Media resource program information processing method and device, computer equipment and readable storage medium

Abstract

The invention discloses a media resource program information processing method and device, computer equipment and a readable storage medium, and the method comprises the steps: firstly carrying out the preprocessing of full-amount media resource information, building a label dictionary library, building a mapping table, and vectorizing and writing the preprocessed data into a vector library; user questions are obtained and subjected to safety examination, and an undetermined retrieval result is obtained in combination with the vector library after examination is passed. And after the result passes the security examination, returning information fields required by program playing to a user interface for display. By means of the design, the accuracy, safety and user experience of media asset retrieval are improved.

Description

Technical Field

[0001] The present invention relates to the field of media resource program processing, and in particular to a method, device, computer equipment and readable storage medium for processing media resource program information. Background Art

[0002] There are many problems with the existing film and television recommendation system, such as the inability to obtain the user's current viewing intention, the single type of recommended programs, the use of the same account by multiple users resulting in inaccurate user portraits, the media programs recommended by the large model do not match the official programs, and the performers are not subject to security checks. Summary of the invention

[0003] The object of the present invention is to provide a method, device, computer equipment and readable storage medium for processing media resource program information.

[0004] In a first aspect, an embodiment of the present invention provides a method for processing media asset program information, including:

[0005] Preprocess the full amount of media information to be retrieved, obtain preprocessed media data, and establish a related tag dictionary library;

[0006] Constructing a vector media asset mapping table, and vectorizing the preprocessed media asset data and writing it into a media asset vector library;

[0007] Obtaining a user question and conducting a security review, and if the user question passes the review, obtaining a pending media resource retrieval result based on the user question and the media resource vector library;

[0008] A security review is performed on the pending media resource search result, and if the pending media resource search result passes the review, the information field required for program playback corresponding to the pending media resource search result is returned to the user interface for display.

[0009] In a possible implementation manner, preprocessing the full amount of media asset information to be retrieved to obtain preprocessed media asset data and establishing a related tag dictionary library includes:

[0010] Clean and format all media asset information to obtain cleaned and formatted media asset data;

[0011] Expanding the media asset data after data cleaning and formatting to obtain media asset description information corresponding to each media asset data as the preprocessed media asset data;

[0012] The related tag dictionary library is established based on the director information and actor information included in the full amount of media asset information.

[0013] In a possible implementation manner, constructing a vector media asset mapping table and vectorizing the preprocessed media asset data and writing the vectorized data into a media asset vector library includes:

[0014] Construct a vector media asset mapping table of vector ID unique identifiers and media asset ID unique identifiers;

[0015] Vectorizing the media asset description information to obtain vector text;

[0016] Converting the media asset ID included in the preprocessed media asset data into a vector ID;

[0017] Extracting the media asset ID, program name, director, actor, and tag information of the preprocessed media asset data as vector metadata;

[0018] The vector text, the vector ID, and the vector metadata are written into a media asset vector library, and the mapping relationship between the media asset ID and the vector ID is written into the vector media asset mapping table.

[0019] In a possible implementation, the method further includes:

[0020] When the full amount of media asset information changes, the changed vector ID corresponding to the changed media asset ID is obtained through the vector media asset mapping table, and the corresponding data in the media asset vector library is updated or deleted.

[0021] In a possible implementation manner, the obtaining of the user question and performing a security review, and obtaining a pending media asset retrieval result based on the user question and the media asset vector library when the user question passes the review, includes:

[0022] Conduct a security review of the user issue;

[0023] If the user question fails to pass the review, a prompt message is returned;

[0024] If the user question passes the review, the user question is vectorized to obtain a question description vector;

[0025] According to the problem description vector, information retrieval is performed in the media resource vector library using an approximate retrieval algorithm to obtain undetermined vector metadata that meets a preset approximate value threshold;

[0026] Based on the relevant tag dictionary library, extract keywords from the user question to obtain pending keywords;

[0027] Filtering according to the pending keywords and the pending vector metadata to obtain candidate vector metadata;

[0028] Performing word frequency statistics on the elements in the candidate vector metadata, taking the element with the highest word frequency ratio in the candidate vector metadata as the public label of the candidate vector metadata, and performing secondary filtering using the public label to obtain a secondary filtering result;

[0029] The pending media resource retrieval result is obtained according to the user question and the secondary filtering result.

[0030] In a possible implementation manner, the performing a security review on the pending media resource search result includes:

[0031] Obtaining the answer description corresponding to the pending media resource retrieval result through the search engine interface corresponding to the pre-trained large model;

[0032] Input the answer description and the pending media resource retrieval result into a pre-trained large model for reasoning to obtain relevant media resource information;

[0033] For the programs included in the media asset information, a data security review is conducted using a vocabulary of artists with bad records and a third-party sensitive vocabulary.

[0034] In a possible implementation, returning the program playback required information field corresponding to the pending media resource search result to the user interface for display includes:

[0035] Perform data completion operations on the program, encapsulate the information fields required for program playback, and return them to the user interface for analysis and display.

[0036] In a second aspect, an embodiment of the present invention provides a media resource program information processing device, including:

[0037] The processing module is used to pre-process the full amount of media information to be retrieved, obtain pre-processed media data, and establish a related label dictionary library; construct a vector media mapping table, and vectorize the pre-processed media data and write it into the media vector library; obtain user questions and perform security review, and if the user questions pass the review, obtain the pending media retrieval results based on the user questions and the media vector library;

[0038] The output module is used to perform a security review on the pending media resource search result, and if the pending media resource search result passes the review, return the program playback required information field corresponding to the pending media resource search result to the user interface for display.

[0039] In a third aspect, an embodiment of the present invention provides a computer device, comprising a processor and a non-volatile memory storing computer instructions, wherein when the computer instructions are executed by the processor, the computer device executes the method described in at least one possible implementation of the first aspect.

[0040] In a fourth aspect, an embodiment of the present invention provides a readable storage medium, wherein the readable storage medium includes a computer program, and when the computer program is run, the computer device where the readable storage medium is located is controlled to execute the method described in at least one possible implementation manner of the first aspect.

[0041] Compared with the prior art, the beneficial effects provided by the present invention include: adopting a media resource program information processing method, device, computer equipment and readable storage medium disclosed in the present invention, preprocessing the full amount of media resource information and establishing a label dictionary library, and constructing a mapping table to vectorize the preprocessed data and write it into the vector library. Obtain user questions and conduct a security review. After the review is passed, the pending search results are obtained in combination with the vector library. Then conduct a security review of the results, and after passing, the information fields required for program playback are returned to the user interface for display. Such a design improves the accuracy, security and user experience of media resource retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be regarded as limiting the scope. For those of ordinary skill in the art, other related drawings can also be obtained based on these drawings without creative work.

[0043] Figure 1 A schematic diagram of the steps of a method for processing media resource program information provided by an embodiment of the present invention;

[0044] Figure 2 A schematic block diagram of the structure of a media resource program information processing device provided by an embodiment of the present invention;

[0045] Figure 3 A schematic block diagram of the structure of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0047] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0048] In order to solve the technical problems in the aforementioned background technology, Figure 1This is a flow chart of a method for processing media asset program information provided by an embodiment of the present disclosure. The method for processing media asset program information is introduced in detail below.

[0049] Step S201, preprocessing the full amount of media asset information to be retrieved, obtaining preprocessed media asset data, and establishing a related tag dictionary library;

[0050] Step S202, constructing a vector media asset mapping table, and vectorizing the pre-processed media asset data and writing it into a media asset vector library;

[0051] Step S203, obtaining the user question and performing a security review, and if the user question passes the review, obtaining a pending media resource retrieval result based on the user question and the media resource vector library;

[0052] Step S204: perform a security review on the pending media resource search result. If the pending media resource search result passes the review, the program playback required information field corresponding to the pending media resource search result is returned to the user interface for display.

[0053] In the embodiment of the present invention, for example, it is assumed that the server receives full media asset information from multiple film and television platforms, and the formats and contents of this information may be different, and there may be data missing, errors or inconsistencies.

[0054] First, the server cleans the data of the full media asset information. For example, it handles the garbled characters, duplicate records, or missing key fields (such as program name, director, etc.) in some media asset data. Assuming that there is a media asset record with garbled characters in the program name field, the server will mark it as invalid data and remove it. For duplicate media asset records, only one complete and accurate record is retained.

[0055] Next, formatting is performed. The media information from different sources is unified into a unified format, for example, the date format is unified into "year-month-day", and the price format is unified into two decimal places. Assume that the original price format of some media information is "￥100" and some is "100 yuan". After formatting, they are all unified into "100.00 yuan".

[0056] Then, the cleaned and formatted media data is expanded. For example, the media data of a movie originally only contains the movie name, director and starring actors. The server obtains more information from other databases and supplements it with the movie's plot summary, release year, awards, etc., thus completing it into a rich media description information.

[0057] At the same time, based on the director information and actor information in the full amount of media information, a related label dictionary is established. Assume that common directors and actors are extracted from a large amount of media data, and labels are established for them, such as "well-known director" and "action movie actor", etc., to facilitate subsequent retrieval and processing.

[0058] The server starts to build a vector media mapping table. A unique vector ID and corresponding media ID are generated for each piece of preprocessed media data. Assuming the media data with media ID 1001, the generated vector ID is 2001 (media ID and vector ID are generally a dozen random characters in random order as unique identifiers, which is only used as an example here. The actual ID generation has a corresponding algorithm).

[0059] Next, a specialized Chinese word vectorization tool or model is used to vectorize the media asset description information to obtain vector text. For example, for a media asset description such as "This is a wonderful action movie, directed by the famous director Zhang XX, starring Li XX and Zhang XX", after vectorization conversion, a vector representation of a specific length is obtained.

[0060] Then, the media asset ID in the preprocessed media asset data is converted into a vector ID. For example, the data with the media asset ID of 1002 corresponds to a vector ID of 2002.

[0061] After that, the key information of the pre-processed media asset data is extracted as vector metadata, such as media asset ID, program name, director, actors, tags, etc. Suppose the vector metadata of a TV series includes media asset ID 1003, program name "Lang XX", director Kong XX, actors Hu XX, Liu XX, etc., and the tag is "costume drama".

[0062] Finally, the vector text, vector ID, and vector metadata are written into the media asset vector library, and the mapping relationship between the media asset ID and the vector ID is written into the vector media asset mapping table. In this way, in the subsequent retrieval process, the corresponding media asset data can be quickly found through the vector ID.

[0063] The server receives questions raised by the user through voice input or text input, such as "I want to watch a comedy movie directed by a famous director."

[0064] First, the user's question is reviewed for security. The server uses the bad artist word library, third-party sensitive word library, etc. to check whether the user's question contains sensitive or illegal content. If the user's question mentions the name of an artist listed as a bad artist, or contains sensitive words that violate laws and regulations, the review will not pass, and the server will return a prompt message to inform the user that the question does not meet the specifications and asks the user to re-enter the question.

[0065] Assuming that the user's question is approved, the server will convert it into a vector. Using the same Chinese word vector tool or model used to process media asset description information, the user's question "I want to watch a comedy movie directed by a famous director" is converted into a vector representation to obtain the question description vector.

[0066] Then, the approximate retrieval algorithm is used to retrieve information in the media vector library based on the question description vector. The server calculates the similarity between the question description vector and the vector text in the media vector library, and finds the pending vector metadata that meets the preset approximate value threshold. For example, the vector metadata of several comedy movies whose directors are well-known is found.

[0067] Next, based on the previously established related tag dictionary library, keywords are extracted from the user's question. Keywords such as "comedy movie" and "famous director" are extracted from the user's question as pending keywords.

[0068] Filtering is performed based on pending keywords and pending vector metadata. The server removes media asset data that does not match the keywords and obtains candidate vector metadata that better meets user needs.

[0069] Perform word frequency statistics on the elements in the candidate vector metadata. Assuming that the candidate vector metadata contains multiple movies, count the frequency of directors, actors, tags and other elements. If the frequency of "Zhou XX" as a director exceeds half, use it as the public tag of the candidate vector metadata.

[0070] Finally, the user question is combined with the public label input into the pre-trained big model for inference. For example, if the public label is Zhou XX, the recommended result may be comedy, but the director and actor that do not include Zhou XX will be eliminated. This is to solve the problem that the results of the big model are sometimes too divergent or the correlation between the recommended results is low. The filtered results, user questions, and the information returned by the search engine are submitted to the big model as input. By enriching the context information of the big model reasoning and improving the accuracy of the information, the rationality and accuracy of the final reasoning are guaranteed.

[0071] After the server obtains the pending media resource retrieval results, it first obtains the answer descriptions corresponding to these results through the search engine interface corresponding to the pre-trained large model. For example, it obtains detailed introductions and comments about recommended movies.

[0072] Extract relevant media information from the answer description, including the movie's name, starring actors, director, plot summary, etc.

[0073] Then, the programs included in the media asset information are reviewed for data security using the bad artist word library and the third-party sensitive word library. If, during the review process, it is found that the lead actor of a movie is listed as a bad artist, the server will remove the movie from the search results.

[0074] If all pending media resource retrieval results have passed the security review, the server will perform data enrichment operations on them, for example, adding movie playback links, ratings, duration and other information.

[0075] Finally, the information fields required for program playback, such as movie name, playback link, introduction, rating, etc., are encapsulated and returned to the user interface for parsing and display. Users can see movie recommendations that meet their needs and are safe and compliant on the interface, and watch the movie by clicking the playback link.

[0076] Suppose the user enters a new question "I want to watch a romantic movie starring Yang X", the server will process it in the same way as above. First, a security review is conducted. If it passes, the question is vectorized, searched in the media vector library, keywords are extracted, filtered and inferred, and the pending media search results are obtained. Then a security review is conducted, the data is improved, and the information fields are encapsulated and returned to the user interface for display.

[0077] Through such a complete set of media program information processing methods, the server can efficiently and accurately provide users with media program recommendations that meet their needs and are safe and compliant, thereby improving users' viewing experience.

[0078] Assume that in the process of processing a large number of user requests, the server continuously optimizes and updates the bad artist word library and the third-party sensitive word library, and adjusts and improves the pre-trained large model to improve the accuracy and security of recommendations. At the same time, as new media data is continuously added, the server can also pre-process and store this data in a timely manner to ensure the timeliness and integrity of the media vector library.

[0079] Assume that during a holiday, the user's demand for media programs increases significantly. The server can quickly respond to user requests with its powerful computing power and optimized processing flow, and provide users with satisfactory media program recommendations. Even in the case of high concurrency, the system can maintain stable operation without freezes or delays.

[0080] In addition, the server can further optimize the recommendation algorithm and model based on user feedback information, such as user clicks, viewing time, and evaluation of the recommended results, so that subsequent recommendations are more in line with the user's real needs and preferences.

[0081] In summary, the above detailed scenario examples demonstrate the entire process of the server executing the media program information processing method, including data preprocessing, vectorization and storage, user question processing and retrieval, result review and display, etc., which can provide users with high-quality, secure and personalized media program services.

[0082] In the embodiment of the present invention, the aforementioned step S201 can be implemented through the following example execution.

[0083] Clean and format all media asset information to obtain cleaned and formatted media asset data;

[0084] Expanding the media asset data after data cleaning and formatting to obtain media asset description information corresponding to each media asset data as the preprocessed media asset data;

[0085] The related tag dictionary library is established based on the director information and actor information included in the full amount of media asset information.

[0086] In the embodiment of the present invention, illustratively, first, the server receives the full amount of media asset information to be retrieved from various channels. Such information comes from a wide range of sources, has various formats, and has varying quality.

[0087] The server starts to clean and format the entire media information. For example, in the received media information, some data field values ​​are not standardized. For example, the program broadcast time is recorded as "2023.5.1", "2023 / 5 / 1", and "May 1, 2023". The server will convert these broadcast times in different formats into the standard format of "2023-05-01".

[0088] For another example, the server processes empty values ​​or abnormal values ​​in some media data. If the episode number field of a TV series is empty, the server will try to supplement the information from other reliable data sources or mark it as pending verification. If a movie score is found to be an obviously abnormal value such as "120 points", the server will treat it as invalid data and process it accordingly.

[0089] After data cleaning and formatting, the server obtains standardized and accurate media asset data after data cleaning and formatting.

[0090] Next, the server expands the information of these cleaned and formatted media data. Assuming that for a TV series "Ideal XX", the initial media data may only contain the basic name, director, starring and a brief plot summary. The server obtains more detailed information about the show through web crawlers and calling other database interfaces, such as the filming location, behind-the-scenes footage, awards, etc. After expansion, the media description information of the show may become: "Ideal XX is an urban workplace drama directed by Liu XX and starring Sun XX and Zhao XX. The show was filmed in a certain place and tells the story of the growth of cost engineer Su Xiao. The plot is ups and downs. The show has won the XX award, and there are many interesting behind-the-scenes stories during the filming process." Such rich description information is used as the pre-processed media data.

[0091] At the same time, the server builds a related tag dictionary based on the director information and actor information in the full media information. For example, from a large amount of media data, the server extracts a series of director names, such as Zheng XX, Guo XX, Kong XX, etc., and labels them as "well-known TV drama director". For actors, such as Huang X, Shen X, Wu X, etc., they are labeled "comedian" and "action actor" according to their performance style and the types of roles they are good at.

[0092] Suppose during the processing, the server finds that a movie is directed by Zhang XX, and starring Gong X and Chen XX. The server will record label information such as "Zhang XX-well-known movie director", "Gong X-acting actor", "Chen XX-senior actor" in the label dictionary. In this way, a comprehensive and accurate relevant label dictionary is established, providing strong support for subsequent media resource retrieval and recommendation.

[0093] In another scenario, the media information received by the server includes a documentary "Beautiful XX". The initial data only contains a brief introduction and a list of the main creators. After expanding the information, the server learns that the documentary has a unique shooting technique, uses advanced photography technology, and has won multiple awards internationally. At the same time, when building the tag dictionary library, the director of the film, Li Chengcai, is marked as a "famous documentary director."

[0094] In short, the server has laid a solid foundation for subsequent media program information processing through a series of processing of the entire media information, including data cleaning and formatting, information expansion, and establishment of a tag dictionary library. It can more accurately meet user needs and provide high-quality media services.

[0095] In the embodiment of the present invention, the aforementioned step S202 may be implemented through the following example execution.

[0096] Construct a vector media asset mapping table of vector ID unique identifiers and media asset ID unique identifiers;

[0097] Vectorizing the media asset description information to obtain vector text;

[0098] Converting the media asset ID included in the preprocessed media asset data into a vector ID;

[0099] Extracting the media asset ID, program name, director, actor, and tag information of the preprocessed media asset data as vector metadata;

[0100] The vector text, the vector ID, and the vector metadata are written into a media asset vector library, and the mapping relationship between the media asset ID and the vector ID is written into the vector media asset mapping table.

[0101] In the embodiment of the present invention, the server first constructs a vector media asset mapping table of vector ID unique identifiers and media asset ID unique identifiers. For example, the server receives a series of pre-processed media asset data, the media asset ID of the first media asset is "001", and generates a corresponding vector ID "1001" for it, and records the corresponding relationship "001-1001" in the mapping table.

[0102] Next, the server vectorizes the media asset description information to obtain vector text. Suppose the description information of a media asset is "This romantic movie "Love XX" is directed by Ni X, starring Rui X and Rui XX. It tells the story of a couple who still stick to their love after going through all kinds of hardships and is deeply loved by the audience." The server uses special vectorization tools and algorithms to convert this description into a vector text with a fixed dimension, such as a vector with a length of 512 dimensions.

[0103] Then, the server converts the media asset ID included in the preprocessed media asset data into a vector ID. For the media asset data with the media asset ID "002", a corresponding vector ID "1002" is generated through a specific conversion rule or algorithm.

[0104] After that, the server extracts the key information of the pre-processed media asset data as vector metadata. Taking a TV series as an example, the server extracts the media asset ID "003", the program name "Lang XX: The Wind Rises in Changlin", the director "Kong XX Li X", the actors "Huang X Liu X Tong X", and the label "ancient costume political drama" from the pre-processed media asset data. These information together constitute the vector metadata of the media asset data.

[0105] Next, the server writes the vector text, vector ID, and vector metadata into the media asset vector library. For example, for the media asset data mentioned above, the server stores the generated vector text, vector ID "1003" and the extracted vector metadata into the media asset vector library.

[0106] At the same time, the server writes the mapping relationship between the media asset ID and the vector ID into the vector media asset mapping table. For example, the corresponding relationship of "003-1003" is recorded in the mapping table (it should be understood that the mapping relationship between the media asset ID and the vector ID is generated because the ID generation rules of the two systems are not exactly the same, and the generated mapping relationship is a one-to-one relationship between key-value pairs, which is simplified to numbers here for example) so that the corresponding vector ID can be quickly found through the media asset ID in the future, thereby quickly obtaining relevant media asset information.

[0107] When processing the media data of another movie, its media ID is "004" and the description information is "This action movie "Fast and Furious" is directed by Lin XX, starring Fan XX, Yue XX, etc. The movie is full of thrilling racing scenes and exciting fighting plots." The server processes it according to the same process, generates a vector ID "1004", converts the description information into vector text, extracts the media ID "004", program name "Fast and Furious", director "Lin XX", actor "Fan XX Yue XX", and label "action adventure movie" as vector metadata, writes the vector text, vector ID, and vector metadata into the media vector library, and records the mapping relationship of "004-1004" in the vector media mapping table.

[0108] For a large amount of media data, the server repeats the above process continuously to ensure that each piece of media data can be accurately vectorized and stored. For example, when processing a documentary, the server also strictly executes various operations to fully and accurately include it in the media vector library and vector media mapping table.

[0109] In actual operation, the server may process tens of thousands of media data at the same time. To improve efficiency, the server may use distributed computing, parallel processing and other technical means. In addition, to ensure the accuracy and integrity of the data, the server will also perform real-time verification and error correction during the data writing process.

[0110] Through such a rigorous and efficient processing flow, the server can build a rich, accurate and easy-to-search media vector library and vector media mapping table, providing a solid foundation for subsequent user problem processing and media retrieval.

[0111] In the embodiments of the present invention, the following implementation modes are also provided.

[0112] When the full amount of media asset information changes, the changed vector ID corresponding to the changed media asset ID is obtained through the vector media asset mapping table, and the corresponding data in the media asset vector library is updated or deleted.

[0113] In the embodiment of the present invention, for example, in daily media asset management, the full amount of media asset information often changes, and the server keeps monitoring and responding to these changes at all times.

[0114] For example, a certain film and television platform decided to remove a TV series "XX Legend". When the server receives this change information, it will first look up the media ID corresponding to the TV series in the previously constructed vector media mapping table. Assuming that its media ID is "005", the server quickly obtains the corresponding vector ID "1005" through the mapping relationship.

[0115] Then, based on the vector ID obtained, the server locates all data related to the TV series in the media vector library, including its vector text, such as the vectorized representation of the plot and cast of the series; extracted vector metadata, such as the media ID "005", the program name "XX Legend", director, actor and other information; and other related associated data.

[0116] After confirming the data location, the server performs a deletion operation to completely clear all relevant information of the TV series in the media vector library to ensure that users will no longer obtain erroneous information about programs that have been removed from the shelves when searching.

[0117] For example, the starring information of another TV series "YY Fengyun" has changed. The original starring actors were "Zhang San and Li Si", and now they are updated to "Zhang San and Wang Wu". When the server receives this change notification, it also finds the corresponding media asset ID of the drama through the vector media asset mapping table, assuming it is "006" and the vector ID "1006".

[0118] Next, the server locates the relevant data in the media vector library. Then, the vector metadata part related to the starring information is updated, and the original "actors: Zhang San, Li Si" is changed to "actors: Zhang San, Wang Wu". At the same time, the relevant tags or categories may be appropriately adjusted according to the updated starring information.

[0119] For example, the plot summary of a movie "Operation ZZ" was optimized and supplemented. After the server obtains this change, it finds the corresponding media ID and vector ID according to the same process, and then updates the vector text and related description information of the movie in the media vector library to make it more accurately reflect the content of the movie.

[0120] There is another situation, if a variety show has added several episodes. When the server processes this change, in addition to updating the relevant description and metadata of the program, it will also vectorize the relevant information of the new content and add it to the corresponding position in the media asset vector library.

[0121] When processing a large amount of media asset information changes, the server will use efficient database operation technology and algorithms to ensure the accuracy and timeliness of the change operation. At the same time, in order to prevent data loss or errors during the change process, the server will back up the data before the operation and perform data verification and consistency checks after the operation is completed.

[0122] Assume that in one day, the server receives dozens of media resource information change requests, including movies, TV series, variety shows and other types of programs. The server methodically processes each change request one by one according to the established process. For some urgent changes, such as the deletion of sensitive content or the update of important information, the server will give priority to processing to ensure user experience and information accuracy.

[0123] Through such a rigorous and efficient processing mechanism, the server can respond to various changes in the entire media information in a timely and accurate manner, always maintain the freshness and accuracy of the data in the media vector library, and provide users with reliable media retrieval services.

[0124] In the embodiment of the present invention, the aforementioned step S203 can be implemented through the following example execution.

[0125] Conduct a security review of the user issue;

[0126] If the user question fails to pass the review, a prompt message is returned;

[0127] If the user question passes the review, the user question is vectorized to obtain a question description vector;

[0128] According to the problem description vector, information retrieval is performed in the media resource vector library using an approximate retrieval algorithm to obtain undetermined vector metadata that meets a preset approximate value threshold;

[0129] Based on the relevant tag dictionary library, extract keywords from the user question to obtain pending keywords;

[0130] Filtering according to the pending keywords and the pending vector metadata to obtain candidate vector metadata;

[0131] Performing word frequency statistics on the elements in the candidate vector metadata, taking the element with the highest word frequency ratio in the candidate vector metadata as the public label of the candidate vector metadata, and performing secondary filtering using the public label to obtain a secondary filtering result;

[0132] The pending media resource retrieval result is obtained according to the user question and the secondary filtering result.

[0133] In the embodiment of the present invention, exemplarily, in the embodiment of the present invention, exemplarily, when a user asks a question to the server, the server first performs a security review on the user's question. Assuming that the question entered by the user is "I want to watch a movie involving a horror plot", the server will compare this question with the pre-set security rules and sensitive word library. Since content such as "horror plot" may not meet security requirements, this user question review fails. The server will return a prompt message, such as "Your question does not meet the content specifications, please re-enter", to guide the user to enter a compliant question.

[0134] For example, another user inputs "I want to watch a science fiction movie starring Wu X". After a security review, the server confirms that this question does not violate the rules and passes the review. Then, the server vectorizes this user question and converts "I want to watch a science fiction movie starring Wu X" into a specific vector representation to obtain the question description vector.

[0135] Next, the server uses an approximate retrieval algorithm to retrieve information in the media vector library based on the question description vector. Assuming that the preset approximate value threshold is 0.8, the server finds the undetermined vector metadata in the media vector library with a similarity higher than 0.8 with the question description vector. These metadata may correspond to multiple science fiction movies, and the starring actors include Wu X.

[0136] Afterwards, the server extracts keywords from the user's question based on the relevant tag dictionary library, and extracts "science fiction film" and "Wu X" from "I want to watch a science fiction film starring Wu X" as pending keywords.

[0137] Then, the server filters the pending vector metadata according to these pending keywords. For example, it removes metadata of some science fiction movies whose starring actors are not Wu X, and obtains candidate vector metadata.

[0138] Next, the server counts the word frequency of the elements in the candidate vector metadata. Assuming that the candidate vector metadata includes several movies, namely "Wandering XX", "Galaxy XX" and "Star XX", the server counts the frequency of occurrence of elements such as directors, actors, and tags. After statistics, it was found that "Liu XX" as the original author appeared frequently, and "Space Adventure" as a tag also appeared frequently. Therefore, the server uses "Liu XX" and "Space Adventure" as public tags for the candidate vector metadata, and uses these two public tags for secondary filtering. For example, some metadata with low correlation with "Liu XX" and "Space Adventure" are eliminated to obtain the secondary filtering results.

[0139] Finally, based on the user's question "I want to watch a science fiction movie starring Wu X" and the secondary filtering results, the server recommends the user a pending media resource retrieval result that better meets his needs. The result may be a science fiction movie adapted from the work of Liu XX, starring Wu X, and with space adventure elements, such as "Wandering XX 2".

[0140] For another example, a user inputs "I want to watch a historical drama directed by Zheng XX". After the server passes the security review, it is vectorized and converted into a question description vector, and the pending vector metadata is retrieved from the media resource vector library. The keywords "ancient costume", "history", and "Zheng XX" are extracted, and the candidate vector metadata is filtered. After the word frequency is counted, it is found that the word "Zhen XX" has a high frequency, and it is used as a public label and filtered twice. The user question and the secondary filtering results are combined and input into the large model, and the pending media resource retrieval results are finally obtained, such as recommending the historical drama "Mi XX" directed by Zheng XX.

[0141] In actual operation, the server may receive a large number of different questions from users in a short period of time. For example, during prime time in the evening, the server received hundreds of question requests from users at the same time. For each user's question, the server strictly follows the above process to ensure that accurate and appropriate pending media resource retrieval results are provided to each user.

[0142] Assume that another user inputs "I want to watch an educational documentary starring Zhong XX", the server also completes the above steps in sequence. After security review, vectorization conversion, approximate search, keyword extraction, filtering, word frequency statistics, secondary filtering using public tags, and combining user questions with secondary filtering results, the server finally recommends pending media resource retrieval results that meet the user's needs, which may be documentaries about Zhong XX and educational, such as "Zhong XX: Guardian of Life".

[0143] Through such a sophisticated and efficient processing flow, the server can more accurately understand the needs of users, filter out the media information that best meets user expectations from the rich media library, and provide users with a better service experience.

[0144] In the embodiment of the present invention, the aforementioned step S204 may be implemented through the following example execution.

[0145] Obtaining the answer description corresponding to the pending media resource retrieval result through the search engine interface corresponding to the pre-trained large model;

[0146] Input the answer description and the pending media resource retrieval result into a pre-trained large model for reasoning to obtain relevant media resource information;

[0147] For the programs included in the media asset information, a data security review is conducted using a vocabulary of artists with bad records and a third-party sensitive vocabulary.

[0148] In the embodiment of the present invention, exemplarily, after the server obtains the pending media resource retrieval result, it obtains the answer description corresponding to the pending media resource retrieval result through the search engine interface corresponding to the pre-trained large model.

[0149] Assume that the pending media resource search results recommend several film and television works, such as "Movie A", "TV Series B" and "Variety Show C". The server obtains the answer description about "Movie A" through the search engine interface, such as "This is a movie directed by a well-known director with an adventure theme and a thrilling plot"; for "TV Series B", the description is obtained as "A drama with family life as the background, showing family affection and growth"; for "Variety Show C", the description is obtained as "A star competition variety show, full of challenges and fun".

[0150] The server then inputs these answer descriptions and pending media asset retrieval results into the pre-trained large model for inference. After analysis and processing, the large model derives more detailed and accurate relevant media asset information from these input information.

[0151] Taking "Movie A" as an example, the big model may infer key information such as the movie's name, director, list of leading actors, specific adventure plot elements, etc. For "TV Series B", it may infer the name of the TV series, the total number of episodes, the personality traits of the main characters and key plot developments, etc.; for "Variety Show C", it may infer the program name, detailed introduction of the host, resident guests, main game session settings and highlights, etc.

[0152] Next, the server conducts a data security review on the programs included in the inferred media information using the bad artist vocabulary and third-party sensitive vocabulary.

[0153] The server compares the extracted director, actor and other information with the bad artist word library. Assume that the bad artist word library contains the actor "Zhang San", and in the inferred media information, if one of the leading actors of "TV Series B" is "Zhang San", then the server will determine that the TV series does not meet the security review requirements.

[0154] At the same time, the server will also compare the program name, plot keywords, etc. with the third-party sensitive word library. For example, the third-party sensitive word library contains the word "terrorism". If it is inferred that the plot description of "Movie A" contains similar sensitive content, the server will also mark it as not meeting the security review requirements.

[0155] For example, the server recommended a newly released movie "Movie D" in the pending media resource search results. The answer about "Movie D" obtained through the search engine interface is described as "a romantic love movie with beautiful pictures". After inputting it together with the pending media resource search results into the big model, it is inferred that the director is "Li Si", the starring actors are "Wang Wu" and "Zhao Liu", and detailed love plot clues.

[0156] The server compares this information with the bad artist word library and the third-party sensitive word library. It is confirmed that "Li Si", "Wang Wu" and "Zhao Liu" are not on the bad artist list, and the movie's name and plot keywords do not match the sensitive content in the third-party sensitive word library. Therefore, it is determined that "Movie D" has passed the data security review.

[0157] For another example, for a popular variety show "Variety E", the server obtains the answer description as "an outdoor reality show where celebrities challenge various tasks", and inputs the large model to infer that the host is "Sun Qi" and the regular guests are "Zhou Ba" and "Wu Jiu", etc. During the security review, if "Zhou Ba" is found in the bad artist dictionary, the server will mark "Variety E" as failing the security review and will not provide it to users as the final recommendation result.

[0158] In actual operation, the server may process security reviews of multiple pending media resource retrieval results at the same time. For example, during a holiday, the server receives a large number of pending media resource retrieval results in a short period of time and needs to conduct security reviews on dozens of film and television works and variety shows.

[0159] The server will perform each step efficiently and accurately to ensure that every media program recommended to users meets the requirements of safety and compliance. Even in the face of complex situations, such as multiple media programs involving many people and rich plot content, the server can complete the review work in an orderly manner to ensure that users obtain healthy, positive and legal media information.

[0160] Through such a strict and comprehensive data security review process, the server can effectively screen out suitable media resources and programs, provide users with a safe and reliable entertainment environment, and enhance user experience and satisfaction.

[0161] In the embodiment of the present invention, the aforementioned step S204 may be implemented through the following example execution.

[0162] Perform data completion operations on the program, encapsulate the information fields required for program playback, and return them to the user interface for analysis and display. In an embodiment of the present invention, illustratively, after the server completes the security review of the pending media resource retrieval results, the information fields required for program playback that meet the requirements are returned to the user interface for display. Assume that the pending media resource retrieval results that have passed the security review include a TV series "Dream XX" and a movie "Star XX".

[0163] The server first performs data enrichment operations on "Dream XX". It will obtain more detailed information about the show, such as the specific duration of each episode, the total number of episodes, the first broadcast time, the update frequency, etc. At the same time, it will also supplement the show's ratings, audience reviews and other related data. As for the broadcast channel, the server will confirm which platforms the show can be watched, such as XXapp or Youku, etc. In addition, the server will also obtain multimedia resources such as high-definition poster images of the show and video clips of highlights.

[0164] After completing the data, the server will package the information. It will organize and combine information such as program name, broadcast platform, episode length, total number of episodes, premiere time, update frequency, ratings, audience reviews, poster images, video clips, etc. according to the pre-set format and field order.

[0165] Similarly, for the movie "Xingchen XX", the server obtains information such as its length, release year, director's description, and interviews with the leading actors. In terms of playback channels, it confirms its schedule in major theaters or playback rights on online video platforms. In addition, it obtains the movie's trailer, theme song, and other related resources.

[0166] After completing the data collection, the server packages the data and integrates information such as the movie name, length, release year, director's explanation, starring interviews, broadcast channels, trailers, theme songs, etc.

[0167] After the encapsulation is completed, the server returns the encapsulated data to the user interface for parsing and display.

[0168] On the user interface, for the TV series "Dream XX", users will see the clearly displayed program name "Dream XX", with a high-definition poster picture next to it to attract the user's attention. Then you will see a detailed description of the show, such as "Total number of episodes: 40 episodes, duration of each episode: 45 minutes, premiere time: May 10, 2023, update frequency: two episodes updated every Tuesday and Wednesday". At the same time, you can also see information such as "Rating: 8.5 points, audience evaluation: wonderful plot, excellent acting skills of the actors". In the playback channel section, it is clearly displayed that "XXapp can be watched". In addition, video clips of highlights will be provided on the user interface, allowing users to experience the highlights of the show in advance.

[0169] For the movie "Xingchen XX", the user interface will display the title "Xingchen XX" and a beautiful movie poster. Then there is a detailed introduction, such as "Length: 120 minutes, Release Year: 2023, Director's Explanation: This is a sci-fi adventure about exploring the unknown". In terms of playback channels, it shows "Major theaters are showing it now, and XXapp will be online in 1 month". At the same time, users can see wonderful trailers and hear moving theme songs.

[0170] For example, there is a variety show "Happy XX" that has passed the review. The server will obtain information such as the guest lineup, theme of each episode, recording location, broadcast time, etc. In terms of the broadcast channel, it is clear which channel of the TV station and the specific link of the network platform. At the same time, it obtains pictures of the wonderful moments and funny clips of the program.

[0171] After the server completes the data improvement and packaging, users can see the eye-catching name "Happy XX" on the interface, and next to it is a poster of the guests' happy group photo. Then there is a detailed introduction, such as "Guest lineup: Zhang San, Li Si, Wang Wu, etc., each theme: different fun challenges, broadcast time: every Saturday night at 8 o'clock." The broadcast channel shows "XX TV station, XXTV can be watched", and there are also pictures of wonderful moments and funny clips for users to click to watch.

[0172] Through such a comprehensive and detailed process of data improvement, packaging and return display, the server can provide users with rich, accurate and attractive media program information, allowing users to easily and quickly understand and select programs of their interest.

[0173] In order to more clearly describe the solution provided by the embodiment of the present application, a more detailed implementation method is provided below.

[0174] 1. Clean and format the data of all media resources to be retrieved, and establish a corresponding dictionary library of director, actor and other dimension labels.

[0175] 2. Expand the media asset data cleaned and formatted in step 1, and complete one media asset data into a media asset description statement. This step uses scripts or tools to batch generate description information for all media assets.

[0176] 3. Construct a vector ID unique identifier automatic generation module and construct a vector ID unique identifier and a media asset ID unique identifier mapping table.

[0177] 4. Use the Chinese word vector tool or model to vectorize the media asset description information generated in step 2 to generate vector text, convert the media asset ID into a vector ID through the vector ID automatic generation module, extract the media asset ID, program name, director, actor, label and other information as vector metadata, write the vector text, vector ID and vector metadata into the media asset vector library; and write the mapping relationship between the media asset ID and the vector ID into the mapping table.

[0178] 5. When the media asset information changes, the vector ID corresponding to the media asset ID is obtained through the ID mapping table generated in step 3, and the data in the media asset vector library generated in step 4 is updated or deleted.

[0179] 6. Conduct security review of user questions (using the vocabulary of bad artists, third-party sensitive vocabulary, etc.). If the security review fails, a prompt message will be returned.

[0180] 7. Use the same Chinese word vector tool or model to vectorize the user's problem description, then use the approximate retrieval algorithm of the vector library to perform information retrieval and return the retrieval results within the threshold setting.

[0181] 8. Use the custom word library tool to load the dimension label dictionary library of directors, actors, etc. generated in step 1, and extract keywords for user questions. If keywords are extracted, filter the media information retrieved from the vector library to eliminate irrelevant ones.

[0182] 9. Calculate the word frequency of the director, actor, label and other dimensions in the vector metadata of all programs in the result set. When the number of a certain type of label accounts for more than half of the programs in the entire result set, it is determined that the label may be a public label for the user's specific question results, and the public labels of all major types are calculated in the same way.

[0183] 10. By filtering the result set with the public tags generated in step 9, you can get a certain type of program or a specific type of program. If the user does not involve a specific type, public tags will generally not be generated, so the algorithm will only correct specific types of questions and will not affect the answer results of other user questions.

[0184] 11. Obtain the answer description corresponding to the user's question through the search engine interface, and extract relevant media information from it.

[0185] 12. Encapsulate the media information and user questions returned from step 10 and step 11 and return them to the big model for knowledge reasoning, and return the media information related to the user question.

[0186] 13. Perform a data security review on the media resource information returned in step 12 (using the bad artist word library, third-party sensitive word library, etc.) to filter out programs related to bad artists. If all programs fail the security review, a prompt message is returned.

[0187] 14. Complete the media resource information after security review, and finally encapsulate all the information fields required for program playback and return them to the user interface for analysis and display.

[0188] Please refer to Figure 2 , Figure 2 A media resource program information processing device 110 provided in an embodiment of the present invention includes:

[0189] The processing module 1101 is used to pre-process the full amount of media asset information to be retrieved, obtain pre-processed media asset data, and establish a related label dictionary library; construct a vector media asset mapping table, and vectorize the pre-processed media asset data and write it into the media asset vector library; obtain user questions and perform security review, and if the user questions pass the review, obtain the pending media asset retrieval results based on the user questions and the media asset vector library;

[0190] The output module 1102 is used to perform a security review on the pending media resource search result, and if the pending media resource search result passes the review, return the program playback required information field corresponding to the pending media resource search result to the user interface for display.

[0191] It should be noted that the implementation principle of the aforementioned media resource program information processing device 110 can refer to the implementation principle of the aforementioned media resource program information processing method, and will not be repeated here. It should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into a physical entity, or physically separated. And these modules can all be implemented in the form of software calling through processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software calling through processing elements, and some modules can be implemented in the form of hardware. For example, the media resource program information processing device 110 can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above media resource program information processing device 110. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element described here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each of the above modules can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0192] For example, the above modules may be one or more integrated circuits configured to implement the above methods, such as one or more application specific integrated circuits (ASIC), or one or more microprocessors (digital signal processors, DSP), or one or more field programmable gate arrays (FPGA), etc. For another example, when a module above is implemented in the form of a processing element scheduling program code, the processing element may be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call program code. For another example, these modules may be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0193] The embodiment of the present invention provides a computer device 100, which includes a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the aforementioned media resource program information processing device 110. Figure 3 As shown, Figure 3 The computer device 100 provided in the embodiment of the present invention is a structural block diagram. The computer device 100 includes a media resource program information processing device 110, a memory 111, a processor 112 and a communication unit 113.

[0194] In order to realize data transmission or interaction, the memory 111, the processor 112 and the communication unit 113 are electrically connected to each other directly or indirectly. For example, the electrical connection between these components can be realized through one or more communication buses or signal lines. The media resource program information processing device 110 includes at least one software function module that can be stored in the memory 111 in the form of software or firmware or fixed in the operating system (OS) of the computer device 100. The processor 112 is used to execute the media resource program information processing device 110 stored in the memory 111, such as the software function modules and computer programs included in the media resource program information processing device 110.

[0195] An embodiment of the present invention provides a readable storage medium, which includes a computer program. When the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the aforementioned media resource program information processing method.

[0196] For illustrative purposes, the foregoing description is made with reference to specific embodiments. However, the above illustrative discussion is not intended to be exhaustive or to limit the present disclosure to the precise form disclosed. Numerous modifications and variations are possible in accordance with the above teachings. These embodiments are selected and described in order to best illustrate the principles of the present disclosure and its practical application, so that those skilled in the art can best utilize the present disclosure and utilize various embodiments with different modifications to suit the intended specific application.

Claims

1. A method for processing media resource program information, characterized in that: include: Preprocess the full amount of media information to be retrieved, obtain preprocessed media data, and establish a related tag dictionary library; Constructing a vector media asset mapping table, and vectorizing the preprocessed media asset data and writing it into a media asset vector library; Obtaining a user question and conducting a security review, and if the user question passes the review, obtaining a pending media resource retrieval result based on the user question and the media resource vector library; A security review is performed on the pending media resource search result, and if the pending media resource search result passes the review, the information field required for program playback corresponding to the pending media resource search result is returned to the user interface for display.

2. The method according to claim 1, characterized in that The preprocessing of the full amount of media asset information to be retrieved to obtain preprocessed media asset data and establishing a related tag dictionary library includes: Clean and format all media asset information to obtain cleaned and formatted media asset data; Expanding the media asset data after data cleaning and formatting to obtain media asset description information corresponding to each media asset data as the preprocessed media asset data; The related tag dictionary library is established based on the director information and actor information included in the full amount of media asset information.

3. The method according to claim 2, characterized in that The step of constructing a vector media asset mapping table and vectorizing the pre-processed media asset data and writing the vectorized data into a media asset vector library includes: Construct a vector media asset mapping table of vector ID unique identifiers and media asset ID unique identifiers; Vectorizing the media asset description information to obtain vector text; Converting the media asset ID included in the preprocessed media asset data into a vector ID; Extracting the media asset ID, program name, director, actor, and tag information of the preprocessed media asset data as vector metadata; The vector text, the vector ID, and the vector metadata are written into a media asset vector library, and the mapping relationship between the media asset ID and the vector ID is written into the vector media asset mapping table.

4. The method according to claim 1, characterized in that: The method further comprises: When the full amount of media asset information changes, the changed vector ID corresponding to the changed media asset ID is obtained through the vector media asset mapping table, and the corresponding data in the media asset vector library is updated or deleted.

5. The method according to claim 1, characterized in that The obtaining of the user question and conducting a security review, and obtaining a pending media resource retrieval result based on the user question and the media resource vector library when the user question passes the review, includes: Conduct a security review of the user issue; If the user question fails to pass the review, a prompt message is returned; If the user question passes the review, the user question is vectorized to obtain a question description vector; According to the problem description vector, information retrieval is performed in the media resource vector library using an approximate retrieval algorithm to obtain undetermined vector metadata that meets a preset approximate value threshold; Based on the relevant tag dictionary library, extract keywords from the user question to obtain pending keywords; Filtering according to the pending keywords and the pending vector metadata to obtain candidate vector metadata; Performing word frequency statistics on the elements in the candidate vector metadata, taking the element with the highest word frequency ratio in the candidate vector metadata as the public label of the candidate vector metadata, and performing secondary filtering using the public label to obtain a secondary filtering result; The pending media resource retrieval result is obtained according to the user question and the secondary filtering result.

6. The method according to claim 5, characterized in that The security review of the pending media resource search result includes: Obtaining the answer description corresponding to the pending media resource retrieval result through the search engine interface corresponding to the pre-trained large model; Input the answer description and the pending media resource retrieval result into a pre-trained large model for reasoning to obtain relevant media resource information; For the programs included in the media asset information, a data security review is conducted using a vocabulary of artists with bad records and a third-party sensitive vocabulary.

7. The method according to claim 6, characterized in that Returning the information fields required for program playback corresponding to the pending media resource search results to the user interface for display, including: Perform data completion operations on the program, encapsulate the information fields required for program playback, and return them to the user interface for analysis and display.

8. A media resource program information processing device, characterized in that: include: The processing module is used to pre-process the full amount of media information to be retrieved, obtain pre-processed media data, and establish a related tag dictionary library; Constructing a vector media asset mapping table, and vectorizing the preprocessed media asset data and writing it into a media asset vector library; obtaining a user question and conducting a security review, and if the user question passes the review, obtaining a pending media asset retrieval result based on the user question and the media asset vector library; The output module is used to perform a security review on the pending media resource search result, and if the pending media resource search result passes the review, return the program playback required information field corresponding to the pending media resource search result to the user interface for display.

9. A computer device, characterized in that: The computer device comprises a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device executes the method according to any one of claims 1 to 7.

10. A readable storage medium, characterized in that: The readable storage medium includes a computer program, and when the computer program is executed, the computer device where the readable storage medium is located is controlled to execute the method according to any one of claims 1 to 7.

Images