Functional map generation method and system

Abstract

The invention discloses a function map generation method and system. The system uses an artificial intelligence model to execute the following steps: obtaining a difficult point and first data related to the difficult point from first reply content of a first inquiry corresponding to the difficult point; obtaining the use target and second data related to the use target from second reply content of the second inquiry corresponding to the use target; obtaining a practical function and a target type based on the difficulty point and the use target; obtaining a required function based on the target type; performing semantic comparison on the practical function and the required function to obtain difference set content; and generating a function map based on the target type, the real function and the difference set content.

Description

Technical Field

[0001] The present invention relates to a human - machine interaction mechanism, and particularly to a method and system for generating a function map based on an artificial intelligence model. Background Art

[0002] In the current era of comprehensive digital transformation, the user interface (UI) and user experience (UX) of websites have become crucial. Besides affecting the direct operation experience, they are directly related to efficiency. For example, for an administrative website, data must be presented according to importance and be quickly linked to related data. Therefore, having a convenient operation process and an intuitive and easy - to - manage dashboard has become the main topic of website screen design. However, to achieve these goals, UI / UX designers must repeatedly interview customers to accurately capture requirements, which undoubtedly increases the time cost of website development.

[0003] In the general UI / UX design and development process, UI / UX designers will conduct an initial requirements interview with customers, sort out the function map of the website according to function categories, such as video monitoring, device control, event notification, etc., and produce the first version of the wireframe for customers to confirm the functions. The wireframe does not have visual elements and is designed to present the architecture, functions, and processes of the website. If the customer has new requirements or opinions, the UI / UX designer will conduct another interview, update the function map, and remake the wireframe, repeating this process until a consensus is reached. Finally, the website development team will use this wireframe as a benchmark to enter the UI design stage and produce visual mockups or even prototypes.

[0004] In the entire design and development cycle, requirements interviews are an important and inevitable process in the whole development process. However, since customers are not clear about their own requirements, it is necessary to guide the core requirements through multiple interviews and supplement necessary relevant information. It is also very common to need to modify and confirm again due to different understandings between the two parties until a consensus is reached before proceeding with UI design and development. Therefore, requirements interviews take up a lot of time in the development cycle, and requirements interviews are a cumbersome and not highly standardized process, making it difficult to improve efficiency. Summary of the Invention

[0005] The present invention provides a method and system for generating a function map based on an artificial intelligence (AI) model, which can effectively reduce the time cost of requirements interviews.

[0006] A method for generating a functional map based on an artificial intelligence model, which is executed by a processor and includes the following steps: a first query step of obtaining a pain point and first data related to the pain point from a first reply content corresponding to a first query corresponding to the pain point by using the artificial intelligence model; a second query step of obtaining a usage target and second data related to the usage target from a second reply content corresponding to a second query corresponding to the usage target by using the artificial intelligence model; a completeness judgment step including: a function induction step of obtaining an implemented function and a target type based on the pain point and the usage target by using the artificial intelligence model; a function association step of obtaining required functions based on the target type by using the artificial intelligence model; and a function comparison step of performing a semantic comparison between the implemented function and the required functions by using the artificial intelligence model to obtain a difference set content; and a functional map generation step of generating a functional map based on the target type, the implemented function, and the difference set content.

[0007] In an embodiment of the present invention, the first query step includes: generating a prompt word corresponding to the first reply content; executing a noun capture program based on the prompt word corresponding to the first reply content by using the artificial intelligence model to obtain a first noun reply result; and executing a semantic capture program by using the artificial intelligence model to obtain the pain point and the first data related to the pain point based on the first reply content and the first noun reply result.

[0008] In an embodiment of the present invention, the second query step includes: generating a prompt word corresponding to the second reply content; executing a noun capture program based on the prompt word corresponding to the second reply content by using the artificial intelligence model to obtain a second noun reply result from the second reply content; and executing a semantic capture program by using the artificial intelligence model to obtain the usage target and the second data related to the usage target based on the second reply content and the second noun reply result.

[0009] In an embodiment of the present invention, the functional map generation method further includes: a third query step including: receiving a third reply content corresponding to a third query corresponding to a data source; generating a prompt word corresponding to the third reply content; executing a noun capture program based on the prompt word corresponding to the third reply content by using the artificial intelligence model to obtain a third noun reply result from the third reply content; and executing a semantic capture program by using the artificial intelligence model to obtain source information based on the third reply content and the third noun reply result.

[0010] In an embodiment of the present invention, the functional map generation step further includes: updating the functional map based on the source information.

[0011] In an embodiment of the present invention, the function map generation step further includes: generating prompt words corresponding to the pain points and usage goals; and using an artificial intelligence model to execute a function induction program based on the prompt words to obtain the implemented functions and goal types.

[0012] In an embodiment of the present invention, the function association step includes: generating prompt words corresponding to the goal type; and using an artificial intelligence model to execute a function association program based on the prompt words to obtain the required functions.

[0013] In an embodiment of the present invention, the function map generation step includes: after the function induction step, generating a function map based on the implemented functions and goal types; and after the function comparison step, updating the function map based on the difference set content.

[0014] The function map generation system based on an artificial intelligence model of the present invention includes: an artificial intelligence model; an interaction interface; and a processor coupled to the artificial intelligence model and the interaction interface. In response to the first reply content received by the interaction interface for the first inquiry corresponding to the pain point, the processor is configured to: use the artificial intelligence model to obtain the pain point and the first data related to the pain point from the first reply content. In response to the second reply content received by the interaction interface for the second inquiry corresponding to the usage goal, the processor is configured to: use the artificial intelligence model to obtain the usage goal and the second data related to the usage goal from the second reply content. The processor is further configured to: use the artificial intelligence model to obtain the implemented functions and goal types based on the pain point and the usage goal; use the artificial intelligence model to obtain the required functions based on the goal type; use the artificial intelligence model to perform semantic comparison between the implemented functions and the required functions to obtain the difference set content; and generate a function map based on the goal type, the implemented functions, and the difference set content.

[0015] Based on the above, by using an artificial intelligence model to obtain important information and organize it into a function map, the customer needs can be quickly and accurately captured, the number of back-and-forth interviews between designers and customers can be reduced, and the time cost of demand interviews can be effectively reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a block diagram of a function map generation system according to an embodiment of the present invention.

[0017] Figure 2 is a schematic diagram of a robot architecture according to an embodiment of the present invention.

[0018] Figure 3 is a flowchart of a function map generation method according to an embodiment of the present invention.

[0019] Figure 4 is a schematic diagram of an architecture for generating a function map according to an embodiment of the present invention.

[0020] Figures 5A to 5C It is a schematic diagram of multiple prompt templates according to an embodiment of the present invention.

[0021] Figure 6 It is a schematic architecture diagram of a problem generation program according to an embodiment of the present invention.

[0022] Figure 7 It is a schematic architecture diagram of a noun capture program according to an embodiment of the present invention.

[0023] Figure 8 It is a schematic architecture diagram of a semantic capture program according to an embodiment of the present invention.

[0024] Figure 9 It is a schematic architecture diagram of a function induction program according to an embodiment of the present invention.

[0025] Figure 10 It is a schematic architecture diagram of a function association program according to an embodiment of the present invention.

[0026] Figure 11A and Figure 11B It is a schematic diagram of the content presented by an interactive interface according to an embodiment of the present invention.

[0027] Explanation of reference numerals:

[0028] 110: Processor

[0029] 120: Interactive interface

[0030] 130: AI model

[0031] 200: Robot

[0032] 210: Question asking module

[0033] 220: Reply parsing module

[0034] 230: Map building module

[0035] 510 - 580: Prompt templates

[0036] 511, 521, 531, 532, 541, 542, 551, 552, 553, 561, 562, 571, 572, 581: Regions

[0037] A1, A2, A3-1, A3-2: Confirmation information

[0038] A4: Functional map information

[0039] D1: Data and source database

[0040] D2: Noun database

[0041] FM: Function map

[0042] Q1: First inquiry

[0043] Q2: Second inquiry

[0044] Q3: Third inquiry

[0045] P0: Problem generation program

[0046] P1: Noun capture program

[0047] P2: Semantic capture program

[0048] P3: Function induction program

[0049] P4: Function association program

[0050] P5: Function comparison program

[0051] R1: First reply content

[0052] R2: Second reply content

[0053] R3: Third reply content

[0054] S310: First inquiry step

[0055] S320: Second inquiry step

[0056] S330: Completeness judgment step

[0057] S331: Function induction step

[0058] S332: Function association step

[0059] S333: Function comparison step

[0060] S340: Function map generation step

[0061] S350: Third inquiry step Detailed implementation manner

[0062] Figure 1 It is a block diagram of a function map generation system according to an embodiment of the present invention. Please refer to Figure 1 , the function map generation system includes a processor 110, an interaction interface 120, and an artificial intelligence (AI) model 130. The processor 110 is coupled to the interaction interface 120 and the AI model 130.

[0063] The processor 110 is, for example, a Central Processing Unit (CPU), a Physics Processing Unit (PPU), a programmable microprocessor, an embedded control chip, a Digital Signal Processor (DSP), an Application-Specific Integrated Circuit (ASIC), or other similar devices.

[0064] The interaction interface 120 is a User Interface (UI), which is a medium for interaction and information exchange between the processor 110 and the user. For example, the interaction interface 120 includes Human Computer Interaction (HCI) and Graphical User Interface (GUI). In one embodiment, the interaction interface 120 can be implemented by a display and an input device, or can be implemented by a touch screen.

[0065] The AI model 130 can be disposed in the same electronic device as the processor 110. Additionally, the AI model 130 can also be disposed in different electronic devices from the processor 110 and communicate with each other through wired or wireless connections. The AI model 130 is, for example, a large language model (LLM), which consists of an artificial neural network with many parameters and is trained on a large amount of unlabeled text using self-supervised learning or semi-supervised learning. LLMs such as ChatGPT, GPT4, LLaMA-65B, PaLM-62B, etc. have achieved excellent results in various Natural Language Processing (NLP) evaluations, including common sense reasoning, closed-book question answering, reading comprehension, mathematical reasoning, code generation, Massive Multitask Language Understanding (MMLU), etc. In recent years, major technology leaders have also been actively shrinking LLMs while endowing them with more powerful capabilities.

[0066] In one embodiment, past interview data can be collected according to the UI / UX designer team, and three key pieces of information (e.g., pain points, usage goals, and relevant data) can be sorted out and related to the type of website and implementation functions finally implemented, to establish the input corpus and output results (including the output corpus). The input for training the AI model 130 is the above three key pieces of information. The output for training the AI model 130 is the type of website and implementation functions. The results obtained from subsequent analysis using the AI model 130 can be fed back to the AI model 130 for retraining to improve the accuracy of the AI model 130.

[0067] The AI model 130 adopted in this embodiment has the following capabilities: sentence deconstruction, which can disassemble the structure of a sentence, e.g., nouns, verbs, etc.; semantic understanding, which can understand the context, parse the semantics and summarize; possess general domain knowledge, learned through training with a large amount of data in a wide range of fields; topic modeling, which can analyze multiple keywords and summarize the corresponding topics; association, which can associate things related to different levels corresponding to the topic; word embedding, which can transform words and sentences into a semantic distance space; sentence generation, which can generate sentences according to the situation, mood, role, etc., with a high degree of anthropomorphism.

[0068] The AI model 130 has a strong generalization ability. When applied to different tasks, it does not need to be retrained. Only by giving relevant prompt words (including task descriptions, examples, output formats, etc.) can the corresponding results be output. This method is called In-context Learning, and the output results will vary according to the input prompt words. The design of prompt words is also called prompt engineering. The execution of prompt engineering is in the processor 110. According to the design and the current state (e.g., function induction step S331, function association step S332, function comparison step S333, third query step S350), one or more corresponding prompt words are selected or generated, and finally, the desired results are obtained by driving the AI model 130 with the prompt words.

[0069] Figure 2 is a schematic diagram of a robot architecture according to an embodiment of the present invention. This embodiment includes three roles, namely, the user, the robot 200, and the AI model 130 (refer to Figure 1) Based on the professional knowledge and experience in UI / UX design, the robot 200 will guide the user to answer questions and capture the key information required for website design with the help of the capabilities of the AI model 130, such as the data to be presented, data sources, usage scenarios, developed functions, etc. It will organize these into a function map and finally generate the final version of the function map through confirmation with the user.

[0070] Please refer to Figure 2 , the robot 200 includes a question asking module 210, a response parsing module 220, and a map building module 230. These modules are, for example, software modules stored in a memory and composed of one or more code snippets. The memory can be, for example, any form of fixed or removable random access memory (RAM), read-only memory (ROM), flash memory, hard disk, or other similar devices or a combination of these devices. Additionally, the question asking module 210, the response parsing module 220, and the map building module 230 can also be implemented by hardware chips or circuits.

[0071] The processor 110 drives the question asking module 210, the response parsing module 220, and the map building module 230 to perform their corresponding functions. The question asking module 210 is used to call the AI model 130 to generate corresponding questions based on the key information. The response parsing module 220 is used to call the AI model 130 to parse the response content to the questions. The map building module 230 is used to call the AI model 130 to find the function list based on the results obtained by the response parsing module 220 and then build the function map.

[0072] For ease of understanding, the following will list Figures 3 to 11B while being paired with Figure 1 and Figure 2 for illustration.

[0073] Figure 3 is a flowchart of a function map generation method according to an embodiment of the present invention. Figure 4 is a schematic structural diagram of generating a function map according to an embodiment of the present invention. Figures 5A to 5C is a schematic diagram of multiple prompt templates according to an embodiment of the present invention. Figure 6 is a schematic structural diagram of a question generation program according to an embodiment of the present invention. Figure 7 is a schematic structural diagram of a noun capture program according to an embodiment of the present invention. Figure 8 is a schematic structural diagram of a semantic capture program according to an embodiment of the present invention.

[0074] Figure 9It is a schematic diagram of the architecture of a function induction program according to an embodiment of the present invention. Figure 10 It is a schematic diagram of the architecture of a function association program according to an embodiment of the present invention. Figure 11A and Figure 11B It is a schematic diagram of the content presented by the interactive interface according to an embodiment of the present invention.

[0075] First, please refer to Figure 3 , the function map generation method includes a first inquiry step S310, a second inquiry step S320, a completeness judgment step S330, and a function map generation step S340. The completeness judgment step S330 includes a function induction step S331, a function association step S332, and a function comparison step S333. In addition, the function map generation method may further include a third inquiry step S350. The third inquiry step S350 is not a necessary step and can be decided whether to add according to the situation.

[0076] Please refer to Figure 4 , in the first inquiry step S310, the second inquiry step S320, and the third inquiry step S350, the first inquiry Q1, the second inquiry Q2, and the third inquiry Q3 used are generated by the question inquiry module 210. The question type of the first inquiry Q1 is the pain point, the question type of the second inquiry Q2 is the usage target (including the usage purpose and object). The question type of the third inquiry Q3 is the data source. Specifically, the question inquiry module 210 will select the corresponding prompt word template according to the question type, and then use the AI model 130 to be responsible for generating the corresponding first inquiry Q1, second inquiry Q2, or third inquiry Q3. The questions generated by using the AI model 130 can be personified, including a dynamic and non-stereotyped, polite, and service-minded tone.

[0077] Please refer to Figure 6 , for the first inquiry Q1 corresponding to the pain point, the question inquiry module 210 adopts the prompt word template 510 according to the question type "the pain points that the website needs to solve", brings in the key information corresponding to the "pain point", and can also supplement the previously collected proper nouns, fields, etc. information in the context to generate the final prompt word. After deciding to adopt the prompt word template 510, fill the content of the question type "the pain points that the website needs to solve" into the area 511 in the prompt word template 510 to obtain the prompt word, and then use the AI model 130 to execute the question generation program P0 to generate the corresponding first inquiry Q1 based on the prompt word. For example, as shown in the first inquiry Q1 in Figure 11A : "To better assist you in developing the website, I need to understand the pain points you are facing so that I can provide the most effective support. Please share some specific challenges that your website needs to solve so that I can provide more specific suggestions and help." The second inquiry Q2 and the third inquiry Q3 are the same by analogy.

[0078] After the question query module 210 generates a query, it presents the query (the first query Q1, the second query Q2, or the third query Q3) to the interaction interface 120, and the interaction interface 120 receives the corresponding response content (the first response content R1, the second response content R2, or the third response content R3). For Figure 11A example, after the interaction interface 120 presents the first query Q1, it then receives the first response content R1 through the interaction interface 120.

[0079] Referring to Figure 3 and Figure 4 , in the first query step S310, the processor 110 uses the AI model 130 to obtain the pain point and the first data related to the pain point from the first response content R1 corresponding to the pain point, and stores the first data in the data and source database D1.

[0080] Specifically, referring to Figure 7 , the response parsing module 220 selects the prompt template 520 based on the first response content R1, then fills the first response content R1 into the area 521 in the prompt template 520 to generate a prompt corresponding to the first response content R1, and then executes the noun capture program P1 based on the prompt corresponding to the first response content R1 through the AI model 130 to obtain the first noun response result.

[0081] The noun capture program P1 is to call the capabilities of the AI model 130 through the prompt, thereby capturing the proper nouns, strange or ambiguous nouns that appear in the first response content R1, and explaining their meanings. Then, during the "parsing" process, the response parsing module 220 will determine whether the noun list listed in the first noun response result is a known noun (a noun known to the AI model 130 and capable of being explained) or an unknown noun. For example, it can be determined whether it is an unknown noun by searching whether the string "search online" or "online search" appears in the first noun response result (as shown in the following example 1-1). If it is an unknown noun, it can be further searched using a search engine to obtain the corresponding explanation. Then, the meanings of the noun and the corresponding explanation are confirmed. For example, the noun and the explanation are presented to the interaction interface 120 for the user to confirm. Finally, the result is stored in the noun database D2. If no relevant explanation can be found, the explanation provided by the user can also be received through the interaction interface 120.

[0082] Example 1-1

[0083]

[0084] That is, if there are unknown nouns that the AI model 130 cannot understand, relevant meanings will be searched through a search engine, and then meaning confirmation will be entered for the user to confirm again. Whether the finally obtained nouns are known or unknown in meaning will be shown to the user for confirmation, including modification.

[0085] In addition, during the "parsing" process, if the reply parsing module 220 confirms that there are no proper nouns, ambiguous or strange nouns currently by searching for preset strings such as "not appeared", "not found", "did not appear", "did not find", etc., it will not proceed to the next step and directly end the process, indicating that it will not enter the meaning confirmation step and the list in the noun database D2 will not be updated. For example, referring to Example 2-1, since "not appeared" appears in the first noun response result, it is determined that there are no proper nouns, ambiguous or strange nouns.

[0086] Example 2-1

[0087]

[0088] Next, referring to Figure 8 , the reply parsing module 220 executes the semantic capture program P2 through the AI model 130 to obtain the difficulties and the first data related to the difficulties based on the first reply content R1 and the first noun response result. The semantic capture program P2 aims to capture key information through semantic understanding by the AI model 130. The "generation of prompt words" and the semantic capture program P2 will have different results based on the "information type" input. For example, the prompt word template is selected according to the "information type", and the AI model 130 will parse the semantic response result to be output according to the "information type".

[0089] For example, referring to Example 2-2 below, the reply parsing module 220 selects the prompt word template 530 based on the information type, then fills the first reply content R1 into the area 531 in the prompt word template 530 and fills the problem type (such as "difficulties to be solved by the website") into the area 532, thereby generating a prompt word for semantic capture. Then, the AI model 130 executes the noun capture program P1 based on the generated prompt word to obtain the first semantic response result (such as including "difficulty:...", "data data:..."). After that, meaning confirmation is performed on the first semantic response result. For example, the first semantic response result is presented on the interaction interface 120 for the user to confirm.

[0090] Example 2-2

[0091]

[0092]

[0093] In another embodiment, referring to the following Examples 1-2, assume that the reply parsing module 220 selects a prompt template 550 based on the information type. Then, the first reply content R1 is filled into the area 551 in the prompt template 550, the question type (such as "difficult points that the website needs to solve") is filled into the area 553, and the proper nouns and their explanations (such as "aaaaaa: refers to the periodic growth rate of tumors, and the volume size is calculated every 1 month") are filled into the area 552, thereby generating a prompt for semantic capture. Then, the AI model 130 executes the noun capture program P1 based on the generated prompt to obtain the first semantic response result (such as including "difficult points:...", "data data:..."). The intention of adding the collected proper nouns to the prompt is to avoid mis-capturing or missing capturing.

[0094] Examples 1-2

[0095]

[0096] Next, referring to Figure 3 and Figure 4 , in the second inquiry step S320, the processor 110 uses the AI model 130 to obtain the usage target and the second data related to the usage target from the second reply content R2 corresponding to the second inquiry Q2 for the usage target, and stores the second data in the data and source database D1.

[0097] The reply parsing module 220 generates a prompt corresponding to the second reply content R2. Then, the AI model 130 executes the noun capture program P1 based on the prompt corresponding to the second reply content R2 to obtain the second noun response result. Then, the reply parsing module 220 executes the semantic capture program P2 through the AI model 130 to obtain the usage target and the second data related to the usage target based on the second reply content R2 and the second noun response result. Here, the usage target includes the usage object and the usage purpose.

[0098] The processes of the noun capture program P1 and the semantic capture program P2 executed in the second inquiry step S320 are similar to those in the first inquiry step S310. First, the noun capture program P1 is used to determine whether there are proper nouns in the second reply content R2. For example, as in the following Example 3, assume that the second reply content R2 is "This website will be used for internal personnel monitoring by us and is not open to the public". In the noun capture program P1, it is determined that there are no proper nouns, nor any strange or ambiguous nouns. The reply parsing module 220 adopts the prompt template 540, fills the second reply content R2 into the area 541 in the prompt template 540, and fills the question type (such as "Which objects use this website? What is its purpose?") into the area 542 to generate the corresponding prompt. Then, the AI model 130 executes the semantic capture program P2 to obtain the usage target and the second data.

[0099] Example 3

[0100]

[0101] In the third query step S350, the processor 110 obtains the data source from the third response content R3 corresponding to the third query Q3 of the data source by using the AI model 130. After receiving the third response content R3 corresponding to the third query Q3 of the data source, a prompt word corresponding to the third response content R3 is generated, and the noun capture program P1 is executed based on the corresponding prompt word by the AI model 130 to obtain the third noun response result from the third response content R3, and the semantic capture program P2 is executed by the AI model 130 to obtain the source information based on the third response content R3 and the third noun response result, and the source information is stored in the data and source database D1.

[0102] In the third query step S350, the processor 110 asks the user about the data source according to the first data obtained in the first query step S310 and the second data obtained in the second query step S320. After obtaining the user's answer, it is judged whether the source has been captured. If it has not been captured yet, the information is first stored in the database, and the question query module 210 generates another query again, and loops in this order until all data sources are captured, as shown in the following Examples 4 and 5.

[0103] Example 4

[0104]

[0105] Example 5

[0106]

[0107] The processes of the noun capture program P1 and the semantic capture program P2 executed in the third query step S350 are similar to those in the first query step S310. First, it is judged whether there are proper nouns in the second response content R2 through the noun capture program P1. Then, in the semantic capture program P2, the response parsing module 220 adopts the prompt word template 550, fills the third response content R3 into the area 561 in the prompt word template 560, and fills the first data and the second data into the area 562, thereby generating the corresponding prompt word. Then, the noun capture program P1 is executed based on the prompt word by the AI model 130 to obtain the source information.

[0108] Next, enter the integrity judgment step S330. In the function induction step S331, the processor 110 uses the AI model 130 to obtain the implemented functions and target types based on the pain points and usage objectives. The map building module 230 generates prompt words corresponding to the pain points and usage objectives, and then uses the AI model 130 to execute the function induction program P3 based on the prompt words to obtain the implemented functions and target types.

[0109] Refer to Figure 9 , the map building module 230 selects the prompt word template 570, fills the pain points into the area 571, and fills the usage objective (usage purpose and object) into the area 572, thereby generating corresponding prompt words. Then, the AI model 130 executes the function induction program P3 based on the prompt words to obtain the implemented functions and target types. The AI model 130 can effectively induce the implemented functions and analyze the target types (such as website types), and the AI model 130 will parse the implemented functions and website types with "Function requirements:..." and "Website type:..." and output them, as shown in Example 6.

[0110] Example 6

[0111]

[0112] In the function association step S332, the processor 110 uses the AI model 130 to obtain the required functions based on the target types. The map building module 230 generates prompt words corresponding to the target types, and uses the AI model 130 to execute the function association program P4 based on the prompt words to obtain the required functions.

[0113] Refer to Figure 10 and Example 7, the map building module 230 selects the prompt word template 580, fills the target type into the area 581, thereby generating corresponding prompt words. Then, using the ability of the AI model 130's general domain knowledge, other required functions are associated based on the target type to complement the professional knowledge that the user does not possess.

[0114] Example 7

[0115]

[0116] The AI model 130 uses professional association to obtain the required functions related to implementation. Since the association ability of the AI model 130 is very strong, among the obtained prompt words, it will be mentioned that the obtained required functions are classified according to the importance level and only the ones with a high importance level are output, but this is not limited to this according to the requirements.

[0117] After that, in the function comparison step S333, the processor 110 executes the function comparison program P5 using the AI model 130 to perform a semantic comparison between the implemented function and the required function to obtain the difference set content. For example, after obtaining the "required function" and the "implemented function", the word embedding ability of the AI model 130 is used for semantic comparison. The main purpose of this process is to take the intersection, that is, to add the "required function" actually used in the implementation in addition to the "implemented function" proposed by the user. To ensure that the "required function" and the "implemented function" do not overlap, in the function comparison step S333, the items in the "required function" that overlap with the "implemented function" are removed, and the difference set content is obtained and output.

[0118] For example, as shown in Example 6, the required functions obtained based on the user's reply include: integrating park information from different departments; tracking the total number of people entering the park, the number of people on site, and the average stay time in the park. The required functions associated by the AI model 130 are shown in Example 7 and include: user authentication and permission management; real-time monitoring data visualization; alarm and notification system; logging and auditing functions. After comparison, the difference set content includes: user authentication and permission management; real-time monitoring data visualization; alarm and notification system; logging and auditing functions.

[0119] After that, in step S340, the processor 110 generates a function map based on the target type, the implemented function, and the difference set content. Further, after obtaining the target type and the implemented function, the processor 110 can initially establish the function map FM. After that, after obtaining the difference set content, the difference set content is further added to update the function map FM. And, after obtaining the source information, the function map FM is updated based on the source information.

[0120] The function map FM is composed of four parts, namely: website type, function, data, existing systems and processes. Presented in the form of a mind map, arranged from abstract to concrete, the website type is at the top layer and is the core of the entire function map FM, determining the direction of the entire website. In addition, the function items required for constructing the website obtained based on the function induction program P3 and the function association program P4 are placed in the second layer.

[0121] Refer to Figure 11A And Figure 11B, this embodiment uses website design to illustrate the content presented by the interactive interface 120. First, the processor 110 uses the question inquiry module 210 to call the AI model 130 to generate a first inquiry Q1 for asking about the pain points that the website needs to solve, and presents it to the interactive interface 120. Then, the first reply content R1 corresponding to the first inquiry Q1 is received through the interactive interface 120. After that, the processor 110 uses the reply analysis module 220 to call the AI model 130 to obtain the pain points and the first data, and generates a corresponding confirmation message A1 based on the pain points and the first data and presents it to the interactive interface 120 for the user to confirm.

[0122] Then, the processor 110 uses the question inquiry module 210 to call the AI model 130 to generate a second inquiry Q2 for asking about the usage goals (usage purposes and objects) of the website, and presents it to the interactive interface 120. Then, the second reply content R2 corresponding to the second inquiry Q2 is received through the interactive interface 120. After that, the processor 110 uses the reply analysis module 220 to call the AI model 130 to obtain the usage goals (usage purposes and objects) and the second data, and generates a corresponding confirmation message A2 based on the usage goals and the second data and presents it to the interactive interface 120 for the user to confirm.

[0123] After that, the processor 110 uses the question inquiry module 210 to call the AI model 130 to generate a third inquiry Q3-1 for asking about the data source, and presents it to the interactive interface 120. Then, the corresponding reply content R3-1 is received through the interactive interface 120. After that, the processor 110 uses the reply analysis module 220 to call the AI model 130 to obtain the source information, and generates a corresponding confirmation message A3-1 based on the source information and presents it to the interactive interface 120 for the user to confirm.

[0124] In addition, since the reply content R3-1 does not fully answer all the data sources, the processor 110 then uses the question inquiry module 210 to call the AI model 130 to generate a fourth inquiry Q3-2 for asking about the data source, and presents it to the interactive interface 120. Then, the corresponding reply content R3-2 is received through the interactive interface 120. After that, the processor 110 uses the reply analysis module 220 to call the AI model 130 to obtain the source information, and generates a corresponding confirmation message A3-2 based on the source information and presents it to the interactive interface 120 for the user to confirm.

[0125] After that, the processor 110 uses the map building module 230 to call the AI model 130 to generate function map information A4 corresponding to the function map, and presents it to the interactive interface 120 for the user to confirm.

[0126] The above embodiments can be divided into the following four stages based on sequence.

[0127] In the first stage, in the first inquiry step S310, the robot 200 inquires about the difficulties that the website needs to solve and captures them. At the same time, it also captures the keywords of the data and files mentioned in the reply content. In the second inquiry step S320, the robot 200 inquires about the users and purposes of the website and captures them. Then, in the function induction step S331, based on the information captured in the first inquiry step S310 and the second inquiry step S320, the robot 200 analyzes the website type (target type) through the AI model 130 to obtain the website type and corresponding implementation functions, and establishes a preliminary function map FM.

[0128] In the second stage, in the function association step S332, based on the website type, the robot 200 calls the AI model 130 to associate the common required functions of this type of website, divides them according to the importance level to obtain the functions with a high importance level, and compares the integrity with the previously obtained implementation functions. Taking the required functions as the main, if it is found that the implementation functions obtained in the function induction step S331 are missing, the function map FM is updated. At this time, the function and data parts of the function map FM have been captured.

[0129] In the third stage, in addition to the key information captured in the second inquiry step S320 in the first stage, the robot 200 captures the keywords of the data and files, and sends them to the third inquiry step S350 to inquire about the sources of these data and files, so as to further capture the source information and update the function map FM.

[0130] In the fourth stage, the generated function map FM is given to the user for confirmation. If there is no problem, it is the final function map FM, and the UI / UX designer will receive the function map FM for subsequent website design.

[0131] In an embodiment, the function map FM generated by the processor 110 can be imported into a prototype website development platform with AI function to generate the prototype website code. Then, the processor 110 packages and deploys the website code to the cloud environment through the compilation engine and records the browsing URL. Then, the robot 200 sends the browsing URL back to the user for the user to further browse the website screen. For example, after the robot 200 records the browsing URL, it encapsulates the browsing URL (digital signal) into a set of logical transmission data that conforms to the standard (i.e., data frame), and drives the communication chip or communication circuit to send the transmission data to the pre-specified electronic device related to the user through email, short message service, push technology, etc.

[0132] In summary, the system of the present invention can operate all-weather, improving the work efficiency of designers. In addition, the AI model adopted by the present invention has the ability of in-depth conversation and a user-friendly way of speaking, providing a better customer experience compared with traditional robots. Based on the professional knowledge of designers, the AI model can effectively capture key information. The proper noun understanding mechanism can quickly eliminate the problem of inconsistent cognition between both parties and improve the efficiency of requirement capture. By directly generating a functional map, designers can skip the initial requirement exploration phase and directly cut into clear requirements, accelerating the overall design and development process.

Claims

1. A method for generating a functional map based on an artificial intelligence model, which is executed by a processor and includes the following steps: A first query step, using the artificial intelligence model to obtain the difficulty point and a first piece of data related to the difficulty point from a first reply content corresponding to a first query corresponding to a difficulty point; A second query step, using the artificial intelligence model to obtain the usage target and a second piece of data related to the usage target from a second reply content corresponding to a second query corresponding to a usage target; A completeness judgment step, including: A function induction step, using the artificial intelligence model to obtain an implementation function and a target type based on the difficulty point and the usage target; A function association step, using the artificial intelligence model to obtain a required function based on the target type; and A function comparison step, using the artificial intelligence model to perform a semantic comparison on the implementation function and the required function to obtain a difference set content; and A functional map generation step, generating a functional map based on the target type, the implementation function, and the difference set content.

2. The functional map generation method according to claim 1, wherein the first query step includes: Generating a prompt corresponding to the first reply content; Executing a noun capture program by the artificial intelligence model based on the prompt corresponding to the first reply content to obtain a first noun response result; And Executing a semantic capture program by the artificial intelligence model to obtain the difficulty point and the first piece of data related to the difficulty point based on the first reply content and the first noun response result.

3. The functional map generation method according to claim 2, wherein the second query step includes: Generating a prompt corresponding to the second reply content; Executing the noun capture program by the artificial intelligence model based on the prompt corresponding to the second reply content to obtain a second noun response result from the second reply content; And Executing the semantic capture program by the artificial intelligence model to obtain the usage target and the second piece of data related to the usage target based on the second reply content and the second noun response result.

4. The functional map generation method according to claim 2, further including: A third query step, including: Receiving a third reply content corresponding to a third query corresponding to a data source; Generating a prompt corresponding to the third reply content; Executing the noun capture program by the artificial intelligence model based on the prompt corresponding to the third reply content to obtain a third noun response result from the third reply content; and Executing the semantic capture program by the artificial intelligence model to obtain a source information based on the third reply content and the third noun response result.

5. The functional map generation method according to claim 4, wherein the functional map generation step further includes: Updating the functional map based on the source information.

6. The functional map generation method according to claim 1, wherein the function induction step includes: Generating a prompt corresponding to the difficulty point and the usage target; And Use the artificial intelligence model to execute a function induction program based on the prompt word to obtain the implemented function and the target type.

7. The function map generation method according to claim 1, wherein the function association step includes: Generate a prompt word corresponding to the target type; And Use the artificial intelligence model to execute a function association program based on the prompt word to obtain a required function.

8. The function map generation method according to claim 1, wherein the function map generation step includes: After the function induction step, generate the function map based on the implemented function and the target type; And After the function comparison step, update the function map based on the difference set content.

9. A function map generation system based on an artificial intelligence model, comprising: An artificial intelligence model; An interactive interface; And A processor coupled to the artificial intelligence model and the interactive interface; Wherein, in response to the interactive interface receiving a first response content corresponding to a first inquiry about a difficulty point, the processor is configured to: use the artificial intelligence model to obtain the difficulty point and a first data related to the difficulty point from the first response content; In response to the interactive interface receiving a second response content corresponding to a second inquiry about a usage target, the processor is configured to: use the artificial intelligence model to obtain a second data representing the usage target and related to the usage target from the second response content; The processor is further configured to: Use the artificial intelligence model to obtain an implemented function and a target type based on the difficulty point and the usage target; Use the artificial intelligence model to obtain a required function based on the target type; Use the artificial intelligence model to perform semantic comparison between the implemented function and the required function to obtain a difference set content; and Generate a function map based on the target type, the implemented function, and the difference set content.

10. The function map generation system according to claim 9, wherein the processor is configured to: Generate a prompt word corresponding to the first response content; Execute a noun capture program based on the prompt word corresponding to the first response content through the artificial intelligence model to obtain a first noun response result; And Execute a semantic capture program through the artificial intelligence model to obtain the difficulty point and the first data related to the difficulty point based on the first response content and the first noun response result.

11. The function map generation system according to claim 10, wherein the processor is configured to: Generate a prompt word corresponding to the second response content; Execute the noun capture program based on the prompt word corresponding to the second response content through the artificial intelligence model to obtain a second noun response result from the second response content; And Execute the semantic capture program through the artificial intelligence model to obtain the usage target and the second data related to the usage target based on the second response content and the second noun response result.

12. The function map generation system according to claim 10, wherein the processor is configured to: Receive a third response content corresponding to a third inquiry about a data source; Generate a prompt corresponding to the third reply content; Execute the noun capture program by the artificial intelligence model based on the prompt corresponding to the third reply content to obtain a third noun response result from the third reply content; and Execute the semantic capture program by the artificial intelligence model to obtain a source information based on the third reply content and the third noun response result.

13. The functional map generation system according to claim 12, wherein the processor is configured to: Update the functional map based on the source information.

14. The functional map generation system according to claim 9, wherein the processor is configured to: Generate a prompt corresponding to the difficulty point and the usage target; and Use the artificial intelligence model to execute a function induction program based on the prompt to obtain the implemented function and the target type.

15. The functional map generation system according to claim 9, wherein the processor is configured to: Generate a prompt corresponding to the target type; and Use the artificial intelligence model to execute a function association program based on the prompt to obtain a required function.

16. The functional map generation system according to claim 9, wherein the processor is configured to: After obtaining the implemented function and the target type, generate the functional map based on the implemented function and the target type; and After obtaining the difference set content, update the functional map based on the difference set content.

Images