Schedule processing method and device

Abstract

The invention discloses a schedule processing method and device, and belongs to the technical field of information. The method comprises the following steps: identifying input information of a user on a first interface; judging whether a schedule processing intention exists or not according to the input information; under the condition that the schedule processing intention exists, acquiring context information associated with the input information; extracting schedule parameter information according to the input information and the context information; and performing schedule processing according to the schedule parameter information, wherein the schedule processing at least comprises any one of the following items: creating a schedule, modifying the schedule or deleting the schedule.

Description

Technical Field

[0001] The present application belongs to the field of information technology, and specifically relates to a schedule processing method and device thereof. Background Art

[0002] Currently, schedule creation is usually done by users manually entering a calendar application or the calendar function of an enterprise office software, requiring users to manually edit the specific events, time, location, people, and other information of the schedule. When users need to change a schedule, they also need to go through multiple steps to enter the corresponding schedule and then modify the schedule information.

[0003] It can be seen that the existing schedule processing method requires manual editing by users, and the operation process is relatively cumbersome, inconvenient and inefficient. Summary of the invention

[0004] The purpose of the embodiments of the present application is to provide a schedule processing method and device thereof, which can solve the problem that the schedule processing method requires manual editing by the user, the operation process is cumbersome, inconvenient and inefficient.

[0005] In a first aspect, an embodiment of the present application provides a schedule processing method, including:

[0006] Identify the user's input information on the first interface;

[0007] Determining whether there is a schedule processing intention based on the input information;

[0008] In the case where there is a schedule processing intention, obtaining context information associated with the input information;

[0009] Extracting schedule parameter information according to the input information and the context information;

[0010] Schedule processing is performed according to the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule, or deleting a schedule.

[0011] In a second aspect, an embodiment of the present application provides a schedule processing device, including:

[0012] An identification module, used to identify the input information of the user on the first interface;

[0013] A judgment module, used for judging whether there is a schedule processing intention according to the input information;

[0014] An acquisition module, used for acquiring context information associated with the input information when there is a schedule processing intention;

[0015] An extraction module, configured to extract schedule parameter information according to the input information and the context information;

[0016] The schedule processing module is used to perform schedule processing according to the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule or deleting a schedule.

[0017] In a third aspect, an embodiment of the present application provides an electronic device, which includes a processor and a memory, wherein the memory stores programs or instructions that can be run on the processor, and when the program or instructions are executed by the processor, the steps of the method described in the first aspect are implemented.

[0018] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a program or instruction is stored, and when the program or instruction is executed by a processor, the steps of the method described in the first aspect are implemented.

[0019] In a fifth aspect, an embodiment of the present application provides a chip, comprising a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the method described in the first aspect.

[0020] In a sixth aspect, an embodiment of the present application provides a computer program product, which is stored in a storage medium and is executed by at least one processor to implement the method described in the first aspect.

[0021] In an embodiment of the present application, the user's input information on the first interface is identified; based on the input information, it is determined whether there is an intention to process a schedule; if there is an intention to process a schedule, context information associated with the input information is obtained; based on the input information and the context information, schedule parameter information is extracted; and schedule processing is performed based on the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule, or deleting a schedule. In this way, by analyzing and determining the user's schedule processing intention based on the user's input information, when the intention is identified, the context information input by the user is contacted, the schedule parameters are extracted, and the schedule creation, modification, deletion, and other processing are automatically completed, thereby realizing the automated processing of the schedule without the need for manual operation by the user, making the creation of the schedule and other processing more convenient and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of a schedule processing method provided in an embodiment of the present application;

[0023] Figure 2a to Figure 2c is a schematic diagram of an interface for different input behaviors provided in an embodiment of the present application;

[0024] Figure 3 is a schematic diagram of a user input matching schedule trigger word library provided in an embodiment of the present application;

[0025] Figure 4 is a schematic diagram of a setting interface for starting a schedule recognition switch provided in an embodiment of the present application;

[0026] Figure 5 It is a schematic diagram of the input method interface provided by the embodiment of the present application showing the recognition of the intention of manually triggering schedule creation;

[0027] Figure 6 is a schematic diagram of an interface for screen content recognition provided in an embodiment of the present application;

[0028] Figure 7 It is a schematic diagram of an interface showing a task viewing button in an input method interface provided in an embodiment of the present application;

[0029] Figure 8 This is a schematic diagram of a display interface of today's schedule list provided in an embodiment of the present application;

[0030] Fig. 9 is a schematic diagram of a display interface of a future schedule list provided in an embodiment of the present application;

[0031] Fig.10 It is a schematic diagram of an interface for modifying the conflicting schedule status provided by an embodiment of the present application;

[0032] Fig.11 It is a schematic diagram of the Transformer model architecture provided in an embodiment of the present application;

[0033] Fig.12 It is a schematic diagram of the expansion of the Transformer model architecture provided in the embodiment of the present application;

[0034] Fig.13 is an interactive flow chart of a method for creating a schedule provided in an embodiment of the present application;

[0035] Fig.14 is a detailed flow chart of triggering schedule creation provided by an embodiment of the present application;

[0036] Fig.15 is a schematic diagram of a chat interface including conversation content provided in an embodiment of the present application;

[0037] Fig.16 is a schematic diagram of a display interface for generating a schedule provided in an embodiment of the present application;

[0038] Fig.17 This is a schematic diagram of an interface for displaying a created schedule list provided in an embodiment of the present application;

[0039] Fig.18 It is a flowchart of a method for creating a schedule in a chat scenario provided by an embodiment of the present application;

[0040] Fig.19 It is a schematic diagram of the interactive process of creating a schedule in a chat scenario provided by an embodiment of the present application;

[0041] Fig. 20 is a structural diagram of a schedule processing device provided in an embodiment of the present application;

[0042] Fig.21 is a structural diagram of an electronic device provided in an embodiment of the present application;

[0043] Fig. 22 It is a hardware structure diagram of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0045] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way are interchangeable where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0046] In order to make the embodiments of the present application clearer, the relevant technical knowledge, technical background, invention objectives and improvements involved in the present application are first introduced as follows:

[0047] T2A: Text2Action, text to action

[0048] OCR: Optical Character Recognition

[0049] NLP: Natural Language Processing

[0050] Large model: refers to a pre-trained deep learning model or a local large model based on the Transform architecture that runs on a local device and is used to process and understand natural language text.

[0051] At present, input methods on mobile devices are mainly used to process user input text and provide functions such as spell checking, auto-completion, and emoticon recommendation. However, with the increase in user needs, traditional input methods have been unable to meet the growing and diverse needs of users, especially in terms of user schedules, task reminders, etc. Therefore, the functions of existing input methods are insufficient. The main defects are as follows:

[0052] It can only process simple text input and cannot deeply understand the user's intention and context; it lacks automated task scheduling and reminder functions, and users need to operate manually; it cannot effectively use the user's historical input data for intelligent prediction and behavior recommendation.

[0053] This application aims to solve the shortcomings of existing input methods in user schedule management experience, and provide a Text2Action method based on a local big model, which automatically generates schedules and reminders by analyzing the user's input content, thereby improving user efficiency and experience. The main improvements include: 1. Achieving in-depth understanding and intent recognition of user input content; 2. Introducing a local big model to improve the intelligence of the input method; 3. Automatically generating and managing user schedules and task reminders.

[0054] The key points of the embodiments of the present application are: when a user inputs content in a chat conversation or other scenario, the continuously input content is triggered to combine the input content with the contextual memory content of the user input scenario that is locally identified and cached, and by understanding the user's input and context, it is predicted whether the user's intention includes the need to create or modify a schedule or other behavior.

[0055] The core point of the embodiment of the present application is: combining the input method scenario and the big model, automatically triggering the method of identifying the user's conversation intention and creating a schedule, providing an improved schedule creation trigger strategy algorithm, accurately identifying the user's intention through natural language processing technology, and generating a suitable schedule, thereby improving the intelligence and automation level of schedule management.

[0056] The embodiment of the present application only needs to predict the content of the conversation through the input content of the user in the input method, so as to identify the user's intention, identify the intention to create a schedule, manually or automatically trigger the screen recognition, obtain the context information of the screen recognition content, aggregate the context information of the screen recognition and the context information of the user input, and the aggregated content needs to be well marked with the source. The aggregated data is submitted to the end-side big model to construct the parameters for extracting the creation of the schedule, and the user's input and the content of the screen recognition are segmented, extracted and summarized in the end-side big model. According to the rules of the defined schedule-related prompt words, the schedule information generated according to the prompt word rules is output. The end-side big model can understand the conversation content and construct a time-series schedule task according to the user's conversation content. In order to predict the intention of the conversation only through the input content of the user, it is necessary to record and recognize the input scene, object, time, location and other information to train the intention recognition model with multi-dimensional data.

[0057] The key points in the embodiments of the present application are described as follows:

[0058] 1. Identify the intention to create a schedule during typing, copying and pasting;

[0059] 2. Preliminarily extract the parameter context required to create a schedule based on the screen content and user input;

[0060] 3. Use the client-side big model to further extract the parameters in point 2 to obtain parameters that conform to the schedule protocol format;

[0061] 4. Generate schedules and differentiate between schedules for the past and future;

[0062] The schedule processing method and related model structure, training method and working principle provided in the embodiment of the present application are described in detail below through specific embodiments and their application scenarios in combination with the accompanying drawings.

[0063] See also Figure 1 , Figure 1 A flowchart of a schedule processing method provided in an embodiment of the present application, such as Figure 1 As shown, the method comprises the following steps:

[0064] Step 101: Identify user input information on a first interface.

[0065] The first interface may be any application interface that supports user input of text, voice and other information. Common application scenarios include a conversation chat interface of an instant messaging application, an editing interface of a memo or notepad, and the like.

[0066] The input information may be text, voice or other information input by the user in the first interface, or may be text information copied and pasted. Figure 2a to 2cAs shown, there are several user input behaviors:

[0067] Typing Input: Text input using 9-key and 26-key keyboards.

[0068] Voice input: converted into text through voice recognition.

[0069] Copy and paste: Text content that users copy and paste from other apps.

[0070] In this step, the user's input information in the first interface can be obtained and recognized, such as obtaining text typed by the user through the input method, voice-to-text, copied and pasted text content, etc., and the obtained input information can be recognized, such as converting non-text information into text information, recognizing the semantics or implied intentions of the text information, etc.

[0071] Step 102: Determine whether there is a schedule processing intention based on the input information.

[0072] In this step, the user's input information can be analyzed to determine whether the information contains schedule processing intentions such as creating a schedule, canceling a schedule, and changing a schedule. Specifically, based on the characteristics or keywords of each schedule processing intention, it can be determined whether the user's input information contains corresponding characteristics or keywords, or whether it matches these characteristics or keywords. Alternatively, the input information can be subjected to natural language understanding, context understanding, etc., and then based on the understanding results, it can be determined whether the above-mentioned schedule processing intention exists.

[0073] It should be noted that the schedule processing intent in the embodiment of the present application can refer to the schedule processing intent in the corresponding scenario according to different application scenarios, such as the scenario for identifying the schedule creation intent can refer to the schedule creation intent, the scenario for identifying the schedule change intent can refer to the schedule change intent, etc. Of course, the schedule processing intent can cover various different schedule processing intents such as creation, cancellation, and change, so when it is determined that any of these intents exists in the user input information, it can be considered that there is a schedule processing intent.

[0074] For example, when the user inputs information such as "going to a meeting tomorrow", since the schedule feature information such as "tomorrow" and "meeting" are recognized, it can be determined that there is currently a schedule processing intention, and specifically a schedule creation intention.

[0075] When the user inputs information such as "I am not going to the meeting tomorrow", since the schedule feature information such as "tomorrow", "not going", and "meeting" are recognized, it can be determined that there is currently a schedule processing intention, and specifically a schedule cancellation intention.

[0076] When the user inputs the information "The meeting tomorrow has been changed to 6 o'clock", since the schedule feature information such as "tomorrow", "changed", "6 o'clock", and "meeting" are recognized, it can be determined that there is currently a schedule processing intention, and specifically a schedule changing intention.

[0077] Optionally, step 102 includes:

[0078] Performing word segmentation processing on the input information to obtain at least one word;

[0079] Matching the at least one word with a trigger word in a first schedule trigger word library;

[0080] When a first word successfully matches a trigger word in the first schedule trigger word library, it is determined that there is a schedule processing intention, wherein the first word is any word among the at least one word.

[0081] The above-mentioned first schedule trigger vocabulary library can be pre-built. In some embodiments, by pre-building some schedule trigger vocabulary libraries, the words in the user input information can be matched with the pre-built schedule trigger vocabulary libraries, and then whether there is a schedule processing intention is determined based on whether it is successfully matched with a trigger word in a schedule trigger vocabulary library.

[0082] Specifically, the user input information may be first segmented to separate the input information into individual words, and then each word may be matched with the constructed first schedule trigger word library to determine whether it matches the trigger word therein. The constructed first schedule trigger word library may be a general word library covering various schedule trigger words, or a trigger word library divided into multiple different parts of speech according to part of speech, or a trigger word library divided into multiple different schedule intent types according to different schedule processing intents.

[0083] For example, when trigger word libraries of different parts of speech are constructed according to the parts of speech, the words of different parts of speech in the user input can be matched with the trigger word libraries of the corresponding parts of speech, such as Figure 3 As shown, the verb "meeting" in the user input can be matched with the verb trigger vocabulary of the schedule, the question word "what" in the user input can be matched with the question word trigger vocabulary of the schedule, and so on.

[0084] When it is determined that a certain word in the user input information successfully matches any trigger word in the first schedule trigger vocabulary, it can be determined that the current user is very likely to have a schedule processing intention. Conversely, when there is no word in the user input information that successfully matches any trigger word in the first schedule trigger vocabulary, it is considered that the current user does not have a schedule processing intention.

[0085] Exemplarily, if the word "attend a meeting" in the user input information "Attend a meeting tomorrow" matches the verb trigger word "attend a meeting" in the schedule trigger word library, it is determined that there is a current schedule processing intention. For the user input information "Are you off work?", since none of the words match the trigger words in the schedule trigger word library, it is determined that there is no current schedule processing intention.

[0086] In this way, in this embodiment, it is possible to accurately and quickly trigger the schedule creation intention by constructing a schedule trigger word library and matching trigger words.

[0087] Optionally, the first schedule trigger word library includes multiple trigger word libraries corresponding to different schedule processing intentions;

[0088] The matching of the at least one word with the trigger words in the first schedule trigger word library includes:

[0089] Matching the at least one word with the trigger words in the multiple trigger word libraries respectively;

[0090] When the first word matches successfully with the trigger words in the first schedule trigger word library, determining that there is a schedule processing intention includes:

[0091] When the first word matches successfully with the trigger words in the target trigger word library, determining that there is the schedule processing intention corresponding to the target trigger word library;

[0092] Wherein, the schedule processing matches the schedule processing intention corresponding to the target trigger word library.

[0093] That is, in some embodiments, multiple trigger word libraries corresponding to different schedule processing intentions can be pre-constructed. For example, the pre-constructed schedule-related trigger word libraries include a schedule creation trigger word library, a schedule deletion trigger word library, a conflicting schedule trigger word library, an updated schedule trigger word library, etc.

[0094] Among them, the schedule creation trigger word library may include the following trigger words:

[0095] 1. Verb trigger: attend a meeting, exercise, arrange, organize, book, plan, confirm, cancel, update, adjust, set, participate, etc.;

[0096] 2. Interrogative word trigger: when, where, which, what time, who, what, why, how, what time, date, etc.;

[0097] 3. Noun trigger: meeting, reminder, activity, seminar, conference call, travel, deadline;

[0098] 4. Geographical location trigger: AB Barbecue, North Ring Road, airport, hotel, school, company, city name, building name, etc.;

[0099] 5. Adjective triggers: good, important, urgent, priority, formal, all day, team, regular, etc.

[0100] The trigger word library for deleting a schedule may include the following trigger words:

[0101] 1. Verb trigger: delete, cancel, remove, revoke, erase, clear, end, terminate, dissolve, withdraw, etc.;

[0102] 2. Question word triggers: when, where, why, how, whether, have, how, what, what time, date, etc.;

[0103] 3. Noun triggers: schedule, meeting, reminder, activity, event, appointment, plan, task, meeting, etc.;

[0104] 4. Geographic location trigger: Log House BBQ, North Ring Road, Airport, Hotel, School, Company, City Name, Building Name, etc.;

[0105] 5. Adjective triggers: urgent, unimportant, low priority, temporary, not required, formally cancelled, canceled, etc.

[0106] The conflicting schedule trigger word library may include the following trigger words:

[0107] 1. Word triggers: conflict, overlap, collision, detect conflict, check conflict, find conflict, resolve conflict, handle conflict, avoid conflict, etc.;

[0108] 2. Question word triggers: when, where, whether, whether there is, how, how, what conflicts, what overlaps, whether there is a time conflict, etc.;

[0109] 3. Word triggers: schedule, meeting, reminder, activity, event, appointment, plan, task, meeting, etc.;

[0110] 4. Geographic location trigger: Log House BBQ, North Ring Road, Airport, Hotel, School, Company, City Name, Building Name, etc.;

[0111] 5. Adjective triggers: conflicting, overlapping, repetitive, competing, temporal, etc.

[0112] The update schedule trigger word library may include the following trigger words:

[0113] 1. Verb triggers: update, modify, change, adjust, rearrange, refresh, edit, revise, etc.;

[0114] 2. Question word triggers: when, where, how, whether, need, can, how, what, what, what time, date, etc.;

[0115] 3. Noun triggers: schedule, meeting, reminder, activity, event, appointment, plan, task, meeting, etc.;

[0116] 4. Geographic location trigger: Log House BBQ, North Ring Road, Airport, Hotel, School, Company, City Name, Building Name, etc.;

[0117] 5. Adjective triggers: updated, modified, changed, adjusted, new, latest, etc.

[0118] In this way, each word in the user input information can be matched with the trigger words in the above multiple trigger word libraries, and when the user input information contains content that matches any of the above trigger word libraries, it is determined that there is a schedule processing intention corresponding to the matching trigger word library, and the subsequent automatic recognition of screen content and extraction of schedule parameters can be triggered to perform corresponding schedule intention processing. For example, when the trigger word "update" in the update schedule trigger word library is matched, it is considered that there is a schedule update intention at present, so the relevant parameters of the schedule update can be extracted and the schedule update processing can be performed.

[0119] In this implementation, by constructing a schedule triggering vocabulary library with different schedule processing intentions, more accurate schedule intention triggering can be achieved, ensuring automated schedule generation and processing based on accurate schedule processing intentions, thereby improving the accuracy and effectiveness of automatic schedule processing.

[0120] Step 103: When there is a schedule processing intention, obtain context information associated with the input information.

[0121] When it is recognized that the user's input information contains a schedule processing intention, the user's schedule processing intention can be further understood and analyzed in conjunction with the context information input by the user to obtain specific schedule parameter information, so as to automatically generate a schedule based on the schedule parameter information and realize the user's corresponding schedule processing intention.

[0122] Specifically, the above-mentioned acquisition of context information associated with the input information can be to acquire the context information associated with the user input information existing on the current interface, such as the chat content between the user and the chat object on the current interface, or the historical chat records related to the current chat topic, such as the chat records related to the same schedule as the current schedule mentioned yesterday. In specific implementation, the context information can be obtained by identifying the screen content, that is, identifying the conversation content displayed on the current interface, or by matching the historical chat records to obtain a certain chat record associated with the current input information as the context information.

[0123] Optionally, the first interface is a chat interface;

[0124] The acquiring of context information associated with the input information includes at least one of the following:

[0125] Extracting the dialog information displayed on the first interface, wherein the context information includes the dialog information;

[0126] Acquire historical chat records with the chat object on the first interface; if there is a target chat record matching the input information in the historical chat records, acquire the target chat record, wherein the context information includes the target chat record.

[0127] In some embodiments, the first interface is a chat interface, that is, the embodiments of the present application can be applied to a chat scenario to identify and process schedule intentions based on the user's input content and conversation content.

[0128] Specifically, in one implementation, when the schedule processing intention is triggered, the conversation information displayed on the current chat interface can be further extracted as the context information of the current input. Specifically, the conversation information can be extracted by triggering screen content recognition. There are many ways to recognize screen content. For example, a screenshot of the current chat interface is first taken, and then the conversation content is extracted through an OCR recognition method.

[0129] Among them, the method of triggering screen content recognition may include manual triggering and automatic triggering, such as Figure 4 As shown, the user can turn on the automatic recognition of schedules through the settings in the input method. If the user turns on the automatic schedule recognition switch 41, the screen content recognition will be directly entered when the current input triggers the schedule processing intention. Otherwise, Figure 5 As shown, the user needs to manually click the "Create Schedule" button 51 to trigger the screen content recognition.

[0130] After triggering screen content recognition, such as Figure 6 As shown, the context of user input can be extracted, and the recognized content includes Figure 6 The name of the boss, the content of the conversation, etc. in the conversation will enter the recognition state at the same time. The recognized content can be stored in a database with a time index established by rules, and the stored content includes people, time, place, content, etc. Specifically, the recorded user input information can also be sorted and classified according to different dimensions, including time, source, place, person and context information. For example:

[0131] Time: Record the specific time of input.

[0132] Source: Record the specific application source of the input, such as keyboard, screenshot, copy and paste, etc.

[0133] Location: The specific location at the time of entry is recorded through the device's geolocation service.

[0134] People: Identify and record chat partners and related people mentioned in the input content.

[0135] Context: Record the specific situation and background information of the input content.

[0136] After the context information input by the user is sorted, a timeline index can be constructed in chronological order to form a comprehensive context information of the user input. For example, the sorted user input and context information can be shown in the following table:

[0137]

[0138] Among them, the main purpose of recording the scene source is to classify the input data source, and different schedule intention trigger strategies can be provided according to different source classifications. For example, typing pre-screening is triggered by words or phrases; recording or cutting is triggered by sentences or phrases, or a piece of text. In other words, in order to classify historical data and match trigger rules, different trigger strategies can be selected according to different sources to improve the trigger accuracy.

[0139] In the second implementation, when the schedule processing intention is triggered, the historical chat records with the current chat object can also be queried based on the current input information to determine whether the historical chat records related to the current input information can be matched. In specific implementation, the query range can be to query the recent chat records cached locally, such as today's chat records, chat records in the past three days, chat records in the past week, etc. If a certain historical chat record is matched and set as the target chat record, the chat record can be obtained as the context information of the current input.

[0140] As another implementation method, you can also first match historical chat records based on the current input information. When relevant historical chat records are matched, the successfully matched target chat records are used as the context information of the current input. When no relevant historical chat records are matched, you can trigger screen content recognition to extract the conversation information displayed on the current chat interface as the context information of the current input.

[0141] It should be noted that the above process of matching historical chat records can also be changed to matching historical schedules, and the successfully matched historical schedule information can be used as the context information of the current input, so that the schedule information in the current input and the historical schedule information can be combined to decide how to update the schedule or recreate the schedule.

[0142] Through this implementation, more complete schedule information can be obtained based on the schedule-related topic currently input by the user, combined with the current conversation content and the same schedule topic discussed in history, to ensure that subsequent related schedule processing is performed accurately.

[0143] Step 104: extract schedule parameter information according to the input information and the context information.

[0144] Specifically, in this step, the input information and the context information may be aggregated, and then the aggregated information may be comprehensively understood and analyzed to determine which are schedule parameters, thereby extracting schedule-related parameter information, such as time, place, person, event, and other element information that a schedule usually has.

[0145] For example, the content currently input by the user "tomorrow at 6 pm" and the conversation context of the sentence content can be aggregated to form a complete chat dialogue process, and then by understanding the content of the entire dialogue process, the schedule parameters required for schedule creation, such as the time "tomorrow at 6 pm", the person "Boss Li", and the event "meeting a client", can be extracted.

[0146] Step 105: Perform schedule processing according to the schedule parameter information, wherein the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule, or deleting a schedule.

[0147] In this step, based on the extracted schedule parameter information, schedule generation processing can be automatically performed, such as creating a schedule, modifying a schedule, deleting a schedule, etc. Specifically, based on the identified corresponding schedule processing intention, the corresponding schedule processing method can be adopted to eliminate the need for manual editing by the user. It should be noted that when multiple schedule processing intentions are identified, multiple schedule processing can be performed accordingly, such as creating a schedule and then modifying or deleting another schedule that conflicts with the currently created schedule.

[0148] For example, if schedule parameters such as time "tomorrow at 6 pm", person "Boss Li", event "meeting a client" are extracted, a schedule reminder for meeting Boss Li at 6 pm tomorrow can be created, and a corresponding schedule reminder can be generated, such as reminding the user of the schedule at 5 pm tomorrow.

[0149] It should be noted that when creating a schedule based on the extracted schedule parameter information, it can also be determined whether it is a past schedule. If it is not a past schedule, the user's schedule is automatically generated based on the extracted schedule parameter information, and the schedule includes the user's tasks and reminders, such as meetings, shopping, appointments, etc. The created schedule can be displayed through the input method interface or related applications, which is convenient for users to view and manage.

[0150] Optionally, the input information is information input through an input method application;

[0151] The step 105 comprises:

[0152] The schedule application programming interface (API) is called to perform schedule processing according to the schedule parameter information.

[0153] Triggering schedule intention recognition through input of input method application is an important application scenario of the embodiments of the present application. That is, in some embodiments, schedule intention recognition can be triggered when the user types text information or converts speech to text through the input method, and when the schedule processing intention is successfully triggered, it is necessary to automatically create, update, delete, etc. the schedule according to the extracted schedule parameters, that is, it is necessary to jump from the current input method application to the calendar or schedule application to create, update, or delete the schedule, etc. At this time, the schedule agent service can be used to call the schedule API interface to execute the steps of schedule processing according to the schedule parameter information. In this way, the automatic generation of schedules based on input method application triggering can be achieved.

[0154] Optionally, a schedule viewing control is displayed on the keyboard interface of the input method application; the method further includes:

[0155] When a touch input to the schedule viewing control is received, a schedule list interface is displayed.

[0156] In some embodiments, in order to facilitate the user to jump from the input method interface to view the created schedule list, a schedule viewing control can be displayed on the keyboard interface of the input method application, and the user can use the control, such as Figure 7 As shown, when a schedule task reminder is currently created, the expanded interface of the input method displays an entry for viewing the task. Click the "View Task" button 71 to view the schedule list and task details.

[0157] When receiving input from the user touching the schedule viewing control, the schedule list can be displayed for the user to view the created schedule. Specifically, the user can jump from the current interface to the interface displaying the schedule list, such as jumping to the schedule application interface, and displaying the created schedule list information in the schedule application interface.

[0158] For example, when you click the "View Tasks" button 71, you can jump to the following Figure 8 In the calendar interface shown, May 10 is today's related schedule task. Of course, you can also improve the task classification, sort and select tasks, define priorities, etc. Fig. 9 As shown, the user can also click on May 11 in the calendar to view future planned or agreed schedule tasks and be reminded at the corresponding time.

[0159] Through this implementation, the user can be supported to jump from the input method interface to view the schedule list.

[0160] Optionally, when the schedule processing includes creating a schedule, the method further includes:

[0161] Check whether the currently created schedule conflicts with the historical schedule;

[0162] When the currently created schedule conflicts with the historical schedule, a schedule conflict reminder is output, or a schedule modification suggestion is output.

[0163] Check whether the currently created schedule conflicts with the historical schedule. If there is a conflict, record the conflict.

[0164] In some embodiments, after creating a schedule based on the currently extracted schedule parameters, schedule conflict detection can be further performed, that is, detecting whether the current schedule conflicts with a historically created schedule, such as time overlap, location conflict, etc. If a conflict is detected, a schedule conflict reminder can be output to prompt the user that the current schedule conflicts with a historical schedule, so that the user can decide whether to modify a certain schedule, or corresponding schedule modification suggestions can be generated based on the current conflict and output for user reference, that is, alternative plans or suggestions can be generated, such as suggesting adjustment of the current time or location, and subsequent interaction with the user can be carried out to finally confirm the new schedule.

[0165] It should be noted that in some embodiments, conflict detection may be performed before creating the current schedule. If a conflict is detected, a schedule conflict reminder or schedule modification suggestion is output first, and then the schedule is created or the creation of the schedule is canceled based on the user's confirmation.

[0166] In the scenario where a historical schedule conflicts with the current schedule, the schedule can be modified according to the content input by the user, and the schedule reminder can be triggered in advance, and the corresponding schedule status can be marked. Fig.10 As shown, for example, when the user inputs "Boss, there is a project meeting tomorrow morning, I need to take a leave to attend a parent-teacher meeting", the leave request is recognized, the schedule parameter extraction is triggered, and the status of the conflicting schedule is modified.

[0167] In this way, in this implementation, conflicting schedule detection can avoid creating conflicting schedules that cause trouble to users, and can also provide users with corresponding schedule adjustment suggestions to help users better manage their schedules.

[0168] Optionally, step 104 includes:

[0169] Sorting the input information and the context information in chronological order to generate model input data;

[0170] Input the model input data and the first prompt word Prompt information into a pre-trained target model, and perform schedule parameter extraction processing on the model input data according to the first prompt word Prompt information through the target model;

[0171] Acquire the schedule parameter information output by the target model.

[0172] In some embodiments, a model may be used to deeply understand user input information and context information, and combined with given prompt word information, schedule parameter information that complies with schedule rules may be extracted.

[0173] The above-mentioned target model can preferably be a large model on the end, that is, a local large model, such as a large-scale language model (LLM). Because the cloud model will obtain user data, the use of a local large model can effectively ensure user privacy and make the end-side implementation feasible.

[0174] The first prompt word Prompt information may be corresponding Prompt information pre-constructed according to the characteristics of the schedule, and is used to guide the end-side big model to extract schedule parameters and output the schedule parameters in a specified format.

[0175] Specifically, the input information and the context information may be sorted in chronological order, such as being combined in chronological order of information issuance. Taking the input conversation text and previous and subsequent conversation records as an example, they may be combined into one conversation text in chronological order of sending / receiving timestamps of each conversation text.

[0176] Among them, prompt word engineering is an important technology and method in the field of natural language processing (NLP), which is mainly used to guide the pre-trained LLM to generate high-quality, accurate and targeted output. Therefore, in this implementation, according to the characteristics of the schedule, the corresponding prompt information can be pre-built to guide the end-side large model to extract schedule parameters and output the schedule parameters in a specified format.

[0177] In this way, the user input information and context information can be aggregated as model input data, combined with the first Prompt information, and input into the end-side large model. The end-side large model extracts schedule parameters such as time, place, event, person and other information from the model input data according to the guidance of the first Prompt prompt, and outputs the extracted schedule parameters in the specified format.

[0178] The following is a simple example to illustrate the application of Prompt. Assume that the Prompt is "Please write a schedule", which includes the following content:

[0179] ·Schedule name

[0180] Schedule

[0181] ·Agenda and location

[0182] Participants

[0183] ·Agenda

[0184] ·Schedule Notes

[0185] The large model is based on such a prompt word, and the output is as follows:

[0186] Example 1:

[0187] {

[0188] "schedule":[

[0189] {

[0190] "meeting_name":"AI Technology Seminar",

[0191] "meeting_time":"2024-07-30 10:00-12:00",

[0192] "meeting_location":"Meeting Room A",

[0193] "participants":[

[0194] "Zhang San",

[0195] "Li Si",

[0196] "Wang Wu",

[0197] "Zhao Liu"

[0198] ],

[0199] "agenda":[

[0200] "10:00-10:10 Welcome Speech",

[0201] "10:10-10:40 Latest progress in AI technology",

[0202] "10:40-11:00 Tea Break",

[0203] "11:00-11:30 AI technology application case sharing",

[0204] "11:30-12:00 Discussion and Exchange"

[0205] ],

[0206] "remarks":"Please attend on time and prepare your sharing content in advance."

[0207] }, Example 2:

[0208] {

[0209] "meeting_name":"Project progress report",

[0210] "meeting_time":"2024-07-30 14:00-16:00",

[0211] "meeting_location":"Meeting Room B",

[0212] "participants":[

[0213] "Project Manager",

[0214] "Development Team",

[0215] "Test Team",

[0216] "Product Manager"

[0217] ],

[0218] "agenda":[

[0219] "14:00-14:15 Project progress report",

[0220] "14:15-14:45 Problem discussion and solutions",

[0221] "14:45-15:00 Tea break",

[0222] "15:00-15:30 Next stage work arrangements",

[0223] "15:30-16:00 Free discussion"

[0224] ],

[0225] "remarks":"Please prepare your report materials in advance."

[0226] } ]

[0228] }

[0229] Through this implementation, it can be ensured that the schedule parameters such as creating and modifying schedules that meet the schedule format specifications are accurately and quickly extracted through the model.

[0230] Furthermore, the target model is a Transform model, and the Transform model includes an N-layer encoder Encoder and an M-layer decoder Decoder, where N and M are both integers greater than 1;

[0231] Each layer of Encoder includes a multi-head self-attention layer, a first splicing and fusion layer, a feedforward network layer, and a second splicing and fusion layer, and the multi-head self-attention layer, the first splicing and fusion layer, the feedforward network layer, and the second splicing and fusion layer are stacked in sequence;

[0232] Each layer of Decoder includes a masked multi-head attention layer, a third splicing fusion layer, a multi-head cross attention layer, a fourth splicing fusion layer, a feedforward network layer, and a fifth splicing fusion layer, and the masked multi-head attention layer, the third splicing fusion layer, the multi-head cross attention layer, the fourth splicing fusion layer, the feedforward network layer, and the fifth splicing fusion layer are stacked in sequence;

[0233] The input of the first layer Encoder includes the embedding code and word position code of the model input data;

[0234] The output of the Nth layer Encoder is the input of the multi-head cross attention layer in each layer Decoder;

[0235] The input of the first layer Decoder includes the target output data predicted based on the model input data.

[0236] In some embodiments, the terminal-side large model for extracting schedule parameters can be used as follows: Fig.11 The Transform model architecture shown is implemented by Fig.11 It can be seen that the Transformer model consists of two parts, the Encoder and the Decoder. First, the Encoder part. The input Embedding enters N Encoders after some processing (Embedding input, position encoding). The N Encoders finally output feature value data, and the result is used as the input of the Multi-Head Cross-Attention layer in each Decoder; then the Decoder part. The output Embedding enters M Decoders after some processing (Embedding input, position encoding). After being processed by M Decoder layers, it is processed by the Linear layer and the Softmax layer in turn to obtain the predicted probability result of the model's final target output. M can be equal to N.

[0237] Next, let's look at the detailed structure of each sub-layer within the Encoder:

[0238] An Encoder layer consists of 4 sequentially stacked sub-layers, namely:

[0239] Multi-Head Self-Attention Layer

[0240] First splicing fusion (Add&Norm) layer

[0241] Feed Forward Layer

[0242] The second splicing fusion (Add&Norm) layer

[0243] The data transmission and processing process between the various sub-layers of the Encoder layer is described as follows:

[0244] 1) Convert the inputs into an input embedding matrix, add the positional embedding of each word in the input, and pass the result as input to the first Encoder layer, ENCODER 1.

[0245] 2) In the ENCODER 1 layer, the input is received and sent to the Multi-Head Self-Attention layer, which outputs the attention matrix after operation;

[0246] 3) The attention matrix is ​​input to the next sublayer, the Feed Forward layer, which takes the attention matrix as input and calculates the eigenvalues ​​as output;

[0247] 4) Next, the output from ENCODER 1 is used as input to the next layer, Encoder 2.

[0248] 5) ENCODER 2 performs the same process and takes the feature value of the given input sentence as output, and finally outputs the final feature value data in ENCODER 1 layer N layer.

[0249] Next, let's look at the detailed structure of each sub-layer within the Decoder:

[0250] A Decoder layer consists of 6 sequentially stacked sub-layers, which are:

[0251] Masked Multi-Head Self-Attention Layer

[0252] Add&Norm layer

[0253] Multi-Head Cross-Attention Layer

[0254] Add&Norm layer

[0255] Feed Forward Layer

[0256] Add&Norm layer

[0257] The data transmission and processing process between the various sub-layers of the Decoder layer is described as follows:

[0258] 1) Convert targets to output embedding matrix, add positional embedding to it, and pass the result as input to the first Decoder layer, namely DECODER1 layer; targets is the target output data predicted by the model input data, such as the next word predicted based on the input "what time does it start today" is "meeting";

[0259] 2) In the DECODER 1 layer, the input is accepted and sent to the Masked Multi-Head Self-Attention sublayer, which outputs the attention matrix after operation;

[0260] 3) The attention matrix is ​​input to the next sub-layer Multi-Head Cross-Attention layer. At the same time, the final feature value data output by the ENCODER N layer is used as the input of the Multi-Head Cross-Attention layer. After processing, the result is output to the next sub-layer, namely the Feed Forward layer, and the result is output;

[0261] 4) Next, the output from the decoder DECODER 1 layer is used as input and passed to the next decoder layer DECODER 2;

[0262] 5) DECODER 2 performs the same process and outputs the feature value of the given input sentence. Finally, DECODER M outputs the final target feature value data.

[0263] 6) Finally, after M decoders, the network needs to be processed by the Linear layer and the Softmax layer in sequence to finally obtain the probability result of predicting the occurrence of the target word.

[0264] To facilitate understanding, we can expand the N encoders and M decoders in the Transformer architecture, fill in the internal sub-layers, and integrate the data interaction between the layers into a diagram, and we can get the following: Fig.12 The expanded Transformer architecture shown.

[0265] In this implementation, by using the Transformer architecture as the end-side large model to extract schedule parameters, high model accuracy and performance can be guaranteed, thereby ensuring the accuracy and reliability of the model output.

[0266] In some embodiments, the following large model training method may be used to train the above-mentioned large model on the terminal side:

[0267] 1. Data simulation and collection

[0268] Data simulation:

[0269] Simulation scenarios: Create a variety of simulation data scenarios, including various formats and contents of user input, such as "the meeting starts at 3 pm", "there is a task at 9 am tomorrow", etc.

[0270] Data Generation Tools: Generate simulated data using Natural Language Generation (NLG) tools. Configure the generative model to produce text containing time expressions, event descriptions, and other relevant information.

[0271] Data collection:

[0272] Real data source: Collect data from real users' input methods, including text input, voice recognition results, etc. Ensure that the collection process complies with privacy protection laws and user consent.

[0273] Data type: The collected data includes text input, keywords, event types, time expressions, etc. The context and related tags of user input are recorded.

[0274] Data Tags:

[0275] Tagging rules: Define tagging rules, including identifying and tagging time, date, event type, etc. Use a rule engine or automatic tagging tool for preliminary tagging.

[0276] Manual verification: Manual verification of automatically marked data to ensure that the marking is accurate. The marking staff makes necessary adjustments and corrections.

[0277] 2. Data Preprocessing

[0278] Data cleaning:

[0279] Noise Removal: Clean the data to remove noise and outliers, such as meaningless text or mislabeling.

[0280] Error correction: Correct spelling errors, inaccurate tags, etc. in the data to improve data quality.

[0281] Data Standardization:

[0282] Unified format: Standardize information such as time, date, and event into a unified format, such as the ISO 8601 standard.

[0283] Text normalization: Perform text normalization processing, such as removing extra spaces, unifying vocabulary and grammar rules.

[0284] Data partitioning:

[0285] Training set, validation set, test set: Divide the processed data set into training set, validation set and test set. Generally, the training set accounts for 70-80% of the total data, and the validation set and test set each account for 10-15%.

[0286] 3. Model training

[0287] Model selection:

[0288] Transformer architecture: Select a suitable Transformer model architecture, such as BERT, GPT, RoBERTa, etc. Select a pre-trained model or train from scratch based on task requirements.

[0289] Model configuration:

[0290] Hyperparameter setting: Set the hyperparameters of the model, including learning rate, batch size, training rounds, etc. Adjust the hyperparameters based on the performance of the validation set.

[0291] Input format: Defines the input format of the model, including how text input is processed and how data is encoded.

[0292] Training process:

[0293] Forward propagation: The input data is forward propagated through the Transformer model to calculate the prediction results.

[0294] Loss calculation: Calculate the loss between the predicted result and the actual label. Use a loss function (such as cross entropy loss) for calculation.

[0295] Back propagation: Use the back propagation algorithm to calculate the gradient and update the model parameters. Select an optimization algorithm (such as Adam, RMS prop) for parameter optimization.

[0296] Model Evaluation:

[0297] Evaluation Metrics: Use evaluation metrics (such as precision, recall, F1 score) to evaluate the performance of the model. Calculate the evaluation metrics using the validation set data and adjust the model based on the results.

[0298] Performance tuning: Perform model tuning based on the evaluation results, optimize hyperparameters, improve model architecture, etc.

[0299] 4. Model Optimization

[0300] Model optimization techniques:

[0301] Quantization: Quantize the model and convert the weights and activation values ​​into low-bit representation to reduce the size of the model and the computational requirements.

[0302] Pruning: Reduce the complexity of the model by removing unimportant neural network connections, improving inference speed and resource utilization efficiency.

[0303] Knowledge distillation: training a smaller student model to mimic the behavior of a larger model, thereby reducing the model size and computational burden.

[0304] Performance Test:

[0305] Inference performance: Test the inference performance of the model on different devices, such as response time, memory usage, etc.

[0306] User experience: Evaluate the user experience of the model in actual applications, including accuracy, response speed, interaction effect, etc.

[0307] 5. Model Deployment

[0308] Deployment environment:

[0309] Platform selection: Deployed on appropriate electronic devices.

[0310] Framework support: Use appropriate deployment frameworks and tools (such as TensorFlow Serving, ONNX Runtime) for model deployment.

[0311] Interface integration:

[0312] API design: Design and implement the model's API interface to support client applications to call the model.

[0313] Client integration: Integrate the model interface into the client application to ensure that the application can interact with the model through the API.

[0314] Monitoring and maintenance:

[0315] Performance monitoring: Set up a monitoring system to track the performance indicators and operating status of the model.

[0316] Model update: Establish a model update mechanism to regularly update the model to improve performance and adapt to new data.

[0317] Combine the following Fig.13 and Fig.14 , taking the application scenario of triggering the creation of a schedule by inputting content on an input method as an example to illustrate the specific implementation method of the embodiment of the present application.

[0318] like Fig.13 As shown, the specific steps include the following:

[0319] Step 1301: Text input: User input behavior, which is identified based on the input content. It includes the following input behaviors:

[0320] Typing input: Text input with 9-key and 26-key keyboards, such as Figure 2a shown.

[0321] Voice input: converted into text through voice recognition, such as Figure 2b shown.

[0322] Copy and paste: Text content that users copy and paste from other applications, such as Figure 2c shown.

[0323] Step 1302: Triggering the intention to create a schedule: based on the content recognition on the screen displayed by the user using the input method, triggering the intention to create a schedule.

[0324] Among them, the methods of intent recognition are as follows:

[0325] 1) Based on trigger word rule matching: triggered when the content contains words such as time, event, place, task, etc.

[0326] 2) Based on NLP natural language processing technology, the text data input by the user is processed by word segmentation, part-of-speech tagging, named entity recognition, etc., and key information (such as time, place, event, person and context) is extracted to trigger.

[0327] There are several methods for content identification:

[0328] 1) Natural language input processing:

[0329] Keyword identification: Describe how to use NLP technology to identify keywords in user input, such as "what time does it start", "what's going on", "there is a meeting", etc. Detailed description of keyword extraction methods, such as regular expression-based, part-of-speech tagging, or deep learning-based sequence tagging models.

[0330] Time Expression Parsing: Describes how the algorithm handles different forms of time expressions, such as "9 am", "2 pm", or "tomorrow afternoon", including time parsing algorithms and time regularization methods.

[0331] Event Type Identification: Explains how to identify event types in user input, such as meetings, reminders, to-do items, and describes the classifier or entity recognition model used, as well as how to handle synonyms and contextual associations.

[0332] 2) Context understanding and completion:

[0333] Context modeling: describes how the algorithm uses contextual information to complete or correct incomplete user input. For example, when the user enters "What time does it open?", how does the algorithm understand the user's intention based on the current time or previous conversation content?

[0334] Fuzzy information processing: explains how to process ambiguous information in user input, such as "there is a meeting", and automatically complete the time, location, participants and other information of the event based on the user's historical schedule or current schedule.

[0335] 3) Event conflict detection and suggestions:

[0336] Conflict Detection: Describe how the algorithm checks for conflicts between the newly generated schedule and the existing schedule, including time overlap, location conflict, etc. Detailed description of the conflict detection process and algorithm.

[0337] Schedule suggestion generation: explains how to generate alternatives or suggestions when conflicts are detected, such as adjusting the time or location, and how to interact with the user through NLP to finally confirm the new schedule.

[0338] Step 1303: Determine whether the schedule is created automatically or manually.

[0339] Users can enable automatic recognition of schedule creation intentions through the settings in the input method, such as Figure 4 As shown, if the user turns on the automatic schedule recognition switch 41, the process will directly proceed to step 1306.

[0340] Step 1304: When manually creating a schedule, a button for triggering screen content recognition is displayed.

[0341] Close Figure 4 The automatic schedule recognition switch 41 is manually triggered, and the manual trigger will pop up a prompt that you need to click the "Create Schedule" button 51, such as Figure 5 shown.

[0342] Step 1305: Click Figure 5 The "Create Schedule" button 51 shown in FIG.

[0343] Clicking the "Create Schedule" button 51 triggers screen recognition. There are many ways to perform screen recognition: taking a screenshot or recording the screen and converting it to OCR to recognize the content on the screen. Then, the recognized text content on the screen is input into the client-side big model to obtain the data structure of the schedule.

[0344] Step 1306: Identify the content of the screen.

[0345] Extract the context of user input, and identify the content including Figure 6 Middle: The name of the boss and the content of the conversation are entered at the same time. The recognized content can be stored in a database with a time index established by rules, and the stored content includes people, time, place, and content. The recorded user input information is sorted and classified according to different dimensions, including time, source, place, person, and context information.

[0346] Step 1307: Aggregate the screen content and the content input by the user for integration.

[0347] Includes data content saved in the local database in the user's historical scenarios. Aggregated content includes information such as time, location, person, event, etc. Time sorting is performed in the internal database.

[0348] Step 1308: sort and categorize the content input by the user and the content recognized by the screen, and input them into the end-side large model.

[0349] Specifically, the data content sorted in step 1307 is input into the local big model service to extract the parameter information for building the schedule, including time, location, task, event, etc. The constructed timeline index is input into the local end-side big model. The big model analyzes and understands the input data through a deep learning algorithm, and compiles the corresponding schedule and precautions based on the analysis results.

[0350] The following is a schematic diagram of the process of generating a local large model schedule:

[0351] The conversation displayed on the screen is as follows Fig.15 , based on the following user input:

[0352] 1. Boss: Accompany me to meet the client tomorrow afternoon.

[0353] 2. Me: Which customer

[0354] 3. Boss: Mr. Li who purchased the equipment last time

[0355] 4. Me: What time tomorrow?

[0356] 5. Boss: Let’s have dinner together tomorrow at 6pm

[0357] 6. Me: OK **Other information:**

[0358] -**Time:** 10:35

[0359] -**Application Name:**vMessage

[0360] -**lbs:**Company

[0361] -**Character:**Boss, I

[0362] The following is the schedule structure generated by the large model: Through the prompt word project, data content in various formats can be generated. The following is the generated schedule:

[0363] **Schedule**

[0364] **Date:** Tomorrow

[0365] **afternoon**

[0366] -**6:00PM**

[0367] -**Event: **Accompanying the boss to meet clients and have dinner together

[0368] -**Location:**To be determined (Confirm the restaurant location)

[0369] -**Character:**Boss, Boss Li

[0370] -**Note: **Boss Li, who purchased the equipment last time, prepares relevant information and questions in advance

[0371] Step 1309: parse the content output by the large model on the end side.

[0372] Determine whether it is a past schedule. If it is not a past schedule, automatically generate the user's schedule based on the analysis results of the big model. The schedule includes the user's tasks and reminders, such as meetings, shopping, appointments, etc. The schedule is displayed through the input method interface or related applications for easy viewing and management by users.

[0373] Step 1310: Generate a schedule by calling the system's API interface.

[0374] The generated schedule is as follows Fig.16 As shown, the created schedule list is as follows Fig.17 shown.

[0375] The detailed algorithm strategy process for triggering schedule creation can be as follows: Fig.14 As shown, the following steps are included:

[0376] Step 1400: In the chat application scenario, pull up the input method.

[0377] Step 1401: The user inputs content.

[0378] Step 1402: Obtain the content input by the user:

[0379] Method 1: Get the content of the input box when it changes.

[0380] Method 2: Trigger after the screen is displayed to obtain content.

[0381] Step 1403: Segment the content displayed on the user's screen.

[0382] Step 1404: Check the content of the word segmentation for matching. Figure 3 As shown, for details on building the trigger word library, please refer to the above related introduction.

[0383] Step 1405: Check whether the trigger word is matched.

[0384] Step 1406: Check the environment to see if it meets the requirements, such as whether the user is authorized.

[0385] Step 1407: Obtain the context: including the input package name, the name of the corresponding chat person, i.e. the chat partner, location information (lbs information), time, etc.

[0386] Step 1408: Based on the information obtained in the previous step, query the locally recorded data.

[0387] Step 1409: Whether the historical context information can be matched, if the context information cannot be matched, trigger automatic screenshot to obtain the context information input by the current user.

[0388] Step 1410: Record the current trigger context into the local database to provide context information for the next trigger.

[0389] Step 1411: After sorting the user input information and context information in time, input them together into the end-side big model through the constructed prompt.

[0390] Step 1412: Continuing from step 1409, the screen content of the current trigger word is obtained. Context information is obtained through the screen content to facilitate the creation of a more accurate schedule.

[0391] Step 1413: Obtain the content of the screen through local OCR text recognition capability.

[0392] Step 1414: extract the content of the screen, such as the title, name, and chat content, to construct a certain record of the chat.

[0393] Step 1415: The content and context information input by the user in step 1411, together with the prompt information, are input to the end-side large model.

[0394] Step 1416: Check whether the schedule information is extracted. The schedule parameters are extracted through the client-side big model: time, location, event, person and other information.

[0395] Step 1417: Check whether the schedule parameters such as time, location, event, etc. extracted by the large model meet the requirements.

[0396] Step 1418: Check whether the currently created schedule conflicts with the historical schedule. If there is a conflict, record the conflict.

[0397] Step 1419: The schedule is created successfully.

[0398] Other additional scenarios are as follows:

[0399] Deletion schedule: Trigger words for deletion.

[0400] Conflict Agenda: Trigger words for conflict.

[0401] Schedule suggestions: Plan the existing schedule, such as meeting conflicts that need to be cancelled, etc.

[0402] Itinerary planning: travel guide schedule arrangement, etc.

[0403] Next, continue to combine Fig.18 , provides an example of how a user can create a schedule in a chat scenario. The specific interaction process is as follows Fig.19 shown.

[0404] In summary, the embodiments of the present application have the following beneficial effects:

[0405] Improve user experience: By recording and analyzing various user inputs, the local large model can automatically generate schedules and task reminders, reducing the tedious steps of manual operations by users and greatly improving the user experience.

[0406] High degree of intelligence: The local large model can deeply understand the user's input content, identify the user's intention and context information, and thus provide more intelligent services.

[0407] Data privacy protection: All data processing and analysis are completed on the local device without uploading to the cloud, ensuring the privacy of users.

[0408] Efficient schedule management: By automatically generating a schedule index list, users can easily manage and view their schedules, improving the efficiency of time management.

[0409] Diverse data sources: Supports multiple input forms, such as typing, voice, handwriting, and copy and paste, making the system more flexible and practical.

[0410] Context and intent recognition: The system not only records the input content, but also conducts in-depth analysis based on the context and intent to provide users with more accurate and personalized services.

[0411] The embodiments of the present application can be applied not only to input methods on mobile devices, but also to other scenarios that require text input and intent recognition, such as smart home control, vehicle systems, office automation, etc. At the same time, with the continuous development of large model technology, the application prospects of the embodiments of the present application will be broader.

[0412] The schedule processing method in the embodiment of the present application identifies the user's input information on the first interface; based on the input information, determines whether there is a schedule processing intention; in the case of a schedule processing intention, obtains context information associated with the input information; extracts schedule parameter information based on the input information and the context information; performs schedule processing based on the schedule parameter information, and the schedule processing includes at least one of the following: creating a schedule, modifying a schedule, or deleting a schedule. In this way, by analyzing and determining the user's schedule processing intention based on the user's input information, when the intention is identified, the context information input by the user is contacted, the schedule parameters are extracted, and the schedule creation, modification, deletion and other processing are automatically completed, thereby realizing the automated processing of the schedule, without the need for manual operation by the user, making the schedule creation and other processing more convenient and efficient.

[0413] The schedule processing method provided in the embodiment of the present application can be executed by a schedule processing device. In the embodiment of the present application, the schedule processing device provided in the embodiment of the present application is described by taking the schedule processing method executed by the schedule processing device as an example.

[0414] See also Fig. 20 , Fig. 20 A schematic diagram of the structure of the schedule processing device provided in the embodiment of the present application is shown in FIG. Fig. 20 As shown, the schedule processing device 2000 includes:

[0415] Identification module 2001, used to identify the user's input information on the first interface;

[0416] A determination module 2002, for determining whether there is a schedule processing intention according to the input information;

[0417] An acquisition module 2003, used for acquiring context information associated with the input information when there is a schedule processing intention;

[0418] An extraction module 2004, configured to extract schedule parameter information according to the input information and the context information;

[0419] The schedule processing module 2005 is used to perform schedule processing according to the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule or deleting a schedule.

[0420] Optionally, the determination module 2002 includes:

[0421] A word segmentation unit, used to perform word segmentation processing on the input information to obtain at least one word;

[0422] a matching unit, configured to match the at least one word with a trigger word in a first schedule trigger word library;

[0423] A determination unit is used to determine the existence of a schedule processing intention when a first word successfully matches a trigger word in the first schedule trigger word library, wherein the first word is any word of the at least one word.

[0424] Optionally, the first schedule trigger word library includes multiple trigger word libraries corresponding to different schedule processing intentions;

[0425] The matching unit is used to match the at least one word with the trigger words in the plurality of trigger word libraries respectively;

[0426] The determining unit is used to determine that there is a schedule processing intention corresponding to the target trigger word library when the first word successfully matches the trigger word in the target trigger word library;

[0427] The schedule processing matches the schedule processing intention corresponding to the target trigger vocabulary.

[0428] Optionally, the first interface is a chat interface;

[0429] The acquisition module 2003 includes at least one of the following:

[0430] an extracting unit, configured to extract the dialog information displayed on the first interface, wherein the context information includes the dialog information;

[0431] The first acquisition unit is used to acquire historical chat records with the chat object on the first interface; if there is a target chat record matching the input information in the historical chat records, acquire the target chat record, wherein the context information includes the target chat record.

[0432] Optionally, the extraction module 2004 includes:

[0433] A generating unit, used for sorting the input information and the context information in chronological order to generate model input data;

[0434] An extraction unit, used for inputting the model input data and the first prompt word Prompt information into a pre-trained target model, and performing schedule parameter extraction processing on the model input data according to the first prompt word Prompt information through the target model;

[0435] The second acquisition unit is used to acquire the schedule parameter information output by the target model.

[0436] Optionally, the target model is a Transform model, and the Transform model includes an N-layer encoder Encoder and an M-layer decoder Decoder, where N and M are both integers greater than 1;

[0437] Each layer of Encoder includes a multi-head self-attention layer, a first splicing and fusion layer, a feedforward network layer, and a second splicing and fusion layer, and the multi-head self-attention layer, the first splicing and fusion layer, the feedforward network layer, and the second splicing and fusion layer are stacked in sequence;

[0438] Each layer of Decoder includes a masked multi-head attention layer, a third splicing fusion layer, a multi-head cross attention layer, a fourth splicing fusion layer, a feedforward network layer, and a fifth splicing fusion layer, and the masked multi-head attention layer, the third splicing fusion layer, the multi-head cross attention layer, the fourth splicing fusion layer, the feedforward network layer, and the fifth splicing fusion layer are stacked in sequence;

[0439] The input of the first layer Encoder includes the embedding code and word position code of the model input data;

[0440] The output of the Nth layer Encoder is the input of the multi-head cross attention layer in each layer Decoder;

[0441] The input of the first layer Decoder includes the target output data predicted based on the model input data.

[0442] Optionally, when the schedule processing includes creating a schedule, the schedule processing device 2000 further includes:

[0443] Check module, used to check whether the currently created schedule conflicts with the historical schedule;

[0444] The output module is used to output schedule conflict reminders or schedule modification suggestions when the currently created schedule conflicts with the historical schedule.

[0445] Optionally, the input information is information input through an input method application;

[0446] The schedule processing module 2005 is used to call the schedule application program interface API and perform schedule processing according to the schedule parameter information.

[0447] Optionally, a schedule viewing control is displayed on the keyboard interface of the input method application; the schedule processing device 2000 further includes:

[0448] The display module is used to display the schedule list interface when receiving a touch input to the schedule viewing control.

[0449] The schedule processing device 2000 in the embodiment of the present application identifies the user's input information on the first interface; determines whether there is a schedule processing intention based on the input information; obtains context information associated with the input information if there is a schedule processing intention; extracts schedule parameter information based on the input information and the context information; performs schedule processing based on the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule, or deleting a schedule. In this way, by analyzing and judging the user's schedule processing intention based on the user's input information, when the intention is identified, the context information input by the user is contacted, the schedule parameters are extracted, and the schedule creation, modification, deletion and other processing are automatically completed, thereby realizing the automated processing of the schedule without the need for manual operation by the user, making the schedule creation and other processing more convenient and efficient.

[0450] The schedule processing device in the embodiment of the present application can be an electronic device, or a component in the electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or it can be other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a vehicle-mounted electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmented reality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personal digital assistant, PDA), etc., and can also be a server, a network attached storage (Network Attached Storage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0451] The schedule processing device in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0452] The schedule processing device provided in the embodiment of the present application can achieve Figures 1 to 19 The various processes implemented by the method embodiment can achieve the same technical effect, and to avoid repetition, they will not be described here.

[0453] Alternatively, if Fig.21 As shown, an embodiment of the present application further provides an electronic device 2100, including a processor 2101 and a memory 2102, wherein the memory 2102 stores programs or instructions that can be executed on the processor 2101, and when the program or instructions are executed by the processor 2101, the various steps of the above-mentioned schedule processing method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, they are not described here.

[0454] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0455] Fig. 22 A schematic diagram of the hardware structure of an electronic device to implement an embodiment of the present application.

[0456] The electronic device 2200 includes but is not limited to: a radio frequency unit 2201, a network module 2202, an audio output unit 2203, an input unit 2204, a sensor 2205, a display unit 2206, a user input unit 2207, an interface unit 2208, a memory 2209, and a processor 2210 and other components.

[0457] Those skilled in the art will appreciate that the electronic device 2200 may also include a power source (such as a battery) for supplying power to each component, and the power source may be logically connected to the processor 2210 through a power management system, thereby implementing functions such as managing charging, discharging, and power consumption management through the power management system. Fig. 22 The electronic device structure shown in the figure does not constitute a limitation on the electronic device. The electronic device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently, which will not be described in detail here.

[0458] The processor 2210 is used for:

[0459] Identify the user's input information on the first interface;

[0460] Determining whether there is a schedule processing intention based on the input information;

[0461] In the case where there is a schedule processing intention, obtaining context information associated with the input information;

[0462] Extracting schedule parameter information according to the input information and the context information;

[0463] Schedule processing is performed according to the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule, or deleting a schedule.

[0464] Optionally, the processor 2210 is further configured to:

[0465] Performing word segmentation processing on the input information to obtain at least one word;

[0466] Matching the at least one word with a trigger word in a first schedule trigger word library;

[0467] When a first word successfully matches a trigger word in the first schedule trigger word library, it is determined that there is a schedule processing intention, wherein the first word is any word among the at least one word.

[0468] Optionally, the first schedule trigger word library includes multiple trigger word libraries corresponding to different schedule processing intentions;

[0469] The processor 2210 is further configured to:

[0470] Matching the at least one word with the trigger words in the plurality of trigger word libraries respectively;

[0471] In the case where the first word successfully matches a trigger word in the target trigger word library, determining that there is a schedule processing intention corresponding to the target trigger word library;

[0472] The schedule processing matches the schedule processing intention corresponding to the target trigger vocabulary.

[0473] Optionally, the first interface is a chat interface;

[0474] The processor 2210 is further configured to:

[0475] Extracting the dialog information displayed on the first interface, wherein the context information includes the dialog information;

[0476] Acquire historical chat records with the chat object on the first interface; if there is a target chat record matching the input information in the historical chat records, acquire the target chat record, wherein the context information includes the target chat record.

[0477] Optionally, the processor 2210 is further configured to:

[0478] Sorting the input information and the context information in chronological order to generate model input data;

[0479] Input the model input data and the first prompt word Prompt information into a pre-trained target model, and perform schedule parameter extraction processing on the model input data according to the first prompt word Prompt information through the target model;

[0480] Acquire the schedule parameter information output by the target model.

[0481] Optionally, the target model is a Transform model, and the Transform model includes an N-layer encoder Encoder and an M-layer decoder Decoder, where N and M are both integers greater than 1;

[0482] Each layer of Encoder includes a multi-head self-attention layer, a first splicing and fusion layer, a feedforward network layer, and a second splicing and fusion layer, and the multi-head self-attention layer, the first splicing and fusion layer, the feedforward network layer, and the second splicing and fusion layer are stacked in sequence;

[0483] Each layer of Decoder includes a masked multi-head attention layer, a third splicing fusion layer, a multi-head cross attention layer, a fourth splicing fusion layer, a feedforward network layer, and a fifth splicing fusion layer, and the masked multi-head attention layer, the third splicing fusion layer, the multi-head cross attention layer, the fourth splicing fusion layer, the feedforward network layer, and the fifth splicing fusion layer are stacked in sequence;

[0484] The input of the first layer Encoder includes the embedding code and word position code of the model input data;

[0485] The output of the Nth layer Encoder is the input of the multi-head cross attention layer in each layer Decoder;

[0486] The input of the first layer Decoder includes the target output data predicted based on the model input data.

[0487] Optionally, the processor 2210 is further configured to:

[0488] In the case where the schedule processing includes creating a schedule, checking whether the currently created schedule conflicts with a historical schedule;

[0489] When the currently created schedule conflicts with the historical schedule, a schedule conflict reminder is output, or a schedule modification suggestion is output.

[0490] Optionally, the input information is information input through an input method application;

[0491] The processor 2210 is further configured to:

[0492] The schedule application program interface API is called to perform schedule processing according to the schedule parameter information.

[0493] Optionally, a schedule viewing control is displayed on the keyboard interface of the input method application;

[0494] The processor 2210 is further configured to:

[0495] When a touch input to the schedule viewing control is received, a schedule list interface is displayed.

[0496] It should be understood that in the embodiment of the present application, the input unit 2204 may include a graphics processor (Graphics Processing Unit, GPU) 22041 and a microphone 22042, and the graphics processor 22041 processes the image data of the static picture or video obtained by the image capture device (such as a camera) in the video capture mode or the image capture mode. The display unit 2206 may include a display panel 22061, and the display panel 22061 may be configured in the form of a liquid crystal display, an organic light emitting diode, etc. The user input unit 2207 includes a touch panel 22071 and at least one of other input devices 22072. The touch panel 22071 is also called a touch screen. The touch panel 22071 may include two parts: a touch detection device and a touch controller. Other input devices 22072 may include, but are not limited to, a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, and a joystick, which will not be repeated here.

[0497] The memory 2209 can be used to store software programs and various data. The memory 2209 may mainly include a first storage area for storing programs or instructions and a second storage area for storing data, wherein the first storage area may store an operating system, an application program or instructions required for at least one function (such as a sound playback function, an image playback function, etc.), etc. In addition, the memory 2209 may include a volatile memory or a non-volatile memory, or the memory 2209 may include both volatile and non-volatile memories. Among them, the non-volatile memory may 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), or a flash memory. The volatile memory may be a random access memory (RAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), a synchronous dynamic random access memory (SDRAM), a double data rate synchronous dynamic random access memory (DDRSDRAM), an enhanced synchronous dynamic random access memory (ESDRAM), a synchronous link dynamic random access memory (SLDRAM) and a direct memory bus random access memory (DRRAM). The memory 2209 in the embodiment of the present application includes but is not limited to these and any other suitable types of memory.

[0498] The processor 2210 may include one or more processing units; optionally, the processor 2210 integrates an application processor and a modem processor, wherein the application processor mainly processes operations related to an operating system, a user interface, and application programs, and the modem processor mainly processes wireless communication signals, such as a baseband processor. It is understandable that the modem processor may not be integrated into the processor 2210.

[0499] The embodiment of the present application also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the above-mentioned schedule processing method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0500] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0501] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned schedule processing method embodiment, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0502] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0503] An embodiment of the present application provides a computer program product, which is stored in a storage medium. The program product is executed by at least one processor to implement the various processes of the above-mentioned schedule processing method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0504] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0505] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0506] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A schedule processing method, characterized in that: include: Identify the user's input information on the first interface; Determining whether there is a schedule processing intention based on the input information; In the case where there is a schedule processing intention, obtaining context information associated with the input information; extracting schedule parameter information according to the input information and the context information; Schedule processing is performed according to the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule, or deleting a schedule.

2. The method according to claim 1, characterized in that The determining, based on the input information, whether there is a schedule processing intention includes: Performing word segmentation processing on the input information to obtain at least one word; Matching the at least one word with a trigger word in a first schedule trigger word library; When a first word successfully matches a trigger word in the first schedule trigger word library, it is determined that there is a schedule processing intention, wherein the first word is any word among the at least one word.

3. The method according to claim 2, characterized in that The first schedule trigger word library includes a plurality of trigger word libraries corresponding to different schedule processing intentions; The matching of the at least one word with a trigger word in the first schedule trigger word library includes: Matching the at least one word with the trigger words in the plurality of trigger word libraries respectively; When the first word successfully matches the trigger word in the first schedule trigger word library, determining that there is a schedule processing intention includes: In the case where the first word successfully matches a trigger word in the target trigger word library, determining that there is a schedule processing intention corresponding to the target trigger word library; The schedule processing matches the schedule processing intention corresponding to the target trigger vocabulary.

4. The method according to claim 1, characterized in that The first interface is a chat interface; The acquiring of context information associated with the input information includes at least one of the following: Extracting the dialog information displayed on the first interface, wherein the context information includes the dialog information; Acquire historical chat records with the chat object on the first interface; if there is a target chat record matching the input information in the historical chat records, acquire the target chat record, wherein the context information includes the target chat record.

5. The method according to claim 1, characterized in that The extracting schedule parameter information according to the input information and the context information includes: Sorting the input information and the context information in chronological order to generate model input data; Input the model input data and the first prompt word Prompt information into a pre-trained target model, and perform schedule parameter extraction processing on the model input data according to the first prompt word Prompt information through the target model; Acquire the schedule parameter information output by the target model.

6. The method according to claim 5, characterized in that The target model is a Transform model, which includes an N-layer encoder Encoder and an M-layer decoder Decoder, where N and M are both integers greater than 1; Each layer of Encoder includes a multi-head self-attention layer, a first splicing and fusion layer, a feedforward network layer, and a second splicing and fusion layer, and the multi-head self-attention layer, the first splicing and fusion layer, the feedforward network layer, and the second splicing and fusion layer are stacked in sequence; Each layer of Decoder includes a masked multi-head attention layer, a third splicing fusion layer, a multi-head cross attention layer, a fourth splicing fusion layer, a feedforward network layer, and a fifth splicing fusion layer, and the masked multi-head attention layer, the third splicing fusion layer, the multi-head cross attention layer, the fourth splicing fusion layer, the feedforward network layer, and the fifth splicing fusion layer are stacked in sequence; The input of the first layer Encoder includes the embedding code and word position code of the model input data; The output of the Nth layer Encoder is the input of the multi-head cross attention layer in each layer Decoder; The input of the first layer Decoder includes the target output data predicted based on the model input data.

7. A schedule processing device, characterized in that: include: An identification module, used to identify the input information of the user on the first interface; A judgment module, used for judging whether there is a schedule processing intention according to the input information; An acquisition module, used for acquiring context information associated with the input information when there is a schedule processing intention; An extraction module, configured to extract schedule parameter information according to the input information and the context information; The schedule processing module is used to perform schedule processing according to the schedule parameter information, and the schedule processing includes at least any one of the following: creating a schedule, modifying a schedule or deleting a schedule.

8. The schedule processing device according to claim 7, characterized in that: The judging module comprises: A word segmentation unit, used to perform word segmentation processing on the input information to obtain at least one word; a matching unit, configured to match the at least one word with a trigger word in a first schedule trigger word library; A determination unit is used to determine the existence of a schedule processing intention when a first word successfully matches a trigger word in the first schedule trigger word library, wherein the first word is any word of the at least one word.

9. The schedule processing device according to claim 8, characterized in that: The first schedule trigger word library includes a plurality of trigger word libraries corresponding to different schedule processing intentions; The matching unit is used to match the at least one word with the trigger words in the plurality of trigger word libraries respectively; The determining unit is used to determine that there is a schedule processing intention corresponding to the target trigger word library when the first word successfully matches the trigger word in the target trigger word library; The schedule processing matches the schedule processing intention corresponding to the target trigger vocabulary.

10. The schedule processing device according to claim 7, characterized in that: The first interface is a chat interface; The acquisition module includes at least one of the following: an extracting unit, configured to extract the dialog information displayed on the first interface, wherein the context information includes the dialog information; The first acquisition unit is used to acquire historical chat records with the chat object on the first interface; if there is a target chat record matching the input information in the historical chat records, acquire the target chat record, wherein the context information includes the target chat record.

11. The schedule processing device according to claim 7, characterized in that: The extraction module comprises: A generating unit, configured to sort the input information and the context information in chronological order to generate model input data; An extraction unit, used for inputting the model input data and the first prompt word Prompt information into a pre-trained target model, and performing schedule parameter extraction processing on the model input data according to the first prompt word Prompt information through the target model; The second acquisition unit is used to acquire the schedule parameter information output by the target model.

12. The schedule processing device according to claim 11, characterized in that: The target model is a Transform model, which includes an N-layer encoder Encoder and an M-layer decoder Decoder, where N and M are both integers greater than 1; Each layer of Encoder includes a multi-head self-attention layer, a first splicing and fusion layer, a feedforward network layer, and a second splicing and fusion layer, and the multi-head self-attention layer, the first splicing and fusion layer, the feedforward network layer, and the second splicing and fusion layer are stacked in sequence; Each layer of Decoder includes a masked multi-head attention layer, a third splicing fusion layer, a multi-head cross attention layer, a fourth splicing fusion layer, a feedforward network layer, and a fifth splicing fusion layer, and the masked multi-head attention layer, the third splicing fusion layer, the multi-head cross attention layer, the fourth splicing fusion layer, the feedforward network layer, and the fifth splicing fusion layer are stacked in sequence; The input of the first layer Encoder includes the embedding code and word position code of the model input data; The output of the Nth layer Encoder is the input of the multi-head cross attention layer in each layer Decoder; The input of the first layer Decoder includes the target output data predicted based on the model input data.

Images