Intelligent sound image file arrangement method based on event knowledge graph

Abstract

The invention provides an intelligent sound image file arrangement method based on an event knowledge graph. The method comprises the following steps: acquiring sound image file data; for structured data in the sound image archive data, directly defining main attributes, other attributes and relationships of archive entities, and completing knowledge extraction; for semi-structured data in the sound image file data, knowledge extraction is completed through rule matching; for unstructured data in the sound image file data, converting audio and video data into unstructured text data, and intelligently extracting entities, relationships and events by utilizing an NLP algorithm and combining with a constructed file event Schema to complete knowledge extraction; and performing knowledge fusion on the knowledge extracted from the structured data, the semi-structured data and the unstructured data, and obtaining a triple based on the knowledge fusion to construct a visual knowledge graph. By constructing the event knowledge graph, archive personnel can be helped to catalog and index the archives more quickly and accurately.

Description

Technical Field

[0001] The present invention relates to the field of audio-visual archives, and in particular to an intelligent arrangement method for audio-visual archives based on event knowledge graphs. Background Art

[0002] Nowadays, people create and accumulate a large number of audio-visual archives, such as audio, video, and images. These archives are huge and diverse. Traditional audio-visual archive sorting methods usually rely on manual operations, which consumes a lot of time and manpower. Moreover, the accuracy and consistency of manual sorting are difficult to guarantee, making traditional manual sorting methods difficult to cope with. The information in audio-visual archives often needs to be labeled and classified for effective management and retrieval. However, due to the complexity and diversity of archival data, manual labeling and classification are inefficient and prone to errors. The most likely consequence of the contradiction between the rapidly growing number of audio-visual archives and the mismatched archivists is the one-sided understanding of archive users when using archives, which will also limit the openness of archive managers to the identification and use of archives.

[0003] At present, in addition to the traditional manual audio-visual archive sorting methods, there are also archive rapid retrieval methods based on face recognition, audio-visual archive recording methods based on image recognition, and audio-visual archive intelligent compilation and research methods based on knowledge graphs. These methods use technologies such as face recognition, OCR recognition, and image classification to analyze the content of archives, obtain corresponding metadata for presentation and retrieval, and have limited effects on the fine-grained sorting, association, and visualization of archives. In an era when the number of audio-visual archives is surging, archive management at the original archive file level can no longer meet the needs of archive users for the use of archive content. Summary of the invention

[0004] In view of the problems existing in the prior art, an intelligent organization method for audio-visual archives based on event knowledge graph is provided, which can help archivists catalog and index audio-visual archives more quickly and accurately.

[0005] The technical solution adopted by the present invention is as follows: a method for intelligently organizing audio-visual archives based on event knowledge graph, comprising: Acquiring audiovisual archive data; For structured data in audio-visual archive data, directly define the main attributes, other attributes and relationships of archive entities to complete knowledge extraction; For the semi-structured data in the audio-visual archive data, knowledge extraction is completed through rule matching; For the unstructured data in the audio-visual archive data, the audio and video data is converted into unstructured text data, and the NLP algorithm is combined with the constructed archive event Schema to intelligently extract entities, relationships and events to complete knowledge extraction; The knowledge extracted from structured data, semi-structured data and unstructured data is fused, and a visual knowledge graph is constructed based on the triples obtained after knowledge fusion.

[0006] As a preferred solution, after acquiring the audio-visual archive data, it also includes repairing the unqualified audio-visual images.

[0007] As a preferred solution, the knowledge extraction process of the structured data in the audio-visual archive data specifically includes: For the archive transfer and receiving directory, define the primary attributes, other attributes and relationships of the archive entity based on metadata; For other archive record excel, combine the main attributes in the transfer receiving directory to complete the definition of other attributes and relationships.

[0008] As a preferred solution, the knowledge extraction process of the semi-structured data in the audio-visual archive data specifically includes: For the XML files, Html files or JSON files in the archive transfer package, they are automatically disassembled into corresponding fields according to the metadata fields, and then combined with the attributes and relationship definitions in the structured data to output the rules.

[0009] As a preferred solution, the knowledge extraction process of the unstructured data in the audio-visual archive data specifically includes: Speech recognition of audio content of audio and video archives in audio and video archives, converting it into unstructured text data; Use the image attribute recognition algorithm to determine whether there is a picture or text screen, combine OCR to obtain the text information in the picture and text screen, and convert it into unstructured text data; The NLP algorithm is combined with the constructed archive event Schema to intelligently extract entity, relationship and event triples.

[0010] As a preferred solution, the unstructured text data can correspond to the time code of the audio and video in the audio-visual archive, and automatically extract the corresponding fragments and marking points. When extracting entities and events, the content fragments or marking points of the archive need to be stored as resource attributes of the entity or event.

[0011] As a preferred solution, after intelligently extracting entities, relationships and events, the entities, relationships and events in the archives professional field are labeled; the labeled triples are provided to the machine for learning to generate a labeling model of the archives field event knowledge graph; the unstructured data of the audio-visual archives is input into the labeling model of the archives field event knowledge graph, and the archives field event knowledge graph can be automatically generated.

[0012] As a preferred solution, the knowledge fusion includes: judging whether disambiguation is needed based on the labeled sample content, and when disambiguating, deduplicating the entity's main attribute values, other attributes and relationships based on the labeled sample content; at the event level, using an event co-reference judgment algorithm to complete event disambiguation.

[0013] As a preferred solution, a visual knowledge graph is constructed to provide associated recommended displays of node-related archives, fragments, and markers.

[0014] As a preferred solution, the specific method of displaying the associated recommendations includes: For audio-visual archives with annotation tasks, we use the relationship between entities in triples to implement entity-oriented recommendation strategies based on sequence and association matching, allowing search users to expand archive browsing according to the timeline or logical relationship of events. For audio-visual archives that have not yet been labeled, a recommendation mechanism based on content features is used to match and associate entities in existing triples to complete archive recommendations.

[0015] Compared with the existing technology, the beneficial effects of adopting the above technical solution are as follows: the present invention utilizes audio and video restoration technologies such as super-resolution and colorization to improve the recognizability of audio and video and increase the completeness of machine-extracted entities and events. In the process of labeling unstructured text data, a pre-trained language model with a small amount of labeled data is used. Under the condition of limited personnel in each unit, the data is processed in a process of machine labeling-manual labeling-model training-automatic labeling-manual correction-model correction-automatic labeling. In the overall knowledge graph construction process, knowledge is obtained to a large extent by applying machine extraction based on rules, and the dependence on machine learning is not strong. For user scenarios that do not have model training capabilities for the time being, this method is still applicable for archive organization. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is a flow chart of the method for intelligent organization of audio-visual archives based on event knowledge graph proposed in the present invention.

[0017] Figure 2 The figure is a flow chart of structured and semi-structured data extraction of audio-visual archives in one embodiment of the present invention.

[0018] Figure 3 The figure is a flow chart of extracting unstructured data from audio and video archives in one embodiment of the present invention. DETAILED DESCRIPTION

[0019] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar modules or modules with the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and cannot be construed as limitations on the present application. On the contrary, the embodiments of the present application include all changes, modifications and equivalents that fall within the spirit and connotation of the appended claims.

[0020] The existing audio-visual archive organization method technology has proposed some solutions to a series of problems such as time-consuming manual organization, error-prone labeling and classification, and limited ability of existing automated methods to process large-scale audio-visual data. However, no effective research and application has been conducted on the extraction of some intrinsically related data between archive contents and the improvement of the depth and breadth of archive utilization. Based on this, the embodiment of the present invention proposes an intelligent organization method for audio-visual archives based on an event knowledge graph. By converting events in the archives into nodes of the knowledge graph and abstracting the relationships and attributes between events into edges, a richly associated and semantically clear archive knowledge graph can be constructed to provide strong support for cataloging and indexing. By performing association analysis on the knowledge graph, the association patterns, trends and evolution laws between archives can be discovered, thereby revealing a deeper meaning and value of the archives. At the same time, after the construction of the archive event knowledge graph is completed, the browsing and navigation of the archives can be realized through the visual display of the knowledge graph, a more fine-grained retrieval method and more comprehensive retrieval results can be obtained, thereby improving the user's archive browsing experience. Please refer to Figure 1 , the specific plan is as follows: Step 1: Obtain audio and video archive data.

[0021] For the audio-visual archive data to be entered into the archive, it is necessary to evaluate the audio-visual quality. Repairing the unqualified audio-visual archive data mainly includes video repair, audio repair and image repair. Specifically, in this embodiment, video repair can adopt repair technologies such as denoising, super-resolution, and coloring; audio repair can adopt repair technologies such as noise reduction, Clicks and PopsRemoval, sound correction, Declicking, equalization (EQ), denoising; image repair can adopt repair technologies such as image enhancement, portrait enhancement, black and white photo coloring, scratch repair, and dark light enhancement. After the repair is completed, the subsequent archive sorting process can be carried out.

[0022] Step 2: Knowledge Extraction The audio and video archive data to be stored can generally be divided into structured data, semi-structured data and unstructured data. In this embodiment, corresponding knowledge extraction methods are provided for different types of data. When analyzing and extracting knowledge from audio and video archives, this embodiment provides support capabilities such as face recognition, face clustering, OCR recognition, entity recognition, scene recognition, speech recognition and voiceprint recognition.

[0023] 1. For structured data Please refer to Figure 2 For the archive transfer and reception directory, the main attributes, other attributes and relationships of the archive entity can be defined directly based on metadata to form triple knowledge. For other archive record excel, the main attributes in the transfer and reception directory are combined to complete the definition of other attributes and relationships to form triple knowledge. Among them, the archive transfer and reception directory, other archive record excel and other information will be provided by default when the docking system is to store archives.

[0024] For example, the main attribute of the archive entity can be "file number" to confirm its uniqueness, and the attributes can be metadata information such as "title", "responsible person", and "time". If the main attribute of the archive entity is "subject", its attributes can be "content description", "content start time" and "content end time", and the relationship between archives can be associated based on the overlap of "content description". The relationship between archives can be defined as the same collection, the same file, the same responsible person and related metadata. The definition here can be set according to the specific use of audio-visual archives by specific users. For example, the urban construction archives are more concerned about the association of archives of the same construction project, so the definition of the relationship can be the same project.

[0025] 2. For semi-structured data Please continue to refer to Figure 2 , for the XML files, Html files or JSON files in the archive transfer package, they are automatically disassembled into corresponding fields according to the metadata fields, and then combined with the attributes and relationship definitions in the structured data, triples are output according to the rules, where the rules are pre-prepared by domain experts. The conversion of semi-structured data and structured data is mainly obtained by mapping based on metadata fields. For example, the main attribute of the archive, "file number", can be obtained in the code corresponding to the XML file. Some fields in the XML data are stored in segments, and they need to be pieced together according to the corresponding rules. For example, the file number may be split into the archive number, category code, year, retention period code, file number, item number, etc., which requires adjustment according to specific rules according to specific circumstances.

[0026] 3. For unstructured data In this embodiment, the audio and video data is converted into unstructured text data, and the NLP algorithm is combined with the constructed archive event Schema to intelligently extract entities, relationships and events to complete knowledge extraction. Figure 3 , specifically: For the unstructured data in the audio and video archives, including audio, video, images, etc., the audio and video will be split into audio content and graphic information for processing. Use speech recognition to convert the audio content of the audio and video archives in the audio and video archives into unstructured text data; use the picture attribute recognition algorithm to determine whether there is a picture and text screen, and combine OCR to obtain the text information in the picture and text screen, and convert it into unstructured text data. In actual applications, text proofreading, scene classification and other technologies can also be used to assist in the conversion of audio and video into unstructured text data.

[0027] After obtaining the converted unstructured text, in this embodiment, the entity extraction, relationship extraction and event extraction algorithms in the NPL algorithm are used to construct the archive event Schema to complete the extraction of entity, relationship and event triples. Among them, the construction of the archive event Schema adopts a top-down strategy, which first requires the ontological definition of the knowledge graph (covering the conceptual model of entities, entity attributes and relationships between entities, that is, the architecture of the Schema); then, for the established ontological framework, the annotation work of the corpus text is performed (involving manual annotation and automatic annotation using information extraction technology, etc.); finally, the conversion of unstructured data to structured data is realized. In this method, the ontology construction of the archive event Schema is based on the general encyclopedia knowledge graph Schema, and the specific attributes of the archive events are integrated, such as the geographical location of the war event, the initiator of the uprising event, the scope of influence of the event, etc., to vertically expand the archive resources and deeply mine the knowledge, and build a knowledge organization system for archive events. The construction process supports dynamic optimization and gradual improvement according to the progress of annotation to form a comprehensive archive event Schema. The advantage of this method lies in its wide applicability, and it can quickly construct the corresponding domain schema for any archival materials containing narrative records.

[0028] The automatic annotation of the domain knowledge graph mainly depends on the model trained by the domain schema, but NLP cannot be directly combined with the schema. In this embodiment, it is proposed to use manual annotation to complete the annotation of entities, relationships and events in the archives professional field, and add some general entities that can be extracted by NLP to complete the process of manual + machine annotation. Finally, the model training platform performs model training based on the corpus content and triples to obtain a domain knowledge graph model based on the schema. After generation, the domain knowledge graph model can automatically annotate the annotated unstructured data and unannotated unstructured data to realize knowledge extraction, but the annotated content will not be annotated again, and only the unannotated content will be supplemented.

[0029] Furthermore, the unstructured text data converted from unstructured audio and video archives can be matched with the time code of the archive audio and video images, and the corresponding segments and markers can be automatically extracted. When extracting entities and events, the content segments or markers of the archives need to be stored as resource attributes of the entity or event, so as to be associated with the archives when the graph is finally displayed.

[0030] In actual applications, for photo files and files without voice data, this embodiment mainly uses face recognition technology or OCR technology to extract entities, and will not generate specific labeling tasks, but will only supplement existing entities.

[0031] Since only the unstructured automatic annotation process of the entire knowledge extraction process uses machine learning, in one embodiment, when there is no machine learning condition, the overall knowledge graph can be presented directly through manual annotation. The knowledge graph is then applied to the knowledge extraction process by machine according to rules.

[0032] Step 3: Knowledge Fusion In this embodiment, two knowledge fusion methods are mainly provided, including manual disambiguation and machine fusion. Manual disambiguation is mainly based on the judgment of the labeled sample content to confirm whether disambiguation is needed. During disambiguation, the main attribute values, other attributes, and relationships of the entity are deduplicated based on the labeled sample content to ensure reasonable disambiguation of the entity.

[0033] Machine fusion mainly focuses on the event disambiguation level, using event co-reference judgment algorithms to assist in confirming whether they are the same event and complete event disambiguation.

[0034] Step 4: Knowledge Application After completing the knowledge fusion, a visual knowledge graph can be constructed based on the obtained ternary elements to display the relationship between entities and support the expansion and exploration of entity relationships.

[0035] On the visualized node, it is possible to support the associated recommendation display of the files, fragments and marking points related to the node. In this embodiment, different recommendation schemes are provided for files with labeling tasks and files without labeling tasks. Specifically, For archives with annotation tasks (i.e., unstructured text is converted through speech recognition technology to form annotation tasks, and then triple knowledge is obtained through manual or automated annotation), the relationship between entities in the triples is used to execute entity-oriented recommendation strategies based on sequence and association matching, thereby allowing search users to conduct archive browsing according to the timeline or logical relationship of events.

[0036] For archives that have not been labeled (such as photo archives or archives without voice data), a recommendation mechanism based on content features is used. This mechanism matches and associates the face data extracted by face recognition technology, the key information identified by OCR technology, and the environmental information such as buildings identified by scene recognition technology with the entities in the existing triples to achieve archive recommendation.

[0037] Since voice annotated entities carry time code information (that is, when users annotate entities, the system not only stores the entity text, but also records the timestamp of the entity in the audio and video archive), the recommendation process can achieve accurate positioning of specific content frames, ensuring that archive recommendations are based on content relevance rather than simply individual archives, thereby improving the accuracy of recommendations and user experience.

[0038] An embodiment of the present invention further proposes a system, including a memory and a processor, wherein the memory stores a computer program corresponding to the aforementioned method for intelligent organization of audio-visual archives based on event knowledge graph, which can be loaded and executed by the processor.

[0039] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a computer readable medium, and the computer program includes a program code for executing the method shown in the flowchart.

[0040] It should be noted that the computer-readable medium shown in the embodiment of the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0041] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0042] The units involved in the embodiments described in this application may be implemented by software or hardware, and the units described may also be set in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0043] As another aspect, the present application also provides a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device reads the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the method for intelligently organizing audio-visual archives based on event knowledge graphs described in the above embodiment.

[0044] As another aspect, the present application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by an electronic device, the electronic device implements the method for intelligently organizing audio-visual archives based on event knowledge graphs described in the above embodiment.

[0045] It should be noted that, although several modules or units of the equipment for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present application, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0046] Through the description of the above implementation methods, it is easy for those skilled in the art to understand that the example implementation methods described here can be implemented by software, or by combining software with necessary hardware. Therefore, the technical solution according to the implementation method of the present application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation method of the present application.

[0047] It should be noted that in the description of the embodiments of the present invention, unless otherwise clearly specified and limited, the terms "setting" and "connection" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a direct connection, or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances; the drawings in the embodiments are used to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings herein can be arranged and designed in various different configurations.

[0048] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for intelligently organizing audio-visual archives based on event knowledge graph, characterized in that: include: Acquiring audiovisual archive data; For structured data in audio-visual archive data, directly define the main attributes, other attributes and relationships of archive entities to complete knowledge extraction; For the semi-structured data in the audio-visual archive data, knowledge extraction is completed through rule matching; For the unstructured data in the audio-visual archive data, the audio and video data is converted into unstructured text data, and the NLP algorithm is combined with the constructed archive event Schema to intelligently extract entities, relationships and events to complete knowledge extraction; The knowledge extracted from structured data, semi-structured data and unstructured data is fused, and a visual knowledge graph is constructed based on the triples obtained after knowledge fusion.

2. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 1 is characterized in that: After obtaining the audio-visual archive data, it also includes repairing the unqualified audio-visual images.

3. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 1 or 2, characterized in that: The knowledge extraction process of the structured data in the audio-visual archive data specifically includes: For the archive transfer and receiving directory, define the primary attributes, other attributes and relationships of the archive entity based on metadata; For other archive record excel, combine the main attributes in the transfer receiving directory to complete the definition of other attributes and relationships.

4. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 3 is characterized in that: The knowledge extraction process of the semi-structured data in the audio-visual archive data specifically includes: For the XML files, Html files or JSON files in the archive transfer package, they are automatically disassembled into corresponding fields according to the metadata fields, and then combined with the attributes and relationship definitions in the structured data to output the rules.

5. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 1 or 2, characterized in that: The knowledge extraction process of the unstructured data in the audio-visual archive data specifically includes: Speech recognition of audio content of audio and video archives in audio and video archives, converting it into unstructured text data; Use the image attribute recognition algorithm to determine whether there is a picture or text screen, combine OCR to obtain the text information in the picture and text screen, and convert it into unstructured text data; The NLP algorithm is combined with the constructed archive event Schema to intelligently extract entity, relationship and event triples.

6. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 5 is characterized in that: The unstructured text data can correspond to the time code of the audio and video in the audio-visual archive, and automatically extract the corresponding fragments and marking points. When extracting entities and events, the content fragments or marking points of the archive need to be stored as resource attributes of the entity or event.

7. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 5 is characterized in that: After intelligently extracting entities, relations and events, the entities, relations and events in the professional field of archives are labeled; The labeled triples are provided to the machine for learning to generate a labeling model for the archives domain event knowledge graph; the unstructured data of the audio-visual archives is input into the labeling model of the archives domain event knowledge graph to automatically generate the archives domain event knowledge graph.

8. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 2 is characterized in that: The knowledge fusion includes: judging whether disambiguation is needed based on the labeled sample content, and when disambiguating, deduplicating the main attribute values, other attributes and relationships of the entity based on the labeled sample content; at the event level, using an event co-reference judgment algorithm to complete event disambiguation.

9. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 1, characterized in that: In constructing a visual knowledge graph, it provides associated recommendations for node-related archives, fragments, and markers.

10. The method for intelligently organizing audio-visual archives based on event knowledge graph according to claim 9, characterized in that: The specific method of displaying the associated recommendation includes: For audio-visual archives with annotation tasks, we use the relationship between entities in triples to implement entity-oriented recommendation strategies based on sequence and association matching, allowing search users to expand archive browsing according to the timeline or logical relationship of events. For audio-visual archives that have not yet been labeled, a recommendation mechanism based on content features is used to match and associate entities in existing triples to complete archive recommendations.

Images