010365 information processing Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 230000000694 effects Effects 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013526 transfer learning Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to the technical field of information processing, and provides a target recommendation method and device. The method comprises the following steps: the large language model and the click prediction model are connected in a bidirectional mode to obtain a target recommendation model, and training is carried out on the target recommendation model based on a target recommendation task; acquiring target user information of a target user; processing the target user information according to the prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into a trained target recommendation model: converting the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model; determining a recommendation candidate sequence through a click prediction model based on the target user feature vector; screening targets to be recommended in the recommended candidate sequence through a large language model to obtain a recommended sequence; and recommending the target to the target user according to the recommendation sequence.
Description
Technical Field
The present application relates to the field of information processing technologies, and in particular, to a target recommendation method and apparatus.
Background
In the information age, a variety of information surrounds users, such as stock information, national policies, entertainment news, sports news, and the like. But users prefer to spend time on information of interest and users do not pay attention to irrelevant information. Recommendation models are widely deployed to automatically infer preferences of people and provide high quality recommendation services. However, there are many information deviating from the preference of the user in the information recommended by the current recommendation model, which wastes time for the user and needs to be further improved.
Disclosure of Invention
In view of the above, the embodiments of the present application provide a target recommendation method, apparatus, electronic device, and computer readable storage medium, so as to solve the problem in the prior art that recommendation information of a recommendation model is inaccurate.
In a first aspect of an embodiment of the present application, there is provided a target recommendation method, including: the large language model and the click prediction model are connected in a bidirectional mode to obtain a target recommendation model, and training is carried out on the target recommendation model based on a target recommendation task; obtaining target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information; processing the target user information according to the prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into a trained target recommendation model: converting the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model; determining a recommendation candidate sequence through a click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended; screening targets to be recommended in a recommended candidate sequence through a large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence; and recommending the target to the target user according to the recommendation sequence.
In a second aspect of the embodiment of the present application, there is provided a target recommendation apparatus, including: the training module is configured to connect the large language model and the click prediction model in a bidirectional manner to obtain a target recommendation model, and train the target recommendation model based on a target recommendation task; the acquisition module is configured to acquire target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information; the processing module is configured to process the target user information according to the prompt word mechanism and input the target user information processed according to the prompt word mechanism into a trained target recommendation model: the conversion module is configured to convert the target user information processed according to the prompt word mechanism into a target user feature vector through the large language model; a determining module configured to determine a recommendation candidate sequence based on the target user feature vector by clicking the prediction model, wherein the recommendation candidate sequence includes a plurality of targets to be recommended; the screening module is configured to screen targets to be recommended in the recommended candidate sequence through the large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence; and the recommendation module is configured to conduct target recommendation to the target user according to the recommendation sequence.
In a third aspect of the embodiments of the present application, there is provided an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the above method when executing the computer program.
In a fourth aspect of the embodiments of the present application, there is provided a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the above method.
Compared with the prior art, the embodiment of the application has the beneficial effects that: because the embodiment of the application obtains the target recommendation model by connecting the large language model and the click prediction model in a bidirectional manner, and trains the target recommendation model based on the target recommendation task; obtaining target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information; processing the target user information according to the prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into a trained target recommendation model: converting the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model; determining a recommendation candidate sequence through a click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended; screening targets to be recommended in a recommended candidate sequence through a large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence; according to the recommendation sequence, target recommendation is carried out on the target user, so that the problem that recommendation information of a recommendation model is inaccurate in the prior art can be solved by adopting the technical means, the accuracy of the recommendation information of the recommendation model is improved, and the user satisfaction is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of a target recommendation method according to an embodiment of the present application;
FIG. 2 is a schematic flow chart of a training method of a target recommendation model according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a target recommendation device according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
In the following description, for purposes of explanation and not limitation, specific details are set forth such as the particular system architecture, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. It will be apparent, however, to one skilled in the art that the present application may be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.
Fig. 1 is a flow chart of a target recommendation method according to an embodiment of the present application. The target recommendation method of fig. 1 may be performed by a computer or a server, or software on a computer or a server. As shown in fig. 1, the target recommendation method includes:
s101, a large language model and a click prediction model are connected in a bidirectional mode to obtain a target recommendation model, and training is carried out on the target recommendation model based on a target recommendation task;
s102, obtaining target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information;
s103, processing the target user information according to a prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into a trained target recommendation model:
s104, converting the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model;
s105, determining a recommendation candidate sequence through a click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended;
s106, screening targets to be recommended in the recommended candidate sequence through a large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence;
s107, performing target recommendation to the target user according to the recommendation sequence.
The click prediction model is a CTR model, CTR, collectively Click Through Rate. The large language model can be a ChatGPT model, which is commonly known as Chat GenerativePre-trained Transformer. The embodiment of the application constructs the target recommendation model by utilizing the large language model and the click prediction model, and can be understood as enhancing the recommendation effect of the click prediction model by utilizing the large language model.
The personal basic information is information such as age, sex, hobbies and the like of the target user, the search word history information is information of search words used by the target user, the history dialogue information is information of dialogue generated by the target user and the large language model, the history interaction behavior information is information of interaction behaviors of the target user and a recommended target, and the interaction behaviors include collection, sharing, clicking, reading duration and the like.
The Prompt word mechanism is a Prompt mechanism. The target user information is processed according to the prompt word mechanism, and an instruction for converting the target user information into a related embedded vector (embedding) can be given, so that the target user information processed according to the prompt word mechanism contains the original target user information and the instruction for converting the target user information into the related embedded vector (embedding). Therefore, the target user information processed according to the prompt word mechanism can be converted into the target user feature vector through the large language model, and the target user feature vector is the relevant embedded vector. The process is an application of the promtt mechanism, which is a common method and will not be described in detail.
According to the embodiment of the application, the target user information processed according to the prompt word mechanism is input into a trained target recommendation model, namely a large language model is input, and the target user feature vector is output; inputting the target user feature vector into a click prediction model, and outputting a recommendation candidate sequence; and inputting the recommended candidate sequence into a large language model, and outputting the recommended sequence. The recommendation effect of the click prediction model is enhanced by using the large language model, and the accuracy of recommendation information of the recommendation model is further improved.
According to the technical scheme provided by the embodiment of the application, a large language model and a click prediction model are connected in a bidirectional manner to obtain a target recommendation model, and the target recommendation model is trained based on a target recommendation task; obtaining target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information; processing the target user information according to the prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into a trained target recommendation model: converting the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model; determining a recommendation candidate sequence through a click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended; screening targets to be recommended in a recommended candidate sequence through a large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence; according to the recommendation sequence, target recommendation is carried out on the target user, so that the problem that recommendation information of a recommendation model is inaccurate in the prior art can be solved by adopting the technical means, the accuracy of the recommendation information of the recommendation model is improved, and the user satisfaction is improved.
Further, after inputting the target user information processed according to the prompt word mechanism into the trained target recommendation model, the method further comprises: the large language model is used as a vector generator to respectively convert the search word history information, the personal basic information, the history dialogue information and the history interaction behavior information processed according to the prompt word mechanism into a search word history feature vector, a personal basic feature vector, a history dialogue feature vector and a history interaction behavior feature vector through the large language model; determining a recommendation candidate sequence through a click prediction model based on the search word history feature vector, the personal basic feature vector, the history dialogue feature vector and the history interaction behavior feature vector; wherein the target user feature vector comprises: search term historical feature vectors, personal basic feature vectors, historical dialogue feature vectors, and historical interaction behavior feature vectors.
In some embodiments: after the large language model is converted to obtain the target user feature vector, the large language model is controlled to transmit the target user feature vector to the click prediction model through unidirectional connection from the large language model to the click prediction model; after the click prediction model determines a recommendation candidate sequence based on the target user feature vector, the click prediction model is controlled to transmit the recommendation candidate sequence to the large language model through unidirectional connection from the click prediction model to the large language model; after the large language model screens the recommended sequence from the recommended candidate sequences, taking the recommended sequence as the output of the target recommended model; a bi-directional connection of a large language model and a click prediction model, comprising: unidirectional connections from the large language model to the click prediction model and unidirectional connections from the click prediction model to the large language model.
That is, the working process inside the large language model is in turn: large language model, click prediction model, large language model. There is a unidirectional connection from the large language model to the click prediction model and a unidirectional connection from the click prediction model to the large language model in the large language model, which is called a bidirectional connection of the language model and the click prediction model.
Further, in the process of screening the target to be recommended in the recommended candidate sequence through the large language model to obtain the recommended sequence: sorting targets to be recommended in a recommendation sequence through a large language model, wherein the targets to be recommended in front of the recommendation sequence are recommended to a target user; and generating recommendation reasons of the targets to be recommended in the recommendation sequence through the large language model.
The large language model can realize various functions or roles, and can sort the targets to be recommended in the recommendation sequence according to the knowledge learned by the large language model and generate the recommendation reason of the targets to be recommended in the recommendation sequence.
Further, training the target recommendation model based on the target recommendation task includes: training the target recommendation model based on a plurality of recommendation tasks at the same time, wherein the target recommendation tasks comprise a plurality of recommendation tasks, each recommendation task corresponds to one recommendation field, and the trained target recommendation model can be used for target recommendation in various recommendation fields.
For example, in the online shopping scene, the recommendation task is a commodity recommendation task, and the recommendation target or the target to be recommended is a commodity; if the recommended task is a text recommended task in the news reading scene, the recommended target or the target to be recommended is text; if the recommended task is a video recommended task in the video watching scene, the recommended target or the target to be recommended is a video; if the recommended task is a music recommended task in a music listening scene, the recommended target or the target to be recommended is music; if the recommended task is a wallpaper recommended task in the wallpaper recommended scene, the recommended target or the target to be recommended is wallpaper, etc.
Further, training the target recommendation model based on the target recommendation task includes: sequencing a plurality of recommended tasks according to a preset sequence, wherein the target recommended tasks comprise a plurality of recommended tasks, each recommended task corresponds to one recommended field, and the preset sequence is a pure text recommended task, a pure audio recommended task, a pure image recommended task and a complex information recommended task; training the target recommendation model according to the ordered plurality of recommendation tasks in sequence, wherein the target recommendation model is trained by adopting a migration learning method in sequence, and the trained target recommendation model can be used for target recommendation in various recommendation fields.
There are four recommended tasks, namely, news recommended task, movie recommended task, music recommended task and wallpaper recommended task. The news recommending task, the music recommending task, the wallpaper recommending task and the film recommending task respectively correspond to a pure text recommending task, a pure audio recommending task, a pure image recommending task and a complex information recommending task, wherein the complex information recommending task means that a recommending target contains audio images and text information. The ordered plurality of recommended tasks are, in turn, news recommended tasks, music recommended tasks, wallpaper recommended tasks, and movie recommended tasks. Training the target recommendation model sequentially by adopting a transfer learning method, and transferring knowledge learned by the target recommendation model after training the target recommendation model according to the news recommendation task to training the target recommendation model according to the music recommendation task; training the target recommendation model according to the news recommendation task and the music recommendation task, and then transferring the knowledge learned by the target recommendation model to training the target recommendation model according to the wallpaper recommendation task; after training the target recommendation model according to the news recommendation task, the music recommendation task and the wallpaper recommendation task, the knowledge learned by the target recommendation model is transferred to training the target recommendation model according to the movie recommendation task.
Fig. 2 is a flow chart of a training method of a target recommendation model according to an embodiment of the present application. As shown in fig. 2, includes:
s201, a large language model and a click prediction model are connected in a bidirectional mode, and a target recommendation model is obtained;
s202, training data is acquired, wherein the training data comprises user information of a plurality of users, and the user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information;
s203, processing each piece of user information according to a prompt word mechanism, and inputting each piece of user information processed according to the prompt word mechanism into a target recommendation model:
s204, converting each piece of user information processed according to a prompt word mechanism into a user feature vector corresponding to each user through a large language model;
s205, determining a recommendation candidate sequence corresponding to each user through a click prediction model based on the user feature vector corresponding to each user, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended;
s206, screening targets to be recommended in a recommendation candidate sequence corresponding to each user through a large language model to obtain a recommendation sequence corresponding to each user, wherein the recommendation sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommendation sequence is smaller than that of the targets to be recommended in the recommendation candidate sequence;
s207, training a target recommendation model based on a target recommendation task according to the target to be recommended and the label in the recommendation sequence corresponding to each user.
The label corresponding to each user is a recommendation target of the user marked in advance.
Further, training the target recommendation model based on the target recommendation task includes: and converting the user information in the training data into user feature vectors through the large language model, transmitting the user feature vectors to the click prediction model, and enhancing the cold starting capability of the click prediction model by introducing a dropout mechanism into the click prediction model.
Any combination of the above optional solutions may be adopted to form an optional embodiment of the present application, which is not described herein.
The following are examples of the apparatus of the present application that may be used to perform the method embodiments of the present application. For details not disclosed in the embodiments of the apparatus of the present application, please refer to the embodiments of the method of the present application.
Fig. 3 is a schematic diagram of a target recommendation device according to an embodiment of the present application. As shown in fig. 3, the target recommendation device includes:
the training module 301 is configured to connect the large language model and the click prediction model in a bidirectional manner to obtain a target recommendation model, and train the target recommendation model based on a target recommendation task;
an obtaining module 302, configured to obtain target user information of a target user, where the target user information includes: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information;
the processing module 303 is configured to process the target user information according to the prompt word mechanism, and input the target user information processed according to the prompt word mechanism into a trained target recommendation model:
the conversion module 304 is configured to convert the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model;
a determining module 305 configured to determine a recommendation candidate sequence by a click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended;
the screening module 306 is configured to screen the targets to be recommended in the recommended candidate sequence through the large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence;
a recommendation module 307 configured to make target recommendations to the target user in accordance with the recommendation sequence.
According to the technical scheme provided by the embodiment of the application, a large language model and a click prediction model are connected in a bidirectional manner to obtain a target recommendation model, and the target recommendation model is trained based on a target recommendation task; obtaining target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information; processing the target user information according to the prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into a trained target recommendation model: converting the target user information processed according to the prompt word mechanism into a target user feature vector through a large language model; determining a recommendation candidate sequence through a click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended; screening targets to be recommended in a recommended candidate sequence through a large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence; according to the recommendation sequence, target recommendation is carried out on the target user, so that the problem that recommendation information of a recommendation model is inaccurate in the prior art can be solved by adopting the technical means, the accuracy of the recommendation information of the recommendation model is improved, and the user satisfaction is improved.
Optionally, the conversion module 304 is further configured to use the large language model as a vector generator to convert the search word history information, the personal basic information, the history dialogue information and the history interaction behavior information processed according to the prompt word mechanism into a search word history feature vector, a personal basic feature vector, a history dialogue feature vector and a history interaction behavior feature vector through the large language model; determining a recommendation candidate sequence through a click prediction model based on the search word history feature vector, the personal basic feature vector, the history dialogue feature vector and the history interaction behavior feature vector; wherein the target user feature vector comprises: search term historical feature vectors, personal basic feature vectors, historical dialogue feature vectors, and historical interaction behavior feature vectors.
Optionally, the transformation module 304 is further configured to control the large language model to transmit the target user feature vector to the click prediction model through unidirectional connection from the large language model to the click prediction model after the large language model is transformed to obtain the target user feature vector; after the click prediction model determines a recommendation candidate sequence based on the target user feature vector, the click prediction model is controlled to transmit the recommendation candidate sequence to the large language model through unidirectional connection from the click prediction model to the large language model; after the large language model screens the recommended sequence from the recommended candidate sequences, taking the recommended sequence as the output of the target recommended model; a bi-directional connection of a large language model and a click prediction model, comprising: unidirectional connections from the large language model to the click prediction model and unidirectional connections from the click prediction model to the large language model.
Optionally, the filtering module 306 is further configured to, in a process of filtering the target to be recommended in the recommended candidate sequence by the large language model, obtain the recommended sequence: sorting targets to be recommended in a recommendation sequence through a large language model, wherein the targets to be recommended in front of the recommendation sequence are recommended to a target user; and generating recommendation reasons of the targets to be recommended in the recommendation sequence through the large language model.
Optionally, the training module 301 is further configured to train the target recommendation model simultaneously based on a plurality of recommendation tasks, where the target recommendation tasks include a plurality of recommendation tasks, each recommendation task corresponding to a recommendation field, and the trained target recommendation model can be used for target recommendation of various recommendation fields.
Optionally, the training module 301 is further configured to sort the plurality of recommended tasks according to a preset order, where the target recommended task includes a plurality of recommended tasks, each recommended task corresponds to a recommended field, and the preset order is a plain text recommended task, a plain audio recommended task, a plain image recommended task, and a complex information recommended task; training the target recommendation model according to the ordered plurality of recommendation tasks in sequence, wherein the target recommendation model is trained by adopting a migration learning method in sequence, and the trained target recommendation model can be used for target recommendation in various recommendation fields.
Optionally, the training module 301 is further configured to bi-directionally connect the large language model and the click prediction model to obtain a target recommendation model; acquiring training data, wherein the training data comprises user information of a plurality of users, the user information comprising: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information; processing each piece of user information according to a prompt word mechanism, and inputting each piece of user information processed according to the prompt word mechanism into a target recommendation model: converting each piece of user information processed according to a prompt word mechanism into a user feature vector corresponding to each user through a large language model; determining a recommendation candidate sequence corresponding to each user through a click prediction model based on the user feature vector corresponding to each user, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended; screening targets to be recommended in a recommendation candidate sequence corresponding to each user through a large language model to obtain a recommendation sequence corresponding to each user, wherein the recommendation sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommendation sequence is smaller than that of the targets to be recommended in the recommendation candidate sequence; and training the target recommendation model based on the target recommendation task according to the target to be recommended and the label in the recommendation sequence corresponding to each user.
Optionally, the training module 301 is further configured to transform the user information in the training data into a user feature vector through the large language model, and transmit the user feature vector to the click prediction model and enhance the cold start capability of the click prediction model by introducing a dropout mechanism in the click prediction model.
It should be understood that the sequence number of each step in the foregoing embodiment does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not limit the implementation process of the embodiment of the present application.
Fig. 4 is a schematic diagram of an electronic device 4 according to an embodiment of the present application. As shown in fig. 4, the electronic apparatus 4 of this embodiment includes: a processor 401, a memory 402 and a computer program 403 stored in the memory 402 and executable on the processor 401. The steps of the various method embodiments described above are implemented by processor 401 when executing computer program 403. Alternatively, the processor 401, when executing the computer program 403, performs the functions of the modules/units in the above-described apparatus embodiments.
The electronic device 4 may be a desktop computer, a notebook computer, a palm computer, a cloud server, or the like. The electronic device 4 may include, but is not limited to, a processor 401 and a memory 402. It will be appreciated by those skilled in the art that fig. 4 is merely an example of the electronic device 4 and is not limiting of the electronic device 4 and may include more or fewer components than shown, or different components.
The processor 401 may be a central processing unit (Central Processing Unit, CPU) or other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application SpecificIntegrated Circuit, ASIC), field programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like.
The memory 402 may be an internal storage unit of the electronic device 4, for example, a hard disk or a memory of the electronic device 4. The memory 402 may also be an external storage device of the electronic device 4, for example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card) or the like, which are provided on the electronic device 4. Memory 402 may also include both internal storage units and external storage devices of electronic device 4. The memory 402 is used to store computer programs and other programs and data required by the electronic device.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit.
The integrated modules/units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the present application may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, and the computer program may be stored in a computer readable storage medium, where the computer program, when executed by a processor, may implement the steps of each of the method embodiments described above. The computer program may comprise computer program code, which may be in source code form, object code form, executable file or in some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), an electrical carrier signal, a telecommunications signal, a software distribution medium, and so forth. It should be noted that the content of the computer readable medium can be appropriately increased or decreased according to the requirements of the jurisdiction's jurisdiction and the patent practice, for example, in some jurisdictions, the computer readable medium does not include electrical carrier signals and telecommunication signals according to the jurisdiction and the patent practice.
The above embodiments are only for illustrating the technical solution of the present application, and are not limiting; although the application has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present application, and are intended to be included in the scope of the present application.
Claims (10)
1. A target recommendation method, comprising:
the large language model and the click prediction model are connected in a bidirectional mode to obtain a target recommendation model, and training is carried out on the target recommendation model based on a target recommendation task;
obtaining target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information;
processing the target user information according to a prompt word mechanism, and inputting the target user information processed according to the prompt word mechanism into the trained target recommendation model:
converting the target user information processed according to the prompt word mechanism into a target user feature vector through the large language model;
determining a recommendation candidate sequence through the click prediction model based on the target user feature vector, wherein the recommendation candidate sequence comprises a plurality of targets to be recommended;
screening targets to be recommended in the recommendation candidate sequence through the large language model to obtain a recommendation sequence, wherein the recommendation sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommendation sequence is smaller than that of the targets to be recommended in the recommendation candidate sequence;
and carrying out target recommendation to the target user according to the recommendation sequence.
2. The method of claim 1, wherein after inputting the target user information processed according to the prompt-word mechanism into the trained target recommendation model, the method further comprises:
the large language model is used as a vector generator to respectively convert the search word historical information, the personal basic information, the historical dialogue information and the historical interaction behavior information processed according to the prompt word mechanism into a search word historical feature vector, a personal basic feature vector, a historical dialogue feature vector and a historical interaction behavior feature vector through the large language model;
determining the recommendation candidate sequence through the click prediction model based on the search term historical feature vector, the personal basic feature vector, the historical dialogue feature vector and the historical interaction behavior feature vector;
wherein the target user feature vector comprises: the search term history feature vector, the personal base feature vector, the history dialogue feature vector, and the history interaction feature vector.
3. The method according to claim 1, wherein the method further comprises:
after the large language model is converted to obtain the target user feature vector, the large language model is controlled to transmit the target user feature vector to the click prediction model through unidirectional connection from the large language model to the click prediction model;
after the click prediction model determines the recommendation candidate sequence based on the target user feature vector, the click prediction model is controlled to transmit the recommendation candidate sequence to the large language model through unidirectional connection from the click prediction model to the large language model;
after the large language model screens the recommendation candidate sequence to obtain the recommendation sequence, the recommendation sequence is used as the output of the target recommendation model;
the bi-directional connection of the large language model and the click prediction model comprises: a unidirectional connection from the large language model to the click prediction model and a unidirectional connection from the click prediction model to the large language model.
4. The method according to claim 1, wherein in the process of screening the target to be recommended in the recommended candidate sequence by the large language model to obtain the recommended sequence:
sorting targets to be recommended in the recommendation sequence through the large language model, wherein the targets to be recommended in the recommendation sequence are recommended to the target user earlier than the targets to be recommended in the recommendation sequence are;
and generating recommendation reasons of the targets to be recommended in the recommendation sequence through the large language model.
5. The method of claim 1, wherein training the target recommendation model based on target recommendation tasks comprises:
and training the target recommendation model based on a plurality of recommendation tasks at the same time, wherein the target recommendation tasks comprise a plurality of recommendation tasks, each recommendation task corresponds to one recommendation field, and the trained target recommendation model can be used for target recommendation in various recommendation fields.
6. The method of claim 1, wherein training the target recommendation model based on target recommendation tasks comprises:
sequencing a plurality of recommended tasks according to a preset sequence, wherein the target recommended tasks comprise a plurality of recommended tasks, each recommended task corresponds to a recommended field, and the preset sequence is a pure text recommended task, a pure audio recommended task, a pure image recommended task and a complex information recommended task;
and training the target recommendation model according to the ordered plurality of recommendation tasks in sequence, wherein the target recommendation model is trained by adopting a migration learning method in sequence, and the trained target recommendation model can be used for target recommendation in various recommendation fields.
7. The method of claim 1, wherein training the target recommendation model based on target recommendation tasks comprises:
and converting user information in training data into user feature vectors through the large language model, transmitting the user feature vectors to the click prediction model, and enhancing cold starting capability of the click prediction model by introducing a dropout mechanism into the click prediction model.
8. A target recommendation device, comprising:
the training module is configured to connect the large language model and the click prediction model in a bidirectional manner to obtain a target recommendation model, and train the target recommendation model based on a target recommendation task;
the acquisition module is configured to acquire target user information of a target user, wherein the target user information comprises: search term history information, personal basic information, historical dialogue information, and historical interaction behavior information;
the processing module is configured to process the target user information according to a prompt word mechanism and input the target user information processed according to the prompt word mechanism into the trained target recommendation model:
the conversion module is configured to convert the target user information processed according to the prompt word mechanism into a target user feature vector through the large language model;
a determining module configured to determine a recommendation candidate sequence based on the target user feature vector through the click prediction model, wherein the recommendation candidate sequence includes a plurality of targets to be recommended;
the screening module is configured to screen targets to be recommended in the recommended candidate sequence through the large language model to obtain a recommended sequence, wherein the recommended sequence comprises a plurality of targets to be recommended, and the number of the targets to be recommended in the recommended sequence is smaller than that of the targets to be recommended in the recommended candidate sequence;
and the recommendation module is configured to recommend targets to the target users according to the recommendation sequence.
9. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the steps of the method according to any one of claims 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310884582.3A CN116703454A (en) | 2023-07-19 | 2023-07-19 | Target recommendation method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310884582.3A CN116703454A (en) | 2023-07-19 | 2023-07-19 | Target recommendation method and device |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116703454A true CN116703454A (en) | 2023-09-05 |
Family
ID=87834173
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310884582.3A Pending CN116703454A (en) | 2023-07-19 | 2023-07-19 | Target recommendation method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116703454A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117094360A (en) * | 2023-10-18 | 2023-11-21 | 杭州同花顺数据开发有限公司 | User characterization extraction method, device, equipment and storage medium |
CN117273868A (en) * | 2023-11-20 | 2023-12-22 | 浙江口碑网络技术有限公司 | Shop recommendation method and device, electronic equipment and storage medium |
-
2023
- 2023-07-19 CN CN202310884582.3A patent/CN116703454A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117094360A (en) * | 2023-10-18 | 2023-11-21 | 杭州同花顺数据开发有限公司 | User characterization extraction method, device, equipment and storage medium |
CN117273868A (en) * | 2023-11-20 | 2023-12-22 | 浙江口碑网络技术有限公司 | Shop recommendation method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110288049B (en) | Method and apparatus for generating image recognition model | |
CN108830235B (en) | Method and apparatus for generating information | |
CN109740018B (en) | Method and device for generating video label model | |
CN108960316B (en) | Method and apparatus for generating a model | |
CN116703454A (en) | Target recommendation method and device | |
CN112905839A (en) | Model training method, model using device, storage medium and equipment | |
CN109981787B (en) | Method and device for displaying information | |
CN107992478A (en) | The method and apparatus for determining focus incident | |
CN116578875B (en) | Click prediction model training method and device based on multiple behaviors | |
CN113395538B (en) | Sound effect rendering method and device, computer readable medium and electronic equipment | |
CN117033799B (en) | Resource recommendation method, device, computer equipment and storage medium | |
CN111897950A (en) | Method and apparatus for generating information | |
CN109348262A (en) | A kind of calculation method, device, equipment and the storage medium of main broadcaster's similarity | |
CN112182281B (en) | Audio recommendation method, device and storage medium | |
CN111026849B (en) | Data processing method and device | |
CN116912734A (en) | Video abstract data set construction method, device, medium and electronic equipment | |
CN112309387A (en) | Method and apparatus for processing information | |
CN112000872A (en) | Recommendation method based on user vector, training method of model and training device | |
CN110580648A (en) | financial information processing method and device based on artificial intelligence | |
CN111125501B (en) | Method and device for processing information | |
CN113837986A (en) | Method, apparatus, electronic device, and medium for recognizing tongue picture | |
CN113392238A (en) | Media file processing method and device, computer readable medium and electronic equipment | |
CN113822324A (en) | Image processing method and device based on multitask model and related equipment | |
CN111611420A (en) | Method and apparatus for generating image description information | |
CN116911954B (en) | Method and device for recommending items based on interests and popularity |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |