Movie recommendation method and system based on knowledge graph and large model

Abstract

The invention is suitable for the technical field of movie recommendation, and provides a movie recommendation method and system based on a knowledge graph and a large model. The input information of the user is received and recognized, and whether the input information is related to movie recommendation or not is judged; when the input information is related to movie recommendation, carrying out movie retrieval based on attribute problems, entity basis or non-entity basis on the basis of the knowledge graph, and obtaining a movie retrieval result; and based on a movie retrieval result, enhancing input information, generating splicing information, importing the splicing information into the universal LLM, and obtaining exported movie recommendation information. According to the method, interests and demands of users can be deeply understood, movie recommendation more conforming to hobbies of the users is provided for the users, the users are helped to quickly find interested works in massive movie resources, time for screening movies by the users is saved, movie watching experience is improved, various information of the movies can be integrated, effective and accurate recommendation is achieved, and user experience is improved. And film related information can be updated in real time, so that the recommendation system can adapt to changes in time.

Description

Technical Field

[0001] The present invention belongs to the technical field of movie recommendation, and in particular to a movie recommendation method and system based on knowledge graph and big model. Background Art

[0002] With the rapid development of Internet technology, the film industry has also ushered in unprecedented prosperity. However, faced with a massive amount of film resources, how users can quickly find the movies they are interested in has become an urgent problem to be solved.

[0003] Traditional movie recommendation methods are mainly based on information such as user historical behavior, ratings, and tags, and often have the following shortcomings:

[0004] (1) Low precision: Traditional recommendation systems cannot accurately capture changes in user interests, resulting in deviations between recommendation results and actual user needs; (2) Poor scalability: As the amount of data increases, the computational complexity and storage cost of traditional recommendation systems increase linearly, making it difficult to cope with big data environments; (3) Lack of deep semantic understanding: Traditional recommendation methods only focus on the surface connection between users and movies and are unable to mine the deep semantic information behind the movie content. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a movie recommendation method and system based on knowledge graph and large model, aiming to solve the technical problems existing in the prior art mentioned in the background technology.

[0006] The embodiment of the present invention is implemented as follows:

[0007] A movie recommendation method based on knowledge graph and big model, the method specifically comprises the following steps:

[0008] Receive input information from the user, and identify the input information based on a preset universal LLM to determine whether it is related to movie recommendations;

[0009] When the input information is related to movie recommendation, the access function and the access parameter are parsed, and movie search based on attribute problem, entity basis or non-entity basis is performed based on the knowledge graph to obtain movie search results;

[0010] Based on the movie search results, the input information is enhanced, splicing information is generated and imported into the general LLM, and the derived movie recommendation information is obtained.

[0011] As a further limitation of the technical solution of the embodiment of the present invention, the receiving of the user's input information, identifying the input information based on the preset universal LLM, and determining whether the input information is related to the movie recommendation specifically includes the following steps:

[0012] Receive user input requests and create an input interface;

[0013] In the input interface, receiving input information;

[0014] Importing the input information into a preset universal LLM;

[0015] The input information is identified through the universal LLM to determine whether it is related to movie recommendations.

[0016] As a further limitation of the technical solution of the embodiment of the present invention, when the input information is related to movie recommendation, parsing the access function and the access parameter, performing attribute problem, entity-based or non-entity-based movie retrieval based on the knowledge graph, and obtaining the movie retrieval results specifically include the following steps:

[0017] When the input information is related to movie recommendation, processing the input information and parsing access functions and access parameters;

[0018] According to the access function and the access parameter, matching the corresponding process branch;

[0019] If the process branch is an attribute problem, the general LLM is used to generate a Cypher query language to obtain an answer and communicate with the movie knowledge graph to obtain movie retrieval results;

[0020] If the process branch is entity-based, node similarity analysis is performed based on the knowledge graph, multiple similar entities are recommended, and movie search results are obtained;

[0021] If the process branch is a non-entity basis, the cosine similarity metric is used to perform a vector similarity search to find the movie and obtain the movie retrieval result.

[0022] As a further limitation of the technical solution of the embodiment of the present invention, the attribute problem includes: movie attribute problem and actor attribute problem.

[0023] As a further limitation of the technical solution of the embodiment of the present invention, the entity basis includes: rating, actor, type, director, similar actor, dissimilar actor, production company, country, and national language.

[0024] As a further limitation of the technical solution of the embodiment of the present invention, the node similarity analysis based on the knowledge graph, recommending multiple similar entities, and obtaining movie search results specifically include the following steps:

[0025] Analyze based on knowledge graph to obtain entity relationships;

[0026] According to the entity relationship, perform node similarity analysis and recommend multiple similar entities;

[0027] Based on multiple similar entities, movie search results are obtained.

[0028] As a further limitation of the technical solution of the embodiment of the present invention, the non-physical basis includes: movie plot and genre issues.

[0029] As a further limitation of the technical solution of the embodiment of the present invention, the use of cosine similarity measurement to perform vector similarity search to find movies and obtain movie search results specifically includes the following steps:

[0030] Create a vector database through the preset Embedding model;

[0031] In the vector database, the cosine similarity metric is used to perform vector similarity search to obtain movie retrieval results.

[0032] As a further limitation of the technical solution of the embodiment of the present invention, the step of enhancing the input information based on the movie search results, generating splicing information and importing it into the general LLM, and obtaining the exported movie recommendation information specifically includes the following steps:

[0033] Based on the movie search results, the input information is enhanced to generate splicing information;

[0034] Importing the splicing information into the universal LLM to perform movie recommendation processing;

[0035] Get the exported movie recommendation information.

[0036] A movie recommendation system based on knowledge graph and big model, the system comprises a movie recommendation related judgment unit, a movie search result acquisition unit and a movie recommendation processing unit, wherein:

[0037] A movie recommendation related judgment unit, used to receive user input information, identify the input information based on a preset universal LLM, and judge whether it is related to movie recommendation;

[0038] A movie search result acquisition unit, used to parse the access function and access parameters when the input information is related to movie recommendation, and to perform attribute problem, entity-based or non-entity-based movie search based on the knowledge graph to obtain movie search results;

[0039] The movie recommendation processing unit is used to enhance the input information based on the movie search results, generate splicing information and import it into the general LLM to obtain the exported movie recommendation information.

[0040] Compared with the prior art, the present invention has the following beneficial effects:

[0041] The embodiment of the present invention receives the input information of the user for identification and determines whether it is related to the movie recommendation; when the input information is related to the movie recommendation, based on the knowledge graph, the movie search is performed based on attribute problems, entity basis or non-entity basis to obtain the movie search results; based on the movie search results, the input information is enhanced, the splicing information is generated and imported into the general LLM to obtain the exported movie recommendation information. It can understand the interests and needs of users more deeply, provide users with movie recommendations that are more in line with their preferences, help users quickly find works of interest in massive movie resources, save users time in screening movies, and improve the viewing experience. It can also integrate various information about movies to achieve effective and accurate recommendations, and can update movie-related information in real time, so that the recommendation system can adapt to changes in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A flowchart of a movie recommendation method based on a knowledge graph and a large model provided by an embodiment of the present invention is shown;

[0043] Figure 2 An application architecture diagram of a movie recommendation system based on a knowledge graph and a large model provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0044] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0045] It is understandable that traditional movie recommendation methods are mainly based on information such as user historical behavior, ratings, and tags, and often have the following shortcomings:

[0046] (1) Low precision: Traditional recommendation systems cannot accurately capture changes in user interests, resulting in deviations between recommendation results and actual user needs; (2) Poor scalability: As the amount of data increases, the computational complexity and storage cost of traditional recommendation systems increase linearly, making it difficult to cope with big data environments; (3) Lack of deep semantic understanding: Traditional recommendation methods only focus on the surface connection between users and movies and are unable to mine the deep semantic information behind the movie content.

[0047] To solve the above problems, the embodiments of the present invention disclose a movie recommendation method and system based on knowledge graph and big model, which receives the user's input information, identifies the input information based on the preset general LLM, and determines whether it is related to the movie recommendation;

[0048] When the input information is related to movie recommendations, the access function and access parameters are parsed, and movie retrieval based on attribute issues, entity basis or non-entity basis is performed based on the knowledge graph to obtain movie retrieval results;

[0049] Based on the movie search results, the input information is enhanced, the splicing information is generated and imported into the general LLM to obtain the exported movie recommendation information. It can understand the user's interests and needs more deeply, provide users with movie recommendations that are more in line with their preferences, help users quickly find interesting works in a large number of movie resources, save users time in screening movies, and improve the viewing experience. It can also integrate various information about movies to achieve effective and accurate recommendations, and can update movie-related information in real time, so that the recommendation system can adapt to changes in a timely manner.

[0050] Specifically, Figure 1 A flowchart of a movie recommendation method based on a knowledge graph and a large model provided by an embodiment of the present invention is shown.

[0051] In a preferred embodiment of the present invention, a movie recommendation method based on a knowledge graph and a large model comprises the following steps:

[0052] Step S101: receiving input information from a user, and identifying the input information based on a preset universal LLM to determine whether the input information is related to movie recommendations.

[0053] In an embodiment of the present invention, an input interface is created by receiving an input request from a user, and then in the input interface, the user's input information is received, and then the input information is imported into a preset universal LLM. The input information is identified through the universal LLM to determine whether it is related to movie recommendations. Specifically: if it is not related to movie recommendations, the user's question is answered directly based on the capabilities of the large model; if it is related to movie recommendations, step S102 is entered.

[0054] Step S102: When the input information is related to movie recommendation, parse the access function and access parameters, perform attribute problem, entity-based or non-entity-based movie search based on the knowledge graph, and obtain movie search results.

[0055] In an embodiment of the present invention, when the input information is related to movie recommendation, the input information is processed to parse the access function and the access parameter, and then the corresponding process branch is matched according to the access function and the access parameter. Different process branches include attribute problems, entity basis and non-entity basis. Specifically: when the process branch is an attribute problem (including movie attribute problems, actor attribute problems, etc.), a Cypher query statement is generated through LLM to obtain the answer to achieve dialogue with the knowledge graph, and then the movie retrieval result is obtained; when the process branch is an entity basis (including ratings, actors, types, directors, similar actors, dissimilar actors, production companies, countries, national languages, etc.), analysis is performed based on the knowledge graph to obtain entity relationships, and then node similarity analysis is performed according to the entity relationship to recommend multiple similar entities, and then the movie retrieval result is obtained based on multiple similar entities; when the process branch is a non-entity basis (including movie plots, type problems, etc.), a vector database is created through a preset Embedding model, and in the vector database, a vector similarity search is performed using the cosine similarity metric to obtain the movie retrieval result.

[0056] It can be understood that when the process branches into entities, basic node similarity can be used to find similarities between individual nodes, such as: there are multiple other nodes around the movie node, and these nodes influence each other based on the meaning of the movie: the genre nodes surrounding the movie will be influenced by the movie node, so we can find similar genres based on the interconnection of the movies: an actor is associated with a movie, so similar actors can be found based on each person's association with the movie. The node similarity algorithm compares a set of nodes based on the nodes they are connected to, and if two nodes share many of the same neighbors, they are considered similar.

[0057] It is understandable that in non-entity evidence, the questions asked by users are related to the plot or genre of the movie, which essentially has no entity or name, but is just a general statement or question about the plot of the movie.

[0058] Step S103: Based on the movie search results, the input information is enhanced, splicing information is generated and imported into the general LLM, and the derived movie recommendation information is obtained.

[0059] In an embodiment of the present invention, based on the movie search results, the input information is enhanced, the input information is spliced ​​with the movie search results to generate spliced ​​information, and then the spliced ​​information is imported into the general LLM to perform movie recommendation processing, thereby obtaining the exported movie recommendation information.

[0060] Furthermore, Figure 2 An application architecture diagram of a movie recommendation system based on a knowledge graph and a large model provided by an embodiment of the present invention is shown.

[0061] Among them, in another preferred embodiment provided by the present invention, a movie recommendation system based on knowledge graph and big model includes:

[0062] The movie recommendation related determination unit 101 is used to receive user input information, identify the input information based on a preset universal LLM, and determine whether it is related to movie recommendation.

[0063] In an embodiment of the present invention, the movie recommendation relevance judgment unit 101 creates an input interface by receiving a user's input request, and then receives the user's input information in the input interface, and then imports the input information into a preset universal LLM, and identifies the input information through the universal LLM to determine whether it is related to the movie recommendation.

[0064] The movie search result acquisition unit 102 is used to parse the access function and access parameters when the input information is related to movie recommendation, and perform attribute problem, entity-based or non-entity-based movie search based on the knowledge graph to obtain movie search results.

[0065] In an embodiment of the present invention, when the input information is related to movie recommendation, the movie search result acquisition unit 102 processes the input information, parses the access function and the access parameter, and then matches the corresponding process branch according to the access function and the access parameter. Different process branches include attribute problems, entity basis and non-entity basis. Specifically: when the process branch is an attribute problem (including movie attribute problems, actor attribute problems, etc.), a Cypher query statement is generated through LLM to obtain the answer to achieve dialogue with the knowledge graph, and then obtain the movie search result; when the process branch is an entity basis (including ratings, actors, types, directors, similar actors, dissimilar actors, production companies, countries, national languages, etc.), analysis is performed based on the knowledge graph to obtain entity relationships, and then node similarity analysis is performed according to the entity relationship to recommend multiple similar entities, and then movie search results are obtained based on multiple similar entities; when the process branch is a non-entity basis (including movie plots, type problems, etc.), a vector database is created through a preset Embedding model, and in the vector database, a vector similarity search is performed using the cosine similarity metric to obtain movie search results.

[0066] The movie recommendation processing unit 103 is used to enhance the input information based on the movie search results, generate splicing information and import it into the general LLM to obtain the derived movie recommendation information.

[0067] In an embodiment of the present invention, the movie recommendation processing unit 103 enhances the input information based on the movie retrieval results, splices the input information with the movie retrieval results to generate spliced ​​information, and then imports the spliced ​​information into the general LLM to perform movie recommendation processing, thereby obtaining the exported movie recommendation information.

[0068] It should be understood that, although each step in the flow chart of each embodiment of the present invention is shown in sequence according to the indication of the arrow, these steps are not necessarily performed in sequence according to the order indicated by the arrow. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be performed in other orders. Moreover, at least a portion of the steps in each embodiment may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily performed at the same time, but can be performed at different times, and the execution order of these sub-steps or stages is not necessarily performed in sequence, but can be performed in turn or alternately with at least a portion of other steps or sub-steps or stages of other steps.

[0069] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0070] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A movie recommendation method based on knowledge graph and big model, characterized in that: The method specifically comprises the following steps: Receive input information from the user, and identify the input information based on a preset universal LLM to determine whether it is related to movie recommendations; When the input information is related to movie recommendation, the access function and the access parameter are parsed, and movie search based on attribute problem, entity basis or non-entity basis is performed based on the knowledge graph to obtain movie search results; Based on the movie search results, the input information is enhanced, splicing information is generated and imported into the general LLM, and the derived movie recommendation information is obtained.

2. The movie recommendation method based on knowledge graph and large model according to claim 1 is characterized in that: The receiving of user input information, identifying the input information based on a preset universal LLM, and determining whether the input information is related to movie recommendation specifically comprises the following steps: Receive user input requests and create an input interface; In the input interface, receiving input information; Importing the input information into a preset universal LLM; The input information is identified through the universal LLM to determine whether it is related to movie recommendations.

3. The movie recommendation method based on knowledge graph and large model according to claim 1 is characterized in that: When the input information is related to movie recommendation, parsing the access function and access parameters, performing attribute problem, entity-based or non-entity-based movie retrieval based on the knowledge graph, and obtaining the movie retrieval results specifically include the following steps: When the input information is related to movie recommendation, processing the input information and parsing access functions and access parameters; According to the access function and the access parameter, matching the corresponding process branch; If the process branch is an attribute problem, the general LLM is used to generate a Cypher query language to obtain an answer and communicate with the movie knowledge graph to obtain movie retrieval results; If the process branch is entity-based, node similarity analysis is performed based on the knowledge graph, multiple similar entities are recommended, and movie search results are obtained; If the process branch is a non-entity basis, the cosine similarity metric is used to perform a vector similarity search to find the movie and obtain the movie retrieval result.

4. The movie recommendation method based on knowledge graph and large model according to claim 3 is characterized in that: The attribute issues include: movie attribute issues and actor attribute issues.

5. The movie recommendation method based on knowledge graph and large model according to claim 3 is characterized in that: The entity basis includes: rating, actor, type, director, similar actors, dissimilar actors, production company, country, and national language.

6. The movie recommendation method based on knowledge graph and large model according to claim 3 is characterized in that: The method of performing node similarity analysis based on the knowledge graph, recommending multiple similar entities, and obtaining movie search results specifically includes the following steps: Analyze based on knowledge graph to obtain entity relationships; According to the entity relationship, perform node similarity analysis and recommend multiple similar entities; Based on multiple similar entities, movie search results are obtained.

7. The movie recommendation method based on knowledge graph and large model according to claim 3 is characterized in that: The non-physical basis includes: movie plot and genre issues.

8. The movie recommendation method based on knowledge graph and large model according to claim 3 is characterized in that: The method of using the cosine similarity metric to perform vector similarity search to find movies and obtain movie search results specifically includes the following steps: Create a vector database through the preset Embedding model; In the vector database, the cosine similarity metric is used to perform vector similarity search to obtain movie retrieval results.

9. The movie recommendation method based on knowledge graph and large model according to claim 1, characterized in that: The method of enhancing the input information based on the movie search results, generating splicing information and importing it into the general LLM, and obtaining the exported movie recommendation information specifically comprises the following steps: Based on the movie search results, the input information is enhanced to generate splicing information; Importing the splicing information into the universal LLM to perform movie recommendation processing; Get the exported movie recommendation information.

10. A movie recommendation system based on knowledge graph and big model, characterized in that: The system includes a movie recommendation related judgment unit, a movie search result acquisition unit and a movie recommendation processing unit, wherein: A movie recommendation related judgment unit, used to receive user input information, identify the input information based on a preset universal LLM, and judge whether it is related to movie recommendation; A movie search result acquisition unit, used to parse the access function and access parameters when the input information is related to movie recommendation, and to perform attribute problem, entity-based or non-entity-based movie search based on the knowledge graph to obtain movie search results; The movie recommendation processing unit is used to enhance the input information based on the movie search results, generate splicing information and import it into the general LLM to obtain the exported movie recommendation information.

Images