CN-122027832-A - Program dynamic arrangement method, system, equipment and medium

CN122027832ACN 122027832 ACN122027832 ACN 122027832ACN-122027832-A

Abstract

The application is suitable for the technical field of computers, and provides a program dynamic programming method which comprises the steps of collecting hot event data from a social media platform and a news platform in real time, collecting viewing behavior data of a user on a television platform at the same time, obtaining viewing preference characteristics, carrying out natural language processing on the hot event data, obtaining hot keywords, carrying out similarity calculation on the hot keywords and metadata information of a program material database, generating a hot matching result, generating an initial population based on the viewing preference characteristics and the hot matching result, optimizing the initial population by adopting a genetic algorithm, generating a program arrangement sequence, monitoring the change condition of the hot events and user preferences in real time, and dynamically adjusting the program arrangement sequence. The application realizes the full-process automation and the intellectualization from the hot spot information acquisition, the program matching to the program list generation and adjustment, reduces the workload and subjectivity of manual programming, and improves the efficiency and the quality of the programming.

Inventors

LIANG ZHEHUI
LI WEIHUA
QUAN DONGMING
FENG YUN
MA XIAO
ZHENG SENYANG
LIU XUN
DAI HAOLIN
XU YI
XU CHUNLIN

Assignees

南方电网数字传媒科技有限公司

Dates

Publication Date: 20260512
Application Date: 20251212

Claims (10)

1. A method of dynamically programming a program, comprising: Collecting hot event data from a social media platform and a news platform in real time, and collecting watching behavior data of a user on a television platform at the same time to obtain watching preference characteristics; performing natural language processing on the hot spot event data to obtain hot spot keywords, performing similarity calculation on the hot spot keywords and metadata information of a program material database, and generating a hot spot matching result; generating an initial population based on the viewing preference characteristics and the hot spot matching result, and optimizing the initial population by adopting a genetic algorithm to generate a program arrangement sequence; and monitoring the change conditions of the hot events and the user preferences in real time, and dynamically adjusting the program arrangement sequence.
2. The method of claim 1, wherein the collecting hotspot event data in real-time from the social media platform and the news platform comprises: capturing hot word data from a social media platform in real time through a distributed web crawler program, and calling a news media API interface to obtain news events; And associating and extracting the hotword data and the news event data to generate structured hot event data, wherein the hot event data comprises time, place and a hot main body.
3. The method of claim 1, wherein the collecting the user's viewing behavior data at the television platform to obtain the viewing preference characteristics comprises: the method comprises the steps of obtaining the watching history, searching record and channel switching behavior of a user on a television platform, subdividing a user group by adopting a machine learning algorithm, constructing user portraits, and obtaining watching preference characteristics of different user groups.
4. The method of claim 2, wherein the performing natural language processing on the hotspot event data to obtain a hotspot keyword comprises: Based on preset scoring parameters, carrying out multi-level cutting on the hot spot main body of the hot spot event data to obtain core words; Combining a weighted fusion formula, and calculating to obtain a comprehensive score of each core word; and sequencing the core words according to the comprehensive scores, and selecting the first N core words as hot keywords.
5. The method of claim 4, wherein the hot subject includes hotword data and event text for a news event, the preset scoring parameters including a paragraph scoring threshold, a sentence scoring threshold, and a keyword scoring coefficient; the multi-level cutting is performed on the hot spot main body of the hot spot event data based on preset scoring parameters to obtain core words, including: text cleaning is carried out on the hot spot main body, and the cleaned hot word data are marked as first core words; performing paragraph cutting on the cleaned event text to obtain a plurality of paragraphs, analyzing the semantic relevance of each paragraph and the event text through a semantic vector model, generating paragraph scores, and determining core paragraphs by combining the paragraph score threshold; sentence segmentation is carried out on each core paragraph to obtain a plurality of sentences, the importance degree of each sentence in the corresponding paragraph is analyzed through an importance scoring model, sentence scores are generated, and the core sentences are determined by combining the sentence scoring threshold values; and performing vocabulary cutting on the core sentences to obtain a plurality of words, generating word scores according to the paragraph scores of the paragraphs where the words are located, the sentence importance scores of the sentences where the words are located and the comprehensive scores of the words, and determining a second core word by combining the keyword score coefficients.
6. The method of claim 1, wherein the program material database comprises a plurality of program materials, each of the program materials being provided with metadata information, the metadata information including program type, subject matter content, playout duration, production time, and copyright information; and performing similarity calculation on the hot spot keywords and metadata information of a program material database to generate a hot spot matching result, wherein the hot spot matching result comprises the following steps: performing semantic similarity calculation on the program types and the subject content of the hot key words and the metadata information to obtain preliminary program materials; Judging the emotion tendencies of the hot events, and adjusting the preliminary program materials by combining the emotion tendencies to obtain associated program materials; performing entity identification on the hot event data by adopting a BERT model to obtain a hot entity; And associating the associated program materials with the hot spot entities to construct an event-program association map, wherein one hot spot entity corresponds to one or more program materials in the event-program association map, and a hot spot matching result is generated.
7. The method of claim 1, wherein generating an initial population based on the viewing preference characteristics and the hotspot matching results and optimizing the initial population using a genetic algorithm to generate a program arrangement sequence comprises: generating an initial population based on the viewing preference characteristics and the hotspot matching results; In the initial population, sequentially encoding a plurality of program arrangement sequences into gene sequences, wherein each gene sequence comprises a program ID, a playing period and a playing priority; And constructing an adaptability function by taking user interest satisfaction, hot spot association and content diversity as optimization targets, and iteratively evolving the initial population through selection, intersection and mutation operations until a preset iteration condition is reached, so as to obtain an optimized program arrangement sequence.
8. A system for dynamic programming of a program, comprising: The data acquisition module is used for acquiring hot event data from the social media platform and the news platform in real time, and simultaneously acquiring watching behavior data of a user on the television platform to obtain watching preference characteristics; The semantic analysis module is used for carrying out natural language processing on the hot event data to obtain hot keywords, carrying out similarity calculation on the hot keywords and metadata information of the program material database, and generating a hot matching result; the genetic algorithm optimization module is used for generating an initial population based on the watching preference characteristics and the hot spot matching result, optimizing the initial population by adopting a genetic algorithm and generating a program arrangement sequence; and the dynamic adjustment module is used for monitoring the change conditions of the hot events and the user preferences in real time and dynamically adjusting the program arrangement sequence.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1 to 7 when executing the computer program.
10. A computer readable storage medium storing a computer program, characterized in that the computer program when executed by a processor implements the method according to any one of claims 1 to 7.

Description

Program dynamic arrangement method, system, equipment and medium Technical Field The application belongs to the technical field of computers, and particularly relates to a method, a system, equipment and a medium for dynamically arranging programs. Background With the rapid development of the Internet and digital media, audience has put forward higher requirements on the real-time property and diversity of television programs, traditional television program arrangement mainly depends on manual experience, the arrangement process is low in efficiency, and the real-time response of social hot events or the change of audience preference is difficult, and although the manual arrangement mode can be integrated into the professional experience of the compiling, the obvious defects of lag response, low individuation degree, insufficient diversity and the like exist when facing massive program contents and rapid changing media environments. In recent years, with the progress of artificial intelligence and big data technology, part of video platforms and intelligent television systems start to adopt recommendation algorithms to realize personalized pushing of program contents. Specifically, existing programming systems mainly employ a rule-based static programming scheme that generates program listings according to preset schedules, program types, and fixed priority rules, which generally lack response capabilities to real-time hotspots and user dynamic demands. (2) Based on a recommendation algorithm of historical behaviors of a user, the user past behavior data are analyzed to be matched with a personalized program list, and real-time adjustment and hot spot fusion of the program list still cannot be realized. (3) Program sequences are generated by applying an optimization algorithm based on the arrangement optimization of the algorithm, so that the matching degree of diversity and user interests is improved to a certain extent, but the algorithm is complex and poor in instantaneity. The conventional program content personalized calculation method has the advantages that personalized recommendation capability is improved to a certain extent, but the conventional program content personalized calculation method still has the defects that (1) real-time performance is insufficient, sudden hot spots in social media and news are difficult to capture, program update is lagged, semantic understanding limitation is achieved, deep understanding of hot event context and semantic association is lacking through depending on keyword matching, (3) a single arrangement strategy is achieved, requirements of masses and masses cannot be dynamically balanced, program content is easy to repeat, aesthetic fatigue of audiences is caused, (4) the arrangement rule and priority are still required to be set manually, automation and intelligence are limited, system generalization capability is poor, adaptability is weak when different field text or emerging program types are faced, processing efficiency is low, particularly, calculation resource consumption based on a deep learning model is large, real-time arrangement requirements are difficult to meet, and (7) the data dependency is strong, a large amount of label data is required by a supervision learning method, labor cost is high, and daily content is not suitable. Disclosure of Invention The embodiment of the application provides a method, a system, equipment and a medium for dynamically arranging programs, which can solve one of the problems in the prior art. In a first aspect, an embodiment of the present application provides a method for dynamically programming a program, including: Collecting hot event data from a social media platform and a news platform in real time, and collecting watching behavior data of a user on a television platform at the same time to obtain watching preference characteristics; performing natural language processing on the hot spot event data to obtain hot spot keywords, performing similarity calculation on the hot spot keywords and metadata information of a program material database, and generating a hot spot matching result; generating an initial population based on the viewing preference characteristics and the hot spot matching result, and optimizing the initial population by adopting a genetic algorithm to generate a program arrangement sequence; and monitoring the change conditions of the hot events and the user preferences in real time, and dynamically adjusting the program arrangement sequence. Further, the collecting, in real time, hot event data from the social media platform and the news platform includes: capturing hot word data from a social media platform in real time through a distributed web crawler program, and calling a news media API interface to obtain news events; And associating and extracting the hotword data and the news event data to generate structured hot event data, wherein the hot event data comprises time, place and a hot main