Navigation instruction analysis method and system, electronic equipment and storage medium

Abstract

The invention provides a navigation instruction analysis method and system, electronic equipment and a storage medium, and relates to the technical field of navigation. The analysis method of the navigation instruction comprises the following steps: determining a navigation event chain according to a received navigation instruction text; according to the navigation event chain and preset navigation element formalized information, performing text segmentation on the navigation event chain to determine a navigation activity composite instruction; according to preset reference formalized information, performing text segmentation on the navigation activity composite instruction to determine preliminary structured instruction information; and complementing the preliminary structured instruction information according to an atomic instruction in the navigation event chain and a preset candidate entity text so as to determine a corresponding navigation instruction knowledge structure when the navigation instruction text reaches the navigation subject. According to the method, the multi-element and complex navigation instruction is analyzed in a hierarchical cooperation mode, the navigation instruction text is converted into the structured instruction, and the instruction execution efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of navigation technology, for example, to a navigation instruction parsing method, system, electronic device and storage medium. Background Art

[0002] In recent years, with the rapid development of unmanned intelligent technology, various types of robots are increasingly used in many livelihood service-related industries such as power inspection, agricultural and forestry monitoring, emergency rescue, housekeeping services, logistics distribution, etc. This technology can improve production efficiency, reduce labor costs, and improve service quality. In order to follow the user's intention to complete a specific task, the robot needs to fully understand the user's navigation instructions, so as to execute the relevant action sequence and complete specific task activities at the designated destination. On the technical level, navigation instruction parsing mainly uses natural language processing (NLP), knowledge graph (KG) and other intelligent technologies to understand and analyze human instruction language, and convert human natural language instructions with rich semantics, complex structure, and flexible expression into machine-readable and executable structured instructions.

[0003] In the related technology, due to the limitation of current usage scenarios, the parsing process of navigation instructions involves fewer parsing elements, and the relationship between the elements is simple. The currently commonly used single models such as rule models or end-to-end sequence labeling models to identify element entities in navigation instructions cannot accurately understand the user's complex task instruction intentions, and thus it is difficult to control the robot to accurately perform specific navigation tasks. Summary of the invention

[0004] The present application aims to provide a navigation instruction parsing method, system, electronic device and storage medium.

[0005] According to one aspect of the present application, a navigation instruction parsing method is proposed, comprising:

[0006] Determine a navigation event chain according to the received navigation instruction text, wherein the navigation event chain is used to represent a series of navigation-related activities performed by the navigation subject;

[0007] According to the navigation event chain and the preset navigation element formal information, the navigation event chain is segmented to determine the navigation activity composite instructions;

[0008] According to the preset reference formal information, the navigation activity composite instruction is segmented to determine the preliminary structured instruction information;

[0009] According to the atomic instructions in the navigation event chain and the preset candidate entity texts, the preliminary structured instruction information is completed to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is sent to the navigation subject.

[0010] According to one aspect of the present application, a navigation instruction parsing system is provided, comprising:

[0011] A navigation event chain determination module, used to determine a navigation event chain according to the received navigation instruction text, wherein the navigation event chain is used to represent a series of navigation-related activities performed by the navigation subject;

[0012] A composite instruction determination module is used to perform text segmentation on the navigation event chain according to the navigation event chain and preset navigation element formalization information to determine the composite instruction of the navigation activity;

[0013] A compound instruction segmentation module is used to segment the navigation activity compound instruction text according to preset reference formal information to determine preliminary structured instruction information;

[0014] The information completion module is used to complete the preliminary structured instruction information according to the atomic instructions in the navigation event chain and the preset candidate entity text, so as to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is sent to the navigation subject.

[0015] According to one aspect of the present application, an electronic device is provided, which includes: a processor; and a memory storing a computer program, wherein when the computer program is executed by the processor, the processor executes the navigation instruction parsing method as described above.

[0016] According to one aspect of the present application, a non-transitory computer-readable medium is provided, on which readable instructions are stored. When the instructions are executed by a processor, the processor executes the navigation instruction parsing method as described above.

[0017] It should be understood that the foregoing general description and the following detailed description are exemplary only and are not restrictive of the present application.

[0018] Beneficial effects:

[0019] Through the above-mentioned embodiment provided by the present application, by receiving and parsing the navigation instruction text, a navigation event chain can be constructed to clearly reflect a series of navigation-related activities that the navigation subject needs to perform. The present application not only considers the overall structure of the navigation event chain, but also further performs text segmentation on the navigation event chain according to the preset navigation element formalization information, thereby identifying the navigation activity composite instruction. This refined processing enables the system to more accurately understand the spatial and temporal elements in the navigation instruction, and the relationship between them. By further performing text segmentation on the navigation activity composite instruction using the preset reference formalization information, the navigation instruction can be converted into preliminary structured instruction information. This structured information is not only convenient for system storage and processing, but also helps to improve the execution efficiency and accuracy of the navigation instruction. Combined with the atomic instructions in the navigation event chain and the preset candidate entity text, the preliminary structured instruction information can be completed, thereby constructing a complete navigation instruction knowledge structure. This completion mechanism ensures the integrity and accuracy of the navigation instruction and reduces the execution errors caused by missing information. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without exceeding the scope of protection required by the present application.

[0021] Figure 1 A schematic diagram of the structure and function of the analysis system provided in the embodiment of the present application;

[0022] Figure 2 A flowchart of a method for parsing navigation instructions provided in an embodiment of the present application;

[0023] Figure 3 A block diagram of a navigation instruction parsing system provided in an embodiment of the present application;

[0024] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0025] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this application will be comprehensive and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar parts, and thus their repeated description will be omitted.

[0026] In addition, described feature, structure or characteristic can be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical scheme of the present application can be put into practice without one or more of the specific details, or other methods, components, devices, steps, etc. can be adopted. In other cases, known methods, devices, realizations or operations are not shown or described in detail to avoid blurring the various aspects of the application.

[0027] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities may be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0028] The flowcharts shown in the accompanying drawings are only exemplary and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps can be decomposed, and some operations / steps can be combined or partially combined, so the actual execution order may change according to actual conditions.

[0029] It should be understood that although the terms first, second, third, etc. may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another component. Therefore, the first component discussed below can be referred to as the second component without departing from the teachings of the concepts of the present application. As used herein, the term "and / or" includes any one of the associated listed items and all combinations of one or more.

[0030] Figure 1 A structural functional diagram of the parsing system provided for the embodiment of the present application. The parsing system can be divided into an information extraction module, an entity linking module and a reasoning completion module from the perspective of detailed function realization. Among them, the information extraction module can identify the navigation event chain, and for the navigation event chain, the action events and skill events therein can be extracted, and then the spatial reference is extracted from the action events and skill events. The entity linking module can link the navigation subject and the spatial reference entity, but the navigation instruction text input by the user will omit some descriptions and need to be completed. At this time, the reasoning completion module can perform knowledge reasoning completion to obtain the final navigation instruction knowledge structure, which is the instruction content that can be understood and executed by the navigation subject. The overall execution process of the navigation event chain determination module, the composite instruction determination module, and the composite instruction segmentation module in the parsing system of the navigation instruction is the same as that of the information extraction module, and the overall execution process of the entity linking module and the reasoning completion module is the same as that of the information completion module.

[0031] The method of this application can be applied to a variety of scenarios involving the analysis of complex navigation instructions with multiple elements. Taking the scenario of agricultural and forestry plant protection as an example, users can control low-altitude unmanned aerial vehicles to perform pest control operations based on the method of this application. The user issues a navigation instruction text: "Green Guardian 1, fly to the farmland area 500 meters east of Heping Village, and spray pesticides at an altitude of 3 meters." This instruction is transmitted through Figure 1 The information extraction module, entity linking module and reasoning completion module are sequentially parsed to obtain structured UAV navigation command information.

[0032] The user conveys the command to the command parsing server through voice, which is processed by ASR to form a text command: "Green Guardian 1, fly to the farmland area 500 meters east of Heping Village and spray pesticides at an altitude of 3 meters."

[0033] The navigation instruction text is automatically annotated by the navigation event chain recognition agent and is divided into three parts:

[0034] "Green Guardian No. 1" (corresponding to Actor, which is an atomic instruction);

[0035] "Fly to the farmland area 500 meters east of Heping Village" (corresponding to MotionEvent, a compound instruction);

[0036] "Perform pesticide spraying at a height of 3 meters" (corresponding to SkillEvent, which is a compound instruction).

[0037] The navigation instruction text is automatically annotated and the navigation action elements are extracted through the action event extraction agent:

[0038] "Fly to" (corresponds to GoTo);

[0039] "The farmland area 500 meters east of Heping Village" (compound instruction).

[0040] The navigation instruction text is automatically annotated and the navigation skill elements are extracted through the skill event extraction agent:

[0041] “Spraying Operation” (corresponding to Spraying Operation);

[0042] "Pesticide" (corresponding to Instrument);

[0043] "Height 3 meters" (corresponding to ActivitySpace, a compound instruction).

[0044] The spatial reference extraction agent automatically labels the navigation activity spatial instructions generated during the parsing process and extracts the navigation geographic elements therein.

[0045] For the navigation activity space instruction "the farmland area 500 meters east of Heping Village", the analysis results are as follows:

[0046] "Peace Village" (corresponding to RefObject, an atomic instruction);

[0047] "East side" (corresponding to East);

[0048] "500 meters" (corresponding to Distance);

[0049] "Farmland" (corresponding to TargetObject, an atomic instruction).

[0050] For the navigation activity space instruction "height 3 meters", the analysis results are as follows:

[0051] "3 meters" (corresponding to Height).

[0052] The entity linking module performs entity linking processing on the active subjects and ground entities appearing in the navigation instructions, and links the relevant text to the correct entity objects in the knowledge base (database).

[0053] The navigation entity linking agent performs entity linking processing on the activity subject instructions generated during the parsing process:

[0054] Considering "Green Guardian 1" as the query text, the candidate entity set (such as "Green Shield 1", "Green Guardian 1 UAV", "Linhai Patrol 3", etc.) is retrieved from the device knowledge base through text matching, and then the semantic similarity score between the query text "Green Guardian 1" and each candidate entity is calculated through the semantic matching model, and the "Green Guardian 1 UAV" with the highest score is set as the correct entity linked to "Green Guardian 1".

[0055] The spatial reference entity linking agent performs entity linking processing on the reference object instructions and target object instructions generated during the parsing process:

[0056] Taking a certain entity link as an example, "Village A" is regarded as the query text, and the candidate entity set (such as "District A", "Town B", "Street C", etc.) is retrieved from the place name knowledge base through text matching. Then the semantic similarity score between the query text "Village A" and each candidate entity is calculated through the semantic matching model, and the "Village A, County A, Province A" with the highest score is set as the correct entity linked to "Village A".

[0057] For specific implementation methods, please refer to the following embodiments.

[0058] Figure 2 This is a flow chart of a method for parsing navigation instructions provided in an embodiment of the present application. The method of this embodiment can be applied to an instruction parsing server. Figure 2As shown, the method includes: step S20, step S21, step S22 and step S23.

[0059] In step S20, a navigation event chain is determined according to the received navigation instruction text, wherein the navigation event chain is used to represent a series of navigation-related activities performed by the navigation subject.

[0060] In this application, the navigation instruction text may be instruction information input by the user to guide the navigation subject to work. The text may be text converted from speech collected by a collection device such as a microphone through automatic speech recognition, or text input by a device such as a touch screen or a keyboard. The navigation subject may be used to represent an unmanned device that executes instructions to complete a specific task activity at a specified destination corresponding to the instruction.

[0061] According to an example embodiment, a navigation instruction text sent by a user in the above manner may be received first. Feature extraction may be performed on the navigation instruction text to obtain a number of navigation events, and the navigation events may be associated according to execution time, space, and logical relationships to obtain a navigation event chain. In some implementations, a generation model of a navigation event chain may be pre-set, and the navigation instruction text may be input to output a navigation event chain.

[0062] In other implementations, the navigation instruction text may be "Green Guardian 1, fly to the farmland area 500 meters east of Heping Village and spray pesticides at an altitude of 3 meters." The navigation event chain is: the navigation subject used is Green Guardian 1, event 1 is: fly to the farmland area 500 meters east of Heping Village, and event 2 is: spray pesticides at an altitude of 3 meters.

[0063] In step S21, the navigation event chain is text segmented according to the navigation event chain and the preset navigation element formalization information to determine the navigation activity composite instruction.

[0064] In this application, the navigation element formalized information can be a pre-set arrangement, order, and other presentation forms for event navigation elements. Events can include action events and skill events, so the navigation activity composite instructions include action event composite instructions and skill event composite instructions. The above-mentioned event 1 and event 2 are both composite instructions.

[0065] In some implementations, the action event compound instruction can be further segmented according to the form of navigation elements, and divided into two categories: embodied mode and mobile mode, see Table 1.

[0066] Skill event instructions can be further segmented according to the form of navigation elements to obtain specific skill types and specific skill parameters. Among them, the skill type needs to be determined based on the actual skill library of the navigation subject itself, and the skill parameters also need to be determined based on the actual skill parameter set of the navigation subject itself.

[0067] In other implementations, the navigation activity composite instruction includes a spatial relationship indicator, and the navigation activity composite instruction can be further segmented according to the form of navigation elements to obtain the spatial relationship indicator in the geographic scope description, which includes three categories: direction relationship, distance relationship, and topological relationship, as shown in Table 2. The target feature object in the geographic scope description can also be obtained, and the text segment corresponding to the navigation element is the atomic instruction.

[0068] The technical implementation models used when extracting elements may include but are not limited to rule models (such as regular pattern matching), sequence labeling models (such as CRF, LSTM, LSTM-CRF, Bert, etc.), and large language generation models (such as Llama, GPT, Qwen, etc.).

[0069] Table 1:

[0070]

[0071] Table 2

[0072]

[0073]

[0074] According to the example embodiment, the navigation instruction text can be matched based on the navigation element formalization information to determine the positions of various elements of different events in the text, and then the text can be segmented to match the navigation activity composite instruction from the segmented segments.

[0075] In step S22, the navigation activity composite instruction is segmented according to the preset reference formal information to determine preliminary structured instruction information.

[0076] In the present application, the navigation activity compound instruction is unstructured instruction information. In order to enable the navigation subject to accurately execute the activity corresponding to the instruction, the unstructured instruction information needs to be adjusted to structured instruction information.

[0077] The reference formal information may be a pre-set reference position used to locate the navigation subject when performing navigation activities. In some implementations, the reference formal information may include spatial reference formal information, which includes multiple geospatial requirements such as altitude and geographic range related to the navigation activities. The spatial reference formal information is:

[0078] <Height,RefObject,SpatialIndicators,TargetObject> , meaning:

[0079] Height: altitude information of navigation activities, such as "100 meters";

[0080] RefObject: the reference feature object in the description of geographic scope, such as "Building 1", "Garden A", etc.;

[0081] SpatialIndicators: Spatial relationship indicators in geographic range descriptions, such as "left", "behind", "outside", etc.;

[0082] TargetObject: The target geographical object in the geographic scope description, such as "forest", "street", etc.

[0083] According to the exemplary embodiment, the corresponding related text words may be matched in the navigation activity composite instruction according to the above-mentioned reference formalized information, and the text segmentation may be performed to obtain preliminary structured instruction information.

[0084] In step S23, the preliminary structured instruction information is completed according to the atomic instructions in the navigation event chain and the preset candidate entity texts to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is sent to the navigation subject.

[0085] In the present application, the text segment corresponding to the navigation body is an atomic instruction, which is used to indicate an instruction that cannot be further decomposed to obtain other useful elements. The candidate entity text may be a text description corresponding to a plurality of preset candidate entities.

[0086] According to the example embodiment, candidate entity texts matching the atomic instruction may be selected from the candidate entity texts, and then the preliminary structured instruction information may be completed based on the matching texts to obtain a navigation instruction knowledge structure.

[0087] By receiving and parsing the navigation instruction text, the present application can construct a navigation event chain, which clearly reflects a series of navigation-related activities that the navigation subject needs to perform. The present application not only considers the overall structure of the navigation event chain, but also further performs text segmentation on the navigation event chain according to the preset navigation element formalization information, thereby identifying the navigation activity composite instruction. This refined processing enables the system to more accurately understand the spatial and temporal elements in the navigation instruction, as well as the relationship between them. By further performing text segmentation on the navigation activity composite instruction using the preset reference formalization information, the navigation instruction can be converted into preliminary structured instruction information. This structured information is not only convenient for system storage and processing, but also helps to improve the execution efficiency and accuracy of the navigation instruction. Combining the atomic instructions in the navigation event chain and the preset candidate entity text, the preliminary structured instruction information can be completed, thereby constructing a complete navigation instruction knowledge structure. This completion mechanism ensures the integrity and accuracy of the navigation instruction and reduces execution errors caused by missing information.

[0088] According to some embodiments, a navigation instruction text sent by a user may be received; the navigation instruction text may be segmented according to preset event chain formalization information to determine a plurality of text segments; and a navigation event chain may be generated according to the plurality of text segments.

[0089] In this application, the event chain formalization information can be:

[0090] <Actor,NaiEvent 1 ,NaiEvent 2 …>, meaning:

[0091] Actor: Navigation subject, such as domestic robots, unmanned delivery vehicles, search and rescue aircraft, etc.

[0092] NaiEvent 1 : Navigation activity event 1;

[0093] NaiEvent 2 : Navigation activity event 2.

[0094] In some implementations, the navigation instruction text sent by the user is first received, and then the event chain formalization information is divided accordingly to obtain multiple text segments, which are then spliced ​​using a preset text segment splicing method to obtain a navigation event chain. In other implementations, a database may be pre-set, and the database may store text segments corresponding to different formalization information for matching.

[0095] This application uses preset event chain formalization information to quickly segment the navigation instruction text, generate multiple text segments, and then construct a navigation event chain. This processing method avoids the one-by-one parsing and reasoning of navigation instructions in traditional methods, greatly shortens the instruction processing time, and improves the response speed of the navigation system. By constructing a navigation event chain, the user's navigation needs can be converted into a series of orderly navigation activities, which not only include movement in physical space, but may also involve time planning, path optimization and other aspects.

[0096] According to some embodiments, action event compound instructions and skill event compound instructions can be extracted from the navigation event chain; the action event compound instructions and the skill event compound instructions are respectively sequence-labeled according to the navigation element formalized information; the navigation event chain is text-segmented according to the annotations to determine multiple text segments; the element granularity of the multiple text segments is detected to determine the navigation activity compound instructions.

[0097] In this application, the action event compound instruction may include multiple navigation elements related to the action, such as forward, backward, etc. The skill event compound instruction may include multiple navigation elements related to the skill task, such as grabbing, extinguishing fire, etc.

[0098] In some implementations, the navigation element formalization information of the action event compound instruction may be:

[0099] <MotionTrigger,ActivitySpace> , meaning:

[0100] MotionTrigger: Navigation action trigger words, such as "forward", "backward", "stay", etc.;

[0101] ActivitySpace: Navigation activity space, used to describe the activity space reference background of the navigation activity subject when performing navigation tasks.

[0102] The navigation element formalization information of the skill event compound instruction can be:

[0103] <SkillTrigger,ActivitySpace,SkillArgs> , meaning:

[0104] SkillTrigger: The skill trigger word of the navigation subject, such as "grab", "extinguish fire", "searchlight", etc.

[0105] ActivitySpace: Navigation activity space, used to describe the activity space reference background of the navigation activity subject when performing navigation skill tasks;

[0106] SkillArgs: The specific execution parameters of the navigation skill. Different skills have different execution parameters.

[0107] The element granularity can be used to indicate whether the element components of a text segment can continue to be effectively segmented. For example, if SkillTrigger is captured, it makes no sense to split it further.

[0108] In some implementations, the action event compound instruction and the skill event compound instruction are first extracted from the navigation event chain. According to the above-mentioned navigation element formalization information, the action event compound instruction and the skill event compound instruction are respectively subjected to text analysis, the corresponding navigation elements are annotated, and then the text is segmented according to the annotations to obtain multiple text fragments. Then, the element granularity corresponding to each of these text fragments can be detected, and the navigation activity compound instruction can be determined according to the association relationship between the preset element granularity and the navigation activity compound instruction.

[0109] This application can more accurately understand the user's navigation needs by clearly distinguishing between action event compound instructions and skill event compound instructions in the navigation event chain. The application of sequence annotation further ensures that each instruction element is accurately identified and labeled, reducing the possibility of misunderstanding and ambiguity. This application can handle navigation event chains containing multiple navigation elements and complex logical structures. Through fine-grained element detection and text segmentation, it can flexibly adapt to the personalized navigation needs of different users. The introduction of skill event compound instructions can respond to users' non-traditional navigation needs, thereby improving the overall flexibility of instruction parsing.

[0110] According to some embodiments, the element granularity of multiple text fragments can be detected; when the element granularity is multiple, the corresponding text fragment is determined as a navigation activity composite instruction; when the element granularity is single, the corresponding text fragment is determined as an atomic instruction.

[0111] In some implementations, the element fine granularity of the text segment can be detected by a preset fine granularity segmentation method. The fine granularity segmentation method can be used to indicate which form of text segment does not need to be segmented, which form of text segment needs to be segmented, and what the segmentation method is.

[0112] When the element granularity is multiple, the corresponding text fragment is a navigation activity composite instruction; when the element granularity is single, the corresponding text fragment is an atomic instruction.

[0113] By detecting the fine-grained elements of text fragments, the present application can more accurately identify key information in navigation instructions and avoid navigation errors caused by element confusion or misunderstanding. For text fragments containing multiple fine-grained elements, they can be correctly identified as navigation activity compound instructions, thereby ensuring the integrity and coherence of the instructions. By distinguishing between navigation activity compound instructions and atomic instructions, corresponding processing strategies can be adopted according to different types of instructions. For atomic instructions, they can be executed quickly; for compound instructions, more complex logical processing can be performed, thereby improving the flexibility and adaptability of instruction processing.

[0114] According to some embodiments, text extraction is performed on the navigation activity compound instructions to determine the element type; based on the reference formal information, the reference extraction element corresponding to the element type is determined; according to the reference extraction element, the navigation activity compound instructions are sequence labeled; according to the annotations, the navigation activity compound instructions are text segmented to determine the text segments corresponding to the reference extraction elements, and preliminary structured instruction information is determined based on the text segments.

[0115] In the present application, referring to step S22, the reference formalized information may include spatial reference formalized information and non-spatial reference formalized information, and correspondingly, the navigation activity composite instruction may include spatial elements or non-spatial elements. The text corresponding to different element types may be pre-set, and the navigation activity composite instruction may be subjected to text extraction, and then text matching may be performed to determine which element type the navigation activity composite instruction contains. If it is a spatial element type, the text segmentation may be performed according to the spatial reference formalized information, and if it is another element type, the corresponding pre-set formalized information may be obtained for text segmentation.

[0116] In some implementations, the spatial reference formalization information is:

[0117] <Height,RefObject,SpatialIndicators,TargetObject>

[0118] Height, RefObject, SpatialIndicators, etc. can all be used as reference extraction elements under the spatial element type. According to the reference extraction elements, the corresponding text fragments are matched in the navigation activity composite instruction, sequence annotation is performed, and then the text is segmented according to the annotation to obtain multiple text fragments, and then the preliminary structured instruction information is obtained.

[0119] This application can accurately identify and extract key reference elements in complex instructions for navigation activities through preset reference formal information. The process of sequence labeling and text segmentation further ensures that each element in the instruction is accurately identified and located, thereby improving the accuracy of instruction parsing. Through the automated sequence labeling and text segmentation process, the time and cost of manual instruction parsing are significantly reduced. The generation of preliminary structured instruction information enables subsequent processing steps to more efficiently utilize instruction information, thereby improving overall processing efficiency.

[0120] According to some embodiments, the navigation activity composite instruction may be segmented according to annotations to determine text segments corresponding to reference extraction elements; and the text segments may be integrated based on preset structured integration information to determine preliminary structured instruction information.

[0121] In the present application, the structured integration information may be pre-set information for structured text integration, and may be a structured template. The segmented text segments may be integrated according to the structured integration information, for example, the text segments may be imported into corresponding positions of the structured template.

[0122] Based on the preset structured integration information, this application can integrate the segmented text fragments into preliminary structured instruction information, which is convenient for subsequent processing and application. The generation of structured information makes the instruction information clearer and more orderly, which is convenient for the navigation subject to perform corresponding tasks.

[0123] According to some embodiments, atomic instructions can be extracted from preliminary structured instruction information and navigation event chains; the atomic instructions are text-matched with candidate entity texts to determine a candidate entity set corresponding to the candidate entity texts; semantic similarity is determined based on the candidate entity set and the atomic instructions to determine a target entity object corresponding to the atomic instructions based on the semantic similarity; the target entity object is associated with a single navigation element corresponding to the atomic instruction to convert the preliminary structured instruction information into target structured instruction information; and information completion is performed on the target structured instruction information based on preset navigation environment context data and preset historical navigation information to determine a navigation instruction knowledge structure.

[0124] In some implementations, atomic instructions can be extracted from the preliminary structured instruction information navigation event chain. Text matching is performed between the atomic instructions and the candidate entity texts, and the candidate entities corresponding to the candidate entity texts that completely contain the atomic instructions can be added to the candidate entity set. The arrangement of different words in the atomic instructions that completely contain the atomic instructions may be different, so the candidate entity set needs to be further screened.

[0125] The semantic similarity between the candidate entity set and the atomic instruction can be calculated through natural language processing techniques, such as word vector similarity calculation, semantic role labeling, etc., and then the candidate entity with the highest semantic similarity is determined as the target entity object.

[0126] According to the semantic similarity, the candidate entity that best matches the target atomic instruction is selected as the target entity object. The atomic instruction cannot be further split, so for a single navigation element, the target entity object can be linked to the corresponding single navigation element to achieve association and obtain the target structured instruction information.

[0127] In some implementations, atomic instructions include multiple types, some of which correspond to single navigation elements that need to be associated with target entity objects. These types of atomic instructions include atomic instructions marked as Actor, in which the navigation elements are active entities, i.e., navigation entities; atomic instructions marked as RefObject and atomic instructions marked as TargetObject, in which the navigation elements are the ground objects in the spatial reference of the navigation activity. In the actual linking and association process, the atomic instructions corresponding to the navigation subject entity are linked to the correct device entity object in the navigation subject knowledge base, and the atomic instructions corresponding to the spatial reference entity are linked to the correct ground object entity object in the place name library.

[0128] When performing entity linking, a two-stage "retrieval-disambiguation" process is technically adopted. First, in the retrieval stage, the atomic instruction text expressing the activity subject or the ground object entity is retrieved from the corresponding knowledge base through text matching, dictionary query and other technologies to retrieve the set of candidate entities that may actually point to. Secondly, in the disambiguation stage, the disambiguation of candidate entities is regarded as a ranking problem, and the semantic similarity between the instruction text and the candidate entity is calculated through a semantic matching model (including but not limited to SentenceBert, Word2Vec, DSSM, etc.), and the one with the highest score is the final correct target entity object.

[0129] The navigation environment context data in this application can be used to indicate the state of the current navigation subject, such as where it has moved to and what kind of event has been completed, so as to facilitate integration when subsequent events are executed. If the current navigation environment context data shows that event N is being executed, then the historical navigation information is used to indicate the relevant information of event N-1, such as the distance the navigation subject has moved when executing the event, the height it has risen, etc. In some implementations, the scene of event N-1 can be assigned to event N. For example, the instruction text is "Green Guardian No. 1, fly to the farmland area 500 meters away from the east side of Heping Village, and spray pesticides at a height of 3 meters." After executing "fly to the farmland area 500 meters away from the east side of Heping Village", the height of 3 meters is unclear. At this time, it can be completed, that is, the farmland area is completed to event 2, and adjusted to "spray pesticides at a height of 3 meters in the farmland area", and finally the navigation instruction knowledge structure is obtained.

[0130] In some implementations, knowledge completion reasoning rules may be pre-set based on navigation environment context data and historical navigation information, and target structured instruction information may be completed based on the pre-set knowledge completion reasoning rules. For example:

[0131] Rule 1: Activity subject completion rules

[0132] If missing: (navigation instruction 1, activity subject 1) ∧ navigation environment context (activity subject 2), then assign (navigation instruction 1, activity subject 2).

[0133] Rule 2: Navigation activity space height information completion rule A

[0134] Missing: (current navigation event E1, altitude information 1) ∧ first (navigation event chain, current navigation event E1) ∧ navigation environment context (altitude information 2), then assign (current navigation event E1, altitude information 2).

[0135] Rule 3: Navigation activity space height information completion rule B

[0136] If missing: (current navigation event E1, altitude information 1) ∧ previous navigation event (altitude information 2), then assign (current navigation event E1, altitude information 2).

[0137] Rule 4: Navigation activity space geographic scope information completion rule A

[0138] Missing: (current navigation event E1, geographic scope 1) ∧ first (navigation event chain, current navigation event E1) ∧ navigation environment context (geographic scope 2), then assign (current navigation event E1, geographic scope 2).

[0139] Rule 5: Navigation activity space reference range information completion rule B

[0140] If missing: (current navigation event E1, geographic scope 1) ∧ previous navigation event (geographic scope 2), then assign (current navigation event E1, geographic scope 2).

[0141] This application can accurately identify key entity objects in navigation instructions, such as places, facilities, etc., by extracting atomic instructions from preliminary structured instruction information and navigation event chains and performing text matching with candidate entity texts. Determining the target entity object based on semantic similarity ensures the precise association between atomic instructions and specific entity objects, thereby improving the accuracy of navigation instructions. Information completion is performed by combining navigation environment context data and historical navigation information, making navigation instructions more complete and including various factors that need to be considered in the actual navigation process.

[0142] The following describes an apparatus embodiment of the present application, which can be used to execute the method embodiment of the present application. For details not disclosed in the apparatus embodiment of the present application, reference can be made to the method embodiment of the present application.

[0143] Figure 3 This is a block diagram of a navigation instruction parsing system provided in an embodiment of the present application. Figure 3 As shown, the navigation instruction parsing system 300 includes a navigation event chain determination module 301 , a compound instruction determination module 302 , a compound instruction segmentation module 303 and an information completion module 304 .

[0144] A navigation event chain determining module 301 is used to determine a navigation event chain according to the received navigation instruction text, wherein the navigation event chain is used to represent a series of navigation-related activities performed by the navigation subject;

[0145] The composite instruction determination module 302 is used to perform text segmentation on the navigation event chain according to the navigation event chain and the preset navigation element formalization information to determine the navigation activity composite instruction;

[0146] The compound instruction segmentation module 303 is used to segment the navigation activity compound instruction text according to the preset reference formal information to determine the preliminary structured instruction information;

[0147] The information completion module 304 is used to complete the preliminary structured instruction information according to the atomic instructions in the navigation event chain and the preset candidate entity texts, so as to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is sent to the navigation body.

[0148] Optionally, the navigation event chain determining module 301 is specifically configured to:

[0149] Receive navigation instruction text sent by the user;

[0150] According to the preset event chain formalization information, the navigation instruction text is segmented to determine a plurality of text segments;

[0151] Generate a navigation event chain based on multiple text fragments.

[0152] Optionally, the compound instruction determination module 302 is specifically configured to:

[0153] Extracting action event compound instructions and skill event compound instructions from the navigation event chain;

[0154] According to the formalized information of navigation elements, the action event compound instructions and skill event compound instructions are sequence labeled respectively;

[0155] Segment the navigation event chain textually according to the annotations to determine multiple text segments;

[0156] Detecting multiple text fragments at a fine-grained level to identify navigation activity composite instructions.

[0157] Optionally, when the compound instruction determination module 302 detects the fine granularity of elements of multiple text segments to determine the composite instruction of the navigation activity, it is specifically used to:

[0158] Detecting fine-grained features of multiple text segments;

[0159] In the case where the element granularity is multiple, the corresponding text fragment is determined as a navigation activity composite instruction;

[0160] When the element granularity is single, the corresponding text segment is determined as an atomic instruction.

[0161] Optionally, the compound instruction segmentation module 303 is specifically used for:

[0162] Extract the text of the navigation activity composite instruction and determine the element type;

[0163] Determine the reference extraction element corresponding to the element type according to the reference formalized information;

[0164] According to the reference extraction elements, the navigation activity composite instructions are marked in sequence;

[0165] The navigation activity composite instructions are segmented according to the annotations to determine the text segments corresponding to the reference extraction elements, and preliminary structured instruction information is determined based on the text segments.

[0166] Optionally, the composite instruction segmentation module 303 is specifically used to segment the composite instruction of the navigation activity according to the annotation to determine the text segment corresponding to the reference extraction element, and determine the preliminary structured instruction information according to the text segment:

[0167] Performing text segmentation on the navigation activity composite instruction according to the annotations to determine the text segments corresponding to the reference extraction elements;

[0168] Based on preset structured integration information, the text fragments are integrated to determine preliminary structured instruction information.

[0169] Optionally, the information completion module 304 is specifically used for:

[0170] Extracting atomic instructions from preliminary structured instruction information and navigation event chains;

[0171] Perform text matching between the atomic instruction and the candidate entity text to determine a candidate entity set corresponding to the candidate entity text;

[0172] Determine semantic similarity based on the candidate entity set and the atomic instruction, so as to determine the target entity object corresponding to the atomic instruction according to the semantic similarity;

[0173] Associating the target entity object with a single navigation element corresponding to the atomic instruction to transform the preliminary structured instruction information into the target structured instruction information;

[0174] According to the preset navigation environment context data and the preset historical navigation information, the target structured instruction information is completed to determine the navigation instruction knowledge structure.

[0175] The device performs functions similar to the method provided above. For other functions, please refer to the previous description and will not be repeated here.

[0176] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 4 As shown, the electronic device 400 of this embodiment may include: a memory 401 and a processor 402 .

[0177] The memory 401 stores a computer program. When the computer program is executed by the processor 402, the processor 402 executes the method in the above embodiment.

[0178] The processor 402 and the memory 401 are connected, for example, via a bus.

[0179] Optionally, the electronic device 400 may further include a transceiver. It should be noted that in actual applications, the number of transceivers is not limited to one, and the structure of the electronic device 400 does not constitute a limitation on the embodiments of the present application.

[0180] Processor 402 may be a CPU (Central Processing Unit), a general purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 402 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0181] The bus may include a path to transmit information between the above components. The bus may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0182] The memory 401 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0183] The memory 401 is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 402. The processor 402 is used to execute the application code stored in the memory 401 to implement the contents shown in the above method embodiment.

[0184] The electronic devices include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and fixed terminals such as digital TVs, desktop computers, etc. It can also be a server, etc. Figure 4 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0185] The electronic device of this embodiment can be used to execute the method of any of the above embodiments, and its implementation principle and technical effect are similar, which will not be described in detail here.

[0186] The present application also provides a non-transitory computer-readable storage medium having computer-readable instructions stored thereon. When the aforementioned instructions are executed by a processor, the processor executes the method in the above embodiment.

[0187] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a non-transient computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0188] The embodiments of the present application are described in detail above. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and its core idea of ​​the present application. At the same time, changes or deformations made by those skilled in the art based on the ideas of the present application, the specific implementation methods and the scope of application of the present application, all belong to the scope of protection of the present application. In summary, the content of this specification should not be construed as a limitation on the present application.

Claims

1. A method for parsing navigation instructions, characterized in that: include: Determining a navigation event chain according to the received navigation instruction text, wherein the navigation event chain is used to represent a series of navigation-related activities performed by the navigation subject; According to the navigation event chain and the preset navigation element formalized information, the navigation event chain is segmented to determine the navigation activity composite instruction; According to preset reference formal information, text segmentation is performed on the navigation activity composite instruction to determine preliminary structured instruction information; The preliminary structured instruction information is completed according to the atomic instructions in the navigation event chain and the preset candidate entity texts to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is delivered to the navigation subject.

2. The method according to claim 1, characterized in that Determining the navigation event chain according to the received navigation instruction text includes: Receiving the navigation instruction text sent by the user; According to the preset event chain formalization information, the navigation instruction text is segmented to determine a plurality of text segments; The navigation event chain is generated according to the multiple text segments.

3. The method according to claim 1, characterized in that The text segmentation of the navigation event chain according to the navigation event chain and the preset navigation element formalization information to determine the navigation activity composite instruction includes: Extracting action event compound instructions and skill event compound instructions from the navigation event chain; According to the navigation element formalized information, the action event compound instruction and the skill event compound instruction are respectively sequence-labeled; Performing text segmentation on the navigation event chain according to the annotations to determine a plurality of text segments; The fine granularity of elements of the plurality of text segments is detected to determine the navigation activity composite instruction.

4. The method according to claim 3, characterized in that The detecting the fine granularity of elements of the plurality of text segments to determine the navigation activity composite instruction comprises: Detecting element granularity of the plurality of text segments; In the case where the element fine-grainedness is multiple, determining the corresponding text fragment as the navigation activity composite instruction; When the element granularity is single, the corresponding text segment is determined as the atomic instruction.

5. The method according to claim 1, characterized in that The step of performing text segmentation on the navigation activity composite instruction according to the preset reference formalized information to determine preliminary structured instruction information includes: Extract text from the navigation activity composite instruction to determine the element type; Determining, according to the reference formalized information, a reference extraction element corresponding to the element type; According to the reference extraction elements, sequence marking is performed on the navigation activity composite instruction; The navigation activity composite instruction is segmented according to the annotations to determine the text segments corresponding to the reference extraction elements, and the preliminary structured instruction information is determined according to the text segments.

6. The method according to claim 5, characterized in that The text segmentation of the navigation activity composite instruction according to the annotation to determine the text segment corresponding to the reference extraction element, and determining the preliminary structured instruction information according to the text segment, includes: Performing text segmentation on the navigation activity composite instruction according to the annotations to determine the text segment corresponding to the reference extraction element; Based on preset structured integration information, the text segments are integrated to determine the preliminary structured instruction information.

7. The method according to claim 1, characterized in that The method of completing the preliminary structured instruction information according to the atomic instructions in the navigation event chain and the preset candidate entity text to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is delivered to the navigation subject includes: extracting the atomic instruction from the preliminary structured instruction information and the navigation event chain; Performing text matching between the atomic instruction and the candidate entity text to determine a candidate entity set corresponding to the candidate entity text; Determine semantic similarity based on the candidate entity set and the atomic instruction, so as to determine a target entity object corresponding to the atomic instruction according to the semantic similarity; Associating the target entity object with a single navigation element corresponding to the atomic instruction to convert the preliminary structured instruction information into target structured instruction information; The target structured instruction information is completed according to preset navigation environment context data and preset historical navigation information to determine the navigation instruction knowledge structure.

8. A navigation instruction parsing system, characterized in that: include: A navigation event chain determination module, used to determine a navigation event chain according to the received navigation instruction text, wherein the navigation event chain is used to represent a series of navigation-related activities performed by the navigation subject; A composite instruction determination module is used to perform text segmentation on the navigation event chain according to the navigation event chain and preset navigation element formalization information to determine the navigation activity composite instruction; A compound instruction segmentation module, used for performing text segmentation on the compound instruction of the navigation activity according to preset reference formal information to determine preliminary structured instruction information; The information completion module is used to complete the preliminary structured instruction information according to the atomic instructions in the navigation event chain and the preset candidate entity text, so as to determine the corresponding navigation instruction knowledge structure when the navigation instruction text is sent to the navigation body.

9. An electronic device, characterized in that: include: processor; A memory storing a computer program, which, when executed by the processor, enables the processor to execute the navigation instruction parsing method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that: Computer-readable instructions are stored thereon, and when the instructions are executed by a processor, the processor executes the navigation instruction parsing method as described in any one of claims 1-7.

Images