Virtual scene level generation method and device, electronic equipment, computer readable storage medium and computer program product

Abstract

The invention provides a level generation method and device of a virtual scene, electronic equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: screening at least one level word meeting a first generation requirement from a plurality of candidate words; performing splitting processing on each level word to obtain a plurality of split words of each level word; performing path exploration processing based on the plurality of split words of each level word to obtain a level of the virtual scene; and performing difficulty assessment processing based on the checkpoint to obtain a difficulty assessment result of the checkpoint. According to the invention, the level with controllable difficulty can be automatically generated, and the generation efficiency and quality of the virtual scene level can be effectively improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, electronic device, computer-readable storage medium, and computer program product for generating levels in a virtual scene. Background Technology

[0002] In related technologies, virtual scene levels on the market are mainly generated manually. This method is inefficient and the quality of generated levels is inconsistent, making it difficult to provide players with a continuous sense of novelty and challenge. As player demands increase and game content becomes more complex, manually generated virtual scene levels often fall short, making it difficult to ensure playability and challenge while generating levels on a large scale. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for generating virtual scene levels, which can automatically generate levels with controllable difficulty and effectively improve the generation efficiency and quality of virtual scene levels.

[0004] The technical solution of this application embodiment is implemented as follows:

[0005] This application provides a method for generating levels in a virtual scene, including:

[0006] Select at least one level word that meets the first generation requirement from multiple candidate words;

[0007] Each of the level words is split into multiple split words for each level word;

[0008] The virtual scene's levels are obtained by performing path exploration processing based on multiple split words of each level word;

[0009] The difficulty assessment of the level is performed to obtain the difficulty assessment result of the level.

[0010] This application provides a virtual scene level generation device, including:

[0011] The filtering module is used to select at least one level word from multiple candidate words that meets the first generation requirement;

[0012] The splitting module is used to split each of the level words to obtain multiple split words for each level word;

[0013] The exploration module is used to perform path exploration processing based on multiple split words of each level word to obtain the level of the virtual scene;

[0014] The evaluation module is used to perform difficulty evaluation processing based on the level to obtain the difficulty evaluation result of the level.

[0015] In the above scheme, the filtering module is further used to extract the difficulty value and theme from the first generated requirements; and to filter at least one level word that meets the difficulty value and the theme from multiple candidate words.

[0016] In the above scheme, the filtering module is further used to determine the word length positively correlated with the difficulty value and the word frequency negatively correlated with the difficulty value; determine the word type corresponding to the difficulty value, wherein different word types correspond to different number of splits; filter out multiple first words that meet the word length, word frequency and word type from multiple candidate words; and filter out level words that meet the theme from the multiple first words.

[0017] In the above scheme, the filtering module is further used to filter out second words that match the theme from the plurality of first words; and to delete words that match the risk type from the plurality of second words to obtain the checkpoint words.

[0018] In the above scheme, the splitting module is further used to identify the word type to which each level word belongs; and to perform the following processing for each word type: based on the splitting rules corresponding to the word type, the level words corresponding to the word type are split to obtain multiple split words of the level words.

[0019] In the above scheme, the splitting module is further configured to perform the following processing for each of the level words: classify the characters in the level words to obtain the predicted position type of each character; when the predicted position type indicates that the character belongs to the boundary character, split the level words with the character as the boundary to obtain multiple split words of the level words.

[0020] In the above scheme, the classification process is implemented by calling a word classification model. The splitting module is further used to classify each character sample in the word sample using the word classification model to obtain the predicted position type of each character sample; determine a first loss based on the actual position type of each character sample in each word sample and the predicted position type of each character sample in each word sample; and train the word classification model based on the first loss to obtain a trained word classification model.

[0021] In the above scheme, the exploration module is further used to explore the initial arrangement of the split words in the level to determine the word elimination path of the level; when the word elimination path of the level is unique, the initial arrangement of the split words in the level is used as the level of the virtual scene; when the word elimination path of the level is not unique, the initial arrangement of the split words in the level is adjusted, and the path exploration is repeated based on the adjusted arrangement until the word elimination path of the level is unique, and the adjusted arrangement is used as the level of the virtual scene.

[0022] In the above scheme, the evaluation module is further configured to determine a first difficulty value of the level based on the word frequency of the level words; determine a second difficulty value of the level based on the number of multiple sub-words in the level; determine a third difficulty value of the level based on the number of level words in the level; determine a fourth difficulty value of the level based on the positional arrangement of multiple sub-words of each level word; merge the first difficulty value, the second difficulty value, the third difficulty value, and the fourth difficulty value, and use the fusion result as the difficulty evaluation result of the level.

[0023] This application provides an electronic device, including:

[0024] Memory is used to store executable instructions or computer programs.

[0025] The processor, when executing computer-executable instructions or computer programs stored in the memory, implements the virtual scene level generation method provided in the embodiments of this application.

[0026] This application provides a computer-readable storage medium storing a computer program or computer-executable instructions, which, when executed by a processor, implements the virtual scene level generation method provided in this application.

[0027] This application provides a computer program product, including a computer program or computer executable instructions. When the computer program or computer executable instructions are executed by a processor, they implement the virtual scene level generation method provided in this application.

[0028] The embodiments of this application have the following beneficial effects:

[0029] First, select level words that meet the initial generation requirements from multiple candidate words to ensure that the selected level words satisfy the initial generation requirements. Then, break down each level word to obtain multiple sub-words, providing the smallest operable unit for subsequent path exploration. Next, perform path exploration based on the multiple sub-words of each level word to obtain the virtual scene level, thereby enhancing the logical depth and strategy of the game. Finally, evaluate the difficulty of the level to obtain the difficulty evaluation result, thereby ensuring the balance and challenge of the game experience. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the architecture of the virtual scene level generation system provided in the embodiments of this application;

[0031] Figure 2 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application;

[0032] Figure 3 This is a flowchart illustrating the method for generating virtual scene levels according to an embodiment of this application;

[0033] Figure 4 This is a flowchart illustrating the method for generating virtual scene levels according to an embodiment of this application;

[0034] Figure 5 This is a flowchart illustrating the method for generating virtual scene levels according to an embodiment of this application;

[0035] Figure 6 This is a schematic diagram of the word elimination process in the monthly theme level provided in this application embodiment;

[0036] Figure 7 This is a schematic diagram of the interface of the shiny theme level provided in the embodiment of this application. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0038] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0039] It is understood that in the embodiments of this application, data such as user information are involved. When the embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with relevant laws, regulations and standards.

[0040] In this application embodiment, the terms "module" or "unit" refer to a computer program or part of a computer program that has a predetermined function and works with other related parts to achieve a predetermined goal, and can be implemented wholly or partially using software, hardware (such as processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that includes the functionality of that module or unit.

[0041] In the following description, the terms “first, second, ...” are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that “first, second, ...” may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0042] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0043] Before providing a further detailed description of the embodiments of this application, the nouns and terms involved in the embodiments of this application will be explained, and the nouns and terms involved in the embodiments of this application shall be interpreted as follows.

[0044] 1) Word splitting: Word splitting refers to using lexical analysis technology to decompose complex words into multiple word roots, affixes or common letter combinations, thereby forming split word combinations. Each split word combination is embedded in a Tile block, and the standard size of the block is 2×2.

[0045] 2) Stacking: Stacking refers to forming a tile stacking area based on split words and combinations, which allows players to interact. This design can optimize space utilization and enhance the player's interactive experience. The hidden blocks show the back and hide all the letters, increasing the challenge of the game.

[0046] 3) Level Theme: The level theme is a preset theme in the level design. It can be used to guide word selection and ensure that all words are related to the theme, which can increase the fun and educational value of the game.

[0047] 4) Two-part or three-part words: In a level, words that are split into two parts are called two-part words, and words that are simultaneously split into three parts are called three-part words.

[0048] 5) Word Elimination: When players click to split and combine words to construct words that match the level's theme, the corresponding tile blocks are removed, revealing the hidden blocks.

[0049] 6) Monte Carlo simulation: Monte Carlo simulation is a randomized statistical method applied to optimization algorithms. By exploring a large number of random nodes and paths, combined with dynamic weight adjustment, it can quickly approach the optimal solution and provide a highly reliable difficulty assessment.

[0050] In related technologies, virtual scene levels on the market are mainly generated manually. This method is inefficient and the quality of generated levels is inconsistent, making it difficult to provide players with a continuous sense of novelty and challenge. As player demands increase and game content becomes more complex, manually generated virtual scene levels often fall short, making it difficult to ensure playability and challenge while generating levels on a large scale.

[0051] Based on this, embodiments of this application provide a method, apparatus, electronic device, computer-readable storage medium, and computer program product for generating virtual scene levels. These methods enable automated generation of levels with controllable difficulty, effectively improving the efficiency and quality of virtual scene level generation. The following describes exemplary applications of the electronic devices provided in this application. These devices can be implemented as various types of terminal devices such as laptops, tablets, desktop computers, set-top boxes, smartphones, smart speakers, smartwatches, smart TVs, and in-vehicle terminals, or as servers.

[0052] See Figure 1 , Figure 1 This is a schematic diagram of the architecture of a virtual scene level generation system provided in an embodiment of this application. To support a level generation application for a virtual scene, such as... Figure 1 As shown, the virtual scene level generation system 100 includes: server 200, network 300, terminal 400, and database 600. Database 600 is connected to server 200, and terminal 400 is connected to server 200 through network 300. Network 300 can be a local area network or a wide area network, or a combination of both.

[0053] In some embodiments, a first generation requirement is input through a terminal 400 and transmitted to a server 200 via a network 300. The server 200 filters multiple candidate words in a database 600 based on the first generation requirement to obtain at least one level word. Subsequently, the server 200 generates a level of a virtual scene based on the at least one level word. After that, the server 200 sends the level of the virtual scene to the terminal 400 via the network 300 and displays it on the client of the terminal 400.

[0054] The virtual scene level generation method provided in this application can be applied to various scenarios that require virtual scene level generation, such as the following scenarios:

[0055] (1) In a children’s literacy app, assuming that the first generation requirement is limited to nouns learned before the second grade of primary school, the server selects multiple words learned before the second grade from multiple candidate words stored in the database as level words, such as “apple”, “schoolbag”, “sun”, etc., and generates corresponding levels based on the selected level words, thereby enhancing the fun and educational significance of the game.

[0056] (2) In the English word elimination mobile game, assuming that the first generation requirement is limited to TOEFL core vocabulary with a length of 5-7 letters, the server selects multiple TOEFL core words with a length of 5-7 letters from the English words stored in the database as level words, such as "planet", "author", "vivid", etc., and generates corresponding elimination levels based on the selected level words, thereby enhancing the fun and educational significance of the game.

[0057] Example, Figure 1 The server 200 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. The terminal 400 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, in-vehicle terminal, etc., but is not limited to these. The terminal 400 and server 200 can be directly or indirectly connected via wired or wireless communication, which is not limited in this embodiment.

[0058] The structure of the electronic device provided in the embodiments of this application will be further described below. Taking the electronic device as a terminal as an example, see... Figure 2 , Figure 2This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Figure 2 The illustrated electronic device 500 includes at least one processor 510, a memory 550, at least one network interface 520, and a user interface 530. The various components in the electronic device 500 are coupled together via a bus system 540. It is understood that the bus system 540 is used to implement communication between these components. In addition to a data bus, the bus system 540 also includes a power bus, a control bus, and a status signal bus. However, for clarity, ... Figure 2 The general labeled all buses as Bus System 540.

[0059] The processor 510 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0060] User interface 530 includes one or more output devices 531 that enable the presentation of media content, including one or more speakers and / or one or more visual displays. User interface 530 also includes one or more input devices 532, including user interface components that facilitate user input, such as a keyboard, mouse, microphone, touch screen display, camera, other input buttons and controls.

[0061] The memory 550 may be removable, non-removable, or a combination thereof. Exemplary hardware devices include solid-state storage, hard disk drives, optical disk drives, etc. The memory 550 may optionally include one or more storage devices physically located away from the processor 510.

[0062] The memory 550 may include volatile memory or non-volatile memory, or both. The non-volatile memory may be read-only memory (ROM), and the volatile memory may be random access memory (RAM). The memory 550 described in this application embodiment is intended to include any suitable type of memory.

[0063] In some embodiments, memory 550 is capable of storing data to support various operations, examples of which include programs, modules, and data structures or subsets or supersets thereof, as illustrated below.

[0064] Operating system 551 includes system programs for handling various basic system services and performing hardware-related tasks, such as the framework layer, core library layer, driver layer, etc., for implementing various basic business functions and handling hardware-based tasks;

[0065] The network communication module 552 is used to reach other computing devices via one or more (wired or wireless) network interfaces 520, exemplary network interfaces 520 including: Bluetooth, WiFi, and Universal Serial Bus (USB), etc.

[0066] Presentation module 553 is configured to enable the presentation of information (e.g., a user interface for operating peripheral devices and displaying content and information) via one or more output devices 531 (e.g., a display screen, a speaker, etc.) associated with user interface 530;

[0067] The input processing module 554 is used to detect and translate one or more user inputs or interactions from one or more input devices 532.

[0068] In some embodiments, the apparatus provided in this application can be implemented in software. Figure 2 A level generation device 555 for a virtual scene stored in memory 550 is shown. It can be software in the form of programs and plug-ins, including the following software modules: a filtering module 5551, a splitting module 5552, an exploration module 5553, and an evaluation module 5554. These modules are logical and can therefore be arbitrarily combined or further split according to the functions they implement.

[0069] The method for generating virtual scene levels provided in this application will be specifically described below with reference to exemplary applications and implementations of the terminals provided in the embodiments of this application.

[0070] See Figure 3 , Figure 3 This is a flowchart illustrating the method for generating virtual scene levels according to an embodiment of this application, which will be combined with... Figure 3 The steps shown are explained.

[0071] It should be noted that, Figure 3 The methods illustrated can be executed by various forms of computer programs running on the terminal, and are not limited to clients. For example, they can also be the operating systems, software modules, scripts, and applets mentioned above. Therefore, the client-side examples used below should not be considered as limiting the embodiments of this application. Furthermore, for ease of description, no specific distinction will be made between the terminal and the client running on the terminal below.

[0072] In step 101, at least one level word that meets the first generation requirement is screened out from multiple candidate words.

[0073] As an example, assume that the multiple candidate words include bat, jade, apple, sun, kitten, etc., and the first generation requirement is common nouns for those below the second grade. According to the first generation requirement for screening, at least one level noun that meets the first generation requirement can include apple, sun, kitten, etc.; assume that the multiple candidate words include tiger, lion, panda, elephant, sun, kitten, puppy, apple, etc., and the first generation requirement is nouns related to animals. According to the first generation requirement for screening, at least one level noun that meets the first generation requirement can include tiger, lion, elephant, kitten, puppy, etc.

[0074] In some embodiments, refer to Figure 4 , Figure 4 which is a schematic flowchart of a method for generating levels of a virtual scene provided by an embodiment of the present application. As shown in Figure 4 shown, Figure 3 step 101 shown can be implemented through Figure 4 steps 1011 to 1012 shown, and will be described in combination with Figure 4 the steps shown.

[0075] In step 1011, the difficulty value and the theme are extracted from the first generation requirement.

[0076] It should be noted that the difficulty value is used to characterize the difficulty level of the levels of the generated virtual scene. It can be measured by the time required for a user to complete the levels of the virtual scene. The longer the time required to complete a level, the greater the difficulty value of the level, that is, the more difficult the level of the generated virtual scene; it can also be measured by the number of specified words. For example, if the specified words include words with greater difficulty such as taotie and maodie, the more the number of specified words included in the levels of the generated virtual scene, the greater the difficulty value of the level, that is, the more difficult the level of the generated virtual scene. It can also be measured according to the usage frequency level of words. For example, the usage frequency of daily nouns such as rain and sunny day is relatively high, and the usage frequency of proper nouns such as mosquito larvae is relatively low. Then, the higher the average usage frequency of the nouns included in the levels of the generated virtual scene, the smaller the difficulty value of the level, that is, the simpler the level of the generated virtual scene. The theme is used to characterize the type of level words in the levels of the generated virtual scene. For example, if the theme is animals, then the level words in the levels of the virtual scene are words related to animals such as lion and tiger.

[0077] As an example, when the content of the first generation request includes a difficulty value and a theme, these can be directly extracted from the content of the first generation request. For example, if the first generation request is that the difficulty value of the level is 3 and the theme is words related to fruit, then the extracted difficulty value is 3 and the theme is words related to fruit. When the content of the first generation request does not include a difficulty value and a theme, then a mapping process can be performed based on the content of the first generation request to obtain the difficulty value and theme corresponding to the first generation request. For example, if the first generation request is common nouns for students below the second grade, through a preset mapping relationship, the difficulty value corresponding to the first generation request can be obtained as 2, and the theme is words related to fruit or animals consisting of 2-3 characters.

[0078] In step 1012, at least one level word that meets the difficulty value and the theme is selected from multiple candidate words.

[0079] As an example, let's take the English word filtering scenario as an example. Suppose the difficulty value is 3, which means filtering words consisting of 5-7 letters. The theme is TOEFL core words. Candidate words include "tiktok", "deepfake", "planet", "author", "vivid", etc. Through filtering, we can get level words that meet the difficulty value and theme, including "planet", "author", "vivid", etc.

[0080] In some embodiments, see Figure 5 , Figure 5 This is a flowchart illustrating the method for generating virtual scene levels according to an embodiment of this application, as shown below. Figure 5 As shown, Figure 4 Step 1012 shown can be achieved through... Figure 5 Steps 10121 to 10124 shown are implemented by combining Figure 5 The steps shown are explained.

[0081] In step 10121, the word lengths that are positively correlated with the difficulty value and the word frequencies that are negatively correlated with the difficulty value are determined.

[0082] It should be noted that word frequency is used to represent how often a word is used. The higher the word frequency, the more common the word is, and the lower the difficulty value of the word. The lower the word frequency, the less common the word is, and the higher the difficulty value of the word. For example, the word frequency of "sun" is relatively high, so the difficulty value of "sun" is relatively low, while the word frequency of "photosynthesis" is relatively low, so the difficulty value of "photosynthesis" is relatively high.

[0083] As an example, assuming the mapping relationship between difficulty value x and word length L is L = 2x, then when the extracted difficulty value is 3, the word length can be calculated to be 6, meaning it contains 6 English words; assuming the mapping relationship between difficulty value x and word frequency w is... When the extracted difficulty value is 2, the word frequency can be obtained as 0.05 according to the mapping relationship, which means that the English word appears 5 times in every 100 words.

[0084] In step 10122, the word type corresponding to the difficulty value is determined.

[0085] Here, different word types correspond to different numbers of splits.

[0086] As an example, suppose the word type A corresponding to a difficulty value of 1 has 2 word segments, the word type B corresponding to a difficulty value of 2 has 3 word segments, and the word type C corresponding to a difficulty value of 3 has 4 word segments. When the extracted difficulty value is 2, the corresponding word type B can be obtained.

[0087] In step 10123, multiple first words that meet the word length, word frequency, and word type are selected from multiple candidate words.

[0088] As an example, suppose candidate word A has a word length of 6, a word frequency of 0.03, and a word type splitting quantity of 3; candidate word B has a word length of 7, a word frequency of 0.05, and a word type splitting quantity of 2; candidate word C has a word length of 5, a word frequency of 0.04, and a word type splitting quantity of 2; and candidate word D has a word length of 8, a word frequency of 0.1, and a word type splitting quantity of 2. The filtering range for word length is 5-7, the filtering range for word frequency is 0.02-0.06, and the word type splitting quantity is 2-3. Through filtering, candidate words A, B, and C meet the filtering requirements, thus obtaining the first words A, B, and C.

[0089] In step 10124, checkpoint words that match the theme are selected from the plurality of first words.

[0090] Continuing with the example above, assuming the theme of the first word A is animals, the theme of the first word B is animals, and the theme of the first word C is fruits, and the filter range corresponding to the theme is animals, then the level words that match the theme are A and B.

[0091] In some embodiments, step 10124 described above can also be implemented by: selecting second words that conform to the theme from the plurality of first words; deleting words that conform to the risk type from the plurality of second words to obtain the checkpoint words. Thus, by using a dual screening process to delete second words that do not conform to the theme and words that conform to the risk type, the obtained checkpoint words are ensured to simultaneously meet the theme relevance and security requirements, reducing manual review costs and significantly lowering the compliance risk of the checkpoint.

[0092] As an example, suppose the theme of the first word A is animals, the theme of the first word B is animals, and the theme of the first word C is fruit. The filter range corresponding to the theme is animals. We can get the second words A and B that match the theme. However, the second word B is a risky word. Therefore, we delete the second word B and get the level word A, that is, the final level word A that matches the theme.

[0093] In this way, by quantifying the word length, word frequency, and word type corresponding to the difficulty value in the filtering range, candidate words can be accurately filtered to obtain words that meet the filtering range. Then, words that meet the filtering range are filtered again according to the theme type to obtain level words that meet the theme type. By using a multi-layer filtering mechanism, efficient and accurate level word filtering is achieved, which can ensure the quality of level word filtering.

[0094] In this way, by extracting the difficulty value and theme according to the generation requirements, the difficulty of the level to be generated can be scientifically determined. Then, by filtering out level words that match the difficulty value and theme from multiple candidate words, the complexity of the level can be precisely controlled to ensure that the difficulty value of the generated level is consistent with the difficulty value corresponding to the generation requirements.

[0095] In step 102, each of the level words is split into multiple split words to obtain each level word.

[0096] It should be noted that there are multiple methods for splitting words in a level. These methods include splitting based on syllables, splitting based on word roots and affixes, splitting based on common spellings, and splitting by evenly dividing the words into segments. The specific method can be determined based on the actual situation and is not specifically limited here.

[0097] As examples, taking word segmentation based on syllables as an example, "magical" can be segmented into three words: "ma", "gi", and "cal"; taking word segmentation based on roots and affixes as an example, "unbreakable" can be segmented into three words: "un", "break", and "able"; taking word segmentation based on common spelling words as an example, "station" can be segmented into two words: "sta" and "tion"; and taking word segmentation based on evenly divided segments as an example, "potion" can be segmented into three words: "po", "ti", and "on".

[0098] In some embodiments, step 102 described above can also be implemented by: identifying the word type to which each level word belongs; and performing the following processing for each word type: based on the splitting rules corresponding to the word type, splitting the level word corresponding to the word type to obtain multiple split words of the level word. Thus, performing different splitting processes for different word types can improve the rationality of the splitting process, and by adopting the method of first identifying the word type and then performing splitting processing according to the type, brute-force word splitting can be avoided, reducing computational costs.

[0099] As an example, suppose the word "artificial intelligence" belongs to the word type "compound new scientific term", and the splitting rule for the word type "compound new scientific term" is "split into 'modifier' + 'central noun'". According to the splitting rule, the word "artificial intelligence" is split into two sub-words: "artificial" and "intelligent". Suppose the word "Peking University" belongs to the word type "property name", and the splitting rule for the word type "property name" is "split into 'regional proper name' + 'generic name'". According to the splitting rule, the word "Peking University" is split into two sub-words: "Beijing" and "university".

[0100] In some embodiments, step 102 described above can also be implemented as follows: For each level word, perform the following processing: classify the characters in the level word to obtain the predicted position type of each character; when the predicted position type indicates that the character belongs to a boundary character, split the level word using the character as a boundary to obtain multiple split words of the level word. Thus, through character boundary prediction, the internal structure of compound words can be accurately captured, transforming the rule dependency of word splitting processing into a data dependency, thereby achieving accurate word splitting.

[0101] It should be noted that the boundary character refers to the boundary for splitting words. For example, in the word "deep learning", the character "学" is the boundary character. Then, splitting is performed with the character "学" as the boundary, and the split words obtained are "深度" and "学习".

[0102] As an example, assume the checkpoint word is "anti - quantum encryption algorithm". Classifying each character in the checkpoint word can obtain the predicted position type of each character. That is, "抗" is a non - boundary character, "量" is a non - boundary character, "子" is a non - boundary character, "加" is a boundary character, "密" is a boundary character, "算" is a boundary character, "法" is a non - boundary character. According to the predicted position types of each character, splitting the checkpoint word can obtain the split words of the checkpoint word as "抗量子", "加密", and "算法".

[0103] In some embodiments, the classification process is implemented by calling a word classification model. Before performing the above - mentioned classification process on the characters in the checkpoint word to obtain the predicted position type of each character, the following processes can also be performed: Classifying each character sample in the word sample through the word classification model to obtain the predicted position type of each character sample; determining the first loss based on the actual position type of each character sample in each word sample and the predicted position type of each character sample in each word sample; training the word classification model based on the first loss to obtain a trained word classification model. In this way, through the model training method, the position type of characters in the checkpoint word can be accurately predicted, and based on the position type of characters, the checkpoint word can be automatically split, which can effectively reduce labor costs and improve the splitting efficiency of the checkpoint word.

[0104] It should be noted that the way to determine the first loss can be calculated through specific loss functions, such as cross - entropy loss function, mean - square error loss function, mean absolute error loss function, etc.

[0105] As an example, taking the cross - entropy loss function as an example, using the word classification model to classify each character sample can obtain the probability distribution of the predicted position type of each character sample. The first loss is calculated through formula (1), and the parameters of the word classification model are updated based on the first loss to obtain a trained word classification model.

[0106]

[0107] where, y i represents the actual position type of the i - th character sample, and represents the probability distribution of the predicted position type of the i - th character sample.

[0108] In practical applications, when the first loss reaches the parameter threshold, backpropagation is performed in the word classification model based on the first loss, and the model parameters of the word classification model are updated during the propagation process.

[0109] Here's an explanation of backpropagation: Character samples are input into the input layer of the word classification model, pass through the hidden layer, and finally reach the output layer to output the result. This is the forward propagation process of the word classification model. Since there is an error between the output result of the word classification model and the actual result, the error between the calculated result and the actual value is propagated back from the output layer to the hidden layer until it reaches the input layer. During the backpropagation process, the values ​​of the model parameters are adjusted according to the error. This process is iterated until convergence.

[0110] Taking the loss function (1) of the word classification model in this application embodiment as an example, the server determines the first loss based on the loss function, and propagates the first loss back from the output layer in the first model. The first loss is propagated back layer by layer. When the first loss reaches each layer, the gradient (that is, the partial derivative of the loss function with respect to the parameters of each layer) is solved in combination with the propagated first loss, and the gradient value of the parameters of each layer is updated.

[0111] In step 103, path exploration is performed based on multiple split words of each level word to obtain the level of the virtual scene.

[0112] It should be noted that path exploration can be implemented using specific algorithms such as recursive backtracking, or it can be implemented using large language model reasoning; no specific limitation is made here.

[0113] In some embodiments, step 103 described above can also be implemented as follows: Path exploration is performed on the initial arrangement of the split words in the level to determine the word elimination path of the level; when the word elimination path of the level is unique, the initial arrangement of the split words in the level is used as the level of the virtual scene; when the word elimination path of the level is not unique, the initial arrangement of the split words in the level is adjusted, and the path exploration is repeated based on the adjusted arrangement until the word elimination path of the level is unique, and the adjusted arrangement is used as the level of the virtual scene. Thus, by adjusting the arrangement of the split words in the level through path exploration, it can be ensured that the generated level is solvable and the solution for each word elimination is unique, thereby enhancing the logical depth and strategic nature of the generated level.

[0114] As an example, firstly, for the initial arrangement of split words in the current level, find all split words that can be combined in the initial arrangement, and select a set of split words that can be combined for elimination according to a preset weight (e.g., the length of the combined word). This results in the eliminated arrangement. Then, find all split words that can be combined in the eliminated arrangement. If there is at least one split word that can be combined, select a set of split words that can be combined for elimination according to a preset weight (e.g., the length of the combined word). This process is iterated until there are no split words that can be combined in the eliminated arrangement. The current elimination path is then recorded as a successful path, and a virtual scene level is generated based on the initial arrangement of split words corresponding to the current elimination path. If there are no split words that can be combined in the eliminated arrangement, the current elimination path is recorded as a failed path, and the initial arrangement of split words in the current level is adjusted. The above path exploration process is then performed on the adjusted arrangement until a successful path is found or all arrangement methods of split words have been traversed.

[0115] In step 104, a difficulty assessment is performed on the level to obtain the difficulty assessment result of the level.

[0116] It should be noted that the difficulty assessment results of a level can include the difficulty value of the level and the estimated completion time of the level, or it can only include the difficulty value of the level; automated playthroughs can be achieved through Monte Carlo simulations, and the difficulty assessment results can be obtained by analyzing the dynamic difficulty curve; alternatively, the difficulty assessment results can be obtained by statistically analyzing different influencing factors in the level according to preset rules.

[0117] In some embodiments, step 104 described above can also be implemented as follows: determining a first difficulty value for the level based on the word frequency of the level words; determining a second difficulty value for the level based on the number of multiple sub-words in the level; determining a third difficulty value for the level based on the number of level words in the level; determining a fourth difficulty value for the level based on the positional arrangement of multiple sub-words for each level word; and fusing the first, second, third, and fourth difficulty values, and using the fusion result as the difficulty assessment result for the level. Thus, by using word frequency, the number of sub-words, the number of level words, and the positional arrangement of sub-words, a precise assessment of level difficulty can be achieved from multiple dimensions, ensuring the balance and challenge of the game experience.

[0118] It should be noted that different difficulty values ​​can be combined using a weighted summation method.

[0119] As an example, count the word frequency 'a' of each level's words contained in the current level. i Calculate the average word frequency of the words in the current level. Average word frequency of words in the level The process involves mapping to obtain the first difficulty value D1 of the level; counting the number of split words n1 in the current level and mapping this number to obtain the second difficulty value D2; counting the number of level words n2 in the current level and mapping this number to obtain the third difficulty value D3; and based on the positional arrangement of the multiple split words corresponding to each level word in the current level, counting the number of sequentially arranged level words n3 and the number of non-sequentially arranged level words n4. The number of sequentially arranged level words (n3) and the number of unsequentially arranged level words (n4) are mapped to obtain the fourth difficulty value D4 of the level. Assuming the weight of the first difficulty value is w1, the weight of the second difficulty value is w2, the weight of the third difficulty value is w3, and the weight of the fourth difficulty value is w4, the first difficulty value D1, the second difficulty value D2, the third difficulty value D3, and the fourth difficulty value D4 are weighted and merged to obtain the overall difficulty value D of the level as w1*D1+w2*D2+w3*D3+w4*D4, thus obtaining the difficulty assessment result.

[0120] In some embodiments, automated trial play can also be achieved through Monte Carlo simulation, a dynamic difficulty curve can be constructed based on the data obtained from the simulated trial play, and the dynamic difficulty curve can be analyzed to obtain the evaluation result of the level difficulty.

[0121] As an example, firstly, a virtual player pool is built on the server side, including 500,000 virtual player profiles. Each profile includes the player's vocabulary level (e.g., basic, intermediate, advanced). Profiles are stored in a cache with serial numbers for the scheduler to randomly select. Next, 100,000 copies of the word elimination level to be tested are cloned. Each instance maintains the same layout, and all instances are mounted to distributed computing nodes. Nodes receive tasks through queues. Subsequently, the scheduler randomly selects virtual players from the pool and injects their profile information into the current level instance. The node then drives the instance to run until the termination condition is met. (Pass or fail); After all trials are completed, a difficulty curve is plotted based on all operation records for the same level; Finally, the level difficulty value is determined based on the pass rate of all virtual players in the "basic vocabulary" category. For example, if the pass rate is below 60%, the level is marked as "hard"; if the pass rate is between 60% and 80%, the level is marked as "medium"; and if the pass rate is above 80%, the level is marked as "easy". The median time taken by all successful pass records is rounded down as the estimated completion time of the level. The level difficulty value and the estimated completion time are combined to obtain the level difficulty assessment result.

[0122] The following describes an exemplary application of the embodiments of this application in a real-world application scenario. This exemplary application describes the specific implementation process of the virtual scene level generation method in a word game level generation scenario.

[0123] Word games are a major branch of casual games. In this era of rapid information technology development, traditional game level design methods mainly rely on manual editing. This method is not only inefficient, but also results in inconsistent level quality, making it difficult to provide players with a continuous sense of novelty and challenge. As player demands increase and game content becomes more complex, manually edited levels often prove insufficient, unable to guarantee playability and challenge while generating a large number of levels.

[0124] Based on this, this application provides a method for generating game levels in virtual scenes. Based on natural language processing and artificial intelligence methods, it can automatically generate game levels with clear solution paths and controllable difficulty, thereby improving the efficiency and quality of level generation and thus enhancing the player's gaming experience.

[0125] In some embodiments, firstly, based on a large-scale corpus, relevant words are selected and preprocessed, sensitive words are filtered out, and words that correspond to the level's theme and meet the difficulty requirements are selected according to a preset difficulty value. Pre-control is mainly performed from two dimensions: word length and word frequency. Here, a sensitive word list is maintained in the database, including multiple sensitive words, which are classified and graded. Sensitive word filtering is performed according to this list to ensure the game's educational and enjoyable aspects. Next, based on the preset level difficulty value, the number of specified words for each level, the range of word usage frequency levels (COCA values), the number of two- and three-part words, and different combinations of two- and three-part words are obtained through mapping, thereby precisely controlling the level's difficulty and complexity. Finally, a level map is generated based on the number of letter combinations and stacking complexity after word splitting. Here, the letter combinations after word splitting are fixed at 1 to 3 ways, mainly based on word roots, affixes, and common letter combinations for automated splitting and storage. During the level map generation process, axial symmetry, central symmetry, or combination symmetry can be set to ensure the playability and aesthetics of the level. Subsequently, words are selected and split through lexical analysis, and the split words are filled into the level structure to ensure the diversity and fun of the game. Then, with the help of artificial intelligence algorithms, word elimination paths are explored and adjusted to ensure that the solution path of the level is complete and the solution of each word combination is unique, which can greatly improve the challenge and fun of the game. Finally, Monte Carlo simulation is used for large-scale automated simulation to analyze the dynamic difficulty curve of the level, and finally evaluate the difficulty of the level and the expected completion time, thereby ensuring the balance and challenge of the game experience.

[0126] In some embodiments, the level generation system mainly includes a topic selection module, a lexical analysis module, a path exploration module, and an automated evaluation module. In the topic selection module, words matching the topic content are extracted and filtered from the corpus, and sensitive words are excluded to ensure that all selected words meet the requirements of the topic content. In the lexical analysis module, lexical analysis algorithms are used to split the selected words, and the split words are filled into the level structure to ensure the randomness and challenge of the game. In the path exploration module, artificial intelligence (AI) algorithms are used to explore and adjust paths. Here, path exploration refers to starting from the original level and performing weighted random attempts on all possible solutions (i.e., possible word combinations) of the current state of the level. The system iterates through the problem and recursively executes the process to ensure that the level is solvable and that each word has a unique solution, thereby enhancing the game's logical depth and strategic complexity. In the automated evaluation module, Monte Carlo simulation is used to assess and adjust the difficulty. The main influencing factors include the frequency of words that can be spelled out, the number of words after being split, the number of words that can be eliminated, and the placement of the split words. Among these factors, the higher the word frequency, the lower the level difficulty; the fewer the number of words after being split, the lower the level difficulty; the more words that can be eliminated, the lower the level difficulty; and the more sequentially arranged split words, the lower the level difficulty.

[0127] In some embodiments, see Figure 6 , Figure 6 This is a schematic diagram of the word elimination process for a month-themed level provided in this application embodiment, such as... Figure 6 As shown, the initial layout of the level includes two word blocks, "AP" and "RIL". After clicking on the two word blocks to find "APRIL", more word blocks will be displayed. Click on the word blocks one by one to eliminate them until there are no more word blocks in the level, thus completing the current level.

[0128] In some embodiments, see Figure 7 , Figure 7 This is a schematic diagram of the interface of the shiny theme level provided in the embodiment of this application, such as... Figure 7 As shown, the initial layout of the level includes six word blocks: "DIA", "RL", "PEA", "CRYS", "MOND", and "TAL". Clicking on word blocks "DIA" and "MOND" will eliminate "DIAMOND", clicking on word blocks "PEA" and "RL" will eliminate "PEARL", and clicking on word blocks "CRYS" and "TAL" will eliminate "CRYSTAL", thus completing the word block layout of the current page. More word blocks will then be displayed, and they will be eliminated by clicking on them one by one until there are no more word blocks left in the level, thus completing the current level.

[0129] In summary, this application provides a method for generating virtual scene levels, which can automate level generation and difficulty assessment, significantly improve the efficiency and quality of game level design, and ensure the aesthetics of level shapes. By combining artificial intelligence and natural language processing technologies, it reduces the reliance on manual editing in game level design and can provide players with a progressive and varied gaming experience.

[0130] The following description continues to illustrate the exemplary structure of the virtual scene level generation device 555 provided in the embodiments of this application as a software module. In some embodiments, such as... Figure 2 As shown, the software modules in the level generation device 555 of the virtual scene stored in the memory 550 may include: a filtering module 5551, a splitting module 5552, an exploration module 5553, and an evaluation module 5554.

[0131] The filtering module 5551 is used to filter at least one level word that meets the first generation requirement from multiple candidate words; the splitting module 5552 is used to split each level word to obtain multiple split words for each level word; the exploration module 5553 is used to perform path exploration processing based on the multiple split words of each level word to obtain the level of the virtual scene; and the evaluation module 5554 is used to perform difficulty evaluation processing based on the level to obtain the difficulty evaluation result of the level.

[0132] In some embodiments, the filtering module 5551 is further configured to extract a difficulty value and a theme from the first generated requirements; and to filter at least one level word from a plurality of candidate words that matches the difficulty value and the theme.

[0133] In some embodiments, the filtering module 5551 is further configured to determine the word length positively correlated with the difficulty value and the word frequency negatively correlated with the difficulty value; determine the word type corresponding to the difficulty value, wherein different word types correspond to different number of splits; filter out multiple first words that meet the word length, the word frequency and the word type from multiple candidate words; and filter out level words that meet the theme from the multiple first words.

[0134] In some embodiments, the filtering module 5551 is further configured to filter out second words that conform to the theme from the plurality of first words; and delete words that conform to the risk type from the plurality of second words to obtain the checkpoint words.

[0135] In some embodiments, the splitting module 5552 is further configured to identify the word type to which each of the level words belongs; and perform the following processing for each word type: based on the splitting rules corresponding to the word type, split the level words corresponding to the word type to obtain multiple split words of the level words.

[0136] In some embodiments, the splitting module 5552 is further configured to perform the following processing for each of the level words: classify the characters in the level words to obtain the predicted position type of each character; when the predicted position type indicates that the character belongs to a boundary character, split the level words with the character as the boundary to obtain multiple split words of the level words.

[0137] In some embodiments, the classification process is implemented by calling a word classification model. The splitting module 5552 is further configured to classify each character sample in the word sample using the word classification model to obtain the predicted position type of each character sample; determine a first loss based on the actual position type of each character sample in each word sample and the predicted position type of each character sample in each word sample; and train the word classification model based on the first loss to obtain a trained word classification model.

[0138] In some embodiments, the exploration module 5553 is further configured to explore the initial arrangement of the split words in the level to determine the word elimination path of the level; when the word elimination path of the level is unique, the initial arrangement of the split words in the level is used as the level of the virtual scene; when the word elimination path of the level is not unique, the initial arrangement of the split words in the level is adjusted, and the path exploration is repeated based on the adjusted arrangement until the word elimination path of the level is unique, and the adjusted arrangement is used as the level of the virtual scene.

[0139] In some embodiments, the evaluation module 5554 is further configured to determine a first difficulty value of the level based on the word frequency of the level words; determine a second difficulty value of the level based on the number of multiple sub-words in the level; determine a third difficulty value of the level based on the number of level words in the level; determine a fourth difficulty value of the level based on the positional arrangement of multiple sub-words of each level word; merge the first difficulty value, the second difficulty value, the third difficulty value, and the fourth difficulty value, and use the fusion result as the difficulty evaluation result of the level.

[0140] It should be noted that the description of the apparatus in this application embodiment is similar to the description of the method embodiment above, and has similar beneficial effects as the method embodiment, therefore it will not be repeated. For any technical details not covered in the virtual scene level generation apparatus provided in this application embodiment, please refer to... Figure 3 , Figure 4 ,or Figure 5 The meaning is understood in accordance with the description of any of the accompanying drawings.

[0141] This application provides a computer program product, which includes a computer program or computer-executable instructions stored in a computer-readable storage medium. The processor of an electronic device reads the computer-executable instructions from the computer-readable storage medium and executes the computer-executable instructions, causing the electronic device to perform the virtual scene level generation method described above in this application.

[0142] This application provides a computer-readable storage medium storing computer-executable instructions or a computer program. When the computer-executable instructions or the computer program are executed by a processor, the processor will execute the virtual scene level generation method provided in this application. For example, ... Figure 3 , Figure 4 ,or Figure 5 The method for generating levels in a virtual scene is shown.

[0143] In some embodiments, the computer-readable storage medium may be a memory such as ferroelectric random access memory (FRAM), ROM, programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic surface memory, optical disc, or compact disc read-only memory (CD-ROM); or it may be a device that includes one or any combination of the above-mentioned memories.

[0144] In some embodiments, computer-executable instructions may take the form of programs, software, software modules, scripts, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as stand-alone programs or as modules, components, subroutines, or other units suitable for use in a computing environment.

[0145] As an example, computer-executable instructions may, but do not necessarily, correspond to files in a file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple co-located files (e.g., files that store one or more modules, subroutines, or code sections).

[0146] As an example, computer-executable instructions can be deployed to execute on a single electronic device, or on multiple electronic devices located at one location, or on multiple electronic devices distributed across multiple locations and interconnected via a communication network.

[0147] In summary, the embodiments of this application first select level words that meet the first generation requirements from multiple candidate words, ensuring that the selected level words can satisfy the first generation requirements; then, each level word is split into multiple sub-words, providing the smallest operable unit for subsequent path exploration; next, path exploration is performed based on the multiple sub-words of each level word to ensure that the level is solvable and the path is unique, resulting in a virtual scene level, thereby enhancing the logical depth and strategy of the game; finally, the difficulty of the level is evaluated to obtain the difficulty evaluation result, thereby ensuring the balance and challenge of the game experience.

[0148] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Any modifications, equivalent substitutions, and improvements made within the spirit and scope of this application are included within the scope of protection of this application.

Claims

1. A method for generating levels in a virtual scene, characterized in that, The method includes: Select at least one level word that meets the first generation requirement from multiple candidate words; Each of the level words is split into multiple split words for each level word; The virtual scene's levels are obtained by performing path exploration processing based on multiple split words of each level word; The difficulty assessment of the level is performed to obtain the difficulty assessment result of the level.

2. The method according to claim 1, characterized in that, The step of selecting at least one level word that meets the first generation requirement from multiple candidate words includes: Extract the difficulty value and theme from the first generated requirement; Select at least one level word from multiple candidate words that matches the difficulty value and the theme.

3. The method according to claim 2, characterized in that, The step of selecting at least one level word from multiple candidate words that matches the difficulty value and the theme includes: Determine the word lengths that are positively correlated with the difficulty value and the word frequencies that are negatively correlated with the difficulty value; Determine the word type corresponding to the difficulty value, wherein different word types correspond to different numbers of splits; Select multiple first words that meet the specified word length, word frequency, and word type from a pool of candidate words; Select the level words that match the theme from the plurality of first words.

4. The method according to claim 3, characterized in that, The step of selecting checkpoint words that match the theme from the plurality of first words includes: Select second words that match the theme from the plurality of first words; The checkpoint words are obtained by deleting words that match the risk type from the plurality of second words.

5. The method according to claim 1, characterized in that, The step of splitting each level term into multiple sub-terms for each level term includes: Identify the word type to which each of the level's words belongs; For each of the word types, the following processing is performed: based on the splitting rules corresponding to the word type, the level words corresponding to the word type are split to obtain multiple split words of the level words.

6. The method according to claim 1, characterized in that, The step of splitting each level term into multiple sub-terms for each level term includes: For each of the aforementioned level words, the following processing is performed: The characters in the checkpoint words are classified to obtain the predicted position type of each character; When the predicted position category indicates that the character belongs to the boundary character, the level word is split using the character as the boundary to obtain multiple split words of the level word.

7. The method according to claim 6, characterized in that, The classification process is implemented by calling a word classification model. Before classifying the characters in the level words to obtain the predicted position type of each character, the method further includes: The word classification model is used to classify each character sample in the word sample to obtain the predicted position type of each character sample; A first loss is determined based on the actual position type of each character sample in each word sample and the predicted position type of each character sample in each word sample; The word classification model is trained based on the first loss to obtain the trained word classification model.

8. The method according to claim 1, characterized in that, The path exploration process based on multiple segmented words of each level word to obtain the level of the virtual scene includes: Explore the initial arrangement of the split words in the level to determine the word elimination path of the level; When the word elimination path of the level is unique, the initial arrangement of the split words in the level is taken as the level of the virtual scene. When the word elimination path of the level is not unique, the initial arrangement of the split words in the level is adjusted, and the path exploration is repeated based on the adjusted arrangement until the word elimination path of the level is unique, and the adjusted arrangement is used as the level of the virtual scene.

9. The method according to claim 1, characterized in that, The process of performing a difficulty assessment based on the level to obtain the difficulty assessment result of the level includes: Based on the word frequency of the words in the level, the first difficulty value of the level is determined; The second difficulty value of the level is determined based on the number of multiple split words in the level; The third difficulty value of the level is determined based on the number of level words in the level. The fourth difficulty value of the level is determined based on the positional arrangement of multiple split words in each level's words. The first difficulty value, the second difficulty value, the third difficulty value, and the fourth difficulty value are merged, and the fusion result is used as the difficulty evaluation result of the level.

10. A virtual scene level generation device, characterized in that, The device includes: The filtering module is used to select at least one level word from multiple candidate words that meets the first generation requirement; The splitting module is used to split each of the level words to obtain multiple split words for each level word; The exploration module is used to perform path exploration processing based on multiple split words of each level word to obtain the level of the virtual scene; The evaluation module is used to perform difficulty evaluation processing based on the level to obtain the difficulty evaluation result of the level.

Images