Information generation method and device, computer readable storage medium and computer program product

Abstract

The embodiment of the invention discloses an information generation method and device, a computer readable storage medium and a computer program product. The method comprises the steps that content retrieval information input by a user is received; determining a matching node from the target structure tree based on the content retrieval information; wherein the target structure tree is a structure tree constructed for the target work; the target structure tree comprises a plurality of nodes and an edge structure representing an association relationship among the nodes; each node comprises an identifier of the node, and the starting time, the ending time and the content summary of the content corresponding to the node; based on the content retrieval information, obtaining a target node having an association relationship with the matching node from the target structure tree; and based on the matching node, the target node and the association relationship, generating target information for the content retrieval information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of information processing, and in particular to an information generation method, device, computer readable storage medium and computer program product. BACKGROUND

[0002] In the technical field of information processing, retrieval and generation of structured content is an important means to improve user interaction experience, especially in content involving complex narrative structure, such as films, literary works, etc. How to quickly and accurately locate and generate relevant information becomes a research focus. In the prior art, a tree structure based on domain knowledge is usually constructed to organize content, and keywords are used to query information in the tree structure, but in actual application, this structure can only retrieve information fragments related to the keywords, and cannot provide complete cause and effect information, reducing retrieval accuracy and retrieval quality. SUMMARY

[0003] To solve the above technical problems, the embodiments of the present application provide an information generation method, device, computer readable storage medium and computer program product, which can solve the problem of low retrieval accuracy and retrieval quality in related technologies.

[0004] The technical solution of the present application is implemented as follows: An information determination generation method, the method comprising: receiving retrieval information input by a user; determining a matching node from a target structure tree based on the retrieval information; wherein the target structure tree is a structure tree constructed for a target work; the target structure tree comprises a plurality of nodes and edge structures representing the association relationship between the nodes; each node comprises an identifier of the node, a start time, an end time and a content summary of the content corresponding to the node; obtaining a target node having the association relationship with the matching node from the target structure tree based on the retrieval information; generating target information for the retrieval information based on the matching node, the target node and the association relationship.

[0005] In the above solution, the determination of the matching node from the target structure tree based on the retrieval information comprises: performing semantic recognition on the retrieval information to obtain a work name, intent information and a keyword; determining the target structure tree from a to-be-screened structure tree based on the work name; wherein the to-be-screened structure tree comprises a plurality of structure trees corresponding to works; determining the matching node from the target structure tree based on the intent information.

[0006] In the scheme, the matching node is determined from the target structure tree based on the intention information, including: matching the intention information with a content outline of each node in the target structure tree to obtain a matching result; determining the matching node from the target structure tree based on the matching result.

[0007] In the scheme, the matching node is determined from the target structure tree based on the matching result, including: if the matching result indicates that there is a target content outline matching the intention information in the target structure tree, determining the node corresponding to the target content outline as the matching node; if the matching result indicates that there is no target content outline in the target structure tree, performing semantic matching in a target database based on the intention information to determine a node corresponding to target description information matching the intention information; wherein the target database includes the identification of each node in the target structure tree and the description information corresponding to each node; determining the node corresponding to the target description information matching the intention information as the matching node.

[0008] In the scheme, the target node having the association relationship with the matching node is obtained from the target structure tree based on the search information, including: traversing the target structure tree from the matching node as a starting point according to a target traversal strategy based on the keywords in the search information; stopping the traversal of the target structure tree according to a traversal termination rule to obtain the target node; wherein the traversal termination rule is used to determine that the obtained node information has constituted a complete explanation of the search information.

[0009] In the scheme, the information generation method further includes: for each play unit in the target work, constructing a plurality of nodes for the play unit; determining the association relationship between the plurality of nodes of the plurality of play units; wherein the association relationship includes at least one of a cause-effect relationship and a time sequence relationship; based on the plurality of nodes of the plurality of play units and the association relationship, constructing a target structure tree for the target work.

[0010] In the scheme, for each play unit in the target work, a plurality of nodes for the play unit are constructed, including: for each play unit, extracting voice information of the play unit and converting it into text information with a timestamp; Split the text information based on content categories to obtain a plurality of sub-text information; For each sub-text information, a node is constructed.

[0011] An information generation device, the device comprising: A receiving unit configured to receive search information input by a user; A processing unit configured to determine a matching node from a target structure tree based on the search information, wherein the target structure tree is a structure tree constructed for a target work; the target structure tree comprises a plurality of nodes and edge structures representing association relationships between the nodes; each node comprises an identifier of the node, a start time, an end time and a content summary of content corresponding to the node; The processing unit is further configured to obtain a target node having the association relationship with the matching node from the target structure tree based on the search information; A generating unit configured to generate target information for the search information based on the matching node, the target node and the association relationship.

[0012] An information generation device, the device comprising: a processor, a memory and a communication bus; The communication bus is configured to realize communication connection between the processor and the memory; The processor is configured to execute an information generation program stored in the memory to realize the steps of the above information generation method.

[0013] A computer readable storage medium, the storage medium storing one or more programs, the one or more programs being executable by one or more processors to realize the steps of the above information generation method.

[0014] A computer program product comprising a computer program, the computer program realizing the steps of the above information generation method when executed by a processor.

[0015] The information generation method, device, computer readable storage medium and computer program product provided by the embodiments of the present application first receive search information input by a user, then determine a matching node from a target structure tree based on the search information, the target structure tree comprising a plurality of nodes and edge structures representing the association relationship between the nodes, each node comprising the identity of the node, the start time and end time of the content corresponding to the node and the content summary, the target structure tree being a structure tree constructed for a target work, further acquire a target node having an association relationship with the matching node from the target structure tree based on the search information, and then generate target information for the search information based on the matching node, the target node and the association relationship. In this way, by constructing the target structure tree comprising the node association relationship, the upstream (antecedent) and downstream (consequence) target nodes associated with the matching node can be automatically traced back, so that complete and coherent content can be generated, rather than only being able to retrieve information fragments related to keywords as in the related art, and being unable to provide complete antecedent and consequence information, thereby not only improving the search precision and search quality, but also enhancing the depth of information acquisition and user experience. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 A flowchart of an information generation method provided by the embodiments of the present application; Figure 2 A flowchart of another information generation method provided by the embodiments of the present application; Figure 3 A structure diagram of an information generation device provided by the embodiments of the present application; Figure 4 A structure diagram of an information generation device provided by the embodiments of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application.

[0018] It should be understood that the specific embodiments described herein are merely intended to explain the present application, and are not intended to limit the present application.

[0019] It should be noted that when the audience is brushing short videos, they can often brush the plot fragments. When interested in the fragments and want to understand the cause and effect, it is found that a large number of TV series episodes need to be watched to understand the complete cause and effect, which will consume a lot of time, which is unacceptable for many audiences. How can we save time and effort to quickly understand the complete plot that the audience wants to query? The related technology mentions a method based on user input keywords and description information, which associates the video index with the description information and returns the indexed video for the user to watch. The above scheme can help users retrieve videos to a certain extent, but it is more of a video retrieval method. Through existing information, the corresponding video information is queried, but if the user wants an introduction to the cause and effect of the specified plot fragment, or why the plot fragment he saw happened, it is not meaningful to find the video corresponding to the keyword (as a user, he prefers to get a plot explanation rather than the plot fragment itself). And due to the limitation of the index itself, an index points to a video, and the fragment information obtained by the user generally belongs to a set of TV series. When querying by this method, it is difficult to achieve cross-set retrieval, resulting in low retrieval accuracy.

[0020] Therefore, in view of the problem in the related art that the complete plot cause and effect cannot be retrieved through video fragment information, the present application splits the target work into multiple nodes to form a complete structure tree for the target work. When a user queries fragment information or asks about the cause of the fragment, a relatively complete cause and effect content tree is given, and a video containing several content nodes is generated, i.e. the video fragment discovered by the user through the short video platform such as TikTok or Kuaishou. By using the method provided by the present application, the complete cause and effect of the content fragment of the target work can be queried from the application containing the target work, as follows: The embodiment of the present application provides an information generation method, which can be applied to an information generation device. Referring to Figure 1 The method includes the following steps: Step 101, receiving the retrieval information input by the user.

[0021] In the embodiment of the present application, the retrieval information refers to the text form of content input by the user to find the related content in the target work, and the retrieval information can specifically include the work name, the intention information and the keyword. The user can input the retrieval information through voice input, keyboard input or touch input, and the input method can be flexibly selected according to different application scenarios.

[0022] In a feasible implementation manner, in a video playing application, the user can click the screen and input the keyword during watching; in a voice recognition system, the user can directly say the keyword, and the system automatically converts the keyword into a text form.

[0023] Step 102, determining a matching node from the target structure tree based on the search information.

[0024] The target structure tree is a structure tree constructed for the target work; the target structure tree includes multiple nodes and edge structures representing the association relationship between the nodes; each node includes the identification of the node, the start time and end time of the content corresponding to the node, and a content summary.

[0025] In the embodiments of the present application, the target structure tree is a structured data model constructed for the target work, used to describe the relationship between the content segments in the target work, and each node in the target structure tree is an abstract description of a content segment in a playback unit in the target work, and can specifically include the start time of the content segment, the end time of the content segment, and a content summary of the content segment, wherein the content summary of the content segment can specifically include the episode number (i.e. which episode it belongs to), the abstract (a textual summary of the content segment), the role list (i.e. the roles involved in the content segment), and the event list (i.e. the key events involved in the content segment); the edge structure in the target structure tree represents the logical relationship between different content segments, such as causal relationship, time sequence relationship, etc.

[0026] In the embodiments of the present application, since each node in the target structure tree has a content summary, the intent information included in the search information and each content summary can be matched, and when the intent information matches a content summary, the node corresponding to the content summary is determined as the matching node; it should be noted that before the user searches, multiple structure trees corresponding to works (each work generates a structure tree) can be constructed in advance to form the structure trees to be screened, for the user to search.

[0027] In a feasible implementation manner, the target work can be a TV series, and the target structure tree can be a plot tree (i.e. a tree-like node structure generated according to the association relationship after the plot is split), then each node (i.e. a plot node) in the target structure tree can correspond to a plot segment in an episode of the TV series, and the node records the time (i.e. the start time and the end time) when the plot segment occurs and a brief description (i.e. a plot summary) of the plot segment; the edge structure represents the association between multiple plot segments, such as a plot segment leading to the occurrence of another plot segment, or two plot segments developing in parallel in the same time period.

[0028] Step 103, obtaining a target node having an association relationship with the matching node from the target structure tree based on the search information.

[0029] In the embodiments of the present application, the target node is another node having an association relationship with the matching node, which is obtained from the target structure tree based on the keywords in the search information on the basis of the matching node; the association relationship refers to a logical connection between different nodes in the target structure tree, such as a cause-effect relationship, a time sequence relationship, and the like.

[0030] In the embodiments of the present application, the target structure tree can be traversed with the matching node as a traversal starting node, and a node having an association relationship with the matching node is obtained as a target node in the process of traversal according to the keywords in the search information; in a feasible implementation manner, if the matching node is a plot segment about a teacher, the target node can be another plot segment having a cause-effect relationship or a time sequence relationship with the plot segment corresponding to the matching node, for example, the teacher assigning homework can be the matching node, and the student completing homework can be the target node, and the two nodes of the teacher assigning homework and the student completing homework have a cause-effect relationship, so the two nodes of the teacher assigning homework and the student completing homework can be taken as a part of the target information together.

[0031] It should be noted that by obtaining the target node having an association relationship with the matching node, more rich and comprehensive information can be provided based on the target nodes to meet the user's demand for in-depth understanding of the content of the target work.

[0032] In step 104, target information for the search information is generated based on the matching node, the target node, and the association relationship.

[0033] In the embodiments of the present application, the target information refers to the final output result generated according to the matching node, the target node, and the association relationship, and the target information can be a link of a video including complete content, or can be text information fed back to the search information; the target information aims to answer the user's query demand and provide detailed related content.

[0034] In the embodiments of the present application, the matching node and the target node obtained above can be out of order in time, so the matching node and the target node can be sorted in a certain logical order according to the association relationship between the nodes and the time stamp (i.e., the start time and the end time) of the nodes, to obtain nodes arranged in a certain logical order, the specific content in the target work is located according to the information (identifier, start time, and end time) included in each node in the sorted list, and the complete content is generated according to the multiple specific contents, and then a piece of text description can be generated for the complete content and fed back to the user, or a link for the complete content can be generated and fed back to the user, of course, the generated complete content can also be directly fed back to the user, to help the user quickly understand the content development context in the target work.

[0035] It should be noted that the user can check the "generate video" option in the interface of the information generation device, if the user checks the "generate video" option, a video is generated according to the start time and end time of the extracted node data, and the link is returned to the client dialog box; if the user does not check the "generate video" option, a text feedback question is generated according to the extracted content node.

[0036] It should be noted that in the case of a target work being a TV series, the present application splits each episode of a TV series into multiple plot nodes, and through the association relationship of each node, the nodes in the previous and next episodes are associated into a complete plot tree, and the present application generates a complete plot video (i.e. a collection of several plot nodes) or a description information of a complete plot when searching, for the user to watch, rather than only some video related to the search information as in the related art, thereby improving the accuracy and relevance of information search.

[0037] The information generation method provided by the embodiment of the present application can automatically trace the associated upstream (cause) and downstream (consequence) target nodes from the matching node by constructing a target structure tree containing node association relationship, thereby generating complete and coherent content, rather than only being able to retrieve information fragments related to keywords as in the related art, which cannot provide complete cause and effect information, thereby not only improving the search precision and search quality, but also enhancing the depth of information acquisition and user experience.

[0038] Based on the foregoing embodiment, the embodiment of the present application provides another information generation method, which refers to Figure 2 The method comprises the following steps: Step 201, the information generation device receives the search information input by the user.

[0039] It should be noted that a search dialog box can be provided in the information generation device for the user to input search information, and after the user inputs the search information, the search result (i.e. target information) is output.

[0040] Step 202, the information generation device performs semantic recognition on the search information to obtain a work name, intent information and keywords.

[0041] In the embodiment of the present application, the semantic recognition module can be used to perform semantic recognition on the search information, and the semantic recognition module is a natural language processing technology, and in the present application, the semantic recognition module is used to extract the work name, intent information and keywords from the search information provided by the user.

[0042] In the embodiments of the present application, the semantic recognition module can analyze the search information input by the user and extract information with semantic meaning (i.e., title, intent information, and keywords). In this way, by converting the original input of the user into structured information, subsequent precise matching operations can be implemented, thereby improving the accuracy and efficiency of system response.

[0043] In a feasible implementation manner, the semantic recognition module can be a deep learning model, such as a bidirectional encoder representations from transformers (BERT) model, a self-attention mechanism model (such as a Transformer), or the like. The semantic recognition module can accurately extract the title of the work, the intent information expressed by the user, and the keywords. For example, when the search information input by the user is “How does Sun Wukong beat the fairy in Journey to the West”, the semantic recognition module can extract “Journey to the West” as the title of the work, “How does Sun Wukong beat the fairy” as the intent information, and “Sun Wukong”, “fairy”, and “how” as the keywords.

[0044] Step 203: The information generation device determines a target structure tree from the to-be-screened structure trees based on the title of the work.

[0045] The to-be-screened structure trees include structure trees corresponding to multiple works.

[0046] In the embodiments of the present application, the to-be-screened structure trees refer to a pre-constructed set of structure trees corresponding to multiple works, and each structure tree represents the content organization manner of a work. When the user provides the title of the work, the corresponding structure tree can be found and selected from the to-be-screened structure trees as the target structure tree based on the title of the work. In this way, the specific work structure that the user is interested in can be quickly located, and invalid search in irrelevant works can be avoided, thereby improving the search efficiency and user interaction experience.

[0047] In a feasible implementation manner, when the to-be-screened structure trees include structure trees corresponding to multiple works such as “Dream of the Red Chamber”, “Romance of the Three Kingdoms”, and “Journey to the West”, the structure tree corresponding to “Journey to the West” (i.e., the target structure tree) can be automatically located when the title of the work “Journey to the West” is obtained from the search information input by the user.

[0048] Step 204: The information generation device determines a matching node from the target structure tree based on the intent information.

[0049] In the embodiment of the present application, the intention information refers to the core demand expressed by the user in the content retrieval process, for example, the behavior of finding a certain role, understanding the development of a certain plot, etc.; based on the intention information of the user, a node in the target structure tree that matches the intention information is found, and the found node is marked as a matching node. Specifically, the matching node can be determined according to the matching degree (i.e., the similarity degree) between the intention information expressed by the user and the content summary of each node of the structure tree in the target structure tree. In this way, the matching node is determined through the intention information of the user, which can ensure that the determined matching node is closer to the real demand of the user, thereby improving the relevance and accuracy of the returned results and improving the user experience.

[0050] In a feasible implementation manner, if the intention information is "how did Sun Wukong defeat Baijuejing", a node related to "Sun Wukong fighting Baijuejing" can be found from the corresponding structure tree of "Journey to the West", and the node with the highest matching degree is taken as the matching node.

[0051] It should be noted that step 204 can be implemented in the following way: Step 204A, the information generation device matches the intention information with the content summary of each node in the target structure tree to obtain a matching result.

[0052] In the embodiment of the present application, the content summary refers to a highly summarized description of the corresponding plot content in each node, which can be used to quickly judge whether the plot corresponding to the content summary matches the intention information described by the current user; the matching result refers to the similarity score or matching state obtained by comparing the intention information with the content summary of each node.

[0053] Step 204B, the information generation device determines a matching node from the target structure tree based on the matching result.

[0054] In the embodiment of the present application, the selection criteria of the matching node can include the highest matching score, the most recent matching time, the highest matching frequency, etc., which can be configured according to the actual application scenario; in a feasible implementation manner, the node corresponding to the content summary with the highest matching score with the intention information can be selected from the target structure tree as the matching node.

[0055] It should be noted that step 204B can be implemented through step 204b1, or through steps 204b2-204b3: Step 204b1, if the matching result indicates that there is a target content summary in the target structure tree that matches the intention information, the information generation device determines the node corresponding to the target content summary as the matching node.

[0056] In the embodiment of the present application, the target content summary refers to the content summary with the highest matching degree with the intention information; when the intention information of the user is determined from the search information input by the user, it can be first checked whether there is a target content summary in the target structure tree that matches the intention information, and if a target content summary that matches the intention information and has the highest matching degree is found in the target structure tree, the node corresponding to the target content summary can be directly marked as a matching node, so as to quickly locate the related content path.

[0057] It should be noted that the introduction of the matching mechanism of the target content summary in the target structure tree can efficiently identify the association between the user intention and the existing content summary, and the introduction of the matching mechanism of the target content summary in the target structure tree can improve the accuracy of intention recognition. By improving the accuracy of intention recognition, unnecessary search paths can be reduced, and by reducing unnecessary search paths, the response speed and user experience can be improved.

[0058] In step 204b2, if the matching result indicates that there is no target content summary in the target structure tree, the information generation device performs semantic matching in the target database based on the intention information to determine the node corresponding to the target description information matching the intention information.

[0059] In the target database, the identification of each node in the target structure tree and the description information corresponding to each node are included.

[0060] In step 204b3, the information generation device determines that the node corresponding to the target description information matching the intention information is a matching node.

[0061] In the embodiment of the present application, the target database is an independent data storage module for saving detailed information of all nodes in the target structure tree, including the unique identification of each node in the target structure tree and the detailed information of each node. These description information is usually in the form of natural language text for describing the plot, theme or function represented by the node. Semantic matching refers to determining whether there is semantic correlation between the user intention and the description information in the target database through natural language processing technology (such as NLP and semantic similarity calculation). The matching node is a node with the highest matching degree, that is, only one matching node can be determined.

[0062] In the embodiment of the present application, when there is no target content summary in the target structure tree that directly matches the intention information, the target database can be turned to for more extensive semantic matching. Once the matching target description information is found in the target database, the node corresponding to the target description information is determined and is determined as a matching node. It should be noted that this way can still provide reasonable plot guidance in the case of no direct matching in the target structure tree.

[0063] It should be noted that the preliminary matching through the combination of the target content summary and the semantic matching in the target database constructs a multi-level and high-precision intent recognition mechanism. Specifically, the first stage uses the target content summary in the target structure tree for rapid rough screening to quickly locate the possible relevant Jireny nodes; in the second stage, when the target content summary fails to be matched, the target database is accessed and the semantic matching technology is used to further refine the matching result, so as to ensure that the finally determined matching node has higher accuracy and rationality. The hierarchical processing manner not only improves the response efficiency of the system, but also enhances the understanding ability of the hierarchical processing manner to complex intentions.

[0064] In step 205, the information generation device traverses the target structure tree according to a target traversal strategy based on the keywords in the search information, taking the matching node as the starting point.

[0065] In the embodiments of the present application, the keywords can include causal keywords and time sequence keywords, wherein the causal keywords refer to words expressing the causal relationship between events, such as because, therefore, lead to, due to, etc., and the causal keywords are used to identify the logical connection between events in the content, so that the node related to the search information can be more accurately located; the time sequence keywords refer to words expressing the time sequence or the chronological relationship of events, such as after, before, at the same time, subsequently, etc., and by identifying these words, the timeline of the content can be better understood, and the traversal path can be adjusted accordingly; the target traversal strategy refers to selecting a suitable traversal method to search the target structure tree according to the content and context of the search information, and specifically can include Depth-First Search (DFS) (i.e. traversing the nodes in the tree vertically), Breadth-First Search (BFS) (i.e. traversing the nodes in the tree horizontally), priority search by weight ordering, etc., and different traversal strategies are suitable for different application scenarios. In a feasible implementation manner, when traversing the target structure tree, breadth-first search or depth-first search can be used, starting from the matching node, and taking the time sequence keywords as the time clues, such as "after", "next", "development" which need subsequent video explanation class semantics keywords, to preferentially traverse the child nodes; breadth-first search or depth-first search can also be used, starting from the matching node, and taking the causal keywords as the causal clues, such as "why", "reason", etc., to preferentially traverse the parent nodes.

[0066] In the embodiments of the present application, by taking the keywords as the guide, the part related to the keywords can be quickly located, and the entire structure tree can be efficiently explored according to the target traversal strategy. In this way, the search efficiency can be improved, and the returned result can be closer to the actual needs of the user.

[0067] Step 206, the information generation device stops the traversal of the target structure tree according to the traversal termination rule, and obtains the target node.

[0068] The traversal termination rule is used to determine that the obtained node information has constituted a complete explanation of the search information.

[0069] In the embodiments of the present application, the target node can refer to a necessary node set determined according to the context, that is, the target node can be one or more; the traversal termination rule is a set of preset conditions or thresholds, which is used to determine whether enough information has been collected to answer the user's query. Specifically, these rules can be set according to various factors, for example, when the number of found nodes reaches a certain upper limit, when all possible causal chains have been traversed, or when some key nodes are found and these key nodes meet certain conditions, and once the condition meeting the termination rule is detected, the further traversal operation is immediately stopped, and all the nodes found at present are determined as the target node; in a feasible implementation manner, when the depth traversal exceeds 5 nodes, the traversal is stopped.

[0070] It should be noted that the mechanism of stopping the traversal when the condition meeting the termination rule is detected helps to prevent resource waste caused by excessive search, reduce unnecessary computing overhead, and at the same time ensure the timeliness and accuracy of the user to obtain the result.

[0071] Step 207, the information generation device generates target information for the search information based on the matching node, the target node and the association relationship.

[0072] It should be noted that the specific construction manner of each target structure tree in the to-be-screened structure tree can be implemented by the following manner: Step 208, for each play unit in the target work, the information generation device constructs a plurality of nodes for the play unit.

[0073] In the embodiments of the present application, the play unit refers to a basic time segment constituting the target work, and the play unit can be independently identified, processed and associated with other play units; each play unit can generate one or more nodes; in a feasible implementation manner, when the target work is a TV series, the play unit can refer to an episode of the TV series, and the role of the node is to structurally represent the plot unit in the target work, and establish association (such as causality, time sequence, etc.) with other nodes through edges (relation), together constituting a plot tree (plot tree) for supporting cross-episode, cross-fragment semantic search and video generation.

[0074] It should be noted that step 208 can be implemented by the following manner: Step 208C, for each play unit, the information generation device extracts the voice information of the play unit and converts it into timestamped text information.

[0075] In the embodiments of the present application, processing the voice information includes converting the voice information into text and adding time to the converted text; that is, the text information is text information with time; the voice information refers to sound data obtained through an audio acquisition device, and then the voice information is further converted into text content through a voice recognition module. After obtaining the text content, the text content needs to be aligned with the actual time point of the play unit to generate text information with a timestamp (including start time and end time).

[0076] In a feasible implementation, when the play unit is an episode of a TV series, the voice extraction can be used to extract the narration information and dialogue information in the episode, and then the voice-to-text processing is performed to obtain the text content. Then, the video commentary tool can be used to match the text content with the time point in the episode to generate details with the time point. It should be noted that in the related art, due to the lack of narration information, the dialogue alone cannot cover all the content. Therefore, the present application inserts the missing narration information into the generated dialogue text according to the script itself.

[0077] Step 208D, the information generation device splits the text information based on the content category to obtain a plurality of sub-text information.

[0078] In the embodiments of the present application, the content category can be classified according to the main content of the content; splitting the text information based on the content category can divide the text information of one play unit into a plurality of sub-text information with finer granularity according to the theme, thereby facilitating the generation and management of subsequent nodes; it should be noted that the text information splitting method based on the content category helps to enhance the context understanding ability.

[0079] Step 208E, the information generation device constructs a node for each sub-text information.

[0080] In the embodiments of the present application, each node can include attributes such as timestamp and content summary. The start time, end time and content summary can be integrated together to create a complete data object (i.e. node) for each sub-text information, thereby realizing the structured organization and dynamic management of the voice information content.

[0081] It should be noted that the above first acquires original voice information through the voice acquisition device, and generates preliminary text information through voice recognition technology; then the content semantics are divided according to the content semantics, and multiple sub-text information is generated; finally, a node is created for each sub-text information to form a structured data unit, so that the entire execution process not only improves the accuracy of voice recognition and text information processing, but also provides clear data support for subsequent semantic analysis and context understanding.

[0082] It should be noted that an index can also be added to each node, which is usually the keyword of the node. Specifically, the keyword of the node can be extracted through an artificial intelligence model after the node is generated. During actual query, only matching with the node index can greatly improve the video retrieval efficiency.

[0083] Step 209, the information generation device determines the association relationship between the multiple nodes of the multiple playing units.

[0084] The association relationship includes at least one of a cause-effect relationship and a time sequence relationship.

[0085] In the embodiment of the present application, the cause-effect relationship refers to that the event of one node is the cause or result of the event of another node; the time sequence relationship refers to that two nodes have a chronological order on the story timeline, but there may be no direct cause and effect; a large language model (LLM) can be used to analyze multiple nodes to determine the association relationship between the nodes. It should be noted that the association relationship can also include reference relationship, parallel relationship (different places at the same time) and other relationships.

[0086] It should be noted that when the target work is a TV series, the playing unit is an episode of the TV series, and in some plot scenes, when there are two episodes of the TV series with large content differences, the node association caused by the two episodes cannot be fully associated, at this time, the association relationship can be manually edited by human.

[0087] Step 210, the information generation device constructs a target structure tree for the target work based on the multiple nodes and the association relationship of the multiple playing units.

[0088] In the embodiment of the present application, the multiple nodes can be connected through the association relationship to form a target structure tree, and the target structure tree not only helps the visual display of the content, but also serves as the basis for subsequent analysis, retrieval and reasoning.

[0089] It should be noted that the descriptions of the same steps and contents in this embodiment and other embodiments can refer to the descriptions in other embodiments, which will not be repeated here.

[0090] The information generation method provided by the embodiments of the present application can automatically trace the associated upstream (antecedent) and downstream (consequence) target nodes of the matching node from the matching node by constructing the target structure tree containing the node association relationship, thereby generating complete and coherent content, rather than only retrieving information fragments related to the keyword as in the related art, which cannot provide complete antecedent and consequence information, thereby not only improving the retrieval accuracy and quality, but also enhancing the depth of information acquisition and user experience.

[0091] Based on the foregoing embodiments, the embodiments of the present application provide an information generation device, which can be applied to Figure 1 and Figure 2 The information generation method provided by the corresponding embodiments can refer to Figure 3 The information generation device 3 can include a receiving unit 31, a processing unit 32, and a generation unit 33, wherein: The receiving unit 31 is configured to receive the search information input by the user; The processing unit 32 is configured to determine the matching node from the target structure tree based on the search information; wherein the target structure tree is a structure tree constructed for a target work; the target structure tree includes a plurality of nodes and edge structures representing the association relationship between the nodes; each node includes the identification of the node, the start time and end time of the content corresponding to the node, and the content summary; The processing unit 32 is further configured to obtain the target node having an association relationship with the matching node from the target structure tree based on the search information; The generation unit 33 is configured to generate the target information for the search information based on the matching node, the target node, and the association relationship.

[0092] In other embodiments of the present application, the processing unit 32 is specifically configured to perform the following steps: Performing semantic recognition on the search information to obtain the work name, the intent information, and the keyword; Determining the target structure tree from the to-be-screened structure tree based on the work name; wherein the to-be-screened structure tree includes a plurality of structure trees corresponding to works; Determining the matching node from the target structure tree based on the intent information.

[0093] In other embodiments of the present application, the processing unit 32 is specifically configured to perform the following steps: Matching the intent information with the content summary of each node in the target structure tree to obtain a matching result; Determining the matching node from the target structure tree based on the matching result.

[0094] In other embodiments of the present application, the processing unit 32 is specifically configured to perform the following steps: If the matching result represents that the target content overview exists in the target structure tree, the node corresponding to the target content overview is determined as the matching node; If the matching result represents that the target content overview does not exist in the target structure tree, semantic matching is performed in the target database based on the intent information, and the node corresponding to the target description information matched with the intent information is determined; wherein, the target database includes the identification of each node in the target structure tree and the description information corresponding to each node; The node corresponding to the target description information matched with the intent information is determined as the matching node.

[0095] In other embodiments of the present application, the processing unit 32 is specifically configured to perform the following steps: Based on the keywords in the search information, the target structure tree is traversed starting from the matching node according to the target traversal strategy; According to the traversal termination rule, the traversal of the target structure tree is stopped, and the target node is obtained; wherein, the traversal termination rule is used to determine that the obtained node information has constituted a complete explanation of the search information.

[0096] In other embodiments of the present application, the processing unit 32 is specifically configured to perform the following steps: For each play unit in the target work, a plurality of nodes for the play unit are constructed; The association relationship between the plurality of nodes of the plurality of play units is determined; wherein, the association relationship includes at least one of a cause-effect relationship and a time sequence relationship; Based on the plurality of nodes of the plurality of play units and the association relationship, a target structure tree for the target work is constructed.

[0097] In other embodiments of the present application, the processing unit 32 is specifically configured to perform the following steps: For each play unit, the voice information of the play unit is extracted and converted into timestamped text information; The text information is split based on the content category to obtain a plurality of sub-text information; For each sub-text information, a node is constructed.

[0098] It should be noted that the specific description of the steps performed by each unit can refer to the information generation method provided by the corresponding embodiments of Figure 1 and Figure 2 The information generation method provided by the corresponding embodiments will not be repeated here.

[0099] The information generation apparatus provided by the embodiments of the present application can automatically trace the associated upstream (antecedent) and downstream (consequence) target nodes of the matching node from the target structure tree containing the node association relationship, so as to generate complete and coherent content, instead of only retrieving the information fragment related to the keyword and failing to provide complete antecedent and consequence information as in the related art, thereby improving the retrieval accuracy and quality and enhancing the depth of information acquisition and user experience.

[0100] Based on the foregoing embodiments, the embodiments of the present application provide an information generation device which can be applied to Figure 1 and Figure 2 The information generation method provided by the corresponding embodiments can refer to Figure 4 The information generation device 4 can include a processor 41, a memory 42 and a communication bus 43, wherein: The communication bus 43 is used to realize the communication connection between the processor 41 and the memory 42; The processor 41 is used to execute the information generation program in the memory 42 to realize the following steps: receiving the retrieval information input by the user; determining the matching node from the target structure tree based on the retrieval information; wherein the target structure tree is a structure tree constructed for a target work; the target structure tree includes a plurality of nodes and edge structures representing the association relationship between the nodes; each node includes the identification of the node, the start time, the end time and the content summary of the content corresponding to the node; obtaining the target node having the association relationship with the matching node from the target structure tree based on the retrieval information; generating the target information for the retrieval information based on the matching node, the target node and the association relationship.

[0101] In other embodiments of the present application, the processor 41 is used to execute the information generation program in the memory 42 to determine the matching node from the target structure tree based on the retrieval information, to realize the following steps: performing semantic recognition on the retrieval information to obtain the work name, the intention information and the keyword; determining the target structure tree from the to-be-screened structure tree based on the work name; wherein the to-be-screened structure tree includes a plurality of structure trees corresponding to works; determining the matching node from the target structure tree based on the intention information.

[0102] In other embodiments of the present application, the processor 41 is used to execute the information generation program in the memory 42 to determine the matching node from the target structure tree based on the intention information, including realizing the following steps: The intention information is matched with the content summary of each node in the target structure tree to obtain a matching result; Based on the matching result, a matching node is determined from the target structure tree.

[0103] In other embodiments of the present application, the processor 41 is configured to execute the information generation program in the memory 42 to determine the matching node from the target structure tree based on the matching result, so as to implement the following steps: If the matching result indicates that there is a target content summary in the target structure tree that matches the intention information, the node corresponding to the target content summary is determined as the matching node; If the matching result indicates that there is no target content summary in the target structure tree, semantic matching is performed in the target database based on the intention information to determine the node corresponding to the target description information that matches the intention information; wherein the target database includes the identification of each node in the target structure tree and the description information corresponding to each node; The node corresponding to the target description information that matches the intention information is determined as the matching node.

[0104] In other embodiments of the present application, the processor 41 is configured to execute the information generation program in the memory 42 to obtain a target node having an association relationship with the matching node from the target structure tree based on the search information, so as to implement the following steps: Based on the keywords in the search information, the target structure tree is traversed according to a target traversal strategy starting from the matching node; The traversal of the target structure tree is stopped according to a traversal termination rule to obtain the target node; wherein the traversal termination rule is used to determine that the obtained node information has constituted a complete explanation of the search information.

[0105] In other embodiments of the present application, the processor 41 is configured to execute the information generation program in the memory 42 to implement the following steps: For each play unit in the target work, a plurality of nodes for the play unit are constructed; The association relationship between the plurality of nodes of the plurality of play units is determined; wherein the association relationship includes at least one of a cause-effect relationship and a time sequence relationship; Based on the plurality of nodes of the plurality of play units and the association relationship, a target structure tree for the target work is constructed.

[0106] In other embodiments of the present application, the processor 41 is configured to execute the information generation program in the memory 42 to construct a plurality of nodes for each play unit in the target work, so as to implement the following steps: For each play unit, the voice information of the play unit is extracted and converted into text information with a timestamp; The text information is split based on content category to obtain multiple sub-text information; For each piece of subtext information, construct a node.

[0107] It should be noted that a detailed description of the steps performed by the processor can be found in [reference needed]. Figure 1 and Figure 2 The information generation method provided in the corresponding embodiments will not be described in detail here.

[0108] The information generation device provided in this application embodiment can automatically trace the upstream (cause) and downstream (effect) target nodes associated with the matching node by constructing a target structure tree containing node association relationships, thereby generating complete and coherent content, instead of only retrieving information fragments related to keywords as in related technologies, which cannot provide complete cause and effect information. This not only improves the retrieval accuracy and retrieval quality, but also enhances the depth of information acquisition and user experience.

[0109] Based on the foregoing embodiments, this application provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to achieve... Figure 1 and Figure 2 The steps in the information generation method provided in the corresponding embodiment.

[0110] A computer program product includes a computer program that is implemented when executed by a processor. Figure 1 and Figure 2 The steps in the information generation method provided in the corresponding embodiment.

[0111] It should be noted that the computer readable storage medium above can be a Read Only Memory (ROM), a Programmable Read-Only Memory (PROM), an Erasable Programmable Read-Only Memory (EPROM), an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Ferromagnetic Random Access Memory (FRAM), a Flash Memory, a magnetic surface memory, an optical disc, or a Compact Disc Read-Only Memory (CD-ROM) memory, etc. It can also be various electronic devices including one or any combination of the above memories, such as a mobile phone, a computer, a tablet device, a personal digital assistant, etc.

[0112] It should be noted that in this paper, the term "include", "contain" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or includes elements inherent to such process, method, article or device. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or device including the element.

[0113] The above sequence number of the embodiments of the present application is only for description, not representing the advantages and disadvantages of the embodiments.

[0114] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product in essence or in the form of a contribution to the prior art. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disc, optical disc), and includes a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) execute the method described in each embodiment of the present application.

[0115] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0116] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the function specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0117] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks. Figure 1 one or more flow or blocks. Figure 1 one or more flow or blocks.

[0118] The above merely provides the preferred embodiment of the present application, and is not intended to limit the patent scope of the present application, and any equivalent structure or equivalent flow transformation made by using the content of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An information generation method characterized by comprising: The method comprises: receiving user input search information; determining a matching node from a target structure tree based on the search information; wherein the target structure tree is a structure tree constructed for a target work; the target structure tree comprises a plurality of nodes and edge structures representing the association relationship between the nodes; each node comprises the identification of the node, the start time, the end time and the content summary of the content corresponding to the node; based on the search information, obtaining a target node having the association relationship with the matching node from the target structure tree; generating target information for the search information based on the matching node, the target node and the association relationship.

2. The method of claim 1, wherein, The method comprises: performing semantic recognition on the search information to obtain a work name, intent information and keywords; determining the target structure tree from a to-be-screened structure tree based on the work name; wherein the to-be-screened structure tree comprises a plurality of structure trees corresponding to works; determining the matching node from the target structure tree based on the intent information.

3. The method of claim 2, wherein, The method comprises: matching the intent information with the content summary of each node in the target structure tree to obtain a matching result; determining the matching node from the target structure tree based on the matching result.

4. The method of claim 3, wherein, The method comprises: if the matching result represents that there is a target content summary matching the intent information in the target structure tree, determining the node corresponding to the target content summary as the matching node; if the matching result represents that there is no target content summary in the target structure tree, performing semantic matching in a target database based on the intent information to determine the node corresponding to the target description information matching the intent information; wherein the target database comprises the identification of each node in the target structure tree and the description information corresponding to each node; determining the node corresponding to the target description information matching the intent information as the matching node.

5. The method of claim 1, wherein, The method comprises: based on the keywords in the search information, traversing the target structure tree starting from the matching node according to a target traversal strategy; stopping the traversal of the target structure tree according to a traversal termination rule to obtain the target node; wherein the traversal termination rule is used to determine that the obtained node information has constituted a complete explanation of the search information.

6. The method of claim 1, wherein, The method further comprises: constructing a plurality of nodes for each playback unit in the target work; determining the association relationship between the plurality of nodes of the plurality of playback units; wherein the association relationship comprises at least one of a cause-effect relationship and a time sequence relationship; constructing a target structure tree for the target work based on the plurality of nodes of the plurality of playback units and the association relationship.

7. The method of claim 6, wherein, The method comprises the following steps: For each play unit in the target work, a plurality of nodes for the play unit are constructed, including: For each play unit, voice information of the play unit is extracted and converted into timestamped text information; The text information is split based on content categories to obtain a plurality of sub-text information; 8. An information generation apparatus characterized by comprising: For each sub-text information, a node is constructed. The device comprises a processor, a memory and a communication bus; The communication bus is used to realize the communication connection between the processor and the memory; 9. A computer-readable storage medium, characterized in that, The processor is used to execute the information generation program in the memory to realize the steps of the information generation method according to any one of claims 1-7.

10. A computer program product comprising a computer program, characterized in that, The computer readable storage medium stores one or more programs, which can be executed by one or more processors to realize the steps of the information generation method according to any one of claims 1-7. The computer program realizes the steps of the information generation method according to any one of claims 1-7 when executed by the processor.

Images