A family service method and system based on artificial intelligence

Abstract

The present invention relates to the field of artificial intelligence technology, specifically to an artificial intelligence-based home service method and system, comprising the following steps: obtaining statement parsing action sequence and task direction, collecting device feedback to generate a state structure, judging instruction trends and reorganizing statements to generate a behavior chain, disassembling action identification conflicts to extract the main control path, analyzing behavior timing to generate prominent trends, and adjusting the display sequence to generate a functional solution. In the present invention, the accuracy of instruction target recognition is improved through semantic component extraction and device state mapping, word order rearrangement and logical connection enhance the coherence of the behavior chain, control paths are optimized and sorted based on verb density and object association, improving task scheduling accuracy, behavior timing analysis identifies the trend of high-frequency operation forward movement, and realizes dynamic adjustment of priority. Recommended content is rearranged in combination with instruction frequency and timing characteristics to enhance response initiative and service matching, and overall achieves the coordinated optimization of semantic understanding, behavior prediction and content recommendation.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based home service method and system. Background Art

[0002] The field of artificial intelligence encompasses algorithm design, knowledge representation, learning methods, reasoning mechanisms, and decision execution based on the simulation of human intelligent behavior. The core of this technology is the use of computer systems to mimic human thinking, enabling autonomous judgment, information processing, and environmentally adaptive behavior. It is widely used in scenarios such as speech recognition, natural language understanding, image recognition, machine learning, and intelligent control. In home service applications, AI is not only used for voice interaction and behavior prediction, but also integrates perception technologies and data processing mechanisms to enable devices to self-learn, adapt, and automatically respond, achieving a deep understanding of and response to home scenarios, user behaviors, and environmental changes.

[0003] Among them, the AI-based home service method and system refers to a home service mechanism that establishes a user behavior model through speech recognition, natural language understanding, and knowledge graphs. The technical matters targeted by this patent subject cover home environment recognition, user demand perception, intelligent task allocation, and service response planning. Specifically, it obtains the semantics of user instructions through semantic analysis, identifies task scenarios through context modeling, and then makes service response decisions based on user historical behavior data and home device status. The system generally uses language recognition methods to extract command information, uses user behavior modeling methods to establish preference patterns, and completes task reasoning and service triggering through rule matching and learning algorithms.

[0004] Existing technologies often rely on preset rules and word-meaning templates during semantic recognition, lacking structural decomposition capabilities and adapting to multi-action commands in free-form expressions. This often leads to one-sided recognition of control content and affects task triggering accuracy. Behavior chain construction fails to incorporate contextual semantic relationship adjustments and lacks logical continuity between behavior nodes, making it prone to step skipping and response confusion within the execution path. Static priority settings are often used to prioritize commands, disregarding actual phrase relevance and verb dominance, and lacking a mechanism for optimizing task execution paths. Temporal processing is limited to operation logs, without dynamic modeling of behavior frequency or forward movement trends. This fails to reflect the time sensitivity of behavior patterns, leading to delayed ranking of responses for key actions. For content recommendation, current ranking is often based on historical usage tags and static preference matching, without incorporating mechanisms to identify current behavior trends. This results in high-frequency commands being deprioritized in the display sequence, reducing the timeliness of system responses and user satisfaction. These deficiencies directly impact the system's ability to understand and adapt to multi-command behaviors in complex scenarios, limiting the personalized and intelligent level of home service responses. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a home service method and system based on artificial intelligence.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solution: a home service method based on artificial intelligence, comprising the following steps:

[0007] S1: Obtain household sentence input, parse the action sequence composed of verbs and conjunctions, determine the task orientation based on the target noun, identify modifiers to construct background semantics, and simultaneously collect device on / off, voltage, and operation feedback to establish a sentence-driven state structure description;

[0008] S2: Based on the action sequence in the state structure description driven by the statement, determine the instruction trend and trend consistency, identify the direction change and control state feedback, restructure the statement and insert linking words to generate an adjusted control behavior chain;

[0009] S3: Decomposing the adjusted control behavior chain, identifying and eliminating conflicting directional verbs, calling the remaining verb phrases and target phrases, analyzing the primary and secondary order and control connection density, and generating a main control instruction path;

[0010] S4: calling the task sequence of the action in the main control instruction path, extracting the time points of the behavioral instructions in the recent interaction, analyzing the interval trend, and generating the time sequence highlight trend of the instruction content;

[0011] S5: Extract key behaviors based on the temporal trend of the instruction content, locate the initial sequence in the recommendation list, sort similar action nodes forward, combine unlabeled behaviors to maintain the original ranking, and generate an artificial intelligence home service plan.

[0012] As a further solution of the present invention, the statement-driven state structure description includes action sequence identification, target object classification, background semantic features, and device state mapping; the adjusted control behavior chain includes rearranged statement structure, connected behavior word set, continuous instruction nodes, and control link path; the main control instruction path includes main control verb phrase, object phrase priority, control connection path, and word frequency density feature; the temporal prominent trend of the instruction content includes the behavior instruction advance rule, instruction frequency distribution curve, and time interval change trend; the artificial intelligence home service plan includes content display sequence, sorting number update, and behavior identification priority.

[0013] As a further solution of the present invention, the specific steps of S1 are:

[0014] S101: Obtaining the command sentence received by the chatbot, extracting verbs and conjunctions, arranging the order of action phrases, identifying and locating the target noun item content, and generating a sentence action ranking value;

[0015] S102: Calling the sentence action ranking value, identifying the time, place, and manner modifiers in each action sentence, extracting the combination structure between the verb and the modifier, calculating the matching frequency and word class association index, obtaining the semantic strength between the combinations, and generating the semantic association degree of the action modifier combination;

[0016] S103: According to the semantic association of the action modification combination, the on-off response value, voltage feedback value and operation response change rate of the current device are collected, and trend ratio conversion is performed on the semantic chain structure and the device response to generate a statement-driven state structure description.

[0017] As a further solution of the present invention, the specific steps of S2 are:

[0018] S201: Obtaining a state structure description of the statement drive, extracting verb items and corresponding direction information in the current and recent instructions, calculating trend consistency based on the ratio of the angles of the direction vectors, and generating a verb trend consistency ratio;

[0019] S202: Calling the verb trend consistency ratio, extracting the device status feedback and return content corresponding to the target noun item, calculating the control response offset value, combining the offset value with the trend ratio, identifying the verb behavior characteristics with abnormal direction, and generating the verb reverse offset rate;

[0020] S203: Identify the order and direction relationship of adjacent instructions in the action chain according to the verb reverse offset rate, rearrange the structural positions and insert conjunctions to construct a continuous behavior sequence, and generate an adjusted control behavior chain.

[0021] As a further solution of the present invention, the specific steps of S3 are:

[0022] S301: Obtaining instruction actions in the adjusted control behavior chain, extracting directional verbs and combining them with instruction positions, calculating directional angle values, screening verbs with angles greater than a directional conflict threshold for tag exclusion, and generating a directional conflict verb tag set;

[0023] S302: calling the direction conflict verb token set, extracting the verb phrases and object phrases in the remaining instructions, counting the frequency of occurrence of the combination, and calculating the word frequency density, constructing a permutation sequence according to the combination order, and generating a master control word frequency density ranking value;

[0024] S303: Locate the control structure between phrases and calculate the path connection strength according to the master control word frequency density ranking value, select the path combination whose connectivity and master control degree both reach the threshold, and generate the master control instruction path.

[0025] As a further solution of the present invention, the word frequency density calculation formula is specifically:

[0026] ;

[0027] in, represents the word frequency density of the combination of the i-th verb phrase and the j-th object phrase, represents the frequency of the i-th verb phrase and the j-th object phrase appearing together in the original corpus, represents the semantic deviation of the corresponding object phrase of the i-th verb phrase in the k-th context, represents the average value of the semantic deviation of the object phrase in all contexts of the i-th verb phrase, represents the frequency of the i-th verb phrase in the m-th context, represents the frequency of the combination of the t-th verb phrase and the j-th object phrase, represents the average frequency of all verb phrases combined with the j-th object phrase, is the total number of contexts.

[0028] As a further solution of the present invention, the specific steps of S4 are:

[0029] S401: calling the action phrase in the main control instruction path, searching for the corresponding time point in the task record, locating the number position of each instruction in chronological order, calculating the interval length between the time points, and generating the instruction occurrence timing interval value;

[0030] S402: Based on the instruction occurrence timing interval value, determine whether the instruction interval is gradually shortened, select segments where the consecutive interval difference values ​​show a decreasing trend, count the number of trend segments, perform threshold comparison, and generate a value for the number of advanced trend segments;

[0031] S403: Call the advance trend segment quantity value, extract action phrases with advance characteristics, calculate the frequency of occurrence in the task record, filter the content with a frequency greater than the average phrase frequency and mark it as forward-moving behavior, and generate a temporal prominent trend of the instruction content.

[0032] As a further solution of the present invention, the calculation formula for the interval duration between the time points is specifically:

[0033] ;

[0034] in, Represents the duration between time points. Represents the triggering time point of the first instruction in the current time period, Represents the triggering time point of the last instruction in this time period, Represents the total number of actions recognized during this time period, represents the time offset weight of each action, Represents the time offset corresponding to each action, Represents the arithmetic mean of all offsets.

[0035] As a further solution of the present invention, the specific steps of S5 are:

[0036] S501: Calling the behavior identifier in the temporal highlight trend of the instruction content, locating the original sort position of the corresponding behavior in the recommended content display list, extracting the initial number and index of the action phrase in the list, establishing a mapping relationship between the number and the behavior phrase, and generating an initial sort index result of the behavior instruction;

[0037] S502: Based on the initial sort index result of the behavior instruction, the sort numbers of the similar action nodes marked as being moved forward are adjusted forward, the corresponding contents are rearranged according to the adjusted number sequence, and the remaining unmarked nodes are continuously filled in according to the original sequence to generate a sorted number set after the move forward;

[0038] S503: Call the forward-shifted sorting number set to reconstruct all behavioral content and function indexes in the display list, match the functional module positions with the main control direction to form a structural combination, integrate the sorting relationship, and generate an artificial intelligence home service plan.

[0039] An artificial intelligence-based home service system, comprising:

[0040] The sentence structure recognition module acquires the command sentences received by the chatbot in the home environment, analyzes the verbs, conjunctions, target nouns, and modifiers in the sentences, extracts the task orientation, collects the current device's on / off response, voltage feedback, and operational response performance, constructs a corresponding relationship between the semantic behavior chain and the device status, and generates a sentence-driven state structure description.

[0041] The instruction tendency analysis module analyzes the consistency between the current instruction action and the recent instruction trend based on the action sequence in the state structure description driven by the statement, identifies the abnormal behavior of the instruction direction, extracts the control state and system feedback of the action target, rearranges the statement structure and inserts connecting action words to generate an adjusted control action chain;

[0042] The control path generation module disassembles the instruction actions from the adjusted control behavior chain, identifies and eliminates conflicting directional verbs, calls the verb phrases and target phrases in the remaining instructions, analyzes the action frequency and sequence relationship, selects the path with the most concentrated control connectivity and master control degree, and generates the master control instruction path;

[0043] The timing trend judgment module calls the action phrases in the main control instruction path, extracts the timing position in the task record, analyzes the time interval trend, determines whether there is a pattern of continuous advancement of multiple behavior contents, and generates a timing prominent trend of the instruction content;

[0044] The sequence priority adjustment module identifies the initial sequence position in the recommended content display list according to the temporal highlight trend of the instruction content, adjusts the sequence numbering of similar action nodes, and generates an artificial intelligence home service plan.

[0045] Compared with the prior art, the advantages and positive effects of the present invention are:

[0046] In the present invention, the accuracy of instruction target recognition is improved through semantic component extraction and device status mapping, the word order rearrangement and logical connection enhance the coherence of the behavior chain, the control path is optimized and sorted according to the verb density and object association, and the task scheduling accuracy is improved. The behavior timing analysis identifies the trend of high-frequency operation forward movement and realizes dynamic adjustment of priority. The recommended content is rearranged in combination with the instruction frequency and timing characteristics to enhance the response initiative and service matching degree, and the overall coordinated optimization of semantic understanding, behavior prediction and content recommendation is realized. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0048] Figure 1 Schematic diagram of the steps of the present invention;

[0049] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0050] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0051] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0052] In the embodiments of the present invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same. The terms "of," "corresponding," and "corresponding" may sometimes be used interchangeably. It should be noted that, when the distinction is not emphasized, the meanings they convey are the same.

[0053] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0054] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0055] See also Figure 1 , an artificial intelligence-based home service method, comprising the following steps:

[0056] S1: Obtain the command sentences received by the chatbot in the home environment, analyze the action sequence through the verb conjunctions in the sentence, extract the target noun to determine the task direction, identify the modifying words to construct the action background semantics, collect the current device's on / off response, voltage feedback, and operational reaction performance, construct the corresponding relationship between the semantic behavior chain and the device status, and generate a sentence-driven state structure description;

[0057] S2: Based on the action sequence structure in the statement-driven state structure description, analyze the consistency between the current instruction action trend and the recent instruction trend, identify the abnormal behavior in the direction instructed by the verb, extract the current control state performance of the action instruction target and the system return action content, identify the continuous instruction position in the action chain, rearrange the sentence structure and insert connecting action words to construct a coherent behavior control chain, and generate an adjusted control behavior chain;

[0058] S3: Disassemble the instruction actions from the adjusted control behavior chain, identify the direction verbs with execution conflicts and mark them for exclusion, call the verb phrases and object phrases in the remaining instructions, analyze the word frequency density and the order of primary and secondary instructions between actions, locate and arrange the direct control connectivity between phrases, select the path content with concentrated connectivity and master control degree, and generate the master control instruction path;

[0059] S4: Call the appearance sequence of the action phrases in the main control instruction path in the task record, extract the location points of the appearance time in the recent interaction, calculate the interval trend of the time points, determine whether there is a pattern of continuous advancement of multiple behavior contents, mark the instruction content as forward-moving performance, and locate the content with prominent control instruction frequency, and generate the temporal prominence trend of the instruction content;

[0060] S5: Call the behavioral identifiers in the temporal highlighting trend of the instruction content, identify the initial sequence position in the recommended content display list, move the sequence numbers of the corresponding similar action nodes forward, arrange the sorting numbers of the corresponding content in the order after the forward movement, and at the same time call the unmarked behaviors to rank according to the remaining order to generate an artificial intelligence family service plan.

[0061] The statement-driven state structure description includes action sequence identification, target object classification, background semantic features, and device state mapping. The adjusted control behavior chain includes rearranged statement structure, connected behavior word set, continuous instruction nodes, and control link path. The main control instruction path includes the main control verb phrase, object phrase priority, control connection path, and word frequency density characteristics. The temporal prominent trend of the instruction content includes the advance rule of behavioral instructions, the instruction frequency distribution curve, and the time interval change trend. The artificial intelligence home service plan includes content display sequence, sorting number update, and behavior identification priority.

[0062] The specific steps of S1 are:

[0063] S101: Obtaining the command sentence received by the chatbot, extracting verbs and conjunctions, arranging the order of action phrases, identifying and locating the target noun item content, and generating a sentence action ranking value;

[0064] First, the entire sentence is preprocessed, all punctuation marks in the sentence are identified and stop words are removed. The remaining morphemes are classified and identified by part of speech. On this basis, each phrase in the sentence is scanned to extract verb items with action characteristics, such as "start", "turn off", "detect", etc., and filtered according to the frequency statistics of the verbs in the sentence, excluding low-frequency verbs with less than 1%. At the same time, the conjunctions adjacent to the verbs, such as "and", "then", and "subsequently", are extracted. The position number of each action and its conjunction in the original sentence is recorded, and then the action phrase structure is formed. For example, for the instruction "detect the temperature and turn on the fan", two action phrases "detect the temperature" and "turn on the fan" are identified, where the verbs "detect" and "turn on" are located in the first and third positions respectively, and the conjunction "and" is located in the second position. Combined with the structured number records, an action sequence table is generated. At the same time, in order to accurately identify the content of the target noun item, the predefined domain entity vocabulary is called to match the "temperature" and "fan" in the action phrase to sensor and actuator class objects respectively. Further mapping is performed according to the position in the sentence to establish a verb-noun correspondence table, such as "detect-temperature" and "turn on-fan", and the mapping result is assigned to the ranking reference value of the action phrase. The action ranking value is based on the position of the verb in the sentence, and the noun position deviation is used as a fine-tuning factor. When the distance between the noun item and the action verb exceeds 3 words, the combination is eliminated. The ranking value is uniformly aggregated into an integer value group, and finally a group of action phrase sequences such as "detect temperature (1) - turn on fan (2)" are output for the next stage of semantic expansion processing.

[0065] S102: Calling the action ranking value of the statement, identifying the time, place, and manner modifiers in each action statement, extracting the combination structure between the verb and the modifier, calculating the matching frequency and word class association index, obtaining the semantic strength between the combinations, and generating the semantic association degree of the action modifier combination;

[0066] After calling the above action phrase ranking value, take each action phrase as the core, extract no more than 5 word units around it, identify the time, place, manner and other modifying components, and filter the components such as "at...", "through..." equipment, "from..." position, etc. through the preset dictionary. The recognition result must match the standard modification template in the dictionary. The matching degree is based on the keyword overlap rate as the standard setting threshold of 0.6. Those below this value are regarded as non-modifiers and eliminated. In the sentence "turn on the fan at night through the remote control", "night" is identified as a time modifier and "remote control" is identified as a manner modifier, which are respectively consistent with the verb " The combination "open" appears 37 times in the database, while the combinations "open-noon" and "open-button" appear 28 and 49 times, respectively, for a total of 160 combinations. Therefore, the frequency of "open-evening" is 23.1%. Taking into account the fixed collocation ratio between word classes, the average co-occurrence ratio of the method type is 38.4%, and the time type is 26.7%. When calculating the correlation, the average ratio of these categories is used as a reference, and the semantic strength of the modification combination of "open-evening" is set to 23.1 / 26.7≈0.865. In this process, a threshold of 0.3 is set; combinations with modification strengths below this value are discarded. The combination modification strength is ultimately used to measure the degree of semantic dependence between the action and the modified item. A higher value indicates that the combination relationship is stable and reliable. For example, the time type modifier "morning" in the sentence "detecting the temperature in the morning through the sensor" has a co-occurrence frequency of 35.2%, which is higher than the average reference ratio of 27%, resulting in a semantic strength of 1.30.

[0067] S103: Based on the semantic association of the action modifier combination, the on / off response value, voltage feedback value, and operation response change rate of the current device are collected, and trend ratio conversion is performed between the semantic chain structure and the device response to generate a statement-driven state structure description;

[0068] Based on the semantic strength combination data obtained above, each semantic chain is associated with the device-side feedback data for analysis to determine the actual response state driven by the semantic chain. The specific method is to read the device status parameters before and after the command is executed, including the switch state value (for example, the fan state 0 is off and 1 is on), the voltage feedback value (for example, for a DC fan, the feedback voltage value is 12.3V), and the reaction rate (for example, the time from receiving the command to the fan starting to rotate is 1.2 seconds). The value difference between the two operations is collected. For example, if the fan changes from 0 to 1, the voltage changes from 0V to 12.3V, and the reaction time changes from no change to 1.2 seconds, the change values ​​of each parameter are calculated as 1, 12.3, and 1.2. Combined with the semantic strength of 0.865 for the semantic chain "Turn on the fan at night with the remote control", the three parameter change values ​​are combined with the semantic strength to generate the semantic chain response coefficient. By equally weighting the three parameters and multiplying them by the semantic strength, the response mapping value is (1 + 12.3 + 1.2) × 0.865 = 14.5 × 0.865. ≈12.52. The mapping coefficient threshold is set to 5. This value is derived from the mean mapping coefficient of all valid execution actions in historical samples, which is 4.6 and has a standard deviation of 1.1. The default threshold is the mean + 0.4 standard deviation, which is approximately 5.04. Therefore, 12.52 is higher than this threshold, indicating that the semantic chain is stable. Records with a response coefficient higher than 5 will be retained in all semantic chains for the next round of semantic iteration.

[0069] The specific steps of S2 are:

[0070] S201: Obtain a statement-driven state structure description, extract verb items and corresponding direction information in current and recent instructions, calculate trend consistency based on the angle ratio of direction vectors, and generate a verb trend consistency ratio;

[0071] First, all recognized action phrases are extracted from the semantic parsing module and their time sequence records are established. Then, the verb items in the current and past 5 instructions are called and their direction descriptions are extracted. The directional morphemes attached to each verb record, such as "left", "right", "close", "away", etc., are matched with the standard vector values ​​in the preset direction library. For example, the corresponding vector for "left" is (-1, 0), and the corresponding vector for "close" is (0, 1). All direction descriptions are mapped to two-dimensional coordinate vectors. Then, they are sorted according to the timestamps of the actions, and a verb sequence and direction vector group are constructed. The angle between each pair of verb direction vectors is calculated, and the cosine value of the angle between each two adjacent action direction vectors is calculated. When the directions are consistent, the cosine value is close to 1, and the opposite direction is -1. The corresponding cosine value of the deviation exceeding 45° is lower than 0.707. The trend consistency ratio is defined as The sum of the direction cosine values ​​between all adjacent actions is divided by the total number of action pairs. For example, if a device record contains five actions, such as "turn on fan forward," "turn off light backward," "move camera left," "adjust volume left," and "detect temperature left," their direction vectors are (0, 1), (0, -1), (-1, 0), (-1, 0), and (-1, 0), respectively. The cosine values ​​of the angles between adjacent vectors are calculated to be -1, 0, 1, and 1, respectively. The average value is 0.25, which is the trend consistency ratio. The judgment threshold is set to 0.6. This threshold is obtained by retrieving the past 100 device operation records, statistically analyzing the distribution range of the trend consistency ratio, and establishing a reasonable boundary interval of [0.4, 0.8] based on the median and standard deviation. Finally, 0.6, which is close to the median, is used as the judgment threshold to output the verb trend consistency ratio.

[0072] S202: Calling the verb trend consistency ratio, extracting the device status feedback and return content corresponding to the target noun item, calculating the control response offset value, combining the offset value with the trend ratio, identifying the verb behavior characteristics with abnormal direction, and generating the verb reverse offset rate;

[0073] Retrieve the target noun item corresponding to each action verb in the structural state record table, and then locate the corresponding device type. Obtain the current state parameter value through the device feedback module, such as whether the device is currently running, whether the voltage feedback is within the standard range, whether the signal delay is abnormal, etc. Then extract the most recent action feedback result of the same device from the historical record, and calculate its deviation from the current feedback. The deviation is defined as the total difference in the change values ​​of the three state parameters. For example, if the current feedback state of the "turn off the light" command after execution is "operating state = 0", "feedback voltage = 0V", and "signal delay = 1.2s", and the previous feedback of the same type of command is "operating state = 1", "feedback voltage = 5V", and "signal delay = 0.8s", then the difference between the three parameters is 1, 5, and 0.4. The total offset value is 6.4. This value is combined with the trend consistency ratio to judge the degree of directional abnormality. The judgment standard is that the offset value is greater than 4 and the trend ratio is less than 0.4, which is an abnormal behavior. Each action is classified and identified, and marked as a reverse offset action. The corresponding action "turn off the light" in the record has an offset value of 6.4 and a trend ratio of 0.25, which meets the conditions. It is given a reverse offset mark of 1, and the rest that do not meet the conditions are marked as 0. Finally, a list of verb behavior offset records is formed. At the same time, the offset value range is defined, and the offset intensity segment is divided by setting a threshold. The offset value below 2 is defined as "weak offset", 2-5 as "medium offset", and greater than 5 as "strong offset". Combined with the trend ratio value below 0.4, it is marked as a reverse action, and finally the verb reverse offset rate of each action is output.

[0074] S203: Identify the order and direction relationship of adjacent instructions in the action chain based on the verb reverse offset rate, rearrange the structural positions and insert connectives to construct a continuous action sequence, and generate an adjusted control action chain;

[0075] According to the verb reverse offset rate generated in the previous stage, the order and direction relationship between each instruction in the action chain is sorted out. In the current action sequence, it is identified whether there is a reverse offset mark between all adjacent instruction pairs. If so, the current action pair is marked as "structure needs to be adjusted", and then the order between the action and the subsequent action is swapped, and connectives are inserted for logical bridging. The choice of connectives is based on the degree of semantic association between actions and the instruction triggering time. Phrases such as "then", "next" or "reversely" that express causal or contrast relationships are preferred. For example, if the original sequence is "turn on the camera → turn off the light → turn to the left", and "turn off the light" is replaced by "turn off the light". If it is marked as a reverse offset, it will be post-positioned as "turn on the camera → turn to the left → then turn off the light". During the process, the timing number of the action will be updated each time the order is exchanged to ensure that the instruction execution logic remains consistent. Then a new behavior chain list is generated based on the rearranged structure. The action nodes record the adjusted position, original sequence number, whether it is rearranged mark and other meta-information in turn. If an action is rearranged more than twice in a row, the conjunction "finally" is inserted after it to indicate the end of the process. During the setting process, the default reordering limit is set to no more than three times. If it exceeds this limit, the action will be fixed and will not move. Finally, the continuous control behavior chain formed by the sorted new structure and its conjunctions in all action chains is output.

[0076] The specific steps of S3 are:

[0077] S301: Obtain the instruction actions in the adjusted control behavior chain, extract directional verbs and calculate the direction angle value based on the instruction position, filter out the verbs with angles greater than the direction conflict threshold, and generate a direction conflict verb tag set;

[0078] After obtaining the instruction actions in the adjusted control behavior chain, read the verb items of each action and their sequential position in the control chain one by one. By looking for directional morphemes carried in the action description, such as "forward", "backward", "turn left", "move right", etc., the directional verbs are mapped to standard vectors on two-dimensional coordinates. For example, "forward" is set to (0, 1), "left" is (-1, 0), and "right" is (1, 0). The direction vectors of two adjacent instructions are paired and numbered in the corresponding order. Then, the directional angle value between each pair of vectors is calculated. The angle value is converted into an angle by calculating the dot product and module length relationship between the two vectors. All angles are presented in the form of angles for easy screening. For example, if instruction 1 is "move the camera forward", the vector is (0, 1), and instruction 2 is "rotate the camera right", the vector is (1, 0), the angle is 90 degrees, and the direction conflict threshold is set to 60 degrees. This threshold is based on the analysis of the median value of 52 degrees and the standard deviation of 6 degrees of the direction change statistics of the normal operation sequences in the past 200 control chains. The reasonable critical point is 60 degrees, which is slightly higher than the median value plus 1.3 times the standard deviation. If the angle between a pair of actions exceeds 60 degrees, the verb is marked as a direction conflict item. For example, the verb "rotate" corresponding to the above 90-degree angle will be marked. All marked directional verbs are recorded in the direction conflict verb marker set and used for screening in the next stage.

[0079] S302: Call the direction conflict verb tag set, extract the verb phrases and object phrases in the remaining instructions, count the frequency of combination occurrences, calculate the word frequency density, construct a permutation sequence according to the combination order, and generate a master control word frequency density ranking value;

[0080] The specific formula for calculating word frequency density is:

[0081] ;

[0082] in, represents the word frequency density of the combination of the i-th verb phrase and the j-th object phrase, represents the frequency of the i-th verb phrase and the j-th object phrase appearing together in the original corpus, represents the semantic deviation of the corresponding object phrase of the i-th verb phrase in the k-th context, represents the average value of the semantic deviation of the object phrase in all contexts of the i-th verb phrase, represents the frequency of the i-th verb phrase in the m-th context, represents the frequency of the combination of the t-th verb phrase and the j-th object phrase, represents the average frequency of all verb phrases combined with the j-th object phrase, is the total number of contexts;

[0083] This formula is designed to calculate the combined frequency-density eigenvalue from known data points and frequency parameters. , which is a numerical value that measures the characteristics of the combination of a verb phrase and an object phrase. For the purpose of specific calculation examples, we select parameter values ​​that are actually monitored or calculated.

[0084] Parameter setting and acquisition:

[0085] represents the frequency of the i-th verb phrase and the j-th object phrase appearing together in the original corpus. For example, if the combination of “development-project” appears 30 times in the dataset, then .

[0086] and Represents the semantic deviation of the object phrase of the i-th verb phrase in the k-th context and its average value. These values ​​are quantified by semantic analysis tools, such as It may vary in different documents. If the five documents are 2.1, 2.0, 1.9, 2.2, and 2.3 respectively, then .

[0087] is the word frequency of the ith verb phrase in the mth context, which can be obtained through text analysis software statistics. For example, if it appears 25 times in a context, then .

[0088] and Also obtained through statistics, assuming The number of occurrences in different documents is 10, 15, 20, then .

[0089] Formula calculation process:

[0090] First calculate The absolute value of and sum, let The values ​​are 2.1, 2.0, 1.9, 2.2, 2.3, then:

[0091] ;

[0092] calculate The sum of , assuming is 25, 30, 35, then:

[0093] ;

[0094] right Take the absolute value of the difference and sum it, then divide it by n, let is 10, 15, 20, then:

[0095] ;

[0096] Finally, substitute the above values ​​into The formula is calculated:

[0097] ;

[0098] This result shows that the calculated frequency feature value is 0.197, which represents the frequency characteristics of the combination of verb phrases and object phrases, and is used for sorting and analysis in subsequent data processing.

[0099] S303: Locate the control structure between phrases and calculate the path connection strength based on the frequency density ranking value of the main control words, select the path combination whose connectivity and main control degree both reach the threshold, and generate the main control instruction path;

[0100] According to the frequency density ranking value of the main control word, the connection relationship between the combination items is sorted out. First, all consecutive phrase pairs are located and a connection path table is established. For each two adjacent combination items, such as "turn on the light → adjust the volume" and "adjust the volume → move the camera", their distance values ​​in the behavior chain are marked. The difference in the sequence number in the path is taken as the basic parameter of the initial connection strength. Then, the average density is calculated by combining the frequency density values ​​of the two, and the path connection strength is set as the product of the inverse of the number difference and the average density. For example, "turn on the light" is numbered 1, and "adjust the volume" is numbered 2. The difference is 1, the density is 0.45 and 0.35 respectively, and the average density is 0.4. The connection strength is set to 1×0.4=0.4. The strength of all paths is calculated and the strength screening lower limit is set to 0.25. This threshold is obtained by analyzing nearly 30 valid The mean path strength in the control chain is 0.22 and the standard deviation is 0.03. The mean plus double the standard deviation is taken as the screening benchmark. Those with a strength of no less than 0.25 are retained for the next round of analysis. Subsequently, the master control degree of the starting phrase and the ending phrase in the path are further identified in the screened path. The master control degree is defined as the proportion of the number of times it is cited as a trigger condition in the control chain. If "turn on the light" is cited 3 times by other instructions and appears in 10 chains, its master control ratio is 0.3. The master control degree threshold is set to 0.25. If it is lower than this value, it is not considered a master control action. Finally, a path combination with a connection strength of no less than 0.25 and a master control degree of no less than 0.25 is screened out. For example, "turn on the light → adjust the volume" has a connection strength of 0.4 and a master control degree of 0.3, which meets the requirements. Finally, the master control instruction path is constructed and the structure list is output.

[0101] The specific steps of S4 are:

[0102] S401: Calling the action phrase in the main control instruction path, searching for the corresponding time point in the task record, locating the number position of each instruction in chronological order, calculating the interval length between the time points, and generating the instruction occurrence timing interval value;

[0103] The calculation formula for the interval between time points is:

[0104] ;

[0105] in, Represents the duration between time points. Represents the triggering time point of the first instruction in the current time period, Represents the triggering time point of the last instruction in this time period, Represents the total number of actions recognized during this time period, represents the time offset weight of each action, Represents the time offset corresponding to each action, Represents the arithmetic mean of all offsets;

[0106] In the main control instruction path, the following action phrases and their associated parameters are identified:

[0107] Number of action phrases (A): 3;

[0108] The time offset (d) of each action phrase:

[0109] Action 1: 0.4 seconds;

[0110] Action 2: 0.6 seconds;

[0111] Action 3: 0.5 seconds;

[0112] The time offset weight (w) of each action phrase:

[0113] Action 1: 1.0;

[0114] Action 2: 1.2;

[0115] Action 3: 0.8;

[0116] The calculation steps are as follows:

[0117] Calculate the average of the time offsets ( ):

[0118] ;

[0119] Compute the average of the weighted time offsets:

[0120] ;

[0121] Compute the standard deviation of the time offset:

[0122] ;

[0123] Set the command trigger time:

[0124] Start time ( ): 10.0 seconds;

[0125] End Time ( ): 12.0 seconds;

[0126] Substitute the above calculation results into the formula to calculate the time interval value ( ):

[0127] ;

[0128] The results show that the timing interval between two consecutive main control instructions is 2.4251 seconds, which reflects the combined influence of the weighted average and standard deviation of the instruction triggering time interval and the action phrase time offset.

[0129] S402: Based on the instruction timing interval values, determine whether the instruction intervals are gradually shortening, select segments where the difference between consecutive intervals shows a decreasing trend, count the number of trend segments, perform threshold comparison, and generate a value for the number of advanced trend segments;

[0130] Compare the differences between each interval value one by one to identify whether there is a phenomenon of continuous decreasing intervals. The judgment standard is that the current interval value is less than the previous value and the next value is still lower than the current one. Only when at least three decreasing trends are formed can it be considered a valid trend segment. For each candidate interval, three intervals are compared. Those that meet the conditions are marked as "trend segments". Take the sequence "25s→20s→15s→12s→9s" as an example. The three consecutive interval values ​​in this sequence are all less than their previous ones. It can be judged as a decreasing trend segment. Perform the same processing on all interval sequences, accumulate the number of trend segments, and set a reasonable threshold to determine whether a trend segment is formed. For the advance trend, the average number of trend segments identified in 50 complete control chains in the history is 2.4 segments, with a standard deviation of 0.6 segments. The average value plus 0.5 times the standard deviation is selected as the judgment limit, and the threshold is 2.7 segments. In the actual calculation, it is rounded to 3 segments. If the number of trend segments identified in a certain task is 3 or more, it is judged that there is an advance behavior. If there are interval groups "30→25→22", "18→14→10", and "9→7→5" in the current instruction chain, a total of 3 decreasing segments, then the task is marked as having an advance trend, and the output value of the number of advance trend segments is 3.

[0131] S403: Calling the number of advance trend segments, extracting action phrases with advance characteristics, calculating the frequency of occurrence in the task record, filtering out content with a frequency greater than the average phrase frequency and marking it as advance behavior, and generating a temporal highlight trend of the instruction content;

[0132] After calling the number of advance trend segments, we backtrack to construct the action phrases in each trend segment and extract all verb phrases with decreasing intervals, such as "adjust the volume", "move the camera", "detect the temperature", etc. Then, we count the frequency of occurrence of these phrases in the entire task record, take the total number of occurrences of all phrases in 50 records and divide it by the total number of phrases to obtain the average frequency value. Assuming that the total number of phrases is 15 and the total number of occurrences is 120 times, the average frequency is 8 times. We compare the occurrence frequencies of the phrases involved in the advance trend segment one by one. If a phrase such as "adjust the volume" appears If the number of occurrences is 12, which is greater than the average frequency of 8, it is determined to be a forward-moving behavior phrase, and its first occurrence position in all instructions is recorded. If the first position is earlier than its preset position in the logical sequence, the phrase is doubly confirmed as a forward-moving behavior phrase. The threshold is set to the average frequency in the judgment, that is, only phrases with a frequency greater than 8 are retained. Finally, a set of forward-moving phrases, such as "adjust the volume" and "detect the temperature", are screened out. Combined with their first occurrence time sequence number and the task behavior chain structure, a temporal position distribution diagram of these phrases is generated, and finally a set of temporal prominent trends of all instruction contents that meet the conditions is output.

[0133] The specific steps of S5 are:

[0134] S501: Calling the behavior identifier in the temporal highlight trend of the instruction content, locating the original ranking position of the corresponding behavior in the recommended content display list, extracting the initial number and index of the action phrase in the list, establishing a mapping relationship between the number and the action phrase, and generating an initial ranking index result of the action instruction;

[0135] After calling the behavior identifier in the temporal salient trend of the instruction content, each action phrase marked as a forward-moving behavior is first extracted and compared with the behavior items in the current recommended content display list in sequence to identify their initial number and physical storage index position in the list. During this operation, the display list is numbered sequentially, for example, "Number 1: Open the curtains," "Number 2: Start the humidifier," "Number 3: Adjust the lights," etc., and a phrase comparison rule is established. When the action phrase completely matches the action name term in the displayed content or the subject and predicate core words are consistent and the semantic similarity is greater than 0.8, it is considered a valid match. The initial number and current storage index are marked in the matching item. For example, if "Adjust the lights" is number 3 and index position 5 in the list, the recorded mapping relationship is "Adjust the lights → Number 3 → Index 5." In this way, a multidimensional mapping structure containing three pieces of information: action phrase, initial number, and index position is constructed. This structure is established one by one for all temporal salient behaviors, ultimately forming a complete initial sorted index result set of action instructions.

[0136] S502: Based on the initial sort index result of the behavior instruction, the sort numbers of the similar action nodes marked as being moved forward are adjusted forward, the corresponding contents are rearranged according to the adjusted number sequence, and the remaining unmarked nodes are filled in according to the original sequence to generate a sorted number set after the forward shift;

[0137] According to the initial sorting index results of the behavioral instructions generated in the previous section, first filter all similar action nodes marked as "forward", remove duplicates according to their original numbers and sort them in ascending order, then advance their numbers in the display list, that is, move the numbers to the front of the display sequence, and sort them according to their temporal prominence. The temporal prominence is determined by the number of forward moves or the number of consecutive appearances in the advance trend segment. For example, if a phrase "adjust the lights" appears three times in three different trend segments, its sorting priority is 3, and another phrase appears only once, so its priority is 1. Set The priority sorting threshold is set to 2. All forward-moving items with a value higher than this value are prioritized to the top, and the rest are arranged in order. After the sorting is completed, the number in the display list is updated. For example, the "Adjust Lights" item originally numbered "No. 6" is updated to "No. 1" after being moved forward. Subsequently, all nodes not marked as forward-moving are arranged in their original order starting from the current largest number. That is, after the forward movement is completed, the remaining items are numbered in sequence starting from "No. 4". All number change records are synchronously updated to the index structure table and used for subsequent path mapping operations, and finally a number set after forward movement is generated.

[0138] S503: Calling the forward and rearranged number set to reconstruct all behavior content and function indexes in the display list, matching the function module positions with the main control direction to form a structural combination, integrating the sorting relationship, and generating an AI-powered home service plan;

[0139] After calling the forward sorting number set, scan the numbers and behavior phrases of all rearranged behavior items in the display list in turn, and compare them with the original function module index structure to extract the function module position coordinates corresponding to each behavior. The coordinate structure is recorded in a two-dimensional index format. For example, the module corresponding to "adjusting the light" is "light control", and the index is (0, 1). The module corresponding to "starting the humidifier" is "environment adjustment", and the index is (1, 0). The behavior number and the function module index are mapped and combined with the main control direction identifier in the forward sorting number, that is, the direction information of the behavior. For example, if the direction of "adjusting the light" is "front of the living room", then its direction main control value is 1, and the direction of "starting the humidifier" is "corner of the living room". ", the main control value is 0.7, and the connectivity and main control degree of each module combination are weighted and multiplied to obtain the structural combination strength. For example, the combination of "adjust the light → start the humidifier" is numbered 1 and 2, the module index distance is 1, and the sum of the main control values ​​is 1.7. The combination strength is defined as the sum of the main control values ​​divided by the module distance, which is 1.7. The screening threshold is set to 1.2, and all combinations with strength values ​​greater than this value are considered valid structural paths. Finally, all valid structural combinations are integrated with their numbering sequences to output an artificial intelligence home service solution composed of the main control path and the behavior module combination. The solution is centered on the structural path and includes sorting numbers, behavior phrases, function index positions, and main control direction values.

[0140] See also Figure 2 , an artificial intelligence-based home service system, comprising:

[0141] The sentence structure recognition module acquires the command sentences received by the chatbot in the home environment, analyzes the verbs, conjunctions, target nouns, and modifiers in the sentences, extracts the task orientation, collects the current device's on / off response, voltage feedback, and operational response performance, constructs a corresponding relationship between the semantic behavior chain and the device status, and generates a sentence-driven state structure description.

[0142] The instruction tendency analysis module analyzes the consistency between the current instruction action and the recent instruction trend based on the action sequence in the statement-driven state structure description, identifies the abnormal behavior in the instruction direction, extracts the control state and system feedback of the action target, rearranges the statement structure and inserts connecting action words to generate the adjusted control action chain;

[0143] The control path generation module disassembles the instruction actions from the adjusted control behavior chain, identifies and eliminates conflicting directional verbs, calls the verb phrases and target phrases in the remaining instructions, analyzes the action frequency and sequence relationship, selects the path with the most concentrated control connectivity and master control degree, and generates the master control instruction path;

[0144] The timing trend judgment module calls the action phrases in the main control instruction path, extracts the timing position in the task record, analyzes the time interval trend, determines whether there is a pattern of continuous advancement of multiple behavior contents, and generates the timing prominent trend of the instruction content;

[0145] The sequence priority adjustment module highlights the trend according to the timing of the instruction content, identifies the initial sequence position in the recommended content display list, adjusts the sequence numbering of similar action nodes, and generates an AI-powered home service plan.

[0146] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A home service method based on artificial intelligence, characterized in that: The following steps are involved: S1: Obtain household sentence input, parse the action sequence composed of verbs and conjunctions, determine the task orientation based on the target noun, identify modifiers to construct background semantics, and simultaneously collect device on / off, voltage, and operation feedback to establish a sentence-driven state structure description; S2: Based on the action sequence in the state structure description driven by the statement, determine the instruction trend and trend consistency, identify the direction change and control state feedback, restructure the statement and insert linking words to generate an adjusted control behavior chain; S3: Decomposing the adjusted control behavior chain, identifying and eliminating conflicting directional verbs, calling the remaining verb phrases and target phrases, analyzing the primary and secondary order and control connection density, and generating a main control instruction path; S4: calling the task sequence of the action in the main control instruction path, extracting the time points of the behavioral instructions in the recent interaction, analyzing the interval trend, and generating the time sequence highlight trend of the instruction content; S5: Extract key behaviors based on the temporal trend of the instruction content, locate the initial sequence in the recommendation list, sort similar action nodes forward, maintain the original order in combination with unlabeled behaviors, and generate an artificial intelligence home service plan.

2. The artificial intelligence-based home service method according to claim 1, characterized in that: The statement-driven state structure description includes action sequence identification, target object classification, background semantic features, and device state mapping. The adjusted control behavior chain includes rearranged statement structure, connected behavior word set, continuous instruction nodes, and control link path. The main control instruction path includes main control verb phrase, object phrase priority, control connection path, and word frequency density features. The temporal prominent trend of the instruction content includes the advance rule of behavior instructions, instruction frequency distribution curve, and time interval change trend. The artificial intelligence home service plan includes content display sequence, sorting number update, and behavior identification priority.

3. The artificial intelligence-based home service method according to claim 1, characterized in that: The specific steps of S1 are: S101: Obtaining the command sentence received by the chatbot, extracting verbs and conjunctions, arranging the order of action phrases, identifying and locating the target noun item content, and generating a sentence action ranking value; S102: Calling the sentence action ranking value, identifying the time, place, and manner modifiers in each action sentence, extracting the combination structure between the verb and the modifier, calculating the matching frequency and word class association index, obtaining the semantic strength between the combinations, and generating the semantic association degree of the action modifier combination; S103: According to the semantic association of the action modification combination, the on-off response value, voltage feedback value and operation response change rate of the current device are collected, and trend ratio conversion is performed on the semantic chain structure and the device response to generate a statement-driven state structure description.

4. The artificial intelligence-based home service method according to claim 1, characterized in that: The specific steps of S2 are: S201: Obtain a state structure description driven by the statement, extract verb items and corresponding direction information in the current and recent instructions, calculate trend consistency based on the angle ratio of the direction vectors, and generate a verb trend consistency ratio; S202: Calling the verb trend consistency ratio, extracting the device status feedback and return content corresponding to the target noun item, calculating the control response offset value, combining the offset value with the trend ratio, identifying the verb behavior characteristics with abnormal direction, and generating the verb reverse offset rate; S203: Identify the order and direction relationship of adjacent instructions in the action chain according to the verb reverse offset rate, rearrange the structural positions and insert conjunctions to construct a continuous behavior sequence, and generate an adjusted control behavior chain.

5. The artificial intelligence-based home service method according to claim 1, characterized in that: The specific steps of S3 are: S301: Obtaining instruction actions in the adjusted control behavior chain, extracting directional verbs and calculating directional angle values ​​based on instruction positions, screening verbs with angle values ​​greater than a directional conflict threshold for tag exclusion, and generating a directional conflict verb tag set; S302: calling the direction conflict verb token set, extracting the verb phrases and object phrases in the remaining instructions, counting the frequency of occurrence of the combination, and calculating the word frequency density, constructing a permutation sequence according to the combination order, and generating a master control word frequency density ranking value; S303: Locate the control structure between phrases and calculate the path connection strength according to the master control word frequency density ranking value, select the path combination whose connectivity and master control degree both reach the threshold, and generate the master control instruction path.

6. The artificial intelligence-based home service method according to claim 1, characterized in that: The specific steps of S4 are: S401: calling the action phrase in the main control instruction path, searching for the corresponding time point in the task record, locating the number position of each instruction in chronological order, calculating the interval length between the time points, and generating the instruction occurrence timing interval value; S402: Based on the instruction occurrence timing interval value, determine whether the instruction interval is gradually shortened, select segments where the consecutive interval difference values ​​show a decreasing trend, count the number of trend segments, perform threshold comparison, and generate a value for the number of advanced trend segments; S403: Call the advance trend segment quantity value, extract action phrases with advance characteristics, calculate the frequency of occurrence in the task record, filter the content with a frequency greater than the average phrase frequency and mark it as forward-moving behavior, and generate a temporal prominent trend of the instruction content.

7. The artificial intelligence-based home service method according to claim 1, characterized in that: The specific steps of S5 are: S501: Calling the behavior identifier in the temporal highlight trend of the instruction content, locating the original sort position of the corresponding behavior in the recommended content display list, extracting the initial number and index of the action phrase in the list, establishing a mapping relationship between the number and the behavior phrase, and generating an initial sort index result of the behavior instruction; S502: Based on the initial sort index result of the behavior instruction, the sort numbers of the similar action nodes marked as being moved forward are adjusted forward, the corresponding contents are rearranged according to the adjusted number sequence, and the remaining unmarked nodes are continuously filled in according to the original sequence to generate a sorted number set after the move forward; S503: Call the forward-shifted sorting number set to reconstruct all behavioral content and function indexes in the display list, match the functional module positions with the main control direction to form a structural combination, integrate the sorting relationship, and generate an artificial intelligence home service plan.

8. An artificial intelligence-based home service system, characterized in that: According to any one of claims 1 to 7, the artificial intelligence-based home service method comprises: The sentence structure recognition module obtains the command sentences received by the chatbot in the home environment, analyzes the verbs, conjunctions, target nouns and modifiers in the sentences, extracts the task directionality, collects the current device's on / off response, voltage feedback and operational response performance, constructs a corresponding relationship between the semantic behavior chain and the device status, and generates a sentence-driven state structure description; The instruction tendency analysis module analyzes the consistency of the current instruction action with the recent instruction trend based on the action sequence in the state structure description driven by the statement, identifies the abnormal behavior of the instruction direction, extracts the control state and system feedback of the action target, rearranges the statement structure and inserts connecting action words to generate the adjusted control action chain; A control path generation module disassembles instruction actions from the adjusted control behavior chain, identifies and eliminates conflicting directional verbs, calls verb phrases and target phrases in the remaining instructions, analyzes the frequency and sequence of actions, selects a path with concentrated control connectivity and master control degree, and generates a master control instruction path; A timing trend judgment module calls the action phrases in the main control instruction path, extracts the timing position in the task record, analyzes the time interval trend, determines whether there is a pattern of continuous advancement of multiple behavior contents, and generates a timing prominent trend of the instruction content; The sequence priority adjustment module identifies the initial sequence position in the recommended content display list based on the temporal highlight trend of the instruction content, adjusts the sequence numbering of similar action nodes, and generates an artificial intelligence home service plan.

Images