KR-102961255-B1 - Apparatus and method for generating content curation information based on language model

KR102961255B1KR 102961255 B1KR102961255 B1KR 102961255B1KR-102961255-B1

Abstract

The present invention relates to a language model-based content curation information generation device and a content curation information generation method. A language model-based content curation information generation device according to one embodiment of the present invention may include: a model learning unit that generates a content recommendation model, a synopsis summary model, and a marketing phrase generation model by performing transfer learning from a large language model (LLM); a content recommendation unit that generates a personalized recommended content list including recommended content recommended to the user among the plurality of content by inputting user characteristic information and metadata of a plurality of content into the content recommendation model; a marketing phrase generation unit that generates personalized marketing phrases for the recommended content by inputting the characteristic information and metadata into the marketing phrase generation model; and a synopsis summary unit that generates a personalized synopsis summary for the user by inputting the synopses of the recommended content, the characteristic information, and the metadata into the synopsis summary model.

Inventors

김은주
김지연
원세연
장두성

Assignees

주식회사 케이티

Dates

Publication Date: 20260508
Application Date: 20230725

Claims (15)

A model learning unit that generates a content recommendation model, a synopsis summary model, and a marketing phrase generation model that perform pre-set tasks by performing transfer learning from a Large Language Model (LLM); A content recommendation unit that inputs user characteristic information and metadata of multiple contents into the above content recommendation model to generate a personalized recommended content list including recommended contents recommended to the user among the multiple contents; A marketing phrase generation unit that inputs the user characteristic information and metadata regarding the recommended content into the marketing phrase generation model to generate a marketing phrase recommending the viewing of the recommended content personalized for each user; and It includes a synopsis summary unit that generates a personalized synopsis summary for the user by inputting the synopses of the recommended content, the user characteristic information, and meta-information about the recommended content into the synopsis summary model. The above characteristic information is Preference information including at least one of the genre, subgenre, actor, theme, time setting, spatial setting, character, director, and sentiment of the content preferred by the user, and It includes history information, which is a viewing list of content watched by the above user, and The above meta-information is It includes at least one of the genre, subgenre, theme, cast, time setting, spatial setting, characters, director, and sentiment corresponding to the above content, and The above content recommendation department A language model-based content curation information generation device that extracts metadata of content viewed by the user from the above history information, and extracts recommended content by comparing the metadata of content viewed by the user and the preference information with the metadata of the plurality of content.
delete
In paragraph 1, the above history information A language model-based content curation information generating device that stores the titles of the contents included in the above viewing list as text.
In paragraph 3, the above content recommendation unit A language model-based content curation information generation device that, if the number of viewed content items included in the user's history information is less than a set number, additionally extracts the viewing lists from other content platforms subscribed to by the user and merges the extracted viewing lists into the history information.
In paragraph 1, the content recommendation unit A language model-based content curation information generating device that uses Explainable Artificial Intelligence (XAI) as the content recommendation model to extract recommended content for a user from among multiple contents and generates a recommendation sentence indicating the reason for the recommendation of the recommended content.
In paragraph 1, the marketing phrase generation unit A language model-based content curation information generation device that, upon receiving a list of contents included in a content container, extracts metadata of the contents included in the list, and generates a container name corresponding to the content container using the extracted metadata and the user's characteristic information.
In paragraph 6, the above marketing phrase generation unit A language model-based content curation information generation device that extracts representative features of a content container based on the frequency of words appearing in the meta-information of the contents included in the content container, and generates a container name using the representative features and the characteristic information.
In a method for generating content curation information based on a language model of a content curation information generating device, A step of generating a content recommendation model, a synopsis summary model, and a marketing phrase generation model that perform pre-set tasks by performing transfer learning from a Large Language Model (LLM); A step of inputting user characteristic information and meta-information of multiple contents into the above content recommendation model to generate a personalized recommendation content list including recommendation content recommended to the user among the multiple contents; A step of inputting the user characteristic information and meta-information regarding the recommended content into the marketing phrase generation model to generate a marketing phrase recommending the viewing of the recommended content personalized for each user; and The method includes the step of generating a personalized synopsis summary for the user by inputting the synopses of the recommended content, the user characteristic information, and meta-information regarding the recommended content into the synopsis summary model. The above characteristic information is Preference information including at least one of the genre, subgenre, actor, theme, time setting, spatial setting, character, director, and sentiment of the content preferred by the user, and It includes history information, which is a viewing list of content watched by the above user, and The above meta-information is It includes at least one of the genre, subgenre, theme, cast, time setting, spatial setting, characters, director, and sentiment corresponding to the above content, and The step of generating the above recommended content list is A method for generating content curation information, wherein the method extracts metadata of content viewed by the user from the above history information, and extracts recommended content by comparing the metadata of content viewed by the user and the above preference information with the metadata of the plurality of content.
delete
In paragraph 8, the above history information A method for generating content curation information, wherein the titles of the contents included in the above viewing list are stored as text.
In item 10, the step of generating the above-mentioned recommended content list A method for generating content curation information, wherein if the number of viewed content items included in the user's history information is less than a set number, additional viewing lists are extracted from other content platforms subscribed to by the user, and the extracted viewing lists are merged into the history information.
In paragraph 8, the step of generating the above-mentioned recommended content list A method for generating content curation information, wherein an Explainable Artificial Intelligence (XAI) is used as the content recommendation model to extract recommended content for a user from among multiple contents and to generate a recommendation sentence indicating the reason for the recommendation of the recommended content.
In paragraph 8, the step of generating the marketing phrase above A method for generating content curation information, wherein, upon receiving a list of contents included in a content container, meta information of the contents included in the list is extracted, and a container name corresponding to the content container is generated using the extracted meta information and the user characteristic information.
In Clause 13, the step of generating the above marketing phrase A method for generating content curation information, wherein the representative feature of the content container is extracted based on the frequency of words appearing in the meta-information of the contents included in the content container, and the container name is generated using the representative feature and the characteristic information.
A computer program stored on a medium to execute a method for generating content curation information according to any one of claims 8, 10 to 14, in combination with hardware.

Description

Apparatus and method for generating content curation information based on language model The present invention relates to a language model-based content curation information generation device and a content curation information generation method capable of providing content curation information based on a language model. Recently, various content such as movies and music is being provided through media platforms, and users can receive content by downloading or streaming it through these platforms. Since media platforms possess a diverse range of content, they can offer users a wide selection. Furthermore, by continuously updating available content, the amount of content available to users can gradually increase. Through this, users can access the media platform with a broad intent, without having specific content in mind, and search for the content they desire. Here, media platforms may implement recommendation algorithms that suggest content the user is likely to like in order to help the user find content. In this case, the recommendation algorithm may be based on the content the user has previously consumed. FIG. 1 is a block diagram showing a content curation system according to one embodiment of the present invention. FIG. 2 is a schematic diagram illustrating the generation of a content recommendation model, a synopsis summary model, and a marketing phrase generation model according to an embodiment of the present invention. FIG. 3 is a schematic diagram illustrating the operation of a content recommendation model according to one embodiment of the present invention. FIG. 4 is a schematic diagram illustrating the operation of a marketing phrase generation model according to one embodiment of the present invention. FIG. 5 is a schematic diagram illustrating the generation of a container name of a marketing phrase generation model according to an embodiment of the present invention. FIG. 6 is an exemplary diagram illustrating the generation of a container name according to an embodiment of the present invention. FIG. 7 is a schematic diagram showing a synopsis summary model according to one embodiment of the present invention. FIG. 8 is a block diagram showing a content curation information generating device according to another embodiment of the present invention. FIG. 9 is a flowchart illustrating a method for generating content curation information according to an embodiment of the present invention. Hereinafter, embodiments disclosed in this specification will be described in detail with reference to the attached drawings. Identical or similar components regardless of drawing symbols will be assigned the same reference number, and redundant descriptions thereof will be omitted. The suffixes "module" and "part" for components used in the following description are assigned or used interchangeably solely for the ease of drafting the specification and do not inherently possess distinct meanings or roles. That is, the term "part" used in this invention refers to a hardware component such as software, FPGA, or ASIC, and the "part" performs certain roles. However, the meaning of "part" is not limited to software or hardware. The "part" may be configured to reside in an addressable storage medium or may be configured to run one or more processors. Accordingly, as an example, a 'part' includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables. The functionality provided within the components and 'parts' may be combined into a smaller number of components and 'parts' or further separated into additional components and 'parts'. In addition, when describing the embodiments disclosed in this specification, if it is determined that a detailed description of related prior art may obscure the essence of the embodiments disclosed in this specification, such detailed description is omitted. Furthermore, the attached drawings are intended only to facilitate understanding of the embodiments disclosed in this specification, and the technical concept disclosed in this specification is not limited by the attached drawings; it should be understood that they include all modifications, equivalents, and substitutions that fall within the spirit and technical scope of the present invention. FIG. 1 is a block diagram showing a content curation system according to one embodiment of the present invention. Referring to FIG. 1, a content curation system according to one embodiment of the present invention may include a content curation information generating device (100) and a media platform (200). A content curation system according to an embodiment of the present invention will be described below with reference to FIG. 1. The content curation information generating devi