User input analysis method and device based on device design intelligent agent

Abstract

The embodiment of the invention provides a user input analysis method and device based on a device design agent, and relates to the technical field of device design. The method comprises the following steps: in response to a natural language instruction input by a user, identifying a device design operation intention represented by the natural language instruction; based on the identified operation intention, determining at least one to-be-analyzed parameter item associated with completion of the operation intention, and forming a parameter extraction framework; analyzing semantics of the natural language instruction to extract or deduce values corresponding to various parameters in the parameter extraction framework; and organizing the parameter items and the corresponding values into structured data conforming to a target device design software interface specification. The implementation mode provided by the invention not only can adapt to diversified expression modes, but also can dynamically adjust the analysis logic in combination with the context, so that the accuracy and adaptability of interaction between the intelligent agent and the user are remarkably improved.

Description

Technical Field

[0001] This application relates to the field of device design technology, specifically to a user input parsing method based on device design intelligent agents, a user input parsing device based on device design intelligent agents, an electronic device, a storage medium, and a computer program product. Background Technology

[0002] In the actual development of intelligent agents for device design, user interaction is usually conducted in the form of natural language dialogue. While this interaction method lowers the barrier to entry, it also brings problems such as ambiguity in input information and non-standard parameter expressions. Since semiconductor device design involves various professional technical parameters such as process parameters, material properties, and performance indicators, users may use diverse expressions to describe the same concept. This requires the intelligent agent to possess accurate semantic parsing and parameter standardization capabilities. Specifically, user input parsing needs to complete three key tasks: first, accurately identify the design goals and constraints in the user's requirements; second, transform unstructured natural language expressions into structured design parameters; and third, handle the correlation and conflict issues between parameters.

[0003] In traditional methods, parsing user input typically relies on rule-based approaches, such as regular expressions (Regex) or keyword matching. These methods extract key parameters from the user's natural language input through predefined patterns or fixed templates. These methods have significant limitations: First, regular expressions and keyword matching have weak generalization capabilities, making it difficult to adapt to the diverse expressions of natural language; second, traditional methods lack contextual understanding and cannot handle parameters with implicit semantics or those dependent on dialogue history. More importantly, semiconductor device design involves numerous technical terms and coupling relationships between parameters (such as the correlation between doping concentration and threshold voltage). Traditional methods require manually maintaining a large rule base, which is not only costly to develop but also difficult to cover all possible user expressions. Ultimately, this results in poor agent interaction flexibility and low fault tolerance, severely limiting their application in complex design scenarios.

[0004] Furthermore, in advanced design scenarios involving multiple rounds of interaction and complex constraints between parameters, the shortcomings of traditional methods become particularly prominent. For example, when a user issues a complex instruction such as "reduce the gate oxide thickness by 20% while ensuring the breakdown voltage is not lower than X volts," rule-based parsers not only struggle to accurately separate the "optimization objective" (thinning the gate oxide) and the "constraint" (maintaining the breakdown voltage), but also fail to understand the inherent physical coupling between these two parameters. This results in potentially contradictory sets of design parameters that cannot be directly used to drive simulation or design tools. Summary of the Invention

[0005] The purpose of this application is to provide a user input parsing method and apparatus based on device design intelligent agents. By combining the contextual understanding advantages of large-scale language models, prompt word templates and parameter extraction rules are constructed to achieve intelligent conversion from user free expression to standardized design parameters, so as to at least solve some of the problems in the background art.

[0006] To achieve the above objectives, this application provides a user input parsing method based on a device design intelligent agent, comprising: responding to a natural language instruction input by a user, identifying the device design operation intention represented by the natural language instruction; based on the identified operation intention, determining at least one parameter item to be parsed associated with completing the operation intention, forming a parameter extraction framework; parsing the semantics of the natural language instruction to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework; and organizing the parameter items and their corresponding values ​​into structured data conforming to the target device design software interface specification.

[0007] Optionally, in response to a natural language instruction input by a user, identifying the device design operation intention represented by the natural language instruction includes: using a large language model to identify the natural language instruction and obtaining an identification result; matching the identification result with a preset set of operation intentions to obtain the confidence level of the identification result for each operation intention category in the preset set of operation intentions, forming a confidence level set; and obtaining the operation intention based on the maximum confidence level in the confidence level set.

[0008] Optionally, the preset set of operation intentions is set according to commonly used instruction types in the semiconductor design field.

[0009] Optionally, the definitions of the operation intent category and the parameter items in the parameter extraction framework are predefined based on knowledge in the field of semiconductor device design.

[0010] Optionally, based on the identified operation intent, at least one parameter item to be parsed is determined to be associated with completing the operation intent, forming a parameter extraction framework, including: determining the corresponding parameter extraction framework from a framework template set according to the operation intent, wherein the parameter extraction frameworks in the framework template set have a corresponding relationship with the operation intent; and using each parameter in the determined parameter extraction framework as a set of parameters to be assigned values.

[0011] Optionally, parsing the semantics of the natural language instruction to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework includes: mapping the non-standard expressions of the natural language instruction regarding device parameters to standardized parameter names and units; parsing the semantic information in the natural language instruction and assigning the parameter values ​​in the semantic information to the corresponding parameters.

[0012] Optionally, the parameters in the parameter extraction framework include necessary parameters and optional parameters; the values ​​of the necessary parameters are determined according to the following order: extracted from natural language instructions, inferred from the semantic information of the natural language instructions, obtained from the dialogue history of the natural language instructions, outputting a separate prompt for the necessary parameters with missing values ​​and obtaining the corresponding input; the values ​​of the optional parameters are determined according to the following order: extracted from natural language instructions, and using default values.

[0013] Optionally, organizing the parameter items and their corresponding values ​​into structured data that conforms to the interface specification of the target device design software includes: determining the structure of the structured data according to the interface specification with the semiconductor design software; and mapping the parameter items and their corresponding values ​​to the structured data in the form of key-value pairs.

[0014] Optionally, the method further includes: performing data verification on the obtained structured data that conforms to the target device design software interface specification; the data verification includes: verifying whether the fields of the structured data are appropriate data types, verifying whether the fields of the structured data conform to numeric or character restrictions, and verifying whether the structured data conforms to custom constraints.

[0015] This application also provides a user input parsing device for a device design intelligent agent. The device includes: an intent recognition module configured to recognize the device design operation intent represented by the natural language instruction input by the user; a frame determination module configured to determine at least one parameter item to be parsed associated with completing the identified operation intent, forming a parameter extraction frame; a semantic parsing and assignment module configured to parse the semantics of the natural language instruction to extract or infer the values ​​corresponding to each parameter in the parameter extraction frame; and a result structuring module configured to organize the parameter items and their corresponding values ​​into structured data conforming to the target device design software interface specification.

[0016] This application also provides an electronic device, including: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the at least one processor implements the aforementioned user input parsing method based on device-designed intelligent agents by executing the instructions stored in the memory.

[0017] This application also provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the aforementioned user input parsing method for a device-based intelligent agent.

[0018] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the aforementioned user input parsing method based on device-designed intelligent agents.

[0019] The above technical solution has the following beneficial effects: This method introduces a large model and designs a hierarchical natural language instruction parsing mechanism to achieve intelligent conversion from user-defined expressions to standardized design parameters. Specifically, the progressive parsing mechanism of "intent-framework-assignment" in this application enables the system to focus on the professional parameters most relevant to the current task and flexibly handle implicit relationships and contextual dependencies between parameters, thereby achieving a reliable conversion from free expression to precise structured data. This method not only adapts to diverse expression methods but also dynamically adjusts the parsing logic based on context, significantly improving the accuracy and adaptability of the agent's interaction with the user.

[0020] Other features and advantages of the embodiments of this application will be described in detail in the following detailed description section. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the following detailed description to explain the embodiments of this application, but do not constitute a limitation on the embodiments of this application. In the drawings: Figure 1 This illustration schematically shows the steps of a user input parsing method based on device-designed intelligent agents according to an embodiment of this application; Figure 2 This illustration schematically shows a structural diagram of a user input parsing device based on device-designed intelligent agents according to an embodiment of this application; Figure 3 The diagram schematically illustrates the internal structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0022] The specific embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the embodiments of this application.

[0023] Figure 1 The illustration shows a schematic diagram of the steps in a user input parsing method based on device-designed intelligent agents according to an embodiment of this application. For example... Figure 1 As shown, a user input parsing method based on device-designed intelligent agents is described, the method comprising: S01. In response to a natural language command input by the user, identify the device design operation intention represented by the natural language command; S02. Based on the identified operational intent, determine at least one parameter item to be parsed that is associated with completing the operational intent, and form a parameter extraction framework; S03. Parse the semantics of the natural language instruction to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework; S04. Organize the parameter items and their corresponding values ​​into structured data that conforms to the target device design software interface specification.

[0024] The core concept of this application lies in decoupling the parsing process of user natural language commands into two stages: "intent-driven" and "dynamic adaptation." First, ambiguous commands are categorized into specific design operation categories through intent recognition, which limits the search space and direction of subsequent parsing. Then, instead of using a fixed parsing template, the system dynamically adapts the set of parameters (i.e., the parameter extraction framework) based on the recognized intent, and performs precise semantic extraction and reasoning under the guidance of this framework. This progressive parsing mechanism of "intent-framework-assignment" allows the system to focus on the most relevant professional parameters for the current task and flexibly handle implicit relationships and contextual dependencies between parameters, thereby achieving a reliable conversion from free expression to precise structured data.

[0025] In some implementations, both the operation intent classification layer for implementing intent-driven operations and the operation parameter extraction layer for implementing dynamic adaptation can be included in the device design agent.

[0026] In this implementation, large language models such as Deepseek / GPT-4 / Claude can be used as tools for recognizing and extracting operational intentions and parameters to achieve better results. In traditional methods, if a user inputs "gate oxide thickness is 5nm," the system uses regular expressions (such as \d+nm) to match the numerical value and unit, and then maps the keyword "gate oxide" to the corresponding design parameters (such as gate_oxide_thickness). Furthermore, some systems combine finite state machines (FSMs) or template-based semantic parsing to gradually fill the structured parameter table. However, these methods have significant limitations: First, regular expressions and keyword matching have weak generalization capabilities, making it difficult to adapt to the diverse expressions of natural language. For example, users may use different expressions such as "5nm oxide layer," "5nm gate oxide thickness," or "GOX=5nm," leading to parameter extraction failures or incorrect mappings. Second, traditional methods lack contextual understanding capabilities and cannot handle parameters with implicit semantics or those dependent on dialogue history (such as "10% reduction in thickness compared to the previous design"), only handling explicit, fixed-format inputs. Furthermore, when user input involves complex constraints (such as "high dielectric constant but low leakage current"), rule-based methods struggle to parse the logical relationships between parameters, easily leading to lost or conflicting design requirements. Using a large language model, on the other hand, avoids these problems. It not only possesses contextual understanding capabilities but can also rely on dialogue history for assisted understanding, accurately translating user instructions and improving design efficiency.

[0027] The following is a specific example of the complete workflow implementation of the above embodiments.

[0028] The operational parameter extraction layer essentially operates on the principle of "intent-driven parameter framework instantiation." Internally, the system maintains an "intent-parameter framework" mapping library. For example, for the intent "execute simulation," its corresponding parameter framework might be predefined to include fields such as [simulation node, simulation type, key process parameters, whether to use caching, etc.]. When the intent classification layer outputs "execute_simulation," the operational parameter extraction layer doesn't simply populate a fixed table. Instead, it first instantiates the parameter framework corresponding to that intent, forming a "set of operational parameters" specific to this parsing task and containing the aforementioned fields. This process ensures the system's targeted and scalable parsing—adding a new design intent only requires defining its corresponding parameter framework, without modifying the core parsing engine.

[0029] def parse_user_input(prompt: str) ->dict: "Two-stage parsing engine" # Phase 1: Intent Recognition intent_response = ollama.generate( prompt=INTENT_CLASS_PROMPT + f"\nUser input: {prompt}" ) # Convert the results returned by the large model to a normalized JSON format intent_data = json.loads(intent_response) # Second Stage: Parameter Extraction detail_response = ollama.generate( prompt = get_detail_prompt(intent_data["intent"]) + f"\nUser input: {prompt}" ) In some embodiments of this application, in response to a user-input natural language command, identifying the device design operation intent represented by the natural language command includes: using a large language model to identify the natural language command and obtain an identification result; matching the identification result with a preset set of operation intents to obtain the confidence level of the identification result for each operation intent category in the preset set of operation intents, forming a confidence level set; and obtaining the operation intent based on the maximum confidence level in the confidence level set. The method includes: first, extracting and understanding semantic features of the user-input command text using a pre-trained large language model to generate an identification result containing semantic information; then, matching the identification result with a preset set of operation intents, determining the confidence level of the identification result corresponding to each operation intent in the set of operation intents through the probability distribution or similarity metric output by the large language model, thereby forming a confidence level set; finally, based on the numerical comparison of each confidence level in the confidence level set, selecting the operation intent corresponding to the maximum confidence level as the finally identified target operation intent. This technical solution achieves accurate identification and classification of user command intents through the powerful natural language understanding capabilities of the large language model. The generation process of the confidence set reflects the quantifiable evaluation of intent matching, while the intent determination mechanism based on the maximum confidence value ensures the reliability and uniqueness of the identification results.

[0030] Furthermore, the preset set of operation intentions is set according to commonly used instruction types in the semiconductor design field. The definitions of the operation intention categories and parameter items in the parameter extraction framework are predefined based on knowledge from the semiconductor device design field. For example, the operation intention classification layer is implemented in the following manner.

[0031] INTENT_CLASS_PROMPT = """ Please analyze the user's requested action intent and select from the following options: 1. upload_script - Upload script / configuration file 2. execute_simulation - Executes the simulation task. 3. modify_parameter - Modify simulation parameters 4. view_report - View the results report Returns in JSON format: {"intent": "operational intent", "confidence": confidence level (0-1)}""" This enables the recognition of operational intent from natural language instructions in the field of semiconductor design.

[0032] In some embodiments of this application, based on the identified operational intent, at least one parameter item to be parsed associated with completing the operational intent is determined to form a parameter extraction framework. This includes: determining the corresponding parameter extraction framework from a framework template set according to the operational intent, wherein the parameter extraction frameworks in the framework template set correspond to the operational intent; and using each parameter in the determined parameter extraction framework as a parameter set to be assigned values. First, a corresponding parameter extraction framework is selected from a pre-constructed framework template set, wherein the framework template set stores multiple parameter extraction frameworks, each of which has a one-to-one correspondence with the operational intent. Subsequently, the selected parameter extraction framework is structurally parsed, and the at least one parameter item contained therein forms a parameter set to be assigned values. This technical solution achieves precise and targeted generation of prompt content by establishing a mapping mechanism between operational intent and parameter extraction framework. The process of identifying the parameter set to be assigned values ​​provides a clear target set for subsequent parameter assignment, ensuring the flexibility and adaptability of the parameter extraction framework. This design enables the system to accurately generate professional prompt content that conforms to the current context based on different interaction scenarios and operational intents. In some alternative implementations, the LangChain framework can be used as the intelligent orchestration engine for prompt word templates. An example of generating corresponding prompts based on the operation intent, where the operation intent is the prompt word template corresponding to the uploaded script / configuration file, is as follows: def get_detail_prompt(intent: str) ->str: prompt_map = { "upload_script": """ Please extract the following information from the user request: - local_path: Local file path (required) - file_types: A list of file extensions (e.g., ["cmd", "tcl"]) - replace_existing: Whether to overwrite an existing file (boolean value) return prompt_map.get(intent, "Please extract key parameters").

[0033] In some embodiments of this application, parsing the semantics of the natural language instructions to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework includes: first, using a pre-trained semantic model or large language model to perform deep semantic parsing and understanding of the user-input instructions, identifying and extracting at least one semantically specific constituent element contained in the instructions; then, performing meaning matching between each parsed element and a predefined set of operation parameters, wherein the meaning matching is based on semantic similarity calculation or predefined mapping rules; finally, assigning the semantic content or value of the successfully matched element to the corresponding operation parameter, completing the conversion from natural language instructions to structured parameters. This technical solution, by utilizing the powerful natural language understanding capabilities of advanced semantic models or large language models, achieves accurate deconstruction of unstructured natural language instructions, mapping the non-standard expressions of the natural language instructions regarding device parameters to standardized parameter names and units, effectively identifying key parameters and operational intentions in the instructions. The processes of "meaning matching" and "assignment" constitute a crucial bridge between natural language and machine-manipulable structured data, significantly improving the system's ability to understand and respond to diverse and colloquial user expressions. Several examples can be used to implement this parameter extraction phase.

[0034] Based on the prompt word template in the previous implementation, the following results are obtained.

[0035] Example input: "Upload ~ / simulations / new.cmd to the server to overwrite the old version" Example output: { "local_path": " / home / user / simulations / new.cmd", "file_types": ["cmd"], "replace_existing": true } """

[0036] In addition, the prompt word templates and input / output examples corresponding to executing simulation tasks and modifying simulation parameters will be explained later.

[0037] In some embodiments of this application, the parameters in the parameter extraction framework include necessary parameters and optional parameters. For the necessary parameters, the system attempts to determine their values ​​sequentially according to a preset priority order: first, it extracts them directly from the displayed content of the current natural language instruction; if extraction fails, it infers based on the deep semantic information of the natural language instruction to generate the parameter; if inference fails, it further retrieves the parameter from the dialogue history context associated with the current natural language instruction; if all the aforementioned steps fail to assign a value, the system outputs a special prompt message for the missing necessary parameter and actively queries the user to obtain the corresponding input value. This process ensures that key parameters are ultimately assigned values, guaranteeing the execution of core functions. For the optional parameters, the assignment order is relatively simplified: they are first extracted directly from the current natural language instruction; if extraction fails, predefined default values ​​are automatically used to fill the gaps without initiating complex inference or interactive query processes. This technical solution, by establishing a hierarchical and orderly variable value determination mechanism, maximizes the optimization of human-computer interaction efficiency and reduces user burden while ensuring the integrity of system functions. This differentiated approach to necessary and optional parameters, along with their respective rigorous internal priorities, reflects the intelligence and precision of resource allocation, effectively enhancing the practicality and user experience of complex dialogue systems or task-oriented robots.

[0038] In some embodiments of this application, parameter items and their corresponding values ​​are organized into structured data conforming to the interface specifications of the target device design software. This includes: determining the structure of the structured data according to the interface specifications with the semiconductor design software; and mapping the parameter items and their corresponding values ​​to the structured data in key-value pairs. Specifically, firstly, based on the application programming interface specifications or data exchange standards provided by the target semiconductor design software, a standardized data structure that the structured data must follow is predefined. This structure clearly defines the data fields, types, and hierarchical relationships. Subsequently, the operation parameters obtained through natural language parsing and their corresponding values ​​after assignment processing are encapsulated and mapped according to the association form of key-value pairs, where the operation parameters serve as keys and their assignment results serve as values. Finally, based on the predefined standardized data structure, the set of key-value pairs is organized and output as structured data conforming to the requirements of the semiconductor design software interface. This method effectively solves the human-computer interaction barrier caused by the ambiguity of natural language input and the high professionalism of operation instructions in the semiconductor design process. This technical solution achieves efficient conversion from unstructured natural language instructions to industry-specific data formats by establishing an adaptation mechanism with the semiconductor design software interface standard.

[0039] In some embodiments of this application, the method further includes: performing data validation on the obtained structured data that conforms to the target device design software interface specification; the data validation includes: verifying whether the fields of the structured data are appropriate data types, verifying whether the fields of the structured data conform to numeric or character restrictions, and verifying whether the structured data conforms to custom constraints. This validation can be performed using Pydantic Validator to complete real-time parameter validation.

[0040] The following examples illustrate the parsing methods used in the application.

[0041] # Generate corresponding prompts based on intent def get_detail_prompt(intent: str) ->str: prompt_map = { "upload_script": """ Please extract the following information from the user request: - local_path: Local file path (required) - file_types: A list of file extensions (e.g., ["cmd", "tcl"]) - replace_existing: Whether to overwrite an existing file (Boolean value) Example input: "Upload ~ / simulations / new.cmd to the server to overwrite the old version" Example output: { "local_path": " / home / user / simulations / new.cmd", "file_types": ["cmd"], "replace_existing": true } """, "execute_simulation": """ Please extract: - node_id: Node number (e.g., n0, n1) - parameters: Special parameter key-value pairs - use_cache: Whether to use caching (default is true) Example input: "Run the n3 node with cached data, Vdd=1.2" Example output: { "node_id": "n3", "parameters": {"Vdd": 1.2}, "use_cache": true } """, "modify_parameter": """ Extraction required: - param_name: Parameter name (e.g., Vth) - new_value: New value - apply_to: Scope of application (all|node_ids) Example input: "Set the threshold voltage of all nodes to 0.5V" Example output: { "param_name": "Vth", "new_value": 0.5, "apply_to": "all } """ } return prompt_map.get(intent, "Please extract key parameters").

[0042] Furthermore, this application also demonstrates good performance in complex interactive embodiments, as illustrated below. To more clearly illustrate how this invention dynamically adjusts the parsing logic based on context, consider the following multi-turn dialogue scenario: - First round of user input: "Design a gate oxide layer with a thickness of 5nm for an NMOS transistor." - System parsing and execution: The operation intent is identified as "define_structure", the parameters {component:"gate_oxide", thickness: 5, unit: "nm"} are extracted, and the corresponding design operation is completed.

[0043] - Second round of user input: "Set its dielectric constant to 25, and then re-evaluate the leakage current." - System parsing process: 1. Intent Recognition: Instructions containing "set to" (modify) and "re-evaluate" (analyze) may be identified as the compound intent "modify_and_analyze".

[0044] 2. Dynamically determine the parameter framework: Based on the dialogue history, the system knows that the current operation target is the "gate oxide layer" defined in the previous round. Therefore, the dynamically generated framework for the "modification" part includes [parameter_to_modify, new_value]; for the "analysis" part, since there is already a complete structure definition in the history, the framework may be simplified to [analysis_type:"leakage_current"].

[0045] 3. Semantic parsing and assignment: Extract the "dielectric constant" from the instruction and map it to parameter_to_modify: "dielectric_constant", and assign the value new_value: 25; extract the "re-evaluate leakage current" and assign the value analysis_type: "leakage_current".

[0046] 4. Structured Output and Verification: The system outputs structured instructions similar to {action_sequence: [{modify: {target:"gate_oxide", param: "dielectric_constant", value: 25}}, {analyze: {target: "gate_oxide", metric: "leakage_current"}}]}. The data verification module checks whether the dielectric constant of 25 is within a reasonable range for the target process.

[0047] This embodiment demonstrates how the system understands the reference of "its," automatically associates the instruction parameters of two rounds of dialogue, and constructs a sequentially executed operation sequence, which significantly surpasses the parsing capabilities based on single rounds and fixed templates, and provides a good user experience.

[0048] Based on the same inventive concept, this application also provides a user input parsing device for device-designed intelligent agents. Figure 2 The diagram illustrates the structure of a user input parsing device for a device-based intelligent agent according to an embodiment of this application. For example... Figure 2As shown, the device includes: an intent recognition module configured to recognize the device design operation intent represented by the natural language command input by the user; a frame determination module configured to determine at least one parameter item to be parsed associated with completing the operation intent based on the recognized operation intent, forming a parameter extraction frame; a semantic parsing and assignment module configured to parse the semantics of the natural language command to extract or infer the values ​​corresponding to each parameter in the parameter extraction frame; and a result structuring module configured to organize the parameter items and their corresponding values ​​into structured data conforming to the target device design software interface specification. The user input parsing device based on the device design agent can be designed with a hierarchical architecture, adapting to diverse expression methods and dynamically adjusting the parsing logic based on context, significantly improving the accuracy and adaptability of the agent's interaction with the user. The intelligent orchestration engine for the prompt word template can adopt the LangChain framework.

[0049] In some alternative implementations, in response to a natural language instruction input by a user, the device design operation intention represented by the natural language instruction is identified; based on the identified operation intention, at least one parameter item to be parsed is determined that is associated with completing the operation intention, forming a parameter extraction framework; the semantics of the natural language instruction is parsed to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework; and the parameter items and their corresponding values ​​are organized into structured data that conforms to the target device design software interface specification.

[0050] In some optional implementations, in response to a natural language instruction input by a user, identifying the device design operation intention represented by the natural language instruction includes: using a large language model to identify the natural language instruction and obtaining an identification result; matching the identification result with a preset set of operation intentions to obtain the confidence level of the identification result for each operation intention category in the preset set of operation intentions, forming a confidence level set; and obtaining the operation intention based on the maximum confidence level in the confidence level set.

[0051] In some alternative implementations, the preset set of operational intents is set according to commonly used instruction types in the semiconductor design field.

[0052] In some alternative implementations, the definitions of the operation intent category and the parameter items in the parameter extraction framework are predefined based on knowledge in the field of semiconductor device design.

[0053] In some alternative implementations, based on the identified operational intent, at least one parameter item to be parsed is determined to be associated with completing the operational intent, forming a parameter extraction framework, including: determining the corresponding parameter extraction framework from a framework template set according to the operational intent, wherein the parameter extraction frameworks in the framework template set have a correspondence with the operational intent; and using each parameter in the determined parameter extraction framework as a set of parameters to be assigned values.

[0054] In some optional implementations, parsing the semantics of the natural language instructions to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework includes: mapping the non-standard expressions of the natural language instructions regarding device parameters to standardized parameter names and units; parsing the semantic information in the natural language instructions and assigning the parameter values ​​in the semantic information to the corresponding parameters.

[0055] In some alternative implementations, the parameters in the parameter extraction framework include necessary parameters and optional parameters; the values ​​of the necessary parameters are determined in the following order: extracted from natural language instructions, inferred from the semantic information of the natural language instructions, obtained from the dialogue history of the natural language instructions, outputting a separate prompt for the necessary parameters with missing values ​​and obtaining the corresponding input; the values ​​of the optional parameters are determined in the following order: extracted from natural language instructions, using default values.

[0056] In some alternative implementations, the parameter items and their corresponding values ​​are organized into structured data that conforms to the interface specification of the target device design software, including: determining the structure of the structured data according to the interface specification with the semiconductor design software; and mapping the parameter items and their corresponding values ​​to the structured data in the form of key-value pairs.

[0057] In some optional embodiments, the apparatus further includes: performing data verification on the obtained structured data that conforms to the target device design software interface specification; the data verification includes: verifying whether the fields of the structured data are appropriate data types, verifying whether the fields of the structured data conform to numeric or character restrictions, and verifying whether the structured data conforms to custom constraints.

[0058] The specific limitations of each functional module in the aforementioned user input parsing device based on device-designed intelligent agents can be found in the limitations of the user input parsing method based on device-designed intelligent agents described above, and will not be repeated here. Each module in the above system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the corresponding operations of each module. This also achieves the beneficial effects of significantly improving the accuracy and adaptability of the interaction between the intelligent agent and the user, and improving the efficiency of device design.

[0059] In some embodiments of this application, an electronic device is also provided, comprising: at least one processor; and a memory connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, which executes the aforementioned user input parsing method based on a device-designed intelligent agent. Its internal structure diagram can be shown as follows. Figure 3 As shown. Figure 3 This schematic diagram illustrates the internal structure of an electronic device according to an embodiment of this application. The electronic device includes a processor A01, a network interface A02, a memory (not shown), and a database (not shown) connected via a system bus. The processor A01 provides computational and control capabilities. The memory includes internal memory A03 and a non-volatile storage medium A04. The non-volatile storage medium A04 stores an operating system B01, a computer program B02, and a database (not shown). The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 stored in the non-volatile storage medium A04. The network interface A02 is used for communication with external terminals via a network connection. When the computer program B02 is executed by the processor A01, it implements a user input parsing method based on a device-designed intelligent agent.

[0060] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the electronic device to which the present application is applied. The specific electronic device may include more or fewer components than shown in the figure, or combine certain components, or have different component arrangements.

[0061] In one embodiment provided in this application, a machine-readable storage medium is provided, on which instructions are stored, which, when executed by a processor, cause the processor to be configured to perform the aforementioned user input parsing method based on device-designed intelligent agents.

[0062] In one embodiment provided in this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the aforementioned user input parsing method based on device-designed intelligent agents.

[0063] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0064] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0065] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0066] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0067] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0068] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0069] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0070] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0071] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A user input parsing method applied to intelligent agents in device design, characterized in that, The method includes: In response to natural language commands input by the user, the device design operation intent represented by the natural language commands is identified; Based on the identified operational intent, at least one parameter item to be parsed is determined that is associated with completing the operational intent, thus forming a parameter extraction framework; The semantics of the natural language instructions are parsed in order to extract or infer the values ​​of each parameter in the parameter extraction framework; Organize the parameter items and their corresponding values ​​into structured data that conforms to the target device design software interface specification.

2. The method according to claim 1, characterized in that, Responding to a user-inputted natural language command, identifying the device design operation intent represented by the natural language command, including: The natural language instructions are identified using a large language model to obtain the recognition results; The recognition results are matched with a preset set of operation intentions to obtain the confidence level of the recognition results for each operation intention category in the preset set of operation intentions, forming a confidence level set. The operational intent is obtained based on the maximum confidence value in the confidence set.

3. The method according to claim 2, characterized in that, The preset set of operation intentions is set according to the commonly used instruction types in the field of semiconductor design.

4. The method according to claim 2, characterized in that, The definitions of the operation intent category and the parameter items in the parameter extraction framework are predefined based on knowledge in the field of semiconductor device design.

5. The method according to claim 1, characterized in that, Based on the identified operational intent, at least one parameter item to be parsed is determined that is associated with completing the operational intent, forming a parameter extraction framework, including: The corresponding parameter extraction framework is determined from the framework template set according to the operation intention, and the parameter extraction framework in the framework template set has a corresponding relationship with the operation intention; The parameters in the determined parameter extraction framework are used as the set of parameters that need to be assigned values.

6. The method according to claim 1, characterized in that, Parsing the semantics of the natural language instructions to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework includes: The non-standard expressions of the natural language instructions regarding device parameters are mapped to standardized parameter names and units; The semantic information in the natural language instruction is parsed, and the parameter values ​​in the semantic information are assigned to the corresponding parameters.

7. The method according to claim 1, characterized in that, The parameters in the parameter extraction framework include necessary parameters and optional parameters; The values ​​of the necessary parameters are determined in the following order: extracted from the natural language instruction, inferred from the semantic information of the natural language instruction, obtained from the dialogue history of the natural language instruction, outputting a separate prompt for the necessary parameter with a missing value and obtaining the corresponding input; The values ​​of the optional parameters are determined in the following order: extracted from natural language instructions, and default values ​​are used.

8. The method according to claim 1, characterized in that, The parameter items and their corresponding values ​​are organized into structured data that conforms to the target device design software interface specification, including: The structure of the structured data is determined according to the interface specifications with the semiconductor design software; The parameter items and their corresponding values ​​are mapped to the structured data in key-value pairs.

9. The method according to claim 1, characterized in that, The method further includes: performing data verification on the structured data that conforms to the target device design software interface specification; The data validation includes: validating whether the fields of the structured data are of appropriate data types, validating whether the fields of the structured data conform to numeric or character restrictions, and validating whether the structured data conforms to custom constraints.

10. A user input parsing device applied to a device design intelligent agent, characterized in that, The device includes: The intent recognition module is configured to recognize the device design operation intent represented by the natural language command input by the user in response to the natural language command. The framework determination module is configured to determine at least one parameter item to be parsed that is associated with completing the identified operation intent based on the identified operation intent, thereby forming a parameter extraction framework; The semantic parsing and assignment module is configured to parse the semantics of the natural language instruction in order to extract or infer the values ​​corresponding to each parameter in the parameter extraction framework; The resulting structured module is configured to organize parameter items and their corresponding values ​​into structured data that conforms to the target device design software interface specification.

11. An electronic device, characterized in that, include: At least one processor; A memory connected to the at least one processor; The memory stores instructions executable by the at least one processor, which implements the user input parsing method for device-based intelligent agents as described in any one of claims 1 to 9 by executing the instructions stored in the memory.

12. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that, When executed by a processor, the computer program / instruction implements the user input parsing method for device-based intelligent agents as described in any one of claims 1 to 9.

13. A computer program product comprising a computer program / instructions, characterized in that, When executed by a processor, the computer program / instruction implements the user input parsing method for device-based intelligent agents as described in any one of claims 1 to 9.

Images