Duplicate multimedia file identification method, device, electronic device and storage medium

Abstract

The embodiments of the present invention provide a method, device, electronic device and storage medium for identifying duplicate multimedia files. The scheme is as follows: obtain the first title of the first multimedia file and the second title of the second multimedia file; obtain the first data to be eliminated in the first title; and / or obtain the second data to be eliminated in the second title; calculate the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated; determine whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair based on the first similarity and a preset similarity threshold. The technical scheme provided by the embodiments of the present invention improves the accuracy of multimedia file similarity calculation, thereby improving the accuracy of duplicate multimedia file identification.

Description

Technical Field

[0001] The present invention relates to the field of multimedia technology, and in particular to a method, device, electronic equipment and storage medium for identifying duplicate multimedia files. Background Art

[0002] In the process of producing and distributing multimedia files, how to accurately identify duplicate multimedia files from a large number of multimedia files is an important means to save hardware resources and improve user experience.

[0003] Taking multimedia files as video files as an example, currently the similarity between every two video files is calculated based on the feature data corresponding to different video files in dimensions such as video title, video description information and video length, so as to determine whether every two video files are duplicate video pairs based on the similarity, thereby realizing file identification of duplicate videos.

[0004] The above relative degree calculation can be specifically expressed as follows: by mapping the feature data of each video file in different dimensions to the same vector space model, the distance between the feature data of each two video files in different dimensions in the vector space model is calculated, and the similarity between the two video files is expressed by the distance. In the similarity calculation process, due to the different weights corresponding to the feature data in different dimensions, the distance between the feature data of each two video files in different dimensions calculated in the same vector space model is directly expressed as the similarity between the two video files, which will affect the accuracy of the calculated similarity, thereby affecting the accuracy of duplicate video file recognition, that is, affecting the accuracy of duplicate multimedia file recognition. Summary of the invention

[0005] The purpose of the embodiments of the present invention is to provide a method, device, electronic device and storage medium for identifying duplicate multimedia files, so as to improve the accuracy of multimedia file similarity calculation, thereby improving the accuracy of duplicate multimedia file identification. The specific technical solution is as follows:

[0006] In a first aspect of the present invention, a method for identifying duplicate multimedia files is provided, the method comprising:

[0007] Obtain a first title of a first multimedia file and a second title of a second multimedia file;

[0008] Acquire the first data to be removed from the first title; and / or,

[0009] Acquire the second data to be eliminated in the second title;

[0010] The first data to be removed is text content in the first title that is associated with the first user identification information corresponding to the first multimedia file, and the second data to be removed is text content in the second title that is associated with the second user identification information corresponding to the second multimedia file;

[0011] Calculating a similarity between a third title and a fourth title as a first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated;

[0012] According to the first similarity and a preset similarity threshold, it is determined whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair.

[0013] Optionally, the step of obtaining the first data to be removed in the first title includes:

[0014] Acquire first user identification information corresponding to the first multimedia file;

[0015] Acquire all multimedia files corresponding to the first user identification information as a first multimedia file set;

[0016] Acquire text content associated with the first user identification information from the title of each multimedia file included in the first multimedia file set as first target text content;

[0017] The first target text content included in the first title is used as the first data to be removed in the first title.

[0018] Optionally, the step of obtaining the second data to be removed in the second title includes:

[0019] Acquire second user identification information corresponding to the second multimedia file;

[0020] Acquire all multimedia files corresponding to the second user identification information as a second multimedia file set;

[0021] Acquire text content associated with the second user identification information from the title of each multimedia file included in the second multimedia file set as second target text content;

[0022] The second target text content included in the second title is used as the second data to be eliminated in the second title.

[0023] Optionally, the step of calculating the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file includes:

[0024] Obtaining the first characteristic data in the third title;

[0025] Get the second characteristic data in the fourth title;

[0026] The similarity between the first feature data and the second feature data is calculated as the first similarity between the first multimedia file and the second multimedia file.

[0027] Optionally, the first feature data includes one or more of sequence words, date keywords or title keywords; the second feature data includes one or more of sequence words, date keywords or title keywords.

[0028] Optionally, the step of calculating the similarity between the first feature data and the second feature data as the first similarity between the first multimedia file and the second multimedia file includes:

[0029] If both the first feature data and the second feature data include the sequence word, then calculating the second similarity between the sequence word in the third title and the sequence word in the fourth title; and / or,

[0030] If both the first feature data and the second feature data include the date keyword, then calculating a third similarity between the date keyword in the third title and the date keyword in the fourth title; and / or,

[0031] If both the first feature data and the second feature data include the title keyword, calculating a fourth similarity between the title keyword in the third title and the title keyword in the fourth title;

[0032] Based on the second similarity and / or the third similarity and / or the fourth similarity, a first similarity between the first multimedia file and the second multimedia file is calculated.

[0033] Optionally, the step of determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold includes:

[0034] If the first similarity is greater than a preset similarity threshold, determining that the first multimedia file and the second multimedia file are a duplicate multimedia file pair;

[0035] If the first similarity is not greater than the preset similarity threshold, it is determined that the first multimedia file and the second multimedia file are not a duplicate file pair.

[0036] In a second aspect of the present invention, a device for identifying duplicate multimedia files is provided, the device comprising:

[0037] A first acquisition module, used for acquiring a first title of a first multimedia file and a second title of a second multimedia file;

[0038] A second acquisition module is used to acquire first data to be eliminated in the first title; and / or to acquire second data to be eliminated in the second title; the first data to be eliminated is text content in the first title associated with the first user identification information corresponding to the first multimedia file, and the second data to be eliminated is text content in the second title associated with the second user identification information corresponding to the second multimedia file;

[0039] a calculation module, configured to calculate a similarity between a third title and a fourth title as a first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated;

[0040] The determination module is used to determine whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold.

[0041] Optionally, the second acquisition module is specifically used to acquire first user identification information corresponding to the first multimedia file;

[0042] Acquire all multimedia files corresponding to the first user identification information as a first multimedia file set;

[0043] Acquire text content associated with the first user identification information from the title of each multimedia file included in the first multimedia file set as first target text content;

[0044] The first target text content included in the first title is used as the first data to be removed in the first title.

[0045] Optionally, the second acquisition module is specifically used to acquire second user identification information corresponding to the second multimedia file;

[0046] Acquire all multimedia files corresponding to the second user identification information as a second multimedia file set;

[0047] Acquire text content associated with the second user identification information from the title of each multimedia file included in the second multimedia file set as second target text content;

[0048] The second target text content included in the second title is used as the second data to be eliminated in the second title.

[0049] Optionally, the computing module includes:

[0050] A first acquisition submodule, used for acquiring first characteristic data in a third title;

[0051] A second acquisition submodule, used for acquiring second characteristic data in the fourth title;

[0052] The calculation submodule is used to calculate the similarity between the first feature data and the second feature data as the first similarity between the first multimedia file and the second multimedia file.

[0053] Optionally, the first feature data includes one or more of sequence words, date keywords or title keywords; the second feature data includes one or more of sequence words, date keywords or title keywords.

[0054] Optionally, the calculation submodule is specifically configured to calculate a second similarity between the sequence words in the third title and the sequence words in the fourth title if both the first feature data and the second feature data include the sequence words; and / or,

[0055] If both the first feature data and the second feature data include the date keyword, then calculating a third similarity between the date keyword in the third title and the date keyword in the fourth title; and / or,

[0056] If both the first feature data and the second feature data include the title keyword, calculating a fourth similarity between the title keyword in the third title and the title keyword in the fourth title;

[0057] Based on the second similarity and / or the third similarity and / or the fourth similarity, a first similarity between the first multimedia file and the second multimedia file is calculated.

[0058] Optionally, the determination module is specifically configured to determine that the first multimedia file and the second multimedia file are a duplicate multimedia file pair if the first similarity is greater than a preset similarity threshold;

[0059] If the first similarity is not greater than the preset similarity threshold, it is determined that the first multimedia file and the second multimedia file are not a duplicate file pair.

[0060] In a third aspect of the present invention, there is also provided an electronic device, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus;

[0061] Memory, used to store computer programs;

[0062] The processor is used to implement any of the above-mentioned steps of the repeated multimedia file identification method when executing the program stored in the memory.

[0063] In a fourth aspect of the present invention, a computer-readable storage medium is provided, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, any of the above-mentioned steps of the method for identifying repeated multimedia files is implemented.

[0064] In a fifth aspect of the implementation of the present invention, a computer program comprising instructions is also provided, which, when executed on a computer, enables the computer to execute any of the above-mentioned duplicate multimedia file identification methods.

[0065] The duplicate multimedia file identification method, device, electronic device and storage medium provided in the embodiment of the present invention can respectively remove the data to be removed from the titles of two multimedia files, that is, remove the text content associated with the user identification information in the multimedia file titles, so as to determine whether the two multimedia files are duplicate media files according to the similarity between the two multimedia file titles after removing the data to be removed. The removal of the data to be removed from the titles of multimedia files can effectively reduce the influence of different user identification information in the titles on the similarity calculation, improve the accuracy of the calculated similarity, and thus improve the accuracy of duplicate multimedia file identification. In addition, since the titles of duplicate multimedia files have a high degree of similarity, compared with the method for duplicate multimedia file identification in the related art that maps feature data on different dimensions to the same vector space model, the accuracy of the similarity calculated only based on the feature data in the title in the embodiment of the present invention is higher, and the recognition result of the duplicate multimedia file determined based on the similarity is more accurate, thereby improving the accuracy of duplicate multimedia file identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below.

[0067] Figure 1 A schematic diagram of a first flow chart of a method for identifying duplicate multimedia files provided by an embodiment of the present invention;

[0068] Figure 2A schematic flow chart of a first method for obtaining data to be eliminated provided in an embodiment of the present invention;

[0069] Figure 3 A schematic flow chart of a second method for obtaining data to be eliminated provided in an embodiment of the present invention;

[0070] Figure 4 A schematic diagram of a first flow chart of a similarity calculation method provided in an embodiment of the present invention;

[0071] Figure 5 A second flow chart of the similarity calculation method provided by an embodiment of the present invention;

[0072] Figure 6 A second flow chart of the method for identifying duplicate multimedia files provided by an embodiment of the present invention;

[0073] Figure 7 A third flow chart of the method for identifying duplicate multimedia files provided by an embodiment of the present invention;

[0074] Figure 8 A schematic diagram of a flow chart of a binary classification model training method provided by an embodiment of the present invention;

[0075] Fig. 9 A schematic diagram of the structure of a device for identifying duplicate multimedia files provided by an embodiment of the present invention;

[0076] Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0077] The technical solutions in the embodiments of the present invention will be described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0078] In order to solve the problem that the accuracy of similarity calculation in related technologies is poor, thereby affecting the accuracy of duplicate multimedia file identification, an embodiment of the present invention provides a method for identifying duplicate multimedia files. The method can be applied to any electronic device, which can be a server for producing and distributing multimedia files. Among them, multimedia file production can include operations such as multimedia file transcoding and storage, and multimedia file distribution can include operations such as multimedia file search and recommendation. Figure 1 As shown, Figure 1 The first flow chart of the method for identifying duplicate multimedia files provided by the embodiment of the present invention is as follows.

[0079] Step S101, obtaining a first title of a first multimedia file and a second title of a second multimedia file.

[0080] Step S102, obtaining first data to be eliminated in the first title; and / or obtaining second data to be eliminated in the second title.

[0081] In an embodiment of the present invention, the first data to be eliminated is the text content in the first title associated with the first user identification information corresponding to the first multimedia file, and the second data to be eliminated is the text content in the second title associated with the second user identification information corresponding to the second multimedia file.

[0082] Step S103, calculating the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated.

[0083] Step S104: determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold.

[0084] Through the method provided by the embodiment of the present invention, the data to be removed from the titles of two multimedia files can be removed respectively, that is, the text content associated with the user identification information in the multimedia file titles can be removed, so as to determine whether the two multimedia files are duplicate media files according to the similarity between the two multimedia file titles after removing the data to be removed. The removal of the data to be removed from the titles of multimedia files can effectively reduce the influence of different user identification information in the titles on the similarity calculation, improve the accuracy of the calculated similarity, and thus improve the accuracy of duplicate multimedia file recognition. In addition, since the titles of duplicate multimedia files have a high degree of similarity, compared with the method for duplicate multimedia file recognition in the related art that maps feature data on different dimensions to the same vector space model, the accuracy of the similarity calculated only based on the feature data in the title in the embodiment of the present invention is higher, and the recognition result of the duplicate multimedia file determined based on the similarity is more accurate, thereby improving the accuracy of duplicate multimedia file recognition.

[0085] The embodiments of the present invention are described below through specific embodiments. For the convenience of description, the following description is based on electronic devices as the execution subject, which does not serve any limiting role.

[0086] With respect to the above step S101, the first title of the first multimedia file and the second title of the second multimedia file are obtained.

[0087] In this step, the electronic device can obtain the first multimedia file and the second multimedia file. After obtaining the first multimedia file and the second multimedia file, the electronic device can obtain the title of the first multimedia file to obtain the first title; obtain the title of the second multimedia file to obtain the second title.

[0088] The first multimedia file and the second multimedia file include, but are not limited to, video files, audio files, and e-book files. In addition, the number of the first multimedia file and the second multimedia file may be one or more. Here, the file category and the number of the first multimedia file and the second multimedia file are not specifically limited. For ease of description, the following description only takes the first multimedia file and the second multimedia file as video files, which does not serve any limiting role.

[0089] In an optional embodiment, when acquiring the first multimedia file and the second multimedia file, the electronic device may acquire the first multimedia file and the second multimedia file from the multimedia files uploaded by the user and the preset multimedia file library. For example, the electronic device may acquire any two multimedia files from the multimedia files uploaded by the user as the first multimedia file and the second multimedia file. For another example, the electronic device may acquire one multimedia file from the multimedia files uploaded by the user and the preset multimedia file library respectively as the first multimedia file and the second multimedia file.

[0090] In another optional embodiment, when acquiring the first multimedia file and the second multimedia file, the electronic device may preliminarily screen the multimedia files uploaded by the user and / or any two multimedia files in the preset multimedia file library, thereby selecting a pair of multimedia files that are suspected to be duplicates. The electronic device may use two multimedia files in the screened out suspected duplicate multimedia files as the first multimedia file and the second multimedia file. The method for screening suspected duplicate multimedia files can be found in the description below and will not be described in detail here.

[0091] In the embodiment of the present invention, the acquisition method of the first multimedia file and the second multimedia file is not specifically limited. For the convenience of description, the following is an example of the first multimedia file being a multimedia file uploaded by a user and the second multimedia file being a multimedia file in a preset multimedia file library, which does not serve any limiting role.

[0092] With respect to the above step S102, the first data to be eliminated in the first title is obtained; and / or the second data to be eliminated in the second title is obtained.

[0093] In this step, when the first title includes text content associated with the first user identification information corresponding to the first multimedia file, the electronic device can obtain the text content associated with the first user identification information corresponding to the first multimedia file from the first title as the first data to be eliminated. When the second title includes text content associated with the second user identification information corresponding to the second multimedia file, the electronic device can obtain the text content associated with the second user identification information corresponding to the second multimedia file from the second title as the second data to be eliminated.

[0094] In an embodiment of the present invention, the first user identification information corresponding to the first multimedia file may be the user identification information of the user who uploaded the first multimedia file or the user identification information of the user who edited the first multimedia file. The second user identification information corresponding to the second multimedia file may be the user identification information of the user who uploaded the second multimedia file or the user identification information of the user who edited the second multimedia file.

[0095] The user identification information includes but is not limited to the user's name, nickname or the identification assigned to the user by the server. Here, the user identification information is not specifically limited.

[0096] For ease of understanding, taking the first multimedia file and the second multimedia file as multimedia files uploaded by the user as an example, when the user uploads a multimedia file, such as video file A, the title of the multimedia file will carry the user identification information of the user, such as the user's identity (ID) account. The electronic device can obtain the ID account from the title of the video file A as the data to be removed.

[0097] In an optional embodiment, in the title of the multimedia file, the user identification information may satisfy a specific format.

[0098] For example, the user identification information may be at a specific position of the title, such as the front end of all text content of the title. Taking the first multimedia file as an example, the electronic device may read the user identification information in the first title from a specific position of the first title of the first multimedia file to obtain the first data to be eliminated. The acquisition of the second data to be eliminated is referred to the acquisition of the first data to be eliminated, and no specific description is given here.

[0099] For another example, the user identification information is the text content included in the preset characters in the title, where the preset characters include but are not limited to [], [], '', "" and []. Still taking the first multimedia file as an example, the electronic device can determine the text content in the preset characters in the first title of the first multimedia file as the first data to be eliminated. The acquisition of the second data to be eliminated is referred to the acquisition of the first data to be eliminated, and no specific description is given here.

[0100] For ease of understanding, take the video title of a certain video file as an example. The video title of the video file is: [Name A] Choose sunglasses according to face shape.

[0101] The electronic device can extract the name A in [ ] as the data to be removed from the video file, that is, the text content associated with the user tag information in the video title of the video file is the name A.

[0102] In another optional embodiment, Figure 2 As shown, Figure 2 A flowchart of a method for obtaining first data to be eliminated provided by an embodiment of the present invention is shown in FIG. Specific examples include obtaining the first data to be eliminated in the first title in step S102 and subdividing the steps into the following steps, namely, step S201 to step S204.

[0103] Step S201: Acquire first user identification information corresponding to a first multimedia file.

[0104] In this step, no matter whether the first title of the first multimedia file contains the first user identification information, the electronic device can obtain the first user identification information corresponding to the first multimedia file.

[0105] In an optional embodiment, when the first title of the first multimedia file includes the first user identification information, the electronic device can obtain the first user identification information of the first multimedia file from the specific position or preset characters.

[0106] In another optional embodiment, when the first title of the first multimedia file does not include the first user identification information, the electronic device determines the user corresponding to the first multimedia file, thereby determining the user identification information used, such as the identification assigned to the user by the server, to obtain the first user identification information corresponding to the first multimedia file.

[0107] Step S202: Acquire all multimedia files corresponding to the first user identification information as a first multimedia file set.

[0108] In this step, for each multimedia file uploaded by the user, the electronic device may associate the user identification information of the user with the multimedia file uploaded by the user, for example, establish a corresponding relationship between the user identification information and the multimedia file, etc. The electronic device may obtain all multimedia files corresponding to the first user identification information according to the first user identification information obtained in step S201 above, and obtain a first multimedia file set.

[0109] Step S203: acquiring text content associated with the first user identification information from the title of each multimedia file included in the first multimedia file set as the first target text content.

[0110] In this step, for each multimedia file included in the first multimedia file set, the electronic device can obtain text content associated with the user identification information from the title of the multimedia file, such as the specific position or the preset characters of the fir tree, as the first target text content.

[0111] Step S204: taking the first target text content included in the first title as the first data to be removed in the first title.

[0112] In this step, the electronic device can match the text content in the first title of the first multimedia file with each first target text content acquired in the above step S203, and determine the first target text content included in the first title as the first data to be removed in the title of the first multimedia file.

[0113] In the embodiment of the present invention, since the first user tag information is user-adjustable, for example, when the first user identification information is a user nickname, the user can adjust the nickname at any time, therefore, in order to ensure the accuracy and completeness of the first data to be eliminated acquired by the electronic device, the electronic device can accurately and completely determine the first data to be eliminated in the first title by matching the text content in the first title with the first target text content of each multimedia file in the first multimedia file set, thereby ensuring the accuracy and completeness of the acquired first data to be eliminated.

[0114] In an optional embodiment, Figure 3 As shown, Figure 3 A flow chart of a method for obtaining the second data to be eliminated provided by an embodiment of the present invention. Specifically, the step of obtaining the second data to be eliminated in the second title in the above step S102 is refined into the following steps, namely, step S301 to step S304.

[0115] Step S301: Acquire second user identification information corresponding to a second multimedia file.

[0116] Step S302: Acquire all multimedia files corresponding to the second user identification information as a second multimedia file set.

[0117] Step S303: acquiring text content associated with the second user identification information from the title of each multimedia file included in the second multimedia file set as the second target text content.

[0118] Step S304: taking the second target text content included in the second title as the second data to be removed in the second title.

[0119] The acquisition of the second data to be eliminated shown in the above steps S301 to S304 can refer to the acquisition of the first data to be eliminated shown in the above steps S201 to S204, and will not be described in detail here.

[0120] In the embodiment of the present invention, since the second user tag information is also adjustable by the user, in order to ensure the accuracy and completeness of the second data to be eliminated acquired by the electronic device, the electronic device can accurately and completely determine the second data to be eliminated in the second title by matching the text content in the second title with the second target text content of each multimedia file in the second multimedia file set, thereby ensuring the accuracy and completeness of the acquired second data to be eliminated.

[0121] With respect to the above step S103, the similarity between the third title and the fourth title is calculated as the first similarity between the first multimedia file and the second multimedia file; the third title is the first title excluding the first data to be eliminated, and the fourth title is the second title excluding the second data to be eliminated.

[0122] In the embodiment of the present invention, the electronic device may remove the first data to be removed from the first title to obtain the third title. The electronic device may remove the second data to be removed from the second title to obtain the fourth title.

[0123] In an optional embodiment, after obtaining the third title and the fourth title, the electronic device may directly calculate the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file.

[0124] For ease of understanding, video A and video B are used as examples for explanation. The video title of video A is: [User A] Choose sunglasses according to face shape | Practical and good-looking sunglasses purchase tips, and the video title of video B is: How to choose suitable sunglasses according to face shape. Although the video titles of the two videos are quite different, the two videos are the same video, that is, video A and video B are a repeated video pair.

[0125] When the data to be removed in the video title of video A is not removed, that is, when user A is not removed, the similarity between video A and video B is calculated to be 0.68, and when user A is removed from the video title of video A, the similarity between video A and video B is calculated to be 0.80, that is, the first similarity between video A and video B is 0.80. Therefore, the similarity calculated when user A is removed from the video title of video A is significantly greater than the similarity calculated when user A is not removed from the video title of video A, which is more consistent with the actual situation.

[0126] In an embodiment of the present invention, by eliminating text content associated with user identification information included in the title of the multimedia file, that is, the data to be eliminated, the impact of the data to be eliminated on the similarity calculation can be effectively reduced, thereby improving the accuracy of the calculated similarity, and further improving the accuracy of duplicate multimedia file identification.

[0127] In another optional embodiment, Figure 4 As shown, Figure 4 The first flow chart of the similarity calculation method provided by the embodiment of the present invention is as follows: Specifically, the above step S103, i.e., calculating the similarity between the third title and the fourth title, is refined into the following steps, i.e., step S401 to step S403, as the first similarity between the first multimedia file and the second multimedia file.

[0128] Step S401, obtaining the first characteristic data in the third title.

[0129] In an optional embodiment, the first characteristic data may include one or more of sequence words, date keywords or title keywords.

[0130] Step S402, obtaining the second characteristic data in the fourth title.

[0131] In an optional embodiment, the second characteristic data may include one or more of sequence words, date keywords or title keywords.

[0132] In the embodiment of the present invention, the execution order of the above step S401 and step S402 is not specifically limited.

[0133] In an optional embodiment, when the electronic device acquires the first characteristic data in the third title and the second characteristic data in the fourth title, at least the following situations may be included.

[0134] In case 1, if the first characteristic data and / or the second characteristic data include the sequence word, the electronic device may extract the sequence word from the third title to obtain the sequence word in the first characteristic data. And / or the electronic device may extract the sequence word from the fourth title to obtain the sequence word in the second characteristic data.

[0135] In the embodiment of the present invention, the sequence words are different according to the different file categories of the first multimedia file and the second multimedia file. Taking the first multimedia file as video A as an example, when video A is a sports event video or a series of variety shows, the sequence words can be the episode number of video A; when video A is a certain episode of a multi-episode TV series or documentary, the sequence words can be the episode number of video A. Here, the sequence words in the first feature data and the second feature data are not specifically limited.

[0136] In case 2, if the first characteristic data and / or the second characteristic data include the date keyword, the electronic device may extract the date keyword from the third title of the first multimedia file to obtain the date keyword in the first characteristic data. And / or the electronic device may extract the date keyword from the third title of the second multimedia file to obtain the date keyword in the second characteristic data.

[0137] For ease of understanding, the extraction of the above date keywords is described by taking the video title of a variety video as an example. The video title of the variety video is: 20191021 Full Version: Name A·Theme B. Among them, 20191021 indicates that the online time of the variety show is October 21, 2019, Name A is the name of the variety show video, and Theme B is the theme name of this variety show. The electronic device can extract 20191021 in the video title as the date keyword in the video title of the variety video.

[0138] Case three, if the above-mentioned first feature data and the second feature data include the above-mentioned title keywords, the electronic device can use a preset word segmentation algorithm to perform word segmentation processing on the third title of the first multimedia file, and select the word with the largest weight from the multiple word segmentations obtained by the word segmentation processing as the title keyword in the first feature data; the electronic device can also use a preset word segmentation algorithm to perform word segmentation processing on the fourth title of the second multimedia file, and select the word with the largest weight from the multiple word segmentations obtained by the word segmentation processing as the title keyword in the second feature data.

[0139] In the embodiment of the present invention, the above-mentioned preset word segmentation algorithm includes but is not limited to a Hidden Markov Model (HMM) word segmentation algorithm and a Conditional Random Fields (CRF) word segmentation algorithm. Here, the above-mentioned preset word segmentation algorithm is not specifically limited.

[0140] For ease of understanding, the extraction of the above title keywords is explained by taking the video title of a certain video file as an example. The video title of the video file is: Stop-motion animation of puppet hands assembling building blocks.

[0141] The electronic device can use the above preset word segmentation algorithm to perform word segmentation processing on the above video title, i.e., the stop-frame animation of puppet hands assembling building blocks, to obtain multiple word segments and the weight corresponding to each word segment, as shown in Table 1.

[0142] Table 1

[0143] Participle puppet hand Assembly Building Blocks Freeze Animation Weight 0.6404055 0.01 0.38056636 0.71411395 0.36493686 0.2512742

[0144] According to the weights corresponding to the participles shown in Table 1, the electronic device can determine that the weight corresponding to the participle "building block" is the largest, that is, 0.71411395. Therefore, the electronic device can determine the word "building block" in the above video title as the title keyword of the video file.

[0145] The weight of each word segment obtained by the above word segmentation process can be calculated by a variety of methods, such as document frequency, inverse document frequency (i.e., TF*IDF), chi-square test, mutual information, information gain, etc. Here, the calculation of the weight of each word segment is not specifically described.

[0146] In an optional embodiment, the number of the extracted title keywords may also be different according to the length of the title of each multimedia file. Still taking the above video title as an example, the electronic device may also select the first two participles with the largest weights as the title keywords of the video file in descending order of weights, i.e., the two participles "building blocks" and "puppet" in the above video title. In the embodiment of the present invention, the number of title keywords included in the above first feature data and the second feature data is not specifically limited.

[0147] In the embodiment of the present invention, since the titles of the first multimedia file and the second multimedia file may be expressed in a variety of forms, other methods may be used to extract the title keywords from the third title and the fourth title.

[0148] For example, a video file is a singing video of two singers adapting two different songs. The video title of the singing video can be "Singer A + Singer B" first collaboration, recorded as Title 1, or the video title can be "Song A + Song B" is so good, recorded as Title 2. When extracting the above title keywords, the electronic device can extract the keywords included in the text content of "" and "", that is, extract singer A and singer B in Title 1 as title keywords, or extract song A and song B in Title 2 as title keywords.

[0149] In the embodiment of the present invention, the first characteristic data and the second characteristic data may include other characteristic data in addition to the sequence words, date keywords and title keywords. Taking the multimedia file as a video file as an example, the first characteristic data and the second characteristic data may also include a video channel or a video tag, etc. Here, the first characteristic data and the second characteristic data are not specifically limited.

[0150] In an optional embodiment, if the first feature data and the second feature data include the video channel, the electronic device can extract the video channel from the third title using a natural language processing (NLP) algorithm to obtain the video channel in the first feature data. The electronic device can also extract the video channel from the fourth title using an NLP algorithm to obtain the video channel in the second feature data.

[0151] In another optional embodiment, if the first feature data and the second feature data include the video tag, the electronic device can extract the video tag from the third title and the fourth title respectively. The extraction of the video tag can refer to the extraction method of the video channel.

[0152] For ease of understanding, the video titles of video A and video B are used as examples to illustrate the extraction of the above video channels and video tags. Among them, the video title of video A is: "XX Highway Gliding Arrives in the Store, the Latest Model of 2020, Really Domineering!", XX represents the car brand. The video title of video B is: "Old Sword Comparison - Li Yunlong Defeats Bantian United".

[0153] The electronic device can use the NLP algorithm to extract features from the video title of video A and obtain the video channel and video label of video A. That is, the video channel of video A is car, and the video labels include: XX, car, new car, motorcycle, American, etc.

[0154] The electronic device can also use the NLP algorithm to extract features from the video title of video B to obtain the video channel and video label of video B. That is, the video channel of video B is a trailer, and the video labels include: Li Youbin, Liang Jian, Li Yunlong, film and television, TV series, plot and clip, etc.

[0155] In the embodiment of the present invention, when extracting the video tag, the electronic device can also determine the weight of each extracted video tag. Taking the above video B as an example, the weights corresponding to Li Youbin, Liang Jian, Li Yunlong, film and television, TV series, plot and clip can be: 0.82, 0.5, 0.48, 0.6, 0.42, 0.42 and 0.19 respectively. The determination of the weight of each video tag is not specifically described here.

[0156] In an optional embodiment, the first feature data and the second feature data are extracted from the titles corresponding to the first multimedia file and the second multimedia file after the data to be removed is removed. In order to further improve the accuracy of the calculated first similarity between the first multimedia file and the second multimedia file, the electronic device can also extract other feature data from the file description information corresponding to the first multimedia file and the second multimedia file. Taking the first multimedia file and the second multimedia file as video files as an example, the other feature data extracted from the file description information includes but is not limited to the actor's name, the director's name, and the video content keywords. The extraction of other feature data from the file description information can refer to the extraction of the first feature data and the second feature data, which will not be described in detail here.

[0157] Step S403: Calculate the similarity between the first feature data and the second feature data as the first similarity between the first multimedia file and the second multimedia file.

[0158] In this step, the electronic device can calculate the similarity between the feature data of the first feature data and the second feature data in the same dimension based on the first feature data extracted in the above step S401 and the second feature data extracted in the above step S402, and thus calculate the similarity between the first feature data and the second feature data based on the similarity between the feature data in each dimension as the similarity between the first multimedia file and the second multimedia file.

[0159] For ease of understanding, video C and video D are used as examples for illustration.

[0160] When video C and video D are two videos from different periods of a variety show, that is, the video titles corresponding to video C and video D include different sequence words or date keywords, but the video titles corresponding to video C and video D include the same other feature data, such as title keywords, at this time, the similarity calculated based on the video titles corresponding to video C and video D will be significantly greater than the similarity calculated based on the sequence words or date keywords, resulting in a misjudgment.

[0161] In an embodiment of the present invention, by extracting the first feature data in the third title and the second feature data in the fourth title, the electronic device can calculate the first similarity between the first multimedia file and the second multimedia file based on the feature data corresponding to each dimension in the first feature data and the second feature data, which will effectively improve the accuracy of the calculated first similarity, thereby improving the accuracy of the duplicate multimedia file recognition result determined based on the first similarity.

[0162] In an optional embodiment, Figure 5 As shown, Figure 5 The second flow chart of the similarity calculation method provided by the embodiment of the present invention is as follows: Specifically, the above step S403, calculating the similarity between the first feature data and the second feature data as the first similarity between the first multimedia file and the second multimedia file, is refined into the following steps, namely, step S501 to step S504.

[0163] Step S501: If both the first feature data and the second feature data include sequence words, then calculate the second similarity between the sequence words in the third title and the sequence words in the fourth title.

[0164] In an optional embodiment, when the above-mentioned first characteristic data and the second characteristic data both include the above-mentioned sequence words, the electronic device can determine whether the sequence words in the third title are the same as the sequence words in the fourth title. If the sequence words in the third title are the same as the sequence words in the fourth title, the electronic device can determine the first preset value as the second similarity between the sequence words in the third title and the sequence words in the fourth title. If the sequence words in the third title are not the same as the sequence words in the fourth title, the electronic device can determine the second preset value as the second similarity between the sequence words in the third title and the sequence words in the fourth title.

[0165] The similarity represented by the first preset value is greater than the similarity represented by the second preset value.

[0166] For ease of understanding, the first preset value is 1 and the second preset value is 0 as an example for explanation.

[0167] In one example, the third title is: Program A, Episode 3, Theme A, and the fourth title is: Program A, Episode 3, Theme B. Since the sequence words in the third title and the fourth title are both Episode 3, the electronic device can determine the second similarity between the sequence words in the third title and the sequence words in the fourth title to be 1.

[0168] In another example, the third title is: Program A, Third Episode, Theme A, and the fourth title is: Program A, Fourth Episode, Theme A. Since the sequence word in the third title is the third episode and the sequence word in the fourth title is the fourth episode, the electronic device can determine the second similarity between the sequence words in the third title and the sequence words in the fourth title to be 0.

[0169] In the embodiment of the present invention, the first preset value and the second preset value are not specifically limited.

[0170] In another optional embodiment, the electronic device may perform binarization processing on the sequence words in the third title and the sequence words in the fourth title to obtain a second similarity between the sequence words in the third title and the sequence words in the fourth title.

[0171] For example, when the sequence words in the first feature data and the sequence words in the second feature data are different, the feature data obtained by the binarization process may be 0, that is, the second similarity between the sequence words in the third title and the sequence words in the fourth title is 0. When the sequence words in the first feature data and the sequence words in the second feature data are the same, the feature data obtained by the binarization process may be 1, that is, the second similarity between the sequence words in the third title and the sequence words in the fourth title is 1.

[0172] In the embodiment of the present invention, the calculation method of the second similarity is not specifically limited.

[0173] Step S502: If both the first feature data and the second feature data include date keywords, a third similarity between the date keywords in the third title and the date keywords in the fourth title is calculated.

[0174] In an optional embodiment, when both the first characteristic data and the second characteristic data include the date keyword, the electronic device can determine whether the date keyword in the third title is the same as the date keyword in the fourth title. If the date keyword in the third title is the same as the date keyword in the fourth title, the electronic device can determine the first preset value as the third similarity between the date keyword in the third title and the date keyword in the fourth title. If the date keyword in the third title is different from the date keyword in the fourth title, the electronic device can determine the second preset value as the third similarity between the date keyword in the third title and the date keyword in the fourth title.

[0175] In another optional embodiment, the electronic device may perform binarization processing on the date keyword in the third title and the date keyword in the fourth title to obtain a third similarity between the sequence words in the third title and the sequence words in the fourth title.

[0176] The method for determining the third similarity may refer to the method for determining the second similarity, and the determination of the third similarity will not be described in detail herein.

[0177] Step S503: If both the first feature data and the second feature data include title keywords, a fourth similarity between the title keywords in the third title and the title keywords in the fourth title is calculated.

[0178] In an optional embodiment, when both the first feature data and the second feature data include the title keywords, the electronic device can calculate the similarity between the title keywords in the third title and the title keywords in the fourth title according to a preset similarity algorithm to obtain a fourth similarity. The preset similarity algorithm includes but is not limited to a Euclidean distance algorithm and a Manhattan distance algorithm.

[0179] For ease of understanding, the preset similarity algorithm is taken as the above-mentioned Euclidean distance algorithm as an example. The electronic device can respectively calculate the Euclidean distance between the first vector representing the title keyword in the third title and the second vector representing the title keyword in the fourth title, thereby determining the reciprocal or negative number of the Euclidean distance as the fourth similarity between the title keyword in the third title and the title keyword in the fourth title. The calculation process of the Euclidean distance is not specifically described here.

[0180] In an embodiment of the present invention, in order to facilitate the subsequent calculation of the first similarity between the first multimedia file and the second multimedia file based on the above-mentioned first feature data and the second feature data, the electronic device can binarize or normalize the above-mentioned first feature data and the second feature data, so as to calculate the above-mentioned first similarity based on the processed feature data.

[0181] In another optional embodiment, the electronic device may perform binarization processing on the title keywords in the third title and the title keywords in the fourth title to obtain a fourth similarity between the title keywords in the third title and the title keywords in the fourth title.

[0182] In this embodiment of the present invention, the calculation of the fourth similarity is not specifically limited.

[0183] In an optional embodiment, when calculating the second similarity, the third similarity and the fourth similarity, if the first multimedia file includes multiple multimedia files, or the second multimedia file includes multiple multimedia files, then for each first multimedia file or each second multimedia file, after calculating the second similarity, the third similarity and the fourth similarity, the calculated second similarity, the third similarity or the fourth similarity can be normalized. That is, the value range of each similarity is adjusted to between 0 and 1 to obtain the normalized similarity.

[0184] For ease of understanding, the normalization process of the second similarity is described by taking the first multimedia file as a plurality of multimedia files as an example.

[0185] Through the above step S501, the electronic device can calculate the second similarity between the sequence word of each first multimedia file and the sequence word of the second multimedia file. At this time, the electronic device can perform normalization processing on each second similarity according to each second similarity and the sum of all second similarities. The normalization processing method is not specifically described here.

[0186] In the embodiment of the present invention, in the process of calculating the second similarity, the third similarity and the fourth similarity, in addition to the binarization and normalization, the electronic device may also perform feature standardization, scaling and other processing. Taking the video duration of a video file as an example, the electronic device may perform a difference operation on the video duration of the first multimedia file and the video duration of the second multimedia file, thereby determining the similarity between the first multimedia file and the second multimedia file in the video duration dimension according to the difference in the video duration.

[0187] Step S504: Calculate the first similarity between the first multimedia file and the second multimedia file based on the second similarity and / or the third similarity and / or the fourth similarity.

[0188] In this step, the electronic device may calculate the similarity between the first multimedia file and the second multimedia file, namely the first similarity, based on the similarities calculated in steps S501 to S503, namely the second similarity, the third similarity and the fourth similarity.

[0189] In an optional embodiment, the electronic device may determine the sum of the second similarity, the third similarity and the fourth similarity as the first similarity between the first multimedia file and the second multimedia file.

[0190] In another optional embodiment, the electronic device may determine a weighted sum of the second similarity, the third similarity and the fourth similarity as the first similarity between the first multimedia file and the second multimedia file.

[0191] For ease of understanding, the calculation of the above-mentioned first similarity is explained by taking the example that the third title only includes title keywords and the fourth title includes title keywords and sequence words.

[0192] In an optional embodiment, since the third title does not include the above-mentioned sequence words and date keywords, the electronic device can directly determine the above-mentioned second preset value as the above-mentioned second similarity and third similarity, and calculate the fourth similarity according to the above-mentioned step S503, thereby calculating the first similarity between the first multimedia file and the second multimedia file according to the above-mentioned step S504.

[0193] In another optional embodiment, since both the third title and the fourth title include title keywords, the electronic device may only perform the above step S503 to calculate the fourth similarity between the title keywords in the third title and the title keywords in the fourth title, thereby determining the first similarity between the first multimedia file and the second multimedia file based on the fourth similarity. For example, the electronic device directly determines the fourth similarity as the first similarity between the first multimedia file and the second multimedia file.

[0194] In an embodiment of the present invention, the above step S504 is explained by taking the example that both the third title and the fourth title include sequence words, date keywords and title keywords. However, due to the diversity of the first title and the second title, the feature data included in the above third title and the fourth title may be different. Therefore, when calculating the first similarity between the above first multimedia file and the second multimedia file, the calculation can be performed based on the similarity corresponding to the feature data on each dimension included in the third title and the fourth title. For example, the electronic device can also calculate the first similarity between the first multimedia file and the second multimedia file by combining the similarities of the above video files in the dimensions of video channels and video tags. Here, the calculation method of the above first similarity is not specifically limited.

[0195] In another optional embodiment, the first similarity between the first multimedia file and the second multimedia file is calculated by the second similarity, the third similarity and the fourth similarity. In addition, the electronic device can also use a pre-trained binary classification model to determine the first similarity between the first multimedia file and the second multimedia file. For a description of the binary classification model, please refer to the following description and will not be repeated here.

[0196] With respect to the above step S104, it is determined whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold.

[0197] In this step, the electronic device may compare the first similarity calculated above with a preset similarity threshold (referred to as the first preset similarity threshold for easy distinction), thereby determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair based on the comparison result.

[0198] In an optional embodiment, if the first similarity is greater than a preset similarity threshold, it is determined that the first multimedia file and the second multimedia file are a duplicate multimedia file pair.

[0199] In another optional embodiment, if the first similarity is not greater than a preset similarity threshold, it is determined that the first multimedia file and the second multimedia file are not a duplicate multimedia file pair.

[0200] In an optional embodiment, if the first similarity between the first multimedia file and the second multimedia file is determined by the pre-trained binary classification model, the electronic device can also use the binary classification model to determine whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair based on the calculated first similarity.

[0201] In an optional embodiment, after determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair, the electronic device may perform production processing or distribution processing on the first multimedia file and the second multimedia file.

[0202] For example, when the first multimedia file is a video file uploaded by the user to the electronic device, and the second video is a video file stored in a preset multimedia file library, there are at least the following two situations:

[0203] In case one, if the first multimedia file and the second multimedia file are a non-duplicate multimedia file pair, that is, the first multimedia file uploaded by the user is not duplicated with the second multimedia file in the preset multimedia file library, at this time, the electronic device can transcode the first multimedia file, and perform operations such as warehousing and recommending the transcoded first multimedia file; the electronic device can search and recommend the second multimedia file.

[0204] Case 2: If the first multimedia file and the second multimedia file are a duplicate multimedia file pair, that is, the first multimedia file uploaded by the user is duplicated with the second multimedia file in the preset multimedia file library, then the electronic device can discard the first multimedia file; the electronic device can search, recommend, and perform other operations on the second multimedia file.

[0205] When the first multimedia file and the second multimedia file are both multimedia files uploaded by the user to the electronic device, there are at least the following two situations:

[0206] In case three, if the first multimedia file and the second multimedia file are a pair of non-repetitive multimedia files, the electronic device can perform operations such as transcoding, warehousing, and recommendation on the first multimedia file and the second multimedia file respectively.

[0207] Case four, if the first multimedia file and the second multimedia file are a duplicate multimedia file pair, the electronic device can discard the multimedia file with shorter file length or smaller file size based on the file length or file size of the first multimedia file and the file length or file size of the second multimedia file, and transcode, store, recommend, and other operations on the multimedia file with longer file length or larger file size.

[0208] When the first multimedia file and the second multimedia file are both videos stored in a preset multimedia file library, at least the following two situations are included:

[0209] Case 5: If the first multimedia file and the second multimedia file are a pair of non-repetitive multimedia files, the electronic device can perform operations such as searching and recommending the first multimedia file and the second multimedia file respectively.

[0210] Case six, if the first multimedia file and the second multimedia file are a duplicate multimedia file pair, the electronic device can discard the multimedia file with a shorter file length or a smaller file size based on the file length of the first multimedia file and the file length or file size of the second multimedia file, and search, recommend, and perform other operations on the multimedia file with a longer file length or a larger file size.

[0211] In an embodiment of the present invention, after determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair, the first multimedia file and the second multimedia file may be processed accordingly according to different user requirements. Here, the processing method of the first multimedia file and the second multimedia file is not specifically limited.

[0212] Through the above embodiment, the electronic device produces and distributes the first multimedia file and the second multimedia file according to the recognition result of the duplicate multimedia file pair, which will effectively reduce the resource waste caused by the duplicate multimedia files in the production and distribution process and improve the user experience.

[0213] In an optional embodiment, according to Figure 1 The method shown in the figure, the embodiment of the present invention also provides a method for identifying repeated multimedia files. Figure 6 As shown, Figure 6 A second flow chart of a method for identifying duplicate multimedia files provided by an embodiment of the present invention. The method comprises the following steps.

[0214] Step S601, obtaining third characteristic data of the first multimedia file, and obtaining fourth characteristic data of the second multimedia file.

[0215] In an embodiment of the present invention, the third feature data and the fourth feature data obtained above are also different according to the different types of the above-mentioned multimedia files. For example, when the above-mentioned multimedia file is a video file, the third feature data and the fourth feature data may include the video title, video description information, video duration, video category, vertical category (i.e. category) of the user who uploaded the video, and whether the video is a horizontal version or a vertical version video. Among them, the user vertical category can be obtained by analyzing the video content uploaded by the user in history, which is not specifically described here. For another example, when the above-mentioned multimedia file is an e-book file, the third feature data and the fourth feature data may include the title, content description information, file size and file category of the e-book file. Here, the third feature data and the fourth feature data are not specifically limited.

[0216] Step S602: Calculate a fifth similarity between the first multimedia file and the second multimedia file based on the third feature data and the fourth feature data.

[0217] For ease of understanding, the calculation of the fifth similarity is described by taking the third feature data and the fourth feature data only including the video title as an example. The electronic device can perform word segmentation processing on the video title of the first multimedia file and the video title of the second multimedia file, respectively, to obtain multiple word segments and the weight corresponding to each word segmentation. The electronic device can vectorize the word segmentation results according to the multiple word segments obtained by the corresponding word segmentation processing of the first multimedia file and the second multimedia file and the weight corresponding to each word segmentation, and calculate the cosine angle, so as to obtain the fifth similarity between the first multimedia file and the second multimedia file.

[0218] In the embodiment of the present invention, the fifth similarity can be calculated by the trained binary classification model. Here, the calculation of the fifth similarity is not specifically limited.

[0219] In an optional embodiment, after executing the above step S602, the electronic device can compare the calculated fifth similarity with a preset similarity threshold (for ease of distinction, recorded as the second preset similarity threshold). If the fifth similarity is greater than the second preset similarity threshold, the electronic device can preliminarily determine that the above first multimedia file and the second multimedia file are suspected duplicate multimedia file pairs. At this time, the electronic device can execute step S603, that is, obtain the first title of the first multimedia file and the second title of the second multimedia file.

[0220] In an optional embodiment, if the fifth similarity is not greater than the second preset similarity threshold, the electronic device may directly determine that the first multimedia file and the second multimedia file are not a suspected duplicate multimedia file pair.

[0221] The above steps S601 to S602 are a preliminary process of screening suspected duplicate multimedia files from the first multimedia file and the second multimedia file. Through the above steps S601 and S602, the electronic device can preliminarily screen out suspected duplicate multimedia files from the first multimedia file and the second multimedia file, thereby improving the recognition efficiency of duplicate multimedia file pairs.

[0222] For ease of understanding, when the second multimedia files are all multimedia files in the preset multimedia file library, if the electronic device uses Figure 1 The method shown in the figure takes a long time to complete the identification process of whether the first multimedia file and the second multimedia file are duplicate multimedia file pairs. However, based on the fifth similarity, the multimedia files in the preset multimedia file library can be preliminarily screened, so as to use Figure 1 The method shown further determines whether the suspected duplicate first multimedia file and the second multimedia file are duplicate multimedia file pairs, which effectively reduces the number of multimedia file pairs to be determined and improves the efficiency of identifying duplicate multimedia file pairs.

[0223] In another optional embodiment, in order to further improve the recognition accuracy of duplicate multimedia file pairs, if the fifth similarity is not greater than the second preset similarity threshold, the electronic device can preliminarily determine that the first multimedia file and the second multimedia file may not be suspected duplicate multimedia file pairs. At this time, the electronic device can execute step S603, that is, obtain the first title of the first multimedia file and the second title of the second multimedia file.

[0224] Step S603: Acquire a first title of the first multimedia file and a second title of the second multimedia file.

[0225] Step S604, obtaining first data to be eliminated in the first title; and / or obtaining second data to be eliminated in the second title.

[0226] Step S605, calculating the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated.

[0227] Step S606: determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold.

[0228] The above steps S603 to S606 are the same as the above steps S101 to S104.

[0229] Based on the same inventive concept, according to the method for identifying duplicate multimedia files provided by the above embodiment of the present invention, the embodiment of the present invention also provides a method for identifying duplicate multimedia files. Figure 7 As shown, Figure 7 A third flow chart of the method for identifying duplicate multimedia files provided by an embodiment of the present invention. The method comprises the following steps.

[0230] Step S701: Obtain a first title of a first multimedia file and a second title of a second multimedia file.

[0231] Step S702, obtaining first data to be eliminated in the first title; and / or obtaining second data to be eliminated in the second title.

[0232] The above steps S701 to S702 are the same as the above steps S101 to S102.

[0233] Step S703, obtaining the first characteristic data in the third title, and obtaining the second characteristic data in the fourth title, the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated.

[0234] Regarding the extraction of the first characteristic data and the second characteristic data, reference may be made to the description of step S401 and step S402, which will not be elaborated herein.

[0235] Step S704: input the first feature data and the second feature data into a pre-trained binary classification model, determine a first similarity between the first multimedia file and the second multimedia file, and determine whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair based on the first similarity.

[0236] In this step, the electronic device can input the above-mentioned first feature data and second feature data as input data into a pre-trained binary classification model, and the binary classification model can calculate the similarity between the third title and the fourth title based on the first feature data and the second feature data, that is, calculate the first similarity between the first multimedia file and the second multimedia file, and based on the first similarity, output an output result indicating whether the first multimedia file and the second multimedia file are duplicate multimedia file pairs.

[0237] The above binary classification model is obtained by training the preset binary classification model using a preset training set. The preset training set includes a plurality of repeated multimedia file pairs, a plurality of non-repeated multimedia file pairs, and identification information indicating whether each multimedia file is a repeated multimedia file pair. The training steps of the above binary classification model are described below, which will not be repeated here.

[0238] In an optional embodiment, the above two-classification model includes but is not limited to a gradient boosting decision tree (GBDT) model, an extreme gradient boosting (eXtreme Gradient Boosting, XGBoost) model and a random forest model. Taking the GBDT model as an example, the GBDT algorithm is an iterative decision tree algorithm, which consists of multiple decision trees, and the prediction results of all trees are accumulated to obtain the output result of the GBDT model.

[0239] exist Figure 7 In the illustrated embodiment, the above steps S702 and S703 may also be processed by the electronic device using the above preset trained binary classification model, that is, the electronic device may use the above trained binary classification model to acquire the above first data to be eliminated, second data to be eliminated, first feature data and second feature data.

[0240] In an embodiment of the present invention, the above-mentioned binary classification model has a high ability to distinguish feature data of different dimensions when processing feature data of different dimensions. Therefore, the above-mentioned pre-trained binary classification model can better distinguish the feature data corresponding to each dimension included in the above-mentioned first feature data and second feature data, thereby overcoming the problem of low accuracy in expressing the distance calculated in the vector space model as the similarity between multimedia files in the related art, and improving the accuracy of duplicate multimedia file recognition. For example, the above-mentioned GBDT algorithm has a natural advantage in distinguishing multidimensional feature data and combining feature data of different dimensions, especially for linearly inseparable data, that is, data that cannot be completely distinguished by a linear function between two types of sample data, such as multimedia files in the embodiment of the present invention, the GBDT algorithm can often achieve better distinguishing effects.

[0241] Based on the same inventive concept, according to the method for identifying repeated multimedia files provided by the above embodiment of the present invention, the embodiment of the present invention also provides a binary classification model training method. Figure 8 As shown, Figure 8 A flowchart of a binary classification model training method provided by an embodiment of the present invention is provided, which specifically includes the following steps.

[0242] Step S801, obtaining a preset training set.

[0243] In an embodiment of the present invention, the preset training set may include a plurality of repeated multimedia file pairs, a plurality of non-repeated multimedia file pairs, and identification information indicating that each multimedia file is a repeated multimedia file pair. The two sample multimedia files included in each multimedia file pair may be obtained from the preset multimedia file library. The number of multimedia file pairs and the number of sample multimedia files included in the preset training set are not specifically limited.

[0244] In an optional embodiment, in order to improve the accuracy of the identification information of each multimedia file pair in the above-mentioned preset training set, thereby improving the accuracy of the trained binary classification model, the identification information of each multimedia file pair can be determined by combining machine labeling and manual labeling.

[0245] For each multimedia file pair, an electronic device can be used to identify whether the two multimedia files included in the multimedia file pair are duplicate multimedia file pairs. For example, for each multimedia file pair, the electronic device can calculate the Secure Hash Algorithm 1 (SHA1) value or the Message-Digest Algorithm (MD5) value corresponding to the two multimedia files included in the multimedia file pair. For a multimedia file pair corresponding to two multimedia files with the same SHA1 value or MD5 value in a preset training set, the electronic device can determine that the multimedia file pair is a duplicate multimedia file pair. For a multimedia file pair corresponding to two multimedia files with different SHA1 values ​​or MD5 values ​​in a preset training set, a manual identification method can be used to determine whether the multimedia file pair is a duplicate multimedia file pair, thereby obtaining identification information for each multimedia file pair in the preset training set.

[0246] In an optional embodiment, in order to improve the accuracy of the binary classification model obtained by training, the ratio between the repeated multimedia file pairs and the non-repeated multimedia file pairs included in the above-mentioned preset training set matches the probability of repeated multimedia files appearing in the actual scene. For example, assuming that the probability of multimedia file duplication in the actual scene is 10%, the ratio between the multimedia file pairs and the non-repeated multimedia file pairs in the preset training set is 1:9, that is, 10%: (1-10%).

[0247] In an optional embodiment, in order to improve the accuracy of the binary classification model obtained through training, before obtaining the above-mentioned preset training set, the electronic device may make a data request for multimedia files in a preset multimedia file library. For example, a multimedia file whose title includes a number of characters greater than a preset character number threshold is selected, so as to obtain the above-mentioned preset training set based on the selected multimedia files.

[0248] Step S802: for each sample multimedia file in the preset training set, obtain the third to-be-eliminated data in the fifth title of the sample multimedia file.

[0249] The method for obtaining the third data to be eliminated in each of the above sample multimedia files may refer to the method for obtaining the first data to be eliminated in the above first multimedia file or the second data to be eliminated in the second multimedia file, and no specific description is given here.

[0250] Step S803, for each sample multimedia file in the preset training set, obtain the fifth feature data in the sixth title corresponding to the sample multimedia file; the sixth title corresponding to each sample multimedia file is the fifth title that does not include the third data to be eliminated in the sample multimedia file.

[0251] In this step, when the fifth title of the above-mentioned sample multimedia file includes text content associated with user identification information, that is, the above-mentioned third data to be eliminated, the electronic device eliminates the third data to be eliminated in the fifth title, obtains the sixth title, and obtains the characteristic data in the sixth title, that is, the fifth characteristic data.

[0252] The method for obtaining the fifth characteristic data in the above sixth title can refer to the method for obtaining the first characteristic data in the above third title or the second characteristic data in the fourth title, and no specific explanation is given here.

[0253] Step S804, for each multimedia file pair in the preset training set, input the fifth feature data of the two sample multimedia files in the multimedia file pair into the preset binary classification model, determine the sixth similarity between the two sample multimedia files in the multimedia file pair, and based on the sixth similarity, obtain a prediction result indicating whether the multimedia file pair is a duplicate multimedia file pair.

[0254] In the embodiment of the present invention, when the preset binary classification model is trained based on the fifth feature data corresponding to each multimedia file in the preset training set, the influence of each dimension feature data on the prediction result is different.

[0255] When the fifth characteristic data includes any one of the sequence words, date keywords or title keywords, the electronic device determines the prediction result based on the fifth characteristic data in combination with other characteristic data.

[0256] Step S805 , calculating the loss value of the preset binary classification model according to the identification information of each multimedia file pair and the prediction result of each multimedia file pair.

[0257] The loss value of the above-mentioned preset binary classification model can be calculated according to a variety of loss value functions, for example, a square error loss function, a mean square error loss function, etc. Here, the calculation of the above-mentioned loss value is not specifically described.

[0258] Step S806, when the loss value is greater than the preset loss value threshold, adjust the parameters of the preset binary classification model and return to execute the above step S804.

[0259] That is, when the loss value is greater than the preset loss value threshold, the step of returning to the above step of inputting the two fifth feature data corresponding to each multimedia file pair in the preset training set into the preset binary classification model, determining the sixth similarity between the two sample multimedia files in the multimedia file pair, and obtaining a prediction result indicating whether the multimedia file pair is a repeated multimedia file pair based on the sixth similarity. Here, the parameter adjustment method of the preset binary classification model is not specifically limited.

[0260] Step S807, when the loss value is not greater than the preset loss value threshold, the preset binary classification model at the current moment is determined as the trained binary classification model.

[0261] pass Figure 8 The method shown, since the above binary classification model can accurately distinguish feature data of multiple dimensions and combine features of different dimensions, this will enable the trained binary classification model to accurately identify duplicate multimedia files, thereby improving the accuracy of duplicate multimedia file identification.

[0262] Based on the same inventive concept, according to the method for identifying duplicate multimedia files provided by the above embodiment of the present invention, the embodiment of the present invention also provides a device for identifying duplicate multimedia files. Fig. 9 As shown, Fig. 9 A schematic diagram of the structure of a device for identifying duplicate multimedia files provided by an embodiment of the present invention. The device includes the following modules.

[0263] A first acquisition module 901, configured to acquire a first title of a first multimedia file and a second title of a second multimedia file;

[0264] The second acquisition module 902 is used to acquire first data to be eliminated in the first title; and / or acquire second data to be eliminated in the second title; the first data to be eliminated is the text content in the first title associated with the first user identification information corresponding to the first multimedia file, and the second data to be eliminated is the text content in the second title associated with the second user identification information corresponding to the second multimedia file;

[0265] A calculation module 903 is used to calculate the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated;

[0266] The determination module 904 is configured to determine whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold.

[0267] Optionally, the second acquisition module 902 may be specifically used to acquire first user identification information corresponding to the first multimedia file;

[0268] Acquire all multimedia files corresponding to the first user identification information as a first multimedia file set;

[0269] Acquire text content associated with the first user identification information from the title of each multimedia file included in the first multimedia file set as the first target text content;

[0270] The first target text content included in the first title is used as the first data to be eliminated in the first title.

[0271] Optionally, the second acquisition module 902 may be specifically used to acquire second user identification information corresponding to the second multimedia file;

[0272] Acquire all multimedia files corresponding to the second user identification information as a second multimedia file set;

[0273] Acquire text content associated with the second user identification information from the title of each multimedia file included in the second multimedia file set as the second target text content;

[0274] The second target text content included in the second title is used as the second data to be eliminated in the second title.

[0275] Optionally, the calculation module 903 may include:

[0276] A first acquisition submodule, used for acquiring first characteristic data in a third title;

[0277] A second acquisition submodule, used for acquiring second characteristic data in the fourth title;

[0278] The calculation submodule is used to calculate the similarity between the first feature data and the second feature data as the first similarity between the first multimedia file and the second multimedia file.

[0279] Optionally, the first characteristic data includes one or more of sequence words, date keywords or title keywords; the second characteristic data includes one or more of sequence words, date keywords or title keywords.

[0280] Optionally, the calculation submodule may be specifically used to calculate the second similarity between the sequence words in the third title and the sequence words in the fourth title if both the first feature data and the second feature data include sequence words; and / or,

[0281] If both the first feature data and the second feature data include date keywords, then calculating a third similarity between the date keywords in the third title and the date keywords in the fourth title; and / or,

[0282] If both the first feature data and the second feature data include title keywords, calculating a fourth similarity between the title keywords in the third title and the title keywords in the fourth title;

[0283] Based on the second similarity and / or the third similarity and / or the fourth similarity, a first similarity between the first multimedia file and the second multimedia file is calculated.

[0284] Optionally, the determination module 904 may be specifically configured to determine that the first multimedia file and the second multimedia file are a duplicate multimedia file pair if the first similarity is greater than a preset similarity threshold;

[0285] If the first similarity is not greater than the preset similarity threshold, it is determined that the first multimedia file and the second multimedia file are not a duplicate file pair.

[0286] Through the device provided by the embodiment of the present invention, the data to be removed from the titles of two multimedia files can be removed respectively, that is, the text content associated with the user identification information in the multimedia file titles can be removed, so as to determine whether the two multimedia files are duplicate media files according to the similarity between the two multimedia file titles after removing the data to be removed. The removal of the data to be removed from the titles of multimedia files can effectively reduce the influence of different user identification information in the titles on the similarity calculation, improve the accuracy of the calculated similarity, and thus improve the accuracy of duplicate multimedia file recognition. In addition, since the titles of duplicate multimedia files have a high degree of similarity, compared with the method for identifying duplicate multimedia files in the related art that maps feature data on different dimensions to the same vector space model, the accuracy of the similarity calculated only based on the feature data in the title in the embodiment of the present invention is higher, and the recognition result of the duplicate multimedia file determined based on the similarity is more accurate, thereby improving the accuracy of duplicate multimedia file recognition.

[0287] Based on the same inventive concept, according to the method for identifying duplicate multimedia files provided by the above-mentioned embodiment of the present invention, the embodiment of the present invention also provides an electronic device, such as Fig.10 As shown, it includes a processor 1001, a communication interface 1002, a memory 603 and a communication bus 1004, wherein the processor 1001, the communication interface 1002, and the memory 1003 communicate with each other through the communication bus 1004;

[0288] Memory 1003, used for storing computer programs;

[0289] The processor 1001 is used to execute the program stored in the memory 1003, and implements the following steps:

[0290] Obtain a first title of a first multimedia file and a second title of a second multimedia file;

[0291] Obtaining the first data to be removed from the first title; and / or,

[0292] Get the second data to be eliminated in the second title;

[0293] The first data to be removed is the text content in the first title associated with the first user identification information corresponding to the first multimedia file, and the second data to be removed is the text content in the second title associated with the second user identification information corresponding to the second multimedia file;

[0294] Calculating the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated;

[0295] According to the first similarity and a preset similarity threshold, it is determined whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair.

[0296] Through the electronic device provided by the embodiment of the present invention, the data to be removed from the titles of two multimedia files can be removed respectively, that is, the text content associated with the user identification information in the multimedia file titles can be removed, so as to determine whether the two multimedia files are duplicate media files according to the similarity between the two multimedia file titles after removing the data to be removed. The removal of the data to be removed from the titles of multimedia files can effectively reduce the influence of different user identification information in the titles on the similarity calculation, improve the accuracy of the calculated similarity, and thus improve the accuracy of duplicate multimedia file recognition. In addition, since the titles of duplicate multimedia files have a high degree of similarity, compared with the method for identifying duplicate multimedia files in the related art that maps feature data on different dimensions to the same vector space model, the accuracy of the similarity calculated only based on the feature data in the title in the embodiment of the present invention is higher, and the recognition result of the duplicate multimedia file determined based on the similarity is more accurate, thereby improving the accuracy of duplicate multimedia file recognition.

[0297] The communication bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0298] The communication interface is used for communication between the above terminal and other devices.

[0299] The memory may include a random access memory (RAM) or a non-volatile memory, such as at least one disk memory. Optionally, the memory may also be at least one storage device located away from the aforementioned processor.

[0300] The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU); or a special-purpose processor, including a network processor (NP), a digital signal processor (DSP), etc.

[0301] Based on the same inventive concept, according to the duplicate multimedia file identification method provided in the above-mentioned embodiments of the present invention, the embodiments of the present invention further provide a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the duplicate multimedia file identification method described in any of the above-mentioned embodiments is implemented.

[0302] Based on the same inventive concept, according to the duplicate multimedia file identification method provided in the above-mentioned embodiments of the present invention, the embodiments of the present invention also provide a computer program containing instructions, which, when executed on a computer, enables the computer to execute any of the duplicate multimedia file identification methods described in the above-mentioned embodiments.

[0303] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented by software, it can be implemented in whole or in part in the form of a computer program. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more available media integrated. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk Solid State Disk (SSD)), etc.

[0304] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0305] Each embodiment in this specification is described in a related manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, electronic device, computer-readable storage medium, and computer program embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0306] The above description is only a preferred embodiment of the present invention and is not intended to limit the protection scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention are included in the protection scope of the present invention.

Claims

1. A method for identifying duplicate multimedia files, characterized in that: The method comprises: Obtain a first title of a first multimedia file and a second title of a second multimedia file; Acquire the first data to be removed from the first title; and / or, Acquire the second data to be eliminated in the second title; The first data to be removed is text content in the first title that is associated with the first user identification information corresponding to the first multimedia file, and the second data to be removed is text content in the second title that is associated with the second user identification information corresponding to the second multimedia file; Calculating a similarity between a third title and a fourth title as a first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated; determining, according to the first similarity and a preset similarity threshold, whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair; The step of obtaining the first data to be removed from the first title includes: Acquire first user identification information corresponding to the first multimedia file; the first user identification information is user-adjustable information; Acquire all multimedia files corresponding to the first user identification information as a first multimedia file set; the multimedia files corresponding to the first user identification information are: multimedia files uploaded by the user indicated by the first user identification information; Acquire text content associated with the first user identification information from the title of each multimedia file included in the first multimedia file set as first target text content; Using the first target text content included in the first title as the first data to be removed in the first title; The step of obtaining the second data to be removed in the second title includes: Acquire second user identification information corresponding to the second multimedia file; the second user identification information is user-adjustable information; Acquire all multimedia files corresponding to the second user identification information as a second multimedia file set; Acquire text content associated with the second user identification information from the title of each multimedia file included in the second multimedia file set as second target text content; The second target text content included in the second title is used as the second data to be eliminated in the second title.

2. The method according to claim 1, characterized in that The step of calculating the similarity between the third title and the fourth title as the first similarity between the first multimedia file and the second multimedia file includes: Obtain the first characteristic data in the third title; Get the second characteristic data in the fourth title; The similarity between the first feature data and the second feature data is calculated as the first similarity between the first multimedia file and the second multimedia file.

3. The method according to claim 2, characterized in that The first characteristic data includes one or more of a sequence word, a date keyword or a title keyword; the second characteristic data includes one or more of a sequence word, a date keyword or a title keyword.

4. The method according to claim 3, characterized in that The step of calculating the similarity between the first feature data and the second feature data as the first similarity between the first multimedia file and the second multimedia file includes: If both the first feature data and the second feature data include the sequence word, then calculating the second similarity between the sequence word in the third title and the sequence word in the fourth title; and / or, If both the first feature data and the second feature data include the date keyword, then calculating a third similarity between the date keyword in the third title and the date keyword in the fourth title; and / or, If both the first feature data and the second feature data include the title keyword, calculating a fourth similarity between the title keyword in the third title and the title keyword in the fourth title; Based on the second similarity and / or the third similarity and / or the fourth similarity, a first similarity between the first multimedia file and the second multimedia file is calculated.

5. The method according to claim 1, characterized in that The step of determining whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold comprises: If the first similarity is greater than a preset similarity threshold, determining that the first multimedia file and the second multimedia file are a duplicate multimedia file pair; If the first similarity is not greater than the preset similarity threshold, it is determined that the first multimedia file and the second multimedia file are not a duplicate file pair.

6. A device for identifying duplicate multimedia files, characterized in that: The device comprises: A first acquisition module, used for acquiring a first title of a first multimedia file and a second title of a second multimedia file; A second acquisition module is used to acquire first data to be eliminated in the first title; and / or to acquire second data to be eliminated in the second title; the first data to be eliminated is text content in the first title associated with the first user identification information corresponding to the first multimedia file, and the second data to be eliminated is text content in the second title associated with the second user identification information corresponding to the second multimedia file; a calculation module, configured to calculate a similarity between a third title and a fourth title as a first similarity between the first multimedia file and the second multimedia file; the third title is the first title that does not include the first data to be eliminated, and the fourth title is the second title that does not include the second data to be eliminated; a determination module, configured to determine whether the first multimedia file and the second multimedia file are a duplicate multimedia file pair according to the first similarity and a preset similarity threshold; The second acquisition module is specifically used to acquire first user identification information corresponding to the first multimedia file; the first user identification information is user-adjustable information; Acquire all multimedia files corresponding to the first user identification information as a first multimedia file set; the multimedia files corresponding to the first user identification information are: multimedia files uploaded by the user indicated by the first user identification information; Acquire text content associated with the first user identification information from the title of each multimedia file included in the first multimedia file set as first target text content; Using the first target text content included in the first title as the first data to be removed in the first title; The second acquisition module is specifically used to acquire second user identification information corresponding to the second multimedia file; the second user identification information is user-adjustable information; Acquire all multimedia files corresponding to the second user identification information as a second multimedia file set; Acquire text content associated with the second user identification information from the title of each multimedia file included in the second multimedia file set as second target text content; The second target text content included in the second title is used as the second data to be eliminated in the second title.

7. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, for implementing the method steps described in any one of claims 1 to 5 when executing a program stored in a memory.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method steps described in any one of claims 1 to 5 are implemented.

Images