CN116595150A - Dialogue recommendation method, device, equipment and storage medium - Google Patents
Dialogue recommendation method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN116595150A CN116595150A CN202310720521.3A CN202310720521A CN116595150A CN 116595150 A CN116595150 A CN 116595150A CN 202310720521 A CN202310720521 A CN 202310720521A CN 116595150 A CN116595150 A CN 116595150A
- Authority
- CN
- China
- Prior art keywords
- user
- target
- shopping guide
- result
- requirement
- 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 51
- 238000000605 extraction Methods 0.000 claims abstract description 42
- 238000012216 screening Methods 0.000 claims description 16
- 238000012986 modification Methods 0.000 claims description 14
- 230000004048 modification Effects 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 13
- 238000005516 engineering process Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 5
- 230000006978 adaptation Effects 0.000 claims description 4
- 238000013473 artificial intelligence Methods 0.000 abstract description 3
- 238000012163 sequencing technique Methods 0.000 abstract description 3
- 238000004891 communication Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000009286 beneficial effect Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 2
- 230000001965 increasing effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000012549 training Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/194—Calculation of difference between files
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0605—Supply or demand aggregation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
Abstract
The application discloses a dialogue recommendation method, a device, equipment and a storage medium, which relate to the technical field of artificial intelligence and comprise the following steps: acquiring initial shopping guide requirements of a user, and judging whether collection is completed or not; if the collection is completed, analyzing the collection as a user target shopping guide requirement by using a target large language model, extracting key information, and fusing the key information to obtain a target extraction result; carrying out semantic recall according to the target extraction result, and determining the similarity between the semantic recall result and the key information, so as to sort the semantic recall result according to the similarity to obtain a sorting result; and generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sequencing result. Shopping guide dialogue management is carried out based on a large language model, recommendation results are more accordant with specific requirements of users through extraction and definition of keywords, explanation is given to final recommendation results based on the large language model, and credibility of dialogue recommendation is enhanced.
Description
Technical Field
The present application relates to the field of artificial intelligence technologies, and in particular, to a method, an apparatus, a device, and a storage medium for dialogue recommendation.
Background
As artificial intelligence technology matures, its application in dialogue recommendation is increasing, for example, corresponding clothing products and recommendations may be given by acquiring the user's needs when the user makes clothing selections. However, in the existing conversational recommendation system, the management of the conversations is usually based on a rule form, and the identification of key information is also a predefined slot, which makes conversations and recommendation results generated in the conversation process too dead and inflexible. The recommendation results are therefore finally rendered unexplained by the black box of the intermediate flow after presenting the recommendation list to the user. The flexibility of use and the interpretability of the recommendation results of the dialogue recommendation system are very important for the permanent development and use of the whole system. Thus, how to increase flexibility in dialogue recommendation and generate interpretable recommendation results is a problem that is highly sought to be solved in the art.
Disclosure of Invention
Accordingly, the present application is directed to a method, apparatus, device, and storage medium for conducting a shopping guide dialogue management based on a large language model, and enhancing credibility of dialogue recommendation by more conforming recommendation results to specific needs of users by extracting and defining keywords, and by interpreting final recommendation results based on the large language model. The specific scheme is as follows:
in a first aspect, the present application provides a dialogue recommendation method, including:
acquiring initial shopping guide requirements of a user, and judging whether the initial shopping guide requirements of the user are collected;
if the collection is completed, analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result;
carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result according to the similarity to obtain a sorting result;
and generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result.
Optionally, after the obtaining the initial shopping guide requirement of the user and determining whether the initial shopping guide requirement of the user is collected, the method further includes:
if the initial user shopping guide requirements are not collected, inquiring the user shopping guide requirements continuously according to a preset rule, and acquiring a corresponding user answer result;
and generating a user target shopping guide requirement according to the user answer result and the initial user shopping guide requirement, so as to analyze the user target shopping guide requirement based on the target large language model.
Optionally, the generating the user target shopping guide requirement according to the user answer result and the initial user shopping guide requirement includes:
judging whether the answer result of the user is the shopping guide requirement modification intention or not;
if yes, the initial user shopping guide requirement is modified according to the shopping guide requirement modification intention, and the modified initial user shopping guide requirement is used as the user target shopping guide requirement.
Optionally, the analyzing the current initial shopping guide requirement of the user as the target shopping guide requirement of the user by using a target large language model includes:
and merging the contexts of the current dialogue, inputting the merged contexts into an initial large language model, screening out clothing industry core words in the current dialogue according to preset word screening rules, fine-tuning the initial large language model by using a low-rank adaptation technology of the large language model based on the contexts and the clothing industry core words, and generating a target large language model so as to analyze the current initial shopping guide requirement of the user as the target shopping guide requirement of the user by using the target large language model.
Optionally, the extracting the key information of the target shopping guide requirement of the user and fusing the key information to obtain a target extraction result includes:
and filtering the key information according to a first preset information label, and screening the filtered key information according to a second preset information label to obtain target key information so as to fuse the target key information to obtain a target extraction result.
Optionally, the generating, based on the sorting result, the commodity recommendation and the commodity recommendation explanation corresponding to the target shopping guide requirement of the user by using the target large language model includes:
acquiring preset self-owned clothing industry shopping guide data, and fusing the preset self-owned clothing industry shopping guide data with clothing industry data in preset third party public customer service data to obtain target data;
and acquiring a historical portrait of the current user, utilizing the target large language model based on the sorting result according to the historical portrait, screening commodity recommendation corresponding to the target shopping guide requirement of the user from the target data, and generating corresponding commodity recommendation explanation.
Optionally, after the generating the commodity recommendation and the commodity recommendation explanation corresponding to the target shopping guide requirement of the user by using the target large language model based on the sorting result, the method further includes:
and generating a satisfaction questionnaire corresponding to the current commodity recommendation and the commodity recommendation interpretation, and acquiring satisfaction degree and corresponding improvement opinion of a user according to the satisfaction degree questionnaire.
In a second aspect, the present application provides a dialogue recommendation device, including:
the demand acquisition module is used for acquiring the initial shopping guide demand of the user and judging whether the initial shopping guide demand of the user is collected;
the information extraction module is used for analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model if the collection is completed so as to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result;
the semantic recall module is used for carrying out semantic recall according to the target extraction result to obtain a semantic recall result, determining the similarity between the semantic recall result and the key information, and sorting the semantic recall result according to the similarity to obtain a sorting result;
and the recommendation generation module is used for generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result.
In a third aspect, the present application provides an electronic device comprising a processor and a memory; the memory is used for storing a computer program, and the computer program is loaded and executed by the processor to realize the conversation recommendation method.
In a fourth aspect, the present application provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the aforementioned dialog recommendation method.
In the application, first, the initial shopping guide requirement of a user is acquired, and whether the initial shopping guide requirement of the user is collected or not is judged; if the collection is completed, analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result; then carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result according to the similarity to obtain a sorting result; and generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result. Through the technical scheme, shopping guide dialogue management can be performed based on the large-scale language model, the recommendation result and the specific requirement of the user are more met through extraction and definition of the keywords, and the final recommendation result is explained based on the large-scale language model, so that the problems of lack of flexibility, excessively discrete recommendation result, poor matching degree with the original requirement of the user, low compactness, poor interpretation of the recommendation result and the like in the dialogue management process in the dialogue type recommendation system are solved, and the credibility of dialogue recommendation is enhanced.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings that are required to be 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 embodiments of the present application, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a dialogue recommendation method provided by the application;
FIG. 2 is a flowchart of a method for recommending conversations in detail according to the present application;
FIG. 3 is a logic diagram of a dialogue recommendation method according to the present application;
fig. 4 is a schematic structural diagram of a dialogue recommendation method device provided by the present application;
fig. 5 is a block diagram of an electronic device according to the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
In the existing conversational recommendation system, the management of conversations is usually based on a rule form, and the identification of key information is based on a predefined slot position, so that the generated conversations and recommendation results lack flexibility and interpretation. The application can conduct shopping guide dialogue management based on the large language model, and the recommendation result and the specific requirement of the user are more accordant through extracting and defining the keywords, and the final recommendation result is interpreted based on the large language model, so that the credibility of dialogue recommendation is enhanced.
Referring to fig. 1, the embodiment of the application discloses a dialogue recommendation method, which comprises the following steps:
and S11, acquiring the initial shopping guide requirement of the user, and judging whether the initial shopping guide requirement of the user is collected.
In the application, before the dialogue recommendation is performed, the shopping requirement of the current user, namely the initial shopping guide requirement, is firstly required to be acquired, and the requirement is usually descriptive text in a text format. It can be understood that when the current system is provided with the image recognition technology, a picture input by a user can be obtained, and related information on the picture is recognized, so that a description of a corresponding text format is further obtained. The shopping guide requirement can represent a requirement label of a user, including but not limited to a relevant label of the clothing industry such as budget, type, color, size, gender, material, style and the like. After the initial shopping guide requirement of the user is acquired, whether the initial shopping guide requirement of the user is collected is required to be judged, and it can be understood that the user may not input the requirement at one time when inputting the requirement, so that the embodiment can judge the requirement of the user and judge whether the current user finishes inputting the requirement, and specific embodiments include, but are not limited to, judging by acquiring an answer result in a question-answer mode. In this way, the recommendation result generated finally is more in line with the specific requirements of the user.
And step S12, if the collection is completed, analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model so as to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result.
In this embodiment, if it is determined that the collection of the initial shopping guide requirement of the user is completed, the context of the current dialogue is merged and then input to the initial large language model (LLMs, large Language Models), and the clothing industry core vocabulary in the current dialogue is screened out according to the preset vocabulary screening rule, and the initial large language model is fine-tuned by using a Low-Rank Adaptation (LORA) of the large language model based on the context and the clothing industry core vocabulary, so as to generate a target large language model, so that the current initial shopping guide requirement of the user is analyzed by using the target large language model as the target shopping guide requirement of the user. It is understood that the core vocabulary of the clothing industry includes, but is not limited to, special vocabulary in the clothing fields such as short sleeves and down jackets.
It should be noted that, in this embodiment, in order to reduce the dependency of resources, the base model of the large language model adopts chatGLM, and performs fine tuning through LORA, and the training mode for chatGLM is modified, so that the fine tuning mode is suitable for multiple rounds of conversations of shopping guide service. Specifically, when organizing data, the context of the current dialogue is merged and input, and when the context is too long, the context is input after the abstract is performed, and the weight of the core vocabulary focused by each round of dialogue is increased by introducing the core vocabulary, so that the shopping guide field chat is finally obtained, and the embodiment is named as a chat guide.
After LLMs are built, labels such as budget, type, color, size, gender, material, style, other and the like are predefined by taking the clothing industry as an example, and input is analyzed and matched to a corresponding type through natural language understanding capability of ChatGLM, so that key information of target shopping guide requirements of users is extracted, and the key information is fused to obtain a target extraction result. In this way, a multi-round dialogue model chatGuide in the shopping guide field can be generated, which is beneficial to flexible control of dialogue management.
And step S13, carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result according to the similarity to obtain a sorting result.
In this embodiment, semantic recall is performed according to the target extraction result to obtain a semantic recall result, and the similarity between the semantic recall result and the key information is determined, so that the semantic recall result is ranked according to the similarity to obtain a ranking result. When the key information of the target shopping guide requirement of the user is extracted and the key information is fused to obtain a target extraction result, the key information can be filtered according to a first preset information label, the filtered key information is filtered according to a second preset information label to obtain target key information, and the target key information is fused to obtain the target extraction result. For example, in performing semantic recall, data filtering is performed by budget, size based on information collected in multiple rounds of conversations; semantic recall is carried out through type, gender and style, then click data are exposed and clicked by a user, a query and the email of a commodity title are finely adjusted, similarity calculation is carried out on recalled commodities and extracted information based on the obtained model, and a sorting result is obtained through sorting. Therefore, the recommendation result can be matched with specific requirements of users as much as possible through the refined extraction and definition of the key information, and the reliability of commodity recommendation is enhanced.
And S14, generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result.
In this embodiment, based on the obtained sorting result, a commodity recommendation and a commodity recommendation explanation corresponding to the target shopping guide requirement of the user are generated by using a target large language model. And generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result, generating a satisfaction questionnaire corresponding to the current commodity recommendation and commodity recommendation explanation, and acquiring satisfaction degree and corresponding improvement opinion of the user according to the satisfaction degree questionnaire. In this way, the recommendation result can be optimized based on the feedback data of the user, which is beneficial to improving the effect of dialogue recommendation.
Through the technical scheme, the embodiment judges whether the initial shopping guide requirement of the user is collected. And if the collection is completed, analyzing by using the target large language model to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result. And then carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result to obtain a sorting result. And generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sequencing result. According to the embodiment, the user needs are judged, whether the current user finishes inputting the needs or not is judged, the finally generated recommendation result is more in line with the specific needs of the user, a multi-round dialogue model chatGuide in the shopping guide field can be generated, flexible control of dialogue management and optimization of the recommendation result based on feedback data of the user are facilitated, and the finally generated recommendation result is more in line with the specific needs of the user.
Based on the above embodiment, the present application can perform shopping guide dialogue management based on a large language model, and in this embodiment, the process of acquiring the user's requirements and the process of generating the explanation of the recommendation result will be described in detail. Referring to fig. 2, the embodiment of the application discloses a specific dialogue recommendation method, which includes:
and S21, acquiring initial shopping guide requirements of the user, judging whether the initial shopping guide requirements of the user are collected, and if the initial shopping guide requirements of the user are not collected, continuously inquiring the shopping guide requirements of the user according to a preset rule to acquire a corresponding user answer result.
In this embodiment, as shown in fig. 3, the initial shopping guide requirement of the user is obtained, and whether the initial shopping guide requirement of the user is collected is judged, if the initial shopping guide requirement of the user is not collected, the user is continuously queried according to a preset rule, and a corresponding user answer result is obtained. And after the corresponding user answer result is obtained, generating a user target shopping guide requirement according to the user answer result and the initial user shopping guide requirement, so as to analyze the user target shopping guide requirement based on the target large language model. According to the method and the device for inquiring the shopping guide requirements of the users, the corresponding questions and answers can be generated before the user requirements are not completely finished, the user shopping guide requirements are continuously inquired according to the preset rules, other requirements of the users are continuously acquired, the user requirements can be more comprehensively known, and the recommendation results are more tightly connected with the user requirements.
Step S22, judging whether the answer result of the user is the shopping guide requirement modification intention; if yes, the initial user shopping guide requirement is modified according to the shopping guide requirement modification intention, and the modified initial user shopping guide requirement is used as the user target shopping guide requirement.
In this embodiment, after obtaining the user answer result, it is determined whether the user answer result is the shopping guide requirement modification intention; if yes, the initial user shopping guide requirement is modified according to the shopping guide requirement modification intention, and the modified initial user shopping guide requirement is used as a user target shopping guide requirement. It can be appreciated that in the process of obtaining the user's requirement, the user may have an error in the process of inputting the requirement, so the embodiment can determine whether the requirement input by the current user is a request for modifying the historical input shopping guide requirement, thereby modifying the initial user shopping guide requirement according to the shopping guide requirement modification intention, so as to more accurately know the user requirement, and be beneficial to making the recommendation result more tightly connected with the user requirement. Thus, the method is favorable for realizing flexible control of dialogue management, can generate a multi-round dialogue model in shopping guide field, and realizes flexible dialogue management based on LLMs.
And S23, if the collection of the initial shopping guide requirement of the user is completed, analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model so as to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result.
And step S24, carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result according to the similarity to obtain a sorting result.
Step S25, acquiring preset self-owned clothing industry shopping guide data, and fusing the preset self-owned clothing industry shopping guide data with clothing industry data in preset third party public customer service data to obtain target data; and acquiring a historical portrait of the current user, utilizing the target large language model based on the sorting result according to the historical portrait, screening commodity recommendation corresponding to the target shopping guide requirement of the user from the target data, and generating corresponding commodity recommendation explanation.
In the embodiment, before commodity recommendation is performed, firstly acquiring preset self-owned clothing industry shopping guide data, and fusing clothing industry data in preset self-owned clothing industry shopping guide data and preset third party public customer service data to obtain target data; and acquiring a historical portrait of the current user, utilizing a target large language model based on the sorting result according to the historical portrait, screening commodity recommendation corresponding to the target shopping guide requirement of the user from target data, and generating corresponding commodity recommendation explanation. Specifically, the target extraction result, the commodity recommendation, and the history image of the user may be input to the chatGLM, and the reason for the corresponding recommendation result, that is, the commodity recommendation interpretation may be generated. Therefore, the recommendation result can be tightly connected with the current requirement of the user, and the interpretability of the recommendation result is enhanced.
For more specific processing procedures in the steps S23 and S24, reference may be made to the corresponding contents disclosed in the foregoing embodiments, and no detailed description is given here.
Through the technical scheme, when the initial user shopping guide requirement is not collected and completed, the method continues to inquire the user shopping guide requirement according to the preset rule, obtains the corresponding user answer result, and judges whether the user answer result is the shopping guide requirement modification intention or not; if yes, the initial user shopping guide requirement is modified according to the shopping guide requirement modification intention, the initial user shopping guide requirement is analyzed by using a target large language model as the user target shopping guide requirement, so that a target extraction result is obtained, semantic recall is carried out, and a sequencing result is further obtained. Acquiring preset self-contained clothing industry shopping guide data, and fusing the preset self-contained clothing industry shopping guide data with clothing industry data in preset third party public customer service data to obtain target data; and acquiring a historical portrait of the current user, utilizing a target large language model based on the sorting result according to the historical portrait, screening commodity recommendation corresponding to the target shopping guide requirement of the user from target data, and generating corresponding commodity recommendation explanation. Therefore, the method is beneficial to flexibly controlling the dialogue recommendation flow, can solve the problems of lack of flexibility, excessively discrete recommendation results, low matching degree and compactness of original requirements of users in the dialogue management process in the dialogue recommendation system, and enhances the credibility of dialogue recommendation.
Referring to fig. 5, the embodiment of the application also discloses a dialogue recommendation device, which comprises:
the demand acquisition module 11 is configured to acquire a user initial shopping guide demand, and determine whether the user initial shopping guide demand is collected;
the information extraction module 12 is configured to analyze the current user initial shopping guide requirement as a user target shopping guide requirement by using a target large language model if the collection is completed, so as to extract key information of the user target shopping guide requirement, and fuse the key information to obtain a target extraction result;
the semantic recall module 13 is configured to perform semantic recall according to the target extraction result to obtain a semantic recall result, and determine a similarity between the semantic recall result and the key information, so as to sort the semantic recall result according to the similarity to obtain a sorted result;
and the recommendation generation module 14 is used for generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result.
In this embodiment, first, a user initial shopping guide requirement is obtained, and whether the user initial shopping guide requirement is collected is judged; if the collection is completed, analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result; then carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result according to the similarity to obtain a sorting result; and generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result. Through the technical scheme, shopping guide dialogue management can be performed based on the large language model, the recommendation result and the specific requirement of a user are more met through extraction and definition of the keywords, and the final recommendation result is explained based on the large language model, so that the problems of lack of flexibility in the dialogue management process and poor interpretation of the recommendation result in a dialogue type recommendation system are solved, and the credibility of dialogue recommendation is enhanced.
In some embodiments, the requirement acquisition module 11 further includes:
the result acquisition unit is used for continuously inquiring the user shopping guide requirement according to a preset rule if the initial user shopping guide requirement is not collected, and acquiring a corresponding user answer result;
and the demand analysis sub-module is used for generating a user target shopping guide demand according to the user answer result and the initial user shopping guide demand so as to analyze the user target shopping guide demand based on the target large language model.
In some embodiments, the requirement parsing sub-module specifically includes:
the result judging unit is used for judging whether the answer result of the user is the shopping guide requirement modification intention or not;
and the demand modification unit is used for modifying the initial user shopping guide demand according to the shopping guide demand modification intention if yes, and taking the modified initial user shopping guide demand as the user target shopping guide demand.
In some embodiments, the information extraction module 12 specifically includes:
the model fine tuning unit is used for merging the contexts of the current dialogue and inputting the merged contexts into the initial large language model, screening out the clothing industry core vocabulary in the current dialogue according to a preset vocabulary screening rule, fine tuning the initial large language model by using a low-rank adaptation technology of the large language model based on the contexts and the clothing industry core vocabulary, generating a target large language model, and analyzing the current initial shopping guide requirement of the user as the target shopping guide requirement of the user by using the target large language model.
In some embodiments, the information extraction module 12 specifically includes:
the information filtering unit is used for filtering the key information according to a first preset information label, screening the filtered key information according to a second preset information label to obtain target key information, and fusing the target key information to obtain a target extraction result.
In some embodiments, the recommendation generation module 14 specifically includes:
the data fusion unit is used for acquiring preset own clothing industry shopping guide data, and fusing the preset own clothing industry shopping guide data with clothing industry data in preset third party public customer service data to obtain target data;
and the interpretation generation unit is used for acquiring a historical portrait of the current user, utilizing the target large language model based on the sorting result according to the historical portrait, screening commodity recommendation corresponding to the target shopping guide requirement of the user from the target data, and generating corresponding commodity recommendation interpretation.
In some specific embodiments, the dialogue recommendation device further includes:
and the opinion acquisition unit is used for generating a satisfaction degree questionnaire corresponding to the current commodity recommendation and the commodity recommendation interpretation, and acquiring satisfaction degree and corresponding improvement opinion of a user according to the satisfaction degree questionnaire.
Further, the embodiment of the present application further discloses an electronic device, and fig. 5 is a block diagram of an electronic device 20 according to an exemplary embodiment, where the content of the figure is not to be considered as any limitation on the scope of use of the present application.
Fig. 5 is a schematic structural diagram of an electronic device 20 according to an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input output interface 25, and a communication bus 26. The memory 22 is configured to store a computer program, which is loaded and executed by the processor 21 to implement relevant steps in the dialog recommendation method disclosed in any of the foregoing embodiments. In addition, the electronic device 20 in the present embodiment may be specifically an electronic computer.
In this embodiment, the power supply 23 is configured to provide an operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and an external device, and the communication protocol to be followed is any communication protocol applicable to the technical solution of the present application, which is not specifically limited herein; the input/output interface 25 is used for acquiring external input data or outputting external output data, and the specific interface type thereof may be selected according to the specific application requirement, which is not limited herein.
The memory 22 may be a carrier for storing resources, such as a read-only memory, a random access memory, a magnetic disk, or an optical disk, and the resources stored thereon may include an operating system 221, a computer program 222, and the like, and the storage may be temporary storage or permanent storage.
The operating system 221 is used for managing and controlling various hardware devices on the electronic device 20 and computer programs 222, which may be Windows Server, netware, unix, linux, etc. The computer program 222 may further include a computer program that can be used to perform other specific tasks in addition to the computer program that can be used to perform the dialog recommendation method performed by the electronic device 20 disclosed in any of the previous embodiments.
Further, the application also discloses a computer readable storage medium for storing a computer program; wherein the computer program, when executed by a processor, implements the previously disclosed dialog recommendation method. For specific steps of the method, reference may be made to the corresponding contents disclosed in the foregoing embodiments, and no further description is given here.
In this specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, so that the same or similar parts between the embodiments are referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
Those of skill would further appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or combinations of both, and that the various illustrative elements and steps are described above generally in terms of functionality in order to clearly illustrate the interchangeability of hardware and software. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present application.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software modules may be disposed in Random Access Memory (RAM), memory, read Only Memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
Finally, it is further noted that relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing has outlined rather broadly the more detailed description of the application in order that the detailed description of the application that follows may be better understood, and in order that the present principles and embodiments may be better understood; meanwhile, as those skilled in the art will have variations in the specific embodiments and application scope in accordance with the ideas of the present application, the present description should not be construed as limiting the present application in view of the above.
Claims (10)
1. A method of conversational recommendation, comprising:
acquiring initial shopping guide requirements of a user, and judging whether the initial shopping guide requirements of the user are collected;
if the collection is completed, analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result;
carrying out semantic recall according to the target extraction result to obtain a semantic recall result, and determining the similarity between the semantic recall result and the key information so as to sort the semantic recall result according to the similarity to obtain a sorting result;
and generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result.
2. The dialogue recommendation method according to claim 1, wherein after the obtaining the initial shopping guide requirement of the user and determining whether the collection of the initial shopping guide requirement of the user is completed, further comprises:
if the initial user shopping guide requirements are not collected, inquiring the user shopping guide requirements continuously according to a preset rule, and acquiring a corresponding user answer result;
and generating a user target shopping guide requirement according to the user answer result and the initial user shopping guide requirement, so as to analyze the user target shopping guide requirement based on the target large language model.
3. The dialogue recommendation method according to claim 2, wherein said generating a user target shopping guide requirement based on said user answer result and said initial user shopping guide requirement comprises:
judging whether the answer result of the user is the shopping guide requirement modification intention or not;
if yes, the initial user shopping guide requirement is modified according to the shopping guide requirement modification intention, and the modified initial user shopping guide requirement is used as the user target shopping guide requirement.
4. The dialogue recommendation method according to claim 1, wherein the parsing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model includes:
and merging the contexts of the current dialogue, inputting the merged contexts into an initial large language model, screening out clothing industry core words in the current dialogue according to preset word screening rules, fine-tuning the initial large language model by using a low-rank adaptation technology of the large language model based on the contexts and the clothing industry core words, and generating a target large language model so as to analyze the current initial shopping guide requirement of the user as the target shopping guide requirement of the user by using the target large language model.
5. The dialogue recommendation method according to claim 1, wherein the extracting key information of the target shopping guide requirement of the user and fusing the key information to obtain a target extraction result comprises:
and filtering the key information according to a first preset information label, and screening the filtered key information according to a second preset information label to obtain target key information so as to fuse the target key information to obtain a target extraction result.
6. The dialogue recommendation method according to claim 1, wherein the generating commodity recommendation and commodity recommendation explanation corresponding to the user target shopping guide requirement using the target large language model based on the ranking result comprises:
acquiring preset self-owned clothing industry shopping guide data, and fusing the preset self-owned clothing industry shopping guide data with clothing industry data in preset third party public customer service data to obtain target data;
and acquiring a historical portrait of the current user, utilizing the target large language model based on the sorting result according to the historical portrait, screening commodity recommendation corresponding to the target shopping guide requirement of the user from the target data, and generating corresponding commodity recommendation explanation.
7. The dialogue recommendation method according to any one of claims 1 to 6, wherein after generating commodity recommendation and commodity recommendation explanation corresponding to the user target shopping guide requirement using the target large language model based on the ranking result, further comprising:
and generating a satisfaction questionnaire corresponding to the current commodity recommendation and the commodity recommendation interpretation, and acquiring satisfaction degree and corresponding improvement opinion of a user according to the satisfaction degree questionnaire.
8. A dialogue recommendation device, comprising:
the demand acquisition module is used for acquiring the initial shopping guide demand of the user and judging whether the initial shopping guide demand of the user is collected;
the information extraction module is used for analyzing the current initial shopping guide requirement of the user as a target shopping guide requirement of the user by using a target large language model if the collection is completed so as to extract key information of the target shopping guide requirement of the user, and fusing the key information to obtain a target extraction result;
the semantic recall module is used for carrying out semantic recall according to the target extraction result to obtain a semantic recall result, determining the similarity between the semantic recall result and the key information, and sorting the semantic recall result according to the similarity to obtain a sorting result;
and the recommendation generation module is used for generating commodity recommendation and commodity recommendation explanation corresponding to the target shopping guide requirement of the user by utilizing the target large language model based on the sorting result.
9. An electronic device comprising a processor and a memory; wherein the memory is for storing a computer program to be loaded and executed by the processor to implement the dialog recommendation method of any of claims 1 to 7.
10. A computer readable storage medium for storing a computer program which when executed by a processor implements the dialog recommendation method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310720521.3A CN116595150A (en) | 2023-06-16 | 2023-06-16 | Dialogue recommendation method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310720521.3A CN116595150A (en) | 2023-06-16 | 2023-06-16 | Dialogue recommendation method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116595150A true CN116595150A (en) | 2023-08-15 |
Family
ID=87593842
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310720521.3A Pending CN116595150A (en) | 2023-06-16 | 2023-06-16 | Dialogue recommendation method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116595150A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116911312A (en) * | 2023-09-12 | 2023-10-20 | 深圳须弥云图空间科技有限公司 | Task type dialogue system and implementation method thereof |
| CN117033641A (en) * | 2023-10-07 | 2023-11-10 | 江苏微皓智能科技有限公司 | Network structure optimization fine tuning method of large-scale pre-training language model |
| CN117350825A (en) * | 2023-12-05 | 2024-01-05 | 浙江口碑网络技术有限公司 | Shopping guide method and system based on large language model |
| CN117391824A (en) * | 2023-12-11 | 2024-01-12 | 深圳须弥云图空间科技有限公司 | Method and device for recommending articles based on large language model and search engine |
| CN117407595A (en) * | 2023-12-14 | 2024-01-16 | 江西财经大学 | Home decoration designer recommendation method integrating large language model and dynamic dialogue intention |
-
2023
- 2023-06-16 CN CN202310720521.3A patent/CN116595150A/en active Pending
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116911312A (en) * | 2023-09-12 | 2023-10-20 | 深圳须弥云图空间科技有限公司 | Task type dialogue system and implementation method thereof |
| CN116911312B (en) * | 2023-09-12 | 2024-01-05 | 深圳须弥云图空间科技有限公司 | Task type dialogue system and implementation method thereof |
| CN117033641A (en) * | 2023-10-07 | 2023-11-10 | 江苏微皓智能科技有限公司 | Network structure optimization fine tuning method of large-scale pre-training language model |
| CN117350825A (en) * | 2023-12-05 | 2024-01-05 | 浙江口碑网络技术有限公司 | Shopping guide method and system based on large language model |
| CN117350825B (en) * | 2023-12-05 | 2024-03-15 | 浙江口碑网络技术有限公司 | Shopping guide method and system based on large language model |
| CN117391824A (en) * | 2023-12-11 | 2024-01-12 | 深圳须弥云图空间科技有限公司 | Method and device for recommending articles based on large language model and search engine |
| CN117391824B (en) * | 2023-12-11 | 2024-04-12 | 深圳须弥云图空间科技有限公司 | Method and device for recommending articles based on large language model and search engine |
| CN117407595A (en) * | 2023-12-14 | 2024-01-16 | 江西财经大学 | Home decoration designer recommendation method integrating large language model and dynamic dialogue intention |
| CN117407595B (en) * | 2023-12-14 | 2024-03-08 | 江西财经大学 | Home decoration designer recommendation method integrating large language model and dynamic dialogue intention |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111026842B (en) | Natural language processing method, natural language processing device and intelligent question-answering system | |
| CN116595150A (en) | Dialogue recommendation method, device, equipment and storage medium | |
| US11003863B2 (en) | Interactive dialog training and communication system using artificial intelligence | |
| CN107784033B (en) | Method and device for recommending based on session | |
| CN111858877A (en) | Multi-type question intelligent question answering method, system, equipment and readable storage medium | |
| CN107291840B (en) | User attribute prediction model construction method and device | |
| CN111429157A (en) | Method, device and equipment for evaluating and processing complaint work order and storage medium | |
| KR101450453B1 (en) | Method and apparatus for recommending contents | |
| CN110399473B (en) | Method and device for determining answers to user questions | |
| CN114742032A (en) | Interactive data analysis method, apparatus, device, medium, and program product | |
| CN112860878A (en) | Service data recommendation method, storage medium and equipment | |
| CN112447173A (en) | Voice interaction method and device and computer storage medium | |
| CN116127011A (en) | Intention recognition method, device, electronic equipment and storage medium | |
| US20230106590A1 (en) | Question-answer expansion | |
| CN114969544A (en) | Hot data-based recommended content generation method, device, equipment and medium | |
| CN113468306A (en) | Voice conversation method, device, electronic equipment and storage medium | |
| CN113688636A (en) | Extended question recommendation method and device, computer equipment and storage medium | |
| CN112131379A (en) | Method, device, electronic equipment and storage medium for identifying problem category | |
| CN112148939A (en) | Data processing method and device and electronic equipment | |
| KR102500962B1 (en) | Video based outsourcing brokerage and match service platform server and operation method of the platform server | |
| CN115905490B (en) | Man-machine interaction dialogue method, device and equipment | |
| CN112925972B (en) | Information pushing method, device, electronic equipment and storage medium | |
| CN113609275B (en) | Information processing method, device, equipment and storage medium | |
| CN116451678B (en) | Data relation recognition and data table integration method | |
| CN117909477A (en) | Business and government affair dialogue management method and system based on emotion analysis |
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 |