A message misdispatching prevention method based on content and object matching analysis

Abstract

This invention provides a method for preventing mis-sent messages based on content and object matching analysis. The method includes acquiring the message to be sent, the recipient, contact tags, and historical context information; automatically generating contact attribute tags using multi-dimensional behavioral analysis; employing natural language processing technology to perform semantic and scenario classification of the message to be sent; and comprehensively determining the risk of mis-sent by combining three dimensions: content type matching with the recipient, content matching with contact attribute tags, and content consistency with historical context. High-risk messages trigger a secondary confirmation pop-up reminder, achieving proactive interception and prevention of mis-sent messages. The beneficial effects of this invention are: it can accurately identify mismatch scenarios between different types of messages (private, venting, group announcements, work, and personal) and their recipients, taking into account semantic understanding, contact attributes, and contextual coherence, effectively reducing the probability of mis-sent messages in instant messaging, and improving user security and experience.

Description

Technical Field

[0001] This invention relates to the field of instant messaging technology, and in particular to a method for preventing missent messages based on content and object matching analysis. Background Technology

[0002] With the widespread use of instant messaging software such as WeChat, QQ, and WeChat Work, message sharing has become one of the core methods of daily social and work communication. In scenarios involving multiple contacts and group chats, users are prone to sending messages incorrectly. For example, private messages may be mistakenly sent to work groups or family / friend groups; sensitive content such as complaints or private conversations may be mistakenly sent to superiors, colleagues, or clients; and group announcements may be mistakenly sent to individuals. Such errors can range from causing awkward communication to leaking work secrets, triggering workplace conflicts, and harming the interests of both individuals and the company.

[0003] In existing technologies, solutions for accidental message sending are limited to the "message recall" function. This function is a post-event remedy and cannot prevent the message from being sent before it is sent. Furthermore, there is a time limit for recall. If the time limit is exceeded or the recipient has already viewed the message, the loss cannot be recovered.

[0004] Currently, the industry lacks an intelligent early warning mechanism that can automatically identify whether the content to be sent is suitable for the current recipient before the message is sent, thus failing to reduce the probability of message misdelivery at the source. Therefore, there is an urgent need for a pre-emptive message misdelivery prevention method that combines content semantics, object attributes, and contextual consistency to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the problems in existing technologies, this invention provides a message missending prevention method based on content and object matching analysis. By acquiring information such as the message to be sent, the recipient, contact tags, and historical context, it automatically generates contact attribute tags using multi-dimensional behavioral analysis. Natural language processing technology is employed to semantically and contextually classify the message to be sent. A comprehensive assessment of missending risk is conducted by combining three dimensions: content type matching with the recipient, content matching with contact attribute tags, and content consistency with historical context. High-risk messages trigger a secondary confirmation pop-up reminder, enabling proactive interception and prevention of message missending. This method accurately identifies mismatches between different types of messages (private, venting, group announcements, work, and personal messages) and their recipients, taking into account semantic understanding, contact attributes, and contextual coherence. It effectively reduces the probability of missending or incorrect message delivery in instant messaging, improving user security and experience. This solves the problem of the lack of an intelligent early warning mechanism in existing technologies that can automatically identify whether the content to be sent is suitable for the current recipient before message delivery.

[0006] The present invention provides a message mistransmission prevention method based on content and object matching analysis, comprising the following steps: Step 1: Fully acquire the message to be sent and context information, including the message to be sent, the recipient information, the contact attribute tags, and historical context information; Step 2: Automatically generate contact attribute tags. The system backend automatically collects the frequency of chats between users and each contact, chat time periods, historical content topics, group chat associations, user notes and related information, and automatically generates contact attribute tags based on these multi-dimensional behavioral data. Step 3: Semantic and contextual analysis of the message to be sent. Using natural language processing technology, accurately identify the content type and sensitivity of the message to be sent, perform semantic and contextual classification of the message to be sent, and determine the core content type. Step 4: Multi-dimensional misdelivery risk assessment. The misdelivery risk is comprehensively assessed from three dimensions: matching content type with recipient, matching content with contact attribute tags, and judging the consistency of content with historical context, to obtain the comprehensive risk level. Step 5: Secondary confirmation pop-up reminder and execution for high-risk messages. A secondary confirmation pop-up reminder is triggered for high-risk messages based on the overall risk level. After the user confirms, the message can be sent or canceled.

[0007] The present invention is further improved in that, in step 1, the acquisition of the message to be sent is specifically triggered when the user clicks the send button. The acquisition process is automatically triggered, the voice message is automatically converted into text, and emoticons, punctuation, and special characters are filtered out, while the core text information is retained. The sending object information includes the unique identifier of the individual contact, the group chat ID and the group chat name. The contact attribute tags are the contact attribute tags stored locally, including family, friends, colleagues, leaders and clients. The historical context information is the 10 most recent valid historical chat records in the current chat window, filtering out pure emoticons and blank invalid messages.

[0008] The present invention is further improved in that, in step 2, when automatically generating contact attribute tags, the tag judgment priority is as follows: user notes keywords > group chat name > historical chat topics > chat frequency and time period. If the conditions are not met, the contact is marked as a stranger. Users can manually edit and modify the tags.

[0009] The present invention is further improved in that, in step 3, the core content types include group announcements, private / complaint / sensitive content, work-related content, and casual conversation content; the priority is: group announcements > private / complaint / sensitive content > work-related content > casual conversation content.

[0010] The present invention is further improved in that, in step 3, the determination rule for group announcement content is that it matches two or more keywords in the group announcement keyword library and is an imperative or directive statement directed at the group; the determination rule for private / complaint / sensitive content is that it matches the sensitive word library or has a negative, intimate or private emotional tendency; the determination rule for work-related content is that it matches two or more keywords in the work keyword library and has a formal and serious tone; the determination rule for casual chat content is that it matches one or more keywords in the life keyword library and has a relaxed and conversational tone.

[0011] In a further improvement to this invention, in step 4, in the dimension of matching content type with recipient, sending group announcement content to individual contacts is considered to have a high risk of missending; sending private chat content to group chats with ≥10 people or those tagged as work groups is also considered to have a high risk of missending.

[0012] In a further improvement to this invention, in step 4, in the dimension of matching content with contact attribute tags, content related to life, family affairs, privacy, complaints, venting, and negative sensitive content sent to objects tagged as leaders, clients, colleagues, and work groups is determined to be of high risk of being sent to the wrong person.

[0013] The present invention is further improved in step 4. In the dimension of content and historical context consistency judgment, keyword overlap <30% is judged as topic jump risk; the historical tone is formal work tone, and the message to be sent is colloquial, emotional, or complaining tone, which is judged as tone abrupt risk; the chat interval is more than 7 days and sensitive, private, or important content is sent, which is judged as wrong window risk; any 2 or more risk points are judged as high error sending risk.

[0014] The present invention is further improved in that, in step 4, the comprehensive risk judgment rule is that if any one of the three dimensions is judged to be of high risk, then the whole is judged to be of high risk; if there are multiple medium risks, then the whole is judged to be of medium risk; if there are no high or medium risks, then the whole is judged to be of no risk.

[0015] The present invention is further improved in that, in step 5, the secondary confirmation pop-up includes a risk warning, a preview of the message to be sent, and confirmation and cancellation buttons; high-risk messages are intercepted, and after cancellation, the message is retained in the input box, supporting modification of content or switching of receiving objects.

[0016] The beneficial effects of this invention are as follows: This invention provides a message misdelivery prevention method based on content and object matching analysis. By acquiring the message to be sent, the recipient, contact tags, and historical context information, it automatically generates contact attribute tags using multi-dimensional behavioral analysis. Natural language processing technology is used to semantically and contextually classify the message to be sent. A comprehensive judgment of misdelivery risk is made by combining three dimensions: content type matching with the recipient, content matching with contact attribute tags, and content consistency with historical context. A secondary confirmation pop-up reminder is triggered for high-risk messages, achieving proactive interception and prevention of message misdelivery. It can accurately identify private messages, complaints, and group messages. This technology addresses scenarios where different message types (announcements, work-related, and personal) do not match the intended recipients. It balances semantic understanding, contact attributes, and contextual coherence to effectively reduce the probability of missent or incorrect messages in instant messaging. By combining message content, contact attributes, and historical context, risk assessment is more accurate. No manual user annotation is required, lowering the barrier to entry. It accurately distinguishes between announcements, private messages, work-related messages, and personal messages, adapting to various missent scenarios. High-risk messages undergo secondary confirmation to prevent missentment at the source, improving user security and experience. This technology solves the problem of the lack of an intelligent early warning mechanism in existing technologies that can automatically identify whether the content to be sent is suitable for the current recipient before sending. Attached Figure Description

[0017] Figure 1 This is a flowchart of a message mistransmission prevention method based on content and object matching analysis according to the present invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments.

[0019] Please see Figure 1 The present invention provides a message mistransmission prevention method based on content and object matching analysis, comprising the following steps: Step 1: Comprehensive acquisition of the message to be sent and context information, including the message to be sent, recipient information, contact attribute tags, and historical context information. Specifically, acquiring the message to be sent is triggered when the user clicks the send button, automatically converting voice messages to text, filtering emojis, punctuation, and special characters, and retaining the core text information. Recipient information includes the unique identifier of the individual contact, group chat ID, and group chat name. Contact attribute tags are locally stored contact attribute tags, including family, friends, colleagues, leaders, and clients. Historical context information consists of the 10 most recent valid chat records in the current chat window, filtering out pure emojis and blank invalid messages.

[0020] Step 2: Automatically generate contact attribute tags. The system backend automatically collects the frequency of chats between the user and each contact, chat time periods, historical content topics, group chat associations, user notes and related information, and automatically generates contact attribute tags based on these multi-dimensional behavioral data. When automatically generating contact attribute tags, the tag judgment priority is as follows: user note keywords > group chat name > historical chat topics > chat frequency and time periods. If the conditions are not met, the user is marked as a stranger. Users can manually edit and modify the tags.

[0021] Step 3: Semantic and Contextual Analysis of Messages to be Sent. Using natural language processing technology, accurately identify the content type and sensitivity of the messages to be sent, and classify them semantically and contextually to determine the core content types. These core content types include group announcements, private / complaint / sensitive messages, work-related messages, and casual conversation messages. The priority is: group announcements > private / complaint / sensitive messages > work-related messages > casual conversation messages. The criteria for group announcements are: matching two or more keywords from the group announcement keyword library, and being an imperative or directive statement directed at a group. The criteria for private / complaint / sensitive messages are: matching a sensitive keyword library or having a negative, intimate, or private emotional tendency. The criteria for work-related messages are: matching two or more keywords from the work keyword library, with a formal and serious tone. The criteria for casual conversation messages are: matching one or more keywords from the life keyword library, with a relaxed and conversational tone.

[0022] Step 4: Multi-dimensional mis-sending risk assessment. A comprehensive risk level is determined by assessing the mis-sending risk across three dimensions: content type and recipient matching, content and contact attribute tags matching, and content consistency with historical context. Specifically, in the content type and recipient matching dimension, sending group announcements to individual contacts is considered a high-risk mis-sending; sending private messages to groups with ≥10 members or those tagged as "work groups" is also considered a high-risk mis-sending. In the content and contact attribute tag matching dimension, sending personal, family, private, venting, complaining, or negative sensitive content to recipients tagged as "leaders," "clients," "colleagues," or "work groups" is considered a high-risk mis-sending. In the content consistency with historical context dimension, keyword overlap <30% is considered a risk of topic jump; a previously formal work tone followed by a conversational, emotional, or complaining tone in the message to be sent is considered a risk of abrupt tone; sending sensitive, private, or important content after a chat interval exceeding 7 days is considered a risk of entering the wrong window; any two or more of these risk points indicate a high-risk mis-sending. The comprehensive risk assessment rule is as follows: if any one of the three dimensions is determined to be of high risk, the overall risk is determined to be high; if multiple dimensions are of medium risk, the overall risk is determined to be medium; if there are no high or medium risks, the overall risk is determined to be no.

[0023] Step 5: A secondary confirmation pop-up is triggered for high-risk messages. Based on the overall risk level, a secondary confirmation pop-up is displayed for high-risk messages. After user confirmation, the message can be sent or canceled. The secondary confirmation pop-up includes a risk warning, a preview of the message to be sent, and confirmation and cancellation buttons. If a high-risk message is blocked and canceled, the message remains in the input box, allowing users to modify its content or switch recipients.

[0024] In this embodiment, a message mistransmission prevention method based on content and object matching analysis includes the following steps: Step 1: Full Acquisition of the Message to be Sent and Context Information 1. Retrieval of messages to be sent: When a user finishes editing a message to be sent in an instant messaging software (such as WeChat or QQ) and clicks the "Send" button, the system automatically triggers the information retrieval process; the system reads the content to be sent edited by the user, and if it is a voice message, it automatically calls the voice-to-text interface to convert the voice into analyzable text content; it filters out irrelevant symbols (such as emoticons, punctuation, and special characters) in the message, retaining the core text information for subsequent semantic analysis; 2. Obtain recipient information: Read the recipient information in the current chat window and determine the recipient type (personal contact, group chat); if it is a personal contact, obtain the unique identifier of the contact (such as WeChat ID, QQ number); if it is a group chat, obtain the group chat ID and group chat name for subsequent attribute matching. 3. Contact Attribute Tag Acquisition: The system retrieves the contact attribute tags stored locally to obtain the attribute tags of the current recipient (e.g., family, friends, colleagues, leaders, clients, etc.). If the contact does not have preset tags, the system automatically triggers a temporary tag determination process (see step 2 for details). 4. Historical context acquisition: Read the 10 most recent historical chat records (text and voice-to-text content) in the current chat window, filter invalid messages (such as pure emoticons and blank messages), and extract the core context content for subsequent topic and tone consistency judgment.

[0025] Step 2: Automatically generate contact attribute tags. Through multi-dimensional behavioral analysis, accurately tag contacts with their attributes, as detailed below: 1. Tag Generation Data Source Collection: The system backend automatically collects data from users and their contacts, including: (1) Chat frequency: The number of chats between the user and the contact in the past 30 days is counted. Chatting ≥3 times per day is considered high frequency, and chatting ≤1 time per week is considered low frequency; (2) Chat time period: Record the peak chat time, such as 9:00-18:00 on weekdays as work time, and 20:00-23:00 as life time; (3) Historical content themes: By extracting keywords, analyze the themes of nearly 30 historical chat contents and distinguish between work themes (such as projects, meetings, clients, quotations) and life themes (such as grocery shopping, eating, household chores). (4) Group chat association: If the contact and the user belong to the same work group or family group, the group chat attributes are associated synchronously (e.g., the work group is associated with the "colleague / leader" tag, and the family group is associated with the "family member" tag). (5) User notes and associated information: Read the notes given by the user to the contact (such as "General Manager Wang", "Mom" or "Customer Sister Li") and extract the attribute keywords in the notes; if it is a WeChat contact, read the department and unit name and associate the "colleague / leader / customer" tags.

[0026] 2. Label determination: (1) Prioritize the determination based on keywords in the remarks. If the remarks contain "Dad, Mom, Grandpa, Grandma" → determine it as a family member; if the remarks contain "General Manager" → determine it as a leader; if the remarks contain "Client, Partner" → client; (2) Secondly, the group chat name is used for judgment: if the group chat name contains "family", "family members" etc., it is judged as family members; if it contains "work group", "department" etc., it is judged as colleagues / leaders; (3) Determine again based on chat topic: If more than 80% of the historical chat content is about life, it is determined to be family / friends; if more than 80% is about work, it is determined to be colleagues / leaders / clients; (4) Determine based on chat frequency and time period: If the chat is frequently during daily life, it is determined to be family / friends; if the chat is frequently during work, it is determined to be colleagues / leaders / clients. The checks are performed sequentially according to the order of the preceding checks. Once a check is successful, no further checks are performed. If a contact fails to meet any of the criteria after all checks have been completed, they are identified as a stranger. Users can manually edit and modify tags.

[0027] Step 3: Semantic and contextual analysis of the message to be sent. Using Natural Language Processing (NLP) technology, accurately identify the content type and sensitivity of the message to be sent, as detailed below: 1. The semantic analysis model is started. The system calls the lightweight NLP semantic analysis model, which is pre-trained with feature libraries for multiple content categories such as "private / complaint / sensitive", "group announcement", "casual chat" and "work-related", which can quickly complete content classification. 2. Content type identification: (1) Identification of private, complaining, sensitive, and indecent content: ① Sensitive word matching: Compare with a preset sensitive word library (including sarcastic words, vulgar words, and privacy-related words, such as "garbage, speechless, annoying, pitfall, salary, private matters"). If any sensitive word is matched, it is marked as a candidate sensitive content. ② Tone analysis: The text tone recognition algorithm analyzes the emotional tendency of the sentence. If it is a negative emotion (complaint, rant) or an intimate tone (only applicable to private scenarios), it is judged as private / rant / sensitive content. (2) Identification of group announcement content: ① Compare with the keyword database of group announcements (such as "please everyone, all, everyone, all, unified, arrangement, notice, reminder, work overtime tonight, meeting tomorrow, etc."). If two or more keywords are matched, proceed to the next step of judgment. ② Identify whether the sentence is an imperative sentence or a directive statement directed to the group (such as "Please attend the meeting at 9 o'clock tomorrow" or "Inform each other"). If so, it is determined to be a group announcement. (3) Recognition of casual conversation content: ① Keyword matching: Compare with a database of lifestyle keywords (such as "grocery shopping, cooking, picking up kids, eating, sleeping, etc.") and match one or more keywords; ②The sentence structure includes first person (I, we) and everyday colloquial language (do you want to, where are you going, what are you doing?), with a relaxed and casual tone and no formality, which is judged as casual conversation content; (4) Identification of job-related professional content: ① Compare with a database of work-related keywords (such as "projects, meetings, clients, quotes, etc.") and find two or more matching keywords; ②The tone is formal and rigorous, without any colloquial language. It mainly consists of descriptions, instructions, and communication of work-related matters, and is therefore classified as work-related content. If a message to be sent falls into multiple categories, the core content type is determined by prioritizing it as follows: "Group Announcement > Sensitive / Private > Work > Personal". This information will be used for subsequent risk assessment.

[0028] Step 4: Multi-dimensional assessment of misdelivery risk Dimension 1: Content type and recipient type matching judgment 1. If step 3 determines that the message to be sent is a "group announcement" message; The current recipient is a "personal contact" → directly judged as high risk of missending; 2. If step 3 determines that the message to be sent is a private message; The current recipient is a "work group or large group" (group chat has ≥10 people, or is tagged "work group") → directly judged as high risk of missending; Dimension 2: Content and Contact Attribute Tags Matching Judgment 1. If the message to be sent is about daily life, family matters, or private matters; Recipients tagged with "leader, client, colleague, work group" are considered high-risk for misdelivery. 2. If the message to be sent contains sensitive content such as complaints, grievances, or negativity; Recipients tagged with "leader, colleague, client, work group" are considered high-risk for misdelivery. Dimension 3: Content and historical context consistency judgment, whether the message to be sent is consistent with the most recent record in the current chat window in terms of topic, and at the same time, whether the chat window has been without dialogue for a long time.

[0029] 1. Topic consistency judgment: ① Extract the core keywords from the 10 historical contexts (such as "project, meeting, client, child, meal") and extract the core keywords for the message to be sent; ② Calculate keyword overlap: Overlap = (Number of common keywords ÷ Total number of context keywords) × 100%; ③ If the overlap is less than 30%, it is judged as "topic jump" and marked as a risk point; 2. Judging the consistency of tone: ①Analyze the overall tone of the historical context (formal work tone, casual conversational tone); ② If the historical context is all in a formal work tone, and the message to be sent suddenly changes to a colloquial, emotional, or sarcastic tone → it is judged as "abrupt tone" and marked as a risk point; ③ If the historical context is casual and the message to be sent suddenly becomes formal and work-related → it is judged as low risk and no risk point is marked; 3. Context length judgment: ① Detect the time of the last historical message in the current chat window. If it is more than 7 days ago (no conversation for a long time); ② If the message to be sent contains sensitive, private, or important content (such as complaints, family matters, or work secrets) → it is judged as "possibly entering the wrong window" and marked as a high-risk point; If any two or more of the above three judgments are marked as risk points, it is judged as a high risk of mis-sending (triggering conditions are met).

[0030] Based on the comprehensive risk assessment, if any one of the three dimensions is determined to be of high risk of missending → the overall risk is determined to be high risk of missending, triggering a secondary confirmation reminder; if only multiple dimensions are determined to be of medium risk → the overall risk is determined to be medium risk, and a slight reminder can be selected (no pop-up window, only a top reminder); if there is no high or medium risk → there is no risk, the message is sent normally, and no reminder is triggered.

[0031] Step 5: A secondary confirmation pop-up notification and execution for high-risk messages. A pop-up notification prompts the user to confirm, preventing accidental sending. Details are as follows: When a step is determined to have a high risk of missending, the system immediately intercepts the message transmission and displays a semi-transparent pop-up window above the current chat interface. The pop-up window includes a risk warning, a preview of the message to be sent, and operation buttons, such as: ① Sending a group announcement to a personal chat: "Content detected as leaning towards group notification. This is currently a personal chat and may have been sent to the wrong person. Confirm sending?"; ② Sending private content to work scenarios: "We have detected that the content you sent is biased towards private / venting, and the recipient is your leader / work group. Do you want to confirm sending?"; ③ Abrupt topic / tone: "The current content is quite different from the chat topic. You may have entered the wrong window. Do you want to confirm sending?"; The pop-up window has two buttons: "Confirm Send" and "Cancel Send". Users can click the "Cancel" button to stop sending, and the message to be sent will be left in the input box for users to modify the content or switch the receiving object. Case Study: 1. User A edits a message in WeChat: "Please attend the project meeting at 9:00 AM tomorrow. Please reply upon receiving this message." The message is about to be sent to a personal contact B (labeled "Friend"). The moment the send button is clicked, the system triggers the information retrieval process. 2. The system retrieves the text of the message to be sent, the recipient is personal contact B, the tag is "friend", and the context of the last 10 messages is casual chat (without work-related content); 3. The semantic analysis model identifies messages containing the keywords "please everyone, attend on time, meeting" and is in the form of an imperative sentence, thus classifying it as a group announcement. 4. Multi-dimensional risk assessment: Dimension 1 (group announcements sent to individuals) → high risk; Dimensions 2 and 3 do not require further assessment, and are judged as high risk of misdelivery. 5. A pop-up window appears: "Content detected as biased towards group notification. This is currently a personal chat and may have been sent to the wrong recipient. Confirm sending?" After seeing the pop-up window, User A realizes that they have mistakenly sent the group announcement to a friend. They click "Cancel Send," change the recipient to the work group, and resend to avoid missending.

[0032] As can be seen from the above, the beneficial effects of the present invention are as follows: The present invention provides a message misdelivery prevention method based on content and object matching analysis. By acquiring the message to be sent, the recipient, contact tags, and historical context information, it automatically generates contact attribute tags using multi-dimensional behavioral analysis. Natural language processing technology is used to semantically and contextually classify the message to be sent. A comprehensive judgment of misdelivery risk is made by combining three dimensions: content type matching with the recipient, content matching with contact attribute tags, and content consistency with historical context. A secondary confirmation pop-up reminder is triggered for high-risk messages, achieving proactive interception and prevention of message misdelivery. It can accurately identify private messages, complaints, and other sensitive messages. In scenarios where different types of messages (group announcements, work-related, personal, etc.) do not match the recipients, this technology effectively reduces the probability of missent or incorrect messages in instant messaging, taking into account semantic understanding, contact attributes, and contextual coherence. By combining message content, contact attributes, and historical context, risk assessment is more accurate. No manual labeling by users is required, lowering the barrier to entry. It accurately distinguishes between announcements, private messages, work-related messages, and personal messages, adapting to various missent scenarios. High-risk messages undergo secondary confirmation to prevent missentment at the source, improving user security and experience. This technology solves the problem of the lack of an intelligent early warning mechanism in existing technologies that can automatically identify whether the content to be sent is suitable for the current recipient before sending.

[0033] The specific embodiments described above are preferred embodiments of the present invention and are not intended to limit the specific scope of the present invention. The scope of the present invention includes, but is not limited to, these specific embodiments. All equivalent changes made in accordance with the present invention are within the protection scope of the present invention.

Claims

1. A method for preventing message mistransmission based on content and object matching analysis, characterized in that, Includes the following steps: Step 1: Fully acquire the message to be sent and context information, including the message to be sent, the recipient information, the contact attribute tags, and historical context information; Step 2: Automatically generate contact attribute tags. The system backend automatically collects the frequency of chats between users and each contact, chat time periods, historical content topics, group chat associations, user notes and related information, and automatically generates contact attribute tags based on these multi-dimensional behavioral data. Step 3: Semantic and contextual analysis of the message to be sent. Using natural language processing technology, accurately identify the content type and sensitivity of the message to be sent, perform semantic and contextual classification of the message to be sent, and determine the core content type. Step 4: Multi-dimensional misdelivery risk assessment. The misdelivery risk is comprehensively assessed from three dimensions: matching content type with recipient, matching content with contact attribute tags, and judging the consistency of content with historical context, to obtain the comprehensive risk level. Step 5: Secondary confirmation pop-up reminder and execution for high-risk messages. A secondary confirmation pop-up reminder is triggered for high-risk messages based on the overall risk level. After the user confirms, the message can be sent or canceled.

2. The message mistransmission prevention method based on content and object matching analysis as described in claim 1, characterized in that: In step 1, obtaining the message to be sent specifically involves triggering the acquisition process when the user clicks the send button. Voice messages are automatically converted to text, filtering out emoticons, punctuation, and special characters while retaining the core text information. The recipient information includes the unique identifier of the individual contact, the group chat ID, and the group chat name. The contact attribute tags are locally stored contact attribute tags, including family, friends, colleagues, leaders, and clients. The historical context information consists of the 10 most recent valid historical chat records in the current chat window, filtering out pure emoticons and blank invalid messages.

3. The message mistransmission prevention method based on content and object matching analysis as described in claim 2, characterized in that: In step 2, when automatically generating contact attribute tags, the tag judgment priority is as follows: user notes keywords > group chat name > historical chat topics > chat frequency and time period. If the conditions are not met, the contact is marked as a stranger. Users can manually edit and modify the tags.

4. The message mistransmission prevention method based on content and object matching analysis as described in claim 3, characterized in that: In step 3, the core content types include group announcements, private / complaint / sensitive content, work-related content, and casual conversation content; the priority is: group announcements > private / complaint / sensitive content > work-related content > casual conversation content.

5. The message mistransmission prevention method based on content and object matching analysis as described in claim 4, characterized in that: In step 3, the rules for determining group announcement content are as follows: it must match two or more keywords in the group announcement keyword library and be an imperative or directive statement directed at the group; the rules for determining private / complaint / sensitive content are as follows: it must match the sensitive word library or have a negative, intimate, or private emotional tendency; the rules for determining work-related content are as follows: it must match two or more keywords in the work keyword library and be formal and serious in tone; the rules for determining casual conversation content are as follows: it must match one or more keywords in the casual keyword library and be casual and conversational in tone.

6. The message mistransmission prevention method based on content and object matching analysis as described in claim 5, characterized in that: In step 4, in the content type and recipient matching dimension, sending group announcement content to individual contacts is considered to have a high risk of being sent to the wrong person; sending private chat content to group chats with ≥10 people or those tagged as work groups is also considered to have a high risk of being sent to the wrong person.

7. The message mistransmission prevention method based on content and object matching analysis as described in claim 6, characterized in that: In step 4, in the content matching dimension with contact attribute tags, sending content related to life, family affairs, privacy, complaints, venting, or negative and sensitive content to individuals tagged with leaders, clients, colleagues, or work groups is considered to have a high risk of missending.

8. The message mistransmission prevention method based on content and object matching analysis as described in claim 7, characterized in that: In step 4, in the dimension of content consistency with historical context, keyword overlap <30% is judged as a risk of topic jump; if the historical tone is formal work tone, and the message to be sent is colloquial, emotional, or complaining tone, it is judged as a risk of abrupt tone; if the chat interval exceeds 7 days and sensitive, private, or important content is sent, it is judged as a risk of entering the wrong window; any two or more risk points are judged as a high risk of wrong message sending.

9. The message mistransmission prevention method based on content and object matching analysis as described in claim 8, characterized in that: In step 4, the comprehensive risk assessment rule is that if any one of the three dimensions is determined to be of high risk, then the overall risk is determined to be high; if multiple dimensions are of medium risk, then the overall risk is determined to be medium. If there are no high or medium risks, the overall risk level is determined to be no risk.

10. The message mistransmission prevention method based on content and object matching analysis as described in claim 9, characterized in that: In step 5, the secondary confirmation pop-up includes a risk warning, a preview of the message to be sent, and confirmation and cancellation buttons. High-risk messages are intercepted, and after cancellation, the message remains in the input box, allowing users to modify the content or switch the recipient.

Images