Method and device for generating summary, equipment and storage medium

Abstract

According to the embodiment of the invention, the invention provides a summary generation method and device, equipment and a storage medium. The method comprises the following steps: in response to a summary generation request, obtaining schedule data and document data within a time range indicated by the summary generation request; determining at least one theme based on the schedule data and the document data; based on the at least one theme, generating at least one content set corresponding to the at least one theme from the schedule data and the document data; and generating summary information corresponding to the time range based on the at least one content set. Therefore, the integrity and the accuracy of generating the summary can be improved.

Description

Technical Field

[0001] The exemplary embodiments disclosed herein generally relate to the field of information technology, and in particular to methods, apparatus, devices, and computer-readable storage media for generating summaries. Background Technology

[0002] With the rapid development of smart office, online collaboration, and information processing technologies, users continuously generate a large amount of semantic content across time and sources in their daily activities, such as meeting minutes, project documents, community exchanges, business communications, or personal notes. This content may be distributed across different data sources or application systems, with high semantic relevance but loose organization.

[0003] To facilitate review, summarization, and information sharing, users typically need to organize and summarize content within a certain timeframe. Therefore, achieving efficient information organization and multi-dimensional semantic summarization has become a significant technical challenge in the field of intelligent text generation. Summary of the Invention

[0004] In a first aspect of this disclosure, a method for generating a summary is provided. The method includes: in response to a summary generation request, acquiring schedule data and document data within a time range indicated by the summary generation request; determining at least one topic based on the schedule data and document data; generating at least one content set corresponding to each of the at least one topic from the schedule data and document data; and generating summary information corresponding to the time range based on the at least one content set.

[0005] In a second aspect of this disclosure, an apparatus for generating a summary is provided. The apparatus includes: an acquisition module configured to acquire schedule data and document data within a time range indicated by a summary generation request in response to such a request; a topic determination module configured to determine at least one topic based on the schedule data and document data; a content segmentation module configured to generate at least one content set corresponding to each of the at least one topic from the schedule data and document data; and a summary module configured to generate summary information corresponding to the time range based on the at least one content set.

[0006] In a third aspect of this disclosure, an electronic device is provided. The electronic device includes at least one processing unit; and at least one memory coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method of the first aspect of this disclosure when executed by the at least one processing unit.

[0007] In a fourth aspect of this disclosure, a computer-readable storage medium is provided. This computer-readable storage medium stores a computer program that can be executed by a processor to perform the method according to a first aspect of this disclosure.

[0008] In a fifth aspect of this disclosure, a computer program product is provided, which is tangibly stored in a computer storage medium and includes computer-executable instructions that, when executed by a device, cause the device to perform the method according to a first aspect of this disclosure.

[0009] It should be understood that the content described in this section is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0010] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. In the drawings, the same or similar reference numerals denote the same or similar elements, wherein: Figure 1 A schematic diagram of an example environment in which embodiments of the present disclosure can be implemented is shown; Figure 2A A schematic diagram illustrating the process of generating a summary according to some embodiments of the present disclosure is shown; Figure 2B A schematic diagram illustrating a content reorganization process according to some embodiments of the present disclosure is shown; Figure 2C A schematic diagram illustrating the process of generating summary information according to some embodiments of the present disclosure is shown; Figure 3 A flowchart of a process for generating a summary according to some embodiments of the present disclosure is provided; Figure 4 A block diagram of an apparatus for generating a summary according to some embodiments of the present disclosure is shown; and Figure 5 A block diagram of an electronic device in which one or more embodiments of the present disclosure may be implemented is shown. Detailed Implementation

[0011] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0012] For example, upon receiving a user's active request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0013] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0014] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0015] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0016] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0017] In the description of embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may also be included below.

[0018] As briefly mentioned earlier, with the rapid development of smart office, online collaboration, and information processing technologies, users continuously generate a large amount of semantic content across time and sources in their daily activities, such as meeting minutes, project documents, group chats, business communication records, or research notes. This content often encompasses the evolution of multiple themes, events, or tasks and is scattered across different documents, message streams, or systems, forming a multi-source collection of long texts. To facilitate periodic reviews, knowledge accumulation, or team synchronization, users typically need to organize and summarize this content to identify key issues, track progress, and plan subsequent actions.

[0019] In this type of summary task, periodic work summaries (such as daily, weekly, and monthly reports) are one of the most common applications. This type of summary process plays a core role in team collaboration and goal management. By systematically reviewing work progress, identifying key issues, and planning subsequent tasks, it directly impacts project schedule control and resource allocation efficiency.

[0020] In related technologies, with the rapid development of Language Models (LLMs), automatic text summarization technology based on deep semantic understanding has demonstrated significant advantages in multiple fields. This type of technology can automatically extract core information from text using semantic reasoning, providing a new approach to summarizing tasks from multiple sources and documents. For example, in office or collaborative scenarios, users can directly invoke the model to automatically generate interim summaries based on the input documents.

[0021] However, traditional language model-based summarization methods have some limitations. On the one hand, in real-world applications, the content to be summarized often arises from multiple days, tasks, or activities involving multiple people, with an overall length far exceeding the model's contextual limitations. Directly inputting this content into the model can easily lead to the truncation of key information or semantic breaks, making it difficult to guarantee the completeness and coherence of the summarized content.

[0022] On the other hand, traditional summarization methods rely on directly controlling the overall output format through prompts, lacking the ability to semantically understand and control the content structure. For example, in work summary scenarios, it is often necessary to generate summary content for different modules. Directly using prompts can easily lead to problems such as module misalignment, omission of key points, or fabricated information.

[0023] In view of this, embodiments of this disclosure provide a scheme for generating a summary. According to embodiments of this disclosure, firstly, in response to a summary generation request, schedule data and document data within the time range indicated by the summary generation request are obtained. Based on the schedule data and document data, at least one topic is determined. Further, based on the at least one topic, at least one content set corresponding to each of the at least one topic is generated from the schedule data and document data. Still further, based on the at least one content set, summary information corresponding to the time range is generated.

[0024] Therefore, a scheme for generating summaries is provided, enabling efficient semantic organization and summary generation of data from multiple sources and spanning multiple time periods. By reorganizing schedule and document data based on topics, semantic integrity and information continuity can be maintained even when the input content exceeds the contextual limitations of the language model, thus avoiding content truncation or semantic breaks that may occur when directly inputting long texts. Furthermore, by generating summary information based on corresponding content sets corresponding to topics, the generation process can output more focused and accurate summary results for different semantic topics, significantly improving the logical consistency and information coverage of the summary content. In this way, well-structured and complete summary information can be generated in multi-document, multi-time, and multi-source scenarios, improving the accuracy and practicality of automated summaries.

[0025] Figure 1 A schematic diagram of an example environment 100 in which embodiments of the present disclosure can be implemented is shown. In this example environment 100, server 130 is capable of establishing a communication connection with one or more terminal devices (collectively or individually referred to as terminal device 110). At least one application 120 may be installed on terminal device 110. User 140 may interact with application 120 via terminal device 110 and / or an attachment device of terminal device 110. Application 120 may be a content presentation application, an online shopping application, or any other suitable application.

[0026] exist Figure 1 In environment 100, application 120 can also be accessed in other ways, such as through a webpage. If application 120 is active, terminal device 110 can display the interface 150 of application 120. Interface 150 may include various interfaces provided by application 120, such as content presentation interface, content creation interface, service request submission interface, request processing result interface, message interface, personal homepage, etc.

[0027] In some embodiments, terminal device 110 communicates with server 130 to provide services to application 120. Terminal device 110 can be any type of mobile terminal, fixed terminal, or portable terminal, including mobile phones, desktop computers, laptop computers, notebook computers, netbook computers, tablet computers, media computers, multimedia tablets, personal communication system (PCS) devices, personal navigation devices, personal digital assistants (PDAs), audio / video players, digital cameras / camcorders, positioning devices, television receivers, radio receivers, e-book devices, gaming devices, or any combination thereof, including accessories and peripherals of these devices or any combination thereof. In some embodiments, terminal device 110 can also support any type of user-facing interface (such as "wearable" circuitry). Server 130 can be various types of computing systems / servers capable of providing computing power, including but not limited to mainframes, edge computing nodes, computing devices in cloud environments, etc.

[0028] In some embodiments of this disclosure, one or more functions of application 120 may be supported based on machine learning model 125. This machine learning model may be deployed locally on or invoked by terminal device 110, or it may be deployed on or invoked by server 130. In embodiments where it is run or invoked by server 130, the server 130 may provide the model's output to terminal device 110.

[0029] In some embodiments, the machine learning model 125 may be based on any suitable model architecture, including but not limited to Transformer models, convolutional neural networks (CNNs), recurrent neural networks (RNNs), deep neural networks (DNNs), and so on. In some embodiments, the machine learning model 125 may be based on a language model (LM). A language model, by learning from a large corpus, is capable of question-answering. The language model can be guided to output the desired answer by inputting prompt words. Prompt words can be in the form of text, images, videos, and other multimodal content. The machine learning model 125 may also be based on other suitable models. The machine learning model 125 may include one or more machine learning models. It should be noted that if the machine learning model 125 includes multiple machine learning models, these multiple machine learning models may have different structures, uses, and functions, and this disclosure does not limit them.

[0030] It should be understood that the structure and function of the various elements in environment 100 are described for illustrative purposes only and do not imply any limitation on the scope of this disclosure.

[0031] Figure 2AA schematic diagram of a process 200 for generating a summary according to some embodiments of the present disclosure is shown. Process 200 can be implemented in environment 100, for example, process 200 can be implemented at server 130 or terminal device 110. It should be understood that the actions described with respect to server 130 can be performed at least partially by entities other than server 130, for example, at least partially by terminal device 110, other terminal devices, or other servers, or by server 130 in collaboration with the aforementioned devices; and vice versa. In the following discussion, reference will be made to… Figure 1 To describe process 200.

[0032] like Figure 2A As shown, server 130 can respond to a summary generation request by retrieving schedule data 202 and document data 204 within the time range indicated by the summary generation request. The summary generation request can be initiated by application 130 and sent to server 130 via terminal device 110. The summary generation request can include various types of data such as text, voice, and images. For example, user 150 can initiate a summary generation request by entering a text message, interacting with the system via voice, or through automatic task triggering.

[0033] In some embodiments, the summary generation request may include a time range parameter, such as "this week," "last month," etc., to indicate the target time range to be summarized. Server 130 may retrieve corresponding schedule data 202 and document data 204 from associated office systems, collaboration platforms, or databases based on the time range indicated by the summary generation request.

[0034] In some embodiments, schedule data 202 includes the name, category, and time information of at least one schedule within a target time range. For example, the schedule name may correspond to the title of a meeting, discussion, task, or event. The schedule category may indicate whether the schedule was accepted (participated in) or created by user 150. The time information may include the start time, end time, or duration. Alternatively or additionally, schedule data 202 may include information such as participants, locations, priorities, or task status related to the schedule to assist in subsequent information summarization and processing.

[0035] In some embodiments, document data 204 includes the content, category, and time information of at least one document within a target time range. Document content may include text, images, speech-to-text, or structured record data, etc. The document category may indicate that the document was opened or created by user 150. The document's time information may indicate the document's creation time, modification time, or opening time. Alternatively or additionally, document data 204 may include attribute information related to the document's creation or use, such as author, recipient, abstract, or source platform, etc.

[0036] In some embodiments, schedule data 202 and document data 204 can be associated data. For example, in an office or collaborative scenario, a schedule event (such as "Project A Weekly Meeting") can correspond to at least one document (such as meeting minutes or discussion records). These documents can reflect the specific content, conclusions, or follow-up tasks of the schedule event. Server 130 can associate at least one document related to the schedule event based on the schedule's time information, event name, participants, or task identifier. Schedule data 202 can provide a semantic timeline and task context, while document data 204 can provide detailed content support. The combination of the two helps improve the semantic consistency and logical coherence of the summary generation process.

[0037] In other types of summary generation scenarios (e.g., summaries of discussions across time periods, summaries of multi-person dialogues, or cross-platform content aggregation), schedule data 202 can be timestamps, dialogue rounds, or other content segmentation identifiers. Document data 204 can be different semantic content. In this case, the relationship between the two can include, but is not limited to, time series correspondence or semantic association, thereby supporting a wider range of multi-source data summary generation tasks.

[0038] Since summary generation tasks typically involve semantic content spanning multiple time periods and sources, the acquired document data 204 may far exceed the context window limit of the language model in length. Directly inputting this data into the model may lead to information truncation or semantic breaks, thereby affecting the completeness and accuracy of the summary results. In this case, the server 130 can segment the document data 204 and reorganize the content based on the extracted topics to ensure the continuity and coverage of the summary generation.

[0039] Continue to refer to Figure 2A In box 210, server 130 can reorganize the acquired data based on topics to obtain at least one content set 215 corresponding to a topic. For example, server 130 can combine schedule data 202 and document data 204 to generate multiple topics, and then re-cluster or group the schedule data 202 and document data 204 based on these topics. After reorganization, server 130 can form a structured content set 215. Each content set 215 can represent a set of content fragments under a specific topic semantics, used to support subsequent summary generation.

[0040] In some embodiments, server 130 can determine at least one topic based on schedule data and document data. For example, server 130 can match text fragments in document data 204 with time, task category, or event description in schedule data 202, and use a semantic analysis model to determine the semantic association between them. Based on this semantic association, server 130 can identify one or more topics under different time periods, task types, or semantic aggregations.

[0041] As an example, Figure 2B A schematic diagram of a content reorganization process 210 according to some embodiments of the present disclosure is shown. Figure 2B As shown in box 211, server 130 can determine multiple content fragments from document data. Summary generation tasks typically require processing multiple documents, long texts, or multi-source corpora; directly using an entire document as input may result in semantic spans that are too large or exceed the model's processing capacity. Server 130 can segment the content in document data 204 to obtain multiple independently processable content fragments. Each content fragment can be considered a semantically relatively complete and appropriately sized text unit for subsequent topic determination and content aggregation.

[0042] In some embodiments, server 130 may identify at least one first document in the document data whose content length is lower than a first threshold length as at least one content fragment among multiple content fragments. For example, the first document itself is short and has a single content theme. In this case, it can directly participate in subsequent theme analysis as an independent content fragment without further splitting. The first threshold length can be set according to the model's processing capacity and scenario requirements, for example, in the range of several thousand to tens of thousands of characters.

[0043] In some embodiments, server 130 may divide at least one second document whose content length exceeds a first threshold length into at least two content segments from a plurality of content segments. For example, for very long documents (such as meeting summaries, project reports, or long conversations), server 130 may divide the document into multiple relatively independent content segments based on document structure, chapter hierarchy, or semantic change locations. In this way, the amount of input per instance can be reduced while maintaining semantic coherence, thereby improving the accuracy of subsequent topic recognition.

[0044] In some embodiments, if at least one second document includes at least one semantic boundary identifier, server 130 may divide at least one second document into at least two content segments according to at least one semantic boundary identifier. For example, a semantic boundary identifier may be an explicit structural identifier in the document used to indicate chapters, paragraphs, or topic changes, such as headings, numbering, table of contents entries, indentations, or section breaks. Server 130 may preferentially utilize these structural identifiers for segmentation to preserve natural semantic boundaries and contextual integrity as much as possible.

[0045] In some embodiments, if at least one second document does not include semantic boundary markers, server 130 may divide at least one second document into at least two content segments based on a second threshold length. For example, server 130 may divide the second document according to a preset second threshold length (e.g., 500 characters per segment). Alternatively or additionally, server 130 may divide the content segments based on linguistic symbols (e.g., punctuation marks).

[0046] In some embodiments, to avoid semantic breaks caused by mechanical segmentation, server 130 may retain a certain length of overlapping content (e.g., 200 characters) between adjacent content segments, thereby ensuring semantic continuity between segments and consistency in subsequent topic analysis. Through this segmentation method, server 130 can automatically determine multiple content segments for documents of different types and structures, providing high-quality basic data units for subsequent topic determination and content reorganization.

[0047] Continue to refer to Figure 2B In box 212, for a content segment among multiple content segments, server 130 can determine at least one topic tag related to that content segment based on schedule data and the content segment. For example, server 130 can treat each content segment as a unit to be labeled, and use a language model or other suitable model to identify the core semantic points of the segment, thereby determining one or more topic tags most relevant to the content segment. Topic tags can identify the semantic or task-oriented affiliation of content segments, facilitating subsequent topic fusion, content aggregation, and structured generation.

[0048] In some embodiments, server 130 may, in response to the content fragment being related to at least one schedule in the schedule data, determine the name of at least one schedule as at least one topic tag. For example, if the timestamp, participants, or text content of a content fragment highly matches "Project A Weekly Meeting" in the schedule data 202, or if the document to which the content fragment belongs is marked as directly related to the schedule "Project A Weekly Meeting," server 130 may use "Project A Weekly Meeting" as one of the topic tags for the content fragment. By using the schedule name as the topic tag, the consistency between the topic tag and the semantics of time or event can be ensured, and it is convenient to organize and trace back by event or task during the content aggregation phase.

[0049] In some embodiments, server 130 may utilize a language model to determine at least one topic tag based on the semantic content of the content fragment and the category of the document to which the content fragment belongs. For example, server 130 may use each content fragment along with the metadata of its document as input to call a multi-tag topic extraction model or a language model to annotate the content fragment.

[0050] As an example, when performing topic tag extraction, server 130 can specify the task objective and judgment criteria for topic extraction to the model by setting prompt words. Server 130 can instruct the model in the prompt words that it needs to extract several topic tags that can represent the core semantics from the document fragment, and specify the corresponding extraction principles in the prompt words.

[0051] For example, the extraction principles may include at least one of the following: If content related to schedule data 202 exists, priority should be given to extracting topic tags corresponding to the schedule name to maintain consistency of the summary results in terms of time and task semantics. Secondly, if multiple candidate topic tags may exist in the same segment, priority should be given to topic tags that reflect the main idea of ​​the content, appear frequently, or account for a significant portion of the text. In addition, the server 130 may also guide the model to refer to document titles, chapter titles, or schedule names in prompts to improve the relevance of topic tags to the overall semantic context.

[0052] In some embodiments, to improve the accuracy of importance identification, server 130 can guide the model to assign different weights to documents from different sources or of different types in the prompt words. For example, documents of the "Created" type (i.e., documents generated or uploaded by user 150) can be assigned higher weights because such documents typically reflect more core or formal content. Documents of the "Open" category are assigned lower weights as supplementary information.

[0053] In some embodiments, to ensure the reliability of topic tags, server 130 can calculate a confidence score for each topic tag of a content fragment and filter or sort the tags based on the confidence score. Furthermore, server 130 can combine schedule priority rules and document category weighting rules to perform weighted calculations on multiple candidate topic tags generated from the same content fragment, thereby determining the final set of topic tags assigned to that content fragment. The confidence score and weighting information can serve as the basis for subsequent topic fusion and content set generation.

[0054] In box 213, server 130 can fuse topic tags related to multiple content fragments based on semantic similarity to determine at least one topic. For example, since topic tags of different content fragments may have synonymous, near-synonymous, or hierarchical relationships in expression, such as "project meeting" and "project weekly meeting" pointing to the same topic, or "budget adjustment" and "cost optimization" being similar concepts, server 130 can analyze the semantic similarity between these tags and fuse identical or similar tags. For example, server 130 can use language models, word vector models, or embedding space similarity calculation methods to measure the semantic distance between tag pairs. If the semantic similarity between two or more tags exceeds a preset threshold, server 130 can merge them into the same topic and generate a topic name as a representative of that topic. In this way, redundancy of topic tags can be reduced while maintaining semantic integrity, thereby forming a compact and unified topic system.

[0055] As an example, server 130 can specify the task objective and judgment criteria for topic fusion to the model by setting prompt words. For instance, prompt words can indicate that if multiple tags are semantically identical, synonymous, or have a hierarchical relationship, they should be grouped into the same category. Furthermore, the most concise, general, or semantically representative words can be selected from the merged tag group as standard topic names. Among multiple candidate topic tags, expressions that appear frequently in the original content fragments or are jointly cited by multiple fragments can be prioritized to improve the representativeness of the topics.

[0056] In box 214, server 130 can generate at least one content set 215 corresponding to each of at least one theme from schedule data and document data. For example, after completing the extraction and fusion of theme tags, server 130 can select content fragments that are semantically related to each determined theme from data from different sources to construct the content set 215 for that theme. Each content set 215 can be understood as a collection of related text blocks aggregated in time and semantics, providing centralized material input for subsequent summary generation.

[0057] In some embodiments, for a topic within at least one subject, server 130 can determine a candidate content set based on multiple content fragments corresponding to that topic in schedule data and document data. For example, if the topic is "Project A Delivery," server 130 can extract meeting, task schedule, or event descriptions related to Project A from schedule data, and extract report, minutes, or record fragments semantically similar to Project A from document data. Server 130 can identify fragments related to the topic using methods such as topic tag similarity, keyword matching, or contextual embedding representation, thereby forming an initial candidate content set. This content set may include multiple content fragments from different documents or different time periods.

[0058] In some embodiments, server 130 may remove content fragments from the candidate content set whose similarity to the topic is below a threshold. For example, server 130 may calculate the similarity between candidate content fragments and topic tags based on a semantic embedding model and retain only fragments with semantic relevance above the threshold. In this way, topic-irrelevant content can be avoided from interfering with the accuracy of subsequent summary information generation.

[0059] Furthermore, server 130 can sort multiple content fragments in the candidate content set according to the corresponding time information of the documents to which the multiple content fragments belong, so as to obtain a content set corresponding to the topic. For example, server 130 can arrange the fragments in chronological order on the timeline, so that the content of the same topic forms a coherent narrative chain in the time dimension, thereby facilitating the identification of the development process of events or the execution status of tasks.

[0060] In some embodiments, server 130 may merge semantically similar parts between adjacent content segments or remove duplicates from highly repetitive content to ensure the conciseness and information integrity of the final content set. Alternatively or additionally, server 130 may further assign priorities to each topic based on the importance or time distribution of different topics to support automatic organization of content by module or time period during subsequent summary generation.

[0061] In the embodiments of this disclosure, content under different topics is aggregated and reorganized to generate several content sets organized by topic. Each content set may include text fragments that have been filtered, sorted, and semantically reconstructed, and these fragments semantically reflect discussions, actions, or results related to that topic. Through this topic-based content reorganization process, not only is the semantic consistency and logical coherence of the input data for the summary generation guaranteed, but the accuracy and readability of the subsequent generated results can also be improved.

[0062] In some embodiments, server 130 may generate summary information corresponding to a time range based on at least one set of content. For example, refer to Figure 2A Server 130 can use agent 220 to analyze the reorganized content to generate summary information 225.

[0063] In some embodiments, the summary information 225 follows a predetermined summary template. To improve generation quality, an agent 220 can be designed to process different types of information separately according to the structure of the predetermined summary template during the generation process. For example, the agent 220 can decompose a complex summary generation task into several sub-tasks, each targeting a module of the summary template. Each template can correspond to different extraction principles, thereby maintaining the stability and accuracy of the generation even with a large amount of input information and many logical branches. This task decomposition mechanism can effectively avoid the problem of model performance degradation caused by excessive input content or overly complex prompt logic.

[0064] In some embodiments, the server 130 may sort at least one set of content in chronological order before generating the summary, to ensure that the presentation order of different topics in the summary is consistent with the actual timeline. This chronological sorting not only helps improve the readability of the summary, but also helps the agent understand the causal relationships of events during the generation process, thereby generating summary content that is more in line with the real context.

[0065] As an example, Figure 2C A schematic diagram of a process 230 for generating summary information according to some embodiments of the present disclosure is shown. For example... Figure 2C As shown in box 231, server 130 can perform task decomposition, dividing the process of generating summary information into multiple sub-tasks.

[0066] In some embodiments, based on at least one sorted set of content, server 130 may utilize agent 220 to perform multiple subtasks. Each subtask is configured to generate information from a template portion of a predetermined summary template. For example, agent 220 may sequentially invoke a language model or other generative model to execute the subtasks, thereby performing separate semantic extraction and natural language generation on the input content of each module. Each subtask may employ different generation strategies or prompting logic for different semantic features.

[0067] In some embodiments, summary information 225 may include pending items. Pending items may include tasks, problems, or issues that are not yet resolved or require further decision-making within a target timeframe. In block 232, agent 220 may invoke a first machine learning model to perform a subtask to generate pending items based on corresponding prompts. This subtask may identify fragments containing semantics of pending confirmation, requiring decision-making, or not yet completed, based on at least one content set 215, and generate a set of pending items corresponding to the current summary period. In this way, task nodes that are still in progress can be automatically filtered during the summary generation process, helping users quickly focus on pending items.

[0068] In some embodiments, the first machine learning model can identify at least one item in at least one set of content that has semantic content to be processed. For example, the model can detect content containing semantic features such as "needs a decision," "to be confirmed," or "to be processed" from document text, meeting minutes, or task descriptions, and mark it as a potential item to be processed.

[0069] Furthermore, for at least one item within a broader category, the first machine learning model can determine whether the corresponding content of that item possesses a processed semantic marker in subsequent times. For example, the first machine learning model can track subsequent document fragments on the same topic based on chronological order and determine whether semantic markers such as "decided," "processed," "approved," or "confirmed" appear, indicating that the item has been processed. If the corresponding content of the item does not possess a processed semantic marker, the first machine learning model can identify the item as a pending item. In this way, redundant summarization or omission of key decision-making nodes can be avoided, thereby improving the logical accuracy and timeliness of the summarized content.

[0070] In some embodiments, the summary information 225 includes key items and other items. Key items may include content that is of high importance or influence within a target timeframe, such as core tasks, key decisions, or key projects. Other items may include relatively minor, auxiliary, or non-core content. In block 233, agent 220 may utilize a second machine learning model to perform a subtask of generating key items and other items. This subtask enables automatic classification and aggregation of tasks at different levels of importance.

[0071] In some embodiments, the second machine learning model can identify at least one item from at least one content set and merge items belonging to multiple phases of the same project within the at least one item. The merged item may include a complete timeline of the task and records of results, facilitating subsequent overall analysis and summarization.

[0072] Furthermore, the second machine learning model can classify items within at least one merged item into key items and other items based on at least one of the following: the corresponding time duration, the number of participants, project deliverables, or predetermined identifiers. For example, the second machine learning model can identify key items as those that take a long time, involve many participants, have clear goals and deliverables, are marked as key projects, or are related to important decisions that have been resolved. Server 130 can output a summary and description of each item, generating a well-structured and organized task overview, thereby providing users with more comprehensive and targeted summary information.

[0073] In some embodiments, summary information 225 includes planned items for the next time frame. Planned items can be used to indicate tasks that need to continue or are planned to be initiated after the current summary period ends. In block 234, agent 220 can utilize a third machine learning model to perform subtasks that generate planned items, enabling automatic extraction of task continuity and future planning.

[0074] In some embodiments, the third machine learning model can identify at least one item from at least one content set that has planning semantic content or semantic content related to the next time frame. For example, the third machine learning model can identify content containing keywords such as "next week," "next step," or "plan" as planned items. Alternatively or additionally, the third machine learning model can also identify items identified as needing to be processed within the target event frame, i.e., unfinished content that needs to be continued, as planned items.

[0075] Furthermore, the third machine learning model can identify at least one item, excluding completed items, as planned items for the next time frame. For example, the third machine learning model can examine the subsequent status of at least one item. If a semantic identifier such as "completed," "finished," or "passed" is detected, the item is removed from the planned items.

[0076] In some embodiments, the organizational structure of the summary information 225 can be flexibly configured according to different application scenarios, user groups, or generation goals. The predetermined summary template is not limited to modules such as "To-do Items," "Key Items," and "Planned Items," but can also be extended to other types of content modules. Different template modules can correspond to different extraction logic, generation prompts, or output formats. The server 130 can automatically select or adjust the template structure according to the target type of the summary generation request, thereby generating summary information that conforms to the target context.

[0077] In some embodiments, server 130 can sort the items in the obtained summary information according to their importance, so as to prioritize the output of summary information on key items. For example, server 130 can perform weighted sorting based on indicators such as scope of impact, number of participants, expected results, or system preset weights, thereby generating clearer and more prioritized summary information.

[0078] In some embodiments, server 130 can determine summary information 225 based on information from multiple template portions generated by multiple subtasks. For example, in block 235, server 130 can splice and merge the content generated by each module after executing the aforementioned multiple subtasks to generate corresponding module content.

[0079] Furthermore, in box 236, server 130 can perform polishing on this content to unify the language style and logical coherence of the summary content, thereby obtaining the final summary information 225 (e.g., a work summary report). The polishing process may include, but is not limited to, semantic completion, sentence optimization, and cross-module logical adjustments to ensure that the summary report is coherent and highlights the key points.

[0080] It should be noted that the language model or machine learning model called by server 130 during the summary generation process can be any suitable model with good generation results. This disclosure does not impose any restrictions on this.

[0081] In summary, the embodiments of this disclosure propose a summary generation method applicable to multi-document, multi-source semantic data. By reorganizing content based on topics, core semantic content can be extracted within contextual constraints, avoiding the omission of key information and semantic fragmentation. Furthermore, by decomposing the summary generation process into multiple independent modular subtasks, and generating them separately for pending matters, key matters, and planned matters, the accuracy, logicality, and readability of the summary content are significantly improved. In this way, while improving the completeness and consistency of summary generation, the performance degradation problem of the model when processing complex long text tasks is further addressed.

[0082] Figure 3 A flowchart of a process 300 for generating a summary according to some embodiments of the present disclosure is shown. Process 300 may be implemented in environment 100, for example, process 300 may be implemented at server 130.

[0083] In box 310, server 130 responds to the summary generation request by retrieving schedule data and document data for the time range indicated by the summary generation request.

[0084] In box 320, server 130 determines at least one topic based on schedule data and document data.

[0085] In box 330, server 130 generates at least one content set corresponding to each of at least one topic from schedule data and document data, based on at least one topic.

[0086] In box 340, server 130 generates summary information corresponding to a time range based on at least one set of content.

[0087] In some embodiments, the schedule data includes the corresponding name, category, and time information of at least one schedule within the target time range, and the document data includes the corresponding content, category, and time information of at least one document within the target time range.

[0088] In some embodiments, determining at least one topic includes: determining a plurality of content fragments from document data; determining at least one topic tag associated with a content fragment among the plurality of content fragments based on schedule data and the content fragment; and fusing topic tags associated with the plurality of content fragments based on semantic similarity to determine at least one topic.

[0089] In some embodiments, determining a plurality of content segments includes at least one of the following: determining at least one first document in the document data whose content length is less than a first threshold length as at least one content segment among a plurality of content segments; or dividing at least one second document in the document data whose content length exceeds the first threshold length into at least two content segments among a plurality of content segments.

[0090] In some embodiments, dividing at least one second document into multiple content segments includes: dividing at least one second document into at least two content segments according to at least one semantic boundary identifier in response to at least one second document including at least one semantic boundary identifier; and dividing at least one second document into at least two content segments based on a second threshold length in response to at least one second document not including a semantic boundary identifier.

[0091] In some embodiments, determining at least one topic tag associated with the content fragment includes at least one of the following: in response to the content fragment being associated with at least one schedule in the schedule data, determining the name of at least one schedule as at least one topic tag; or using a language model to determine at least one topic tag based on the semantic content of the content fragment and the category of the document to which the content fragment belongs.

[0092] In some embodiments, generating at least one content set corresponding to at least one topic includes: for a topic in at least one topic, determining a candidate content set based on multiple content fragments corresponding to the topic in schedule data and document data; deleting content fragments in the candidate content set whose similarity to the topic is lower than a threshold from the candidate content set; and sorting multiple content fragments in the candidate content set according to the corresponding time information of the documents to which the multiple content fragments belong, so as to obtain a content set corresponding to the topic.

[0093] In some embodiments, the summary information follows a predetermined summary template, and generating the summary information includes: sorting at least one set of content by time; performing multiple subtasks using an agent based on the sorted at least one set of content, each subtask being configured to generate information in a template portion of the predetermined summary template; and determining the summary information based on the information from the multiple template portions generated by the multiple subtasks.

[0094] In some embodiments, the summary information includes pending items, wherein performing multiple subtasks includes: using a first machine learning model to identify at least one item in at least one content set that has pending semantic content; for an item in the at least one item, determining whether the corresponding content of the item has a processed semantic identifier in a subsequent time; and in response to the corresponding content of the item not having a processed semantic identifier, determining the item as a pending item.

[0095] In some embodiments, the summary information includes key items and other items, wherein performing multiple sub-tasks includes: using a second machine learning model to identify at least one item in at least one content set; merging items belonging to multiple phases of the same project in at least one item; and classifying items in at least one item into key items and other items based on at least one of the corresponding time length, number of participants, project deliverables, or predetermined identifiers of the merged at least one item.

[0096] In some embodiments, the summary information includes planned items for the next time frame, wherein performing multiple subtasks includes: using a third machine learning model to identify at least one item in at least one content set that has planned semantic content or semantic content related to the next time frame; and identifying other items in the at least one item besides those identified as needing to be processed within the target event frame and completed items as planned items for the next time frame.

[0097] Figure 4 A block diagram of an apparatus 400 for generating a summary according to some embodiments of the present disclosure is shown. The apparatus 400 may be implemented as or include a server 130.

[0098] The apparatus 400 includes an acquisition module 410 configured to acquire schedule data and document data within a time range indicated by a summary generation request in response to the summary generation request; a topic determination module 420 configured to determine at least one topic based on the schedule data and document data; a content segmentation module 430 configured to generate at least one content set corresponding to each of the at least one topic from the schedule data and document data; and a summary module 440 configured to generate summary information corresponding to the time range based on at least one content set.

[0099] In some embodiments, the schedule data includes the corresponding name, category, and time information of at least one schedule within the target time range, and the document data includes the corresponding content, category, and time information of at least one document within the target time range.

[0100] In some embodiments, the topic determination module 420 is further configured to: determine multiple content fragments from document data; determine at least one topic tag related to a content fragment based on schedule data and the content fragment; and fuse topic tags related to the multiple content fragments based on semantic similarity to determine at least one topic.

[0101] In some embodiments, the topic determination module 420 is further configured to: determine at least one first document in the document data whose content length is less than a first threshold length as at least one content segment among a plurality of content segments; or divide at least one second document in the document data whose content length exceeds the first threshold length into at least two content segments among a plurality of content segments.

[0102] In some embodiments, the topic determination module 420 is further configured to: in response to at least one second document including at least one semantic boundary identifier, divide at least one second document into at least two content segments according to at least one semantic boundary identifier; and in response to at least one second document not including a semantic boundary identifier, divide at least one second document into at least two content segments based on a second threshold length.

[0103] In some embodiments, the topic determination module 420 is further configured to: determine the name of at least one schedule as at least one topic tag in response to the content fragment being related to at least one schedule in the schedule data; or determine at least one topic tag based on the semantic content of the content fragment and the category of the document to which the content fragment belongs using a language model.

[0104] In some embodiments, the content segmentation module 430 is further configured to: determine a candidate content set based on multiple content fragments corresponding to the topic in the schedule data and document data for at least one topic; delete content fragments in the candidate content set whose similarity to the topic is lower than a threshold from the candidate content set; and sort multiple content fragments in the candidate content set according to the corresponding time information of the documents to which the multiple content fragments belong, so as to obtain a content set corresponding to the topic.

[0105] In some embodiments, the summary information follows a predetermined summary template, and the summary module 440 is further configured to: sort at least one set of content by time; based on the sorted at least one set of content, use an agent to perform multiple subtasks, each subtask being configured to generate information in a template portion of the predetermined summary template; and determine the summary information based on the information from the multiple template portions generated by the multiple subtasks.

[0106] In some embodiments, the summary information includes pending items, and the summary module 440 is further configured to: use a first machine learning model to identify at least one item in at least one content set that has pending semantic content; for an item in the at least one item, determine whether the corresponding content of the item has a processed semantic identifier in a subsequent time; and in response to the corresponding content of the item not having a processed semantic identifier, determine the item as a pending item.

[0107] In some embodiments, the summary information includes key items and other items, and the summary module 440 is further configured to: use a second machine learning model to identify at least one item in at least one content set; merge items belonging to multiple phases of the same project in at least one item; and classify items in at least one item into key items and other items based on at least one of the corresponding time length, number of participants, project deliverables, or predetermined identifiers of the merged at least one item.

[0108] In some embodiments, the summary information includes planned items for the next time frame, and the summary module 440 is further configured to: use a third machine learning model to identify at least one item in at least one content set that has planned semantic content or semantic content related to the next time frame; and identify other items in the at least one item, other than those identified as needing to be processed within the target event frame and completed items, as planned items for the next time frame.

[0109] The modules included in device 400 can be implemented in various ways, including software, hardware, firmware, or any combination thereof. In some embodiments, one or more units may be implemented using software and / or firmware, such as machine-executable instructions stored on a storage medium. In addition to or as an alternative to machine-executable instructions, some or all of the units in device 400 may be implemented at least partially by one or more hardware logic components. By way of example and not limitation, exemplary types of hardware logic components that may be used include field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-chips (SoCs), complex programmable logic devices (CPLDs), and so on.

[0110] Figure 5 A block diagram of an electronic device 500 in which one or more embodiments of the present disclosure may be implemented is shown. It should be understood that... Figure 5 The electronic device 500 shown is merely exemplary and should not be construed as limiting the functionality and scope of the embodiments described herein.

[0111] like Figure 5 As shown, electronic device 500 is in the form of a general-purpose electronic device. Components of electronic device 500 may include, but are not limited to, one or more processors or processing units 510, memory 520, storage device 530, one or more communication units 540, one or more input devices 550, and one or more output devices 560. Processing unit 510 may be a physical or virtual processor and is capable of performing various processes according to programs stored in memory 520. In a multiprocessor system, multiple processing units execute computer-executable instructions in parallel to improve the parallel processing capability of electronic device 500.

[0112] Electronic device 500 typically includes multiple computer storage media. Such media can be any available media accessible to electronic device 500, including but not limited to volatile and non-volatile media, removable and non-removable media. Memory 520 can be volatile memory (e.g., registers, cache, random access memory (RAM)), non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory), or some combination thereof. Storage device 530 can be removable or non-removable media and can include machine-readable media, such as flash drives, disks, or any other media capable of storing information and / or data and accessible within electronic device 500.

[0113] Electronic device 500 may further include additional removable / non-removable, volatile / non-volatile storage media. Although not explicitly stated... Figure 5As shown, disk drives for reading from or writing to removable, non-volatile disks (e.g., "floppy disks") and optical disk drives for reading from or writing to removable, non-volatile optical disks can be provided. In these cases, each drive can be connected to a bus (not shown) via one or more data media interfaces. Memory 520 may include computer program product 525 having one or more program modules configured to perform various methods or actions of various embodiments of this disclosure.

[0114] Communication unit 540 enables communication with other electronic devices via a communication medium. Additionally, the functionality of components of electronic device 500 can be implemented using a single computing cluster or multiple computing machines capable of communicating via communication connections. Therefore, electronic device 500 can operate in a networked environment using logical connections to one or more other servers, network personal computers (PCs), or another network node.

[0115] Input device 550 can be one or more input devices, such as a mouse, keyboard, trackball, etc. Output device 560 can be one or more output devices, such as a monitor, speaker, printer, etc. Electronic device 500 can also communicate with one or more external devices (not shown) via communication unit 540 as needed. These external devices include storage devices, display devices, etc., and can communicate with one or more devices that enable user interaction with electronic device 500, or with any device that enables electronic device 500 to communicate with one or more other electronic devices (e.g., network card, modem, etc.). Such communication can be performed via input / output (I / O) interface (not shown).

[0116] According to an exemplary implementation of this disclosure, a computer-readable storage medium is provided that stores one or more computer instructions, wherein the one or more computer instructions are executed by a processor to implement the methods described above. According to an exemplary implementation of this disclosure, a computer program product is also provided, which is tangibly stored on a non-transient computer-readable medium and includes computer-executable instructions that are executed by a processor to implement the methods described above.

[0117] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products implemented according to this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0118] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processing unit of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner. Thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0119] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions that execute on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0120] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction, which contains one or more executable instructions for implementing the specified logical function. In some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0121] Various implementations of this disclosure have been described above. The foregoing description is exemplary and not exhaustive, nor is it limited to the disclosed implementations. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described implementations. The terminology used herein is chosen to best explain the principles, practical applications, or improvements to technology in the market, or to enable others skilled in the art to understand the implementations disclosed herein.

Claims

1. A method for generating a summary, comprising: In response to a summary generation request, obtain schedule data and document data within the time range indicated by the summary generation request; Based on the schedule data and the document data, at least one topic is identified; Based on the at least one topic, generate at least one content set corresponding to each of the at least one topic from the schedule data and the document data; as well as Based on the at least one set of content, summary information corresponding to the time range is generated.

2. The method of claim 1, wherein the schedule data includes the corresponding name, category, and time information of at least one schedule within the target time range, and The document data includes the content, category, and time information of at least one document within the target time range.

3. The method of claim 1, wherein determining at least one subject comprises: Multiple content fragments are determined from the document data; For each of the multiple content segments, based on the schedule data and the content segment, at least one topic tag related to the content segment is determined; as well as To determine at least one topic, topic tags related to the multiple content fragments are fused based on semantic similarity.

4. The method of claim 3, wherein determining the plurality of content fragments includes at least one of the following: At least one first document in the document data whose content length is lower than a first threshold length is identified as at least one content fragment among the plurality of content fragments; or At least one second document whose content length exceeds the first threshold length is divided into at least two content segments from the plurality of content segments.

5. The method of claim 4, wherein dividing the at least one second document into multiple content segments comprises: In response to the inclusion of at least one semantic boundary identifier in the at least one second document, the at least one second document is divided into the at least two content segments according to the at least one semantic boundary identifier; as well as In response to the absence of the semantic boundary identifier in the at least one second document, the at least one second document is divided into the at least two content segments based on a second threshold length.

6. The method of claim 3, wherein determining at least one hashtag associated with the content fragment includes at least one of the following: In response to the content fragment being related to at least one event in the schedule data, the name of the at least one event is determined as the at least one topic tag; or Using a language model, at least one topic tag is determined based on the semantic content of the content fragment and the category of the document to which the content fragment belongs.

7. The method according to claim 1, wherein generating at least one content set corresponding to each of the at least one topic comprises: For the topic of at least one of the topics, Based on the schedule data and multiple content fragments in the document data that correspond to the topic, a set of candidate content is determined; Remove content segments from the candidate content set whose similarity to the topic is lower than a threshold from the candidate content set. as well as The multiple content fragments in the candidate content set are sorted according to the corresponding time information of the documents to which the multiple content fragments belong, so as to obtain the content set corresponding to the topic.

8. The method according to claim 1, wherein the summary information follows a predetermined summary template, and generating the summary information includes: Sort the at least one content set by time; Based on at least one sorted set of content, an agent performs multiple subtasks, each subtask being configured to generate information in a template portion of the predetermined summary template; as well as The summary information is determined based on information from multiple template portions generated from the multiple subtasks.

9. The method of claim 8, wherein the summary information includes pending items, wherein performing the plurality of subtasks includes: Using the first machine learning model, Identify at least one item in the at least one content set that has semantic content to be processed; For any of the at least one of the matters mentioned above. Determine whether the relevant content of this matter has a processed semantic identifier in subsequent times; as well as If the corresponding content of the matter does not have the processed semantic identifier, the matter is identified as the pending matter.

10. The method of claim 8, wherein the summary information includes key points and other matters, wherein performing the plurality of sub-tasks includes: Using a second machine learning model, Identify at least one item from the at least one content set; Merge items belonging to multiple phases of the same project from at least one of the items; as well as Based on at least one of the following: the corresponding time length, number of participants, project outcome, or predetermined identifier of the merged at least one item, the items in the at least one item are divided into the key items and the other items.

11. The method of claim 8, wherein the summary information includes planned items for the next time frame, wherein performing the plurality of subtasks includes: Using a third machine learning model, Identify at least one item in the at least one content set that has planning semantic content or semantic content related to the next time range; as well as The items other than those that have been completed are identified as planned items for the next time frame.

12. An apparatus for generating a summary, comprising: The acquisition module is configured to acquire schedule data and document data within the time range indicated by the summary generation request in response to the summary generation request; The topic determination module is configured to determine at least one topic based on the schedule data and the document data; The content segmentation module is configured to generate at least one content set corresponding to each of the at least one theme from the schedule data and the document data, based on the at least one theme. as well as The summary module is configured to generate summary information corresponding to the time range based on the at least one set of content.

13. An electronic device, comprising: At least one processing unit; as well as At least one memory is coupled to at least one processing unit and stores instructions for execution by the at least one processing unit, which, when executed by the at least one processing unit, cause the electronic device to perform the method according to any one of claims 1 to 11.

14. A computer-readable storage medium having a computer program stored thereon, the computer program being executable by a processor to implement the method according to any one of claims 1 to 11.

15. A computer program product tangibly stored in a computer storage medium and comprising computer-executable instructions that, when executed by a device, cause the device to perform the method according to any one of claims 1 to 11.

Images