Method for pricing game items and computing apparatus for performing the same

Abstract

The present invention relates to a method for determining the price of in-game items and a computing device for performing the same. The present invention relates to a method for determining the price of in-game items, wherein a merchant NPC (Non-player Character) is created to trade items within a game, and the merchant NPC predicts the price of a new item based on information about the new item that the player intends to trade, using an appraiser model.

Description

Technology Field

[0001] The present invention relates to the pricing of items by merchant NPCs (Non-player Characters) within a game. Background Technology

[0002] In games, item trading is a crucial element in enhancing gameplay and item utility. Therefore, game developers aim to increase this utility by implementing trading systems in open-world games.

[0003] However, merchant NPCs are limited to merely mediating item exchanges; interactions with players generally follow a script, and merchant NPCs buy and sell items at fixed prices. Consequently, a problem exists in the industry where conventional merchant NPCs are unable to adjust prices or communicate freely, performing only the role of presenting a fixed price list. Prior art literature

[0004] Republic of Korea Published Patent No. 10-2024-0044830 The problem to be solved

[0005] The present disclosure is derived in response to the aforementioned background technology and aims to enable negotiation between a merchant NPC and a player by utilizing a Large Language Model (LLM). means of solving the problem

[0006] According to one aspect of the present invention, as a means for achieving the technical problem described above, a method for determining the price of an item in a game is performed in a computing device having a memory that stores one or more programs executed by one or more processors, comprising: a step of creating a merchant NPC (Non-player Character) for trading items in a game; and a step in which the merchant NPC predicts the price of a new item based on information of the new item that a player intends to trade using an appraiser model.

[0007] According to one embodiment, the step of predicting the price of the new item comprises: a step of obtaining an item dataset including item description information and item price information crawled from one or more other games; a step of training the appraisal model to predict the price of the new item based on the item dataset; and a step of predicting the price of the new item that the player intends to trade with the NPC using the trained appraisal model; a method for determining the price of an in-game item.

[0008] According to one embodiment, the step of training the appraisal model comprises: a step of inputting the item dataset into the appraisal model; and a step of causing the appraisal model to learn the relationship between one or more item description information included in the item dataset and the price information of the item; a method for determining the price of an in-game item.

[0009] According to one embodiment, the step of predicting the price of the new item comprises: the step of obtaining descriptive information of the new item that the player intends to trade; and the step of the learned appraisal model predicting the price of the new item based thereon; a method for determining the price of an in-game item.

[0010] According to one embodiment, the step of predicting the price of the new item comprises: obtaining descriptive information of the new item that the player intends to trade; extracting one or more items similar to the item descriptive information of the new item from item datasets of one or more other games; sorting items containing item descriptive information similar to the new item in order of high similarity of strings based on the extracted items; and predicting the price of the new item based on the sorted item price information in the appraisal model; a method for determining the price of an in-game item.

[0011] According to one embodiment, the pricing step further comprises the step of having the merchant NPC negotiate the price of the new item with the player using the negotiator model based on the price of the new item predicted by the appraiser model; a method for pricing an in-game item.

[0012] According to one embodiment, the step of negotiating the price of the new item comprises: a step of creating a virtual NPC and a virtual player; a step of obtaining a negotiation dataset by generating a price negotiation conversation for an item in an item dataset crawled from any game between a virtual merchant NPC and a virtual player based on the negotiator model; a step of training the negotiator model to enable transaction negotiation of the new item based on the negotiation dataset; and a step in which the merchant NPC negotiates the transaction price of the new item with the player by means of the trained negotiator model; a method for determining the price of an in-game item.

[0013] According to one embodiment, the step of acquiring the negotiation dataset comprises: providing information on any negotiation item included in an item dataset crawled from any game to a virtual merchant NPC and a virtual player; inputting one or more negotiation strategy prompts pre-set in the negotiator model; inputting the price of the virtual negotiation item predicted by the appraisal model; inputting one or more previous conversation contents between the virtual merchant NPC and the virtual player as prompts; generating a conversation between the virtual merchant NPC and the virtual player through a negotiation simulation based on the negotiator model; and acquiring the generated conversation as the negotiation dataset for training the negotiator model; a method for determining the price of an in-game item.

[0014] According to one embodiment, the step of negotiating the transaction price comprises: a step of inputting the price of the new item predicted by the appraisal model; a step of inputting one or more previous conversation contents between the merchant NPC and the player as a prompt; a step of inputting item information of the new item that the player intends to trade; and a step of conducting the negotiation by the learned negotiator model; a method for determining the price of an in-game item.

[0015] According to one embodiment, the step of negotiating the transaction price comprises: inputting one or more negotiation strategy prompts pre-set in the negotiator model; inputting the price of the new item predicted by the appraisal model; inputting one or more previous conversation contents between the merchant NPC and the player as prompts; inputting information of the new item that the player intends to trade; and, based on the negotiation strategy prompts, causing the negotiator model to have the NPC merchant proceed with price negotiation with the player regarding the new item; a method for determining the price of an in-game item.

[0016] According to one embodiment, the negotiation strategy prompt comprises a first negotiation strategy prompt and a second negotiation strategy prompt, wherein the first negotiation strategy prompt comprises at least one of the following: a Liking strategy that builds a trust relationship by finding common ground or through praise; a Reciprocity strategy that utilizes the psychology of the other party to reciprocate when a favor is shown; a Social Proof strategy that gains trust by emphasizing the fact that others have also chosen this item; a Consistency strategy that induces maintaining consistency with previous behavior; an Authority strategy that gains trust by emphasizing expert opinions or authority; and a Scarcity strategy that stimulates the desire to purchase by emphasizing that the item is limited in quantity or difficult to obtain.

[0017] According to one embodiment, the second negotiation strategy prompt is a method for pricing in-game items, wherein the second negotiation strategy prompt includes at least one of the following: an Enjoyment strategy that emphasizes elements that enhance the game experience; a Character Competency strategy that emphasizes elements such as ability enhancement; a Visual Authority strategy that emphasizes elements that can decorate the appearance of a character or attract attention; and a Monetary Value strategy that induces purchase by emphasizing price-to-performance ratio.

[0018] According to one embodiment, the pricing method further includes a step of evaluating whether the negotiation result of a new item through the negotiator model is appropriate, wherein the evaluation of appropriateness evaluates the performance of the negotiator model by at least one indicator among a persuasiveness evaluation, which is the ability of a merchant NPC to persuade a player; a dominance evaluation, which is the degree of power held; and an agreement evaluation, which is the rate at which an agreement is successfully concluded.

[0019] According to another aspect of the present invention, a computing device comprising: a processor; and a memory storing one or more programs executed by the processor, wherein the processor performs the operations of: generating a merchant NPC (Non-player Character) for trading items within a game; causing the merchant NPC to predict the price of a new item based on information of the new item that a player intends to trade using an Appraiser Model; and causing the merchant NPC to conduct a negotiation with the player regarding the price of the new item using a Negotiator Model.

[0020] According to another aspect of the present invention, a computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises one or more instructions, and when the instructions are executed by a computing device having one or more processors, the computing device performs the steps of: generating a merchant NPC (Non-player Character) for trading an item in a game; causing the merchant NPC to predict the price of a new item based on information of the new item that a player intends to trade using an Appraiser Model; and causing the merchant NPC to conduct a negotiation with the player regarding the price of the new item using a Negotiator Model. Effects of the invention

[0021] According to one embodiment, a method can be provided to generate a merchant NPC capable of interacting with a player through an appraiser model and a negotiator model based on a Large Language Model (LLM). Brief explanation of the drawing

[0022] The drawings attached below for use in describing embodiments of the present invention are merely some of the embodiments of the present invention, and other drawings can be obtained based on these drawings without effort to arrive at a novel invention by a person skilled in the art to which the present invention pertains (hereinafter referred to as "person skilled in the art"). FIG. 1 is a block diagram showing the configuration of a computing device according to one embodiment of the present invention. FIG. 2 is a flowchart regarding item pricing by a merchant NPC (Non-player Character) according to one embodiment of the present invention. FIG. 3 is a flowchart regarding item price prediction by an emotion model according to one embodiment of the present invention. FIG. 4 is a flowchart regarding the progress of item price negotiation by a negotiator model according to one embodiment of the present invention. FIG. 5 is a diagram showing examples of 10 negotiation strategies for proceeding with negotiation in a negotiator model according to one embodiment of the present invention. FIG. 6 is a diagram relating to an example of generating a negotiation dataset by at least one of a negotiation strategy according to one embodiment of the present invention. FIG. 7 is a diagram illustrating examples of conventional passive interaction, active price prediction, and active negotiation according to one embodiment of the present invention. Specific details for implementing the invention

[0023] The following detailed description is provided to facilitate a comprehensive understanding of the methods, devices, and / or systems described herein. However, this is merely illustrative and the invention is not limited thereto.

[0024] In describing the embodiments of the present invention, detailed descriptions of known technologies related to the present invention are omitted if it is determined that such descriptions may unnecessarily obscure the essence of the invention. Furthermore, the terms described below are defined in consideration of their functions within the present invention, and these definitions may vary depending on the intentions or practices of the user or operator. Therefore, such definitions should be based on the content throughout this specification. Terms used in the detailed description are intended merely to describe the embodiments of the present invention and should not be limiting in any way. Unless explicitly stated otherwise, expressions in the singular form include the meaning of the plural form. In this description, expressions such as "include" or "comprise" are intended to refer to certain characteristics, numbers, steps, actions, elements, parts thereof, or combinations thereof, and should not be interpreted to exclude the existence or possibility of one or more other characteristics, numbers, steps, actions, elements, parts thereof, or combinations thereof other than those described.

[0025] Additionally, terms such as "first," "second," etc., may be used to describe various components, but the components should not be limited by these terms. The terms may be used for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component.

[0026] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings attached to this specification.

[0027] FIG. 1 is a block diagram showing the configuration of a computing device for pricing in-game items according to one embodiment.

[0028] The computing device (12) includes a processor (14) that executes an appraisal model and a negotiator model. Through the processor (14), the merchant NPC can predict the appraisal value of a new item that a player intends to trade, and can trade the new item through negotiation based on the predicted appraisal value. The appraisal model and the negotiator model operate using a language model, and details regarding this will be described later.

[0029] The computing device (12) can receive or output conversation content from a player who wants to trade with a merchant NPC for a new item through an input / output device (24).

[0030] In the illustrated embodiment, each component may have different functions and capabilities in addition to those described below, and may include additional components in addition to those described below. The illustrated computing environment (10) includes a computing device (12). In one embodiment, the computing device (12) may perform a method for pricing in-game items.

[0031] The computing device (12) includes at least one processor (14), a computer-readable storage medium (16), and a communication bus (18). The processor (14) may enable the computing device (12) to operate according to exemplary embodiments described below. For example, the processor (14) may execute one or more programs stored in the computer-readable storage medium (16). The one or more programs may include one or more computer-executable instructions, and the computer-executable instructions may be configured to enable the computing device (12) to perform operations according to exemplary embodiments when executed by the processor (14).

[0032] A computer-readable storage medium (16) is configured to store computer-executable instructions or program code, program data and / or other suitable forms of information. A program (20) stored in the computer-readable storage medium (16) includes a set of instructions executable by a processor (14). In one embodiment, the computer-readable storage medium (16) may be memory (volatile memory such as random access memory, non-volatile memory, or a suitable combination thereof), one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or other forms of storage media that are accessed by a computing device (12) and capable of storing desired information, or a suitable combination thereof.

[0033] The communication bus (18) interconnects various other components of the computing device (12), including the processor (14) and the computer-readable storage medium (16).

[0034] The computing device (12) may also include one or more input / output interfaces (22) and one or more network communication interfaces (26) that provide interfaces for one or more input / output devices (24). The input / output interfaces (22) and network communication interfaces (26) are connected to a communication bus (18). The input / output devices (24) may be connected to other components of the computing device (12) through the input / output interfaces (22). An exemplary input / output device (24) may include an input device such as a pointing device (such as a mouse or trackpad), a keyboard, a touch input device (such as a touchpad or touchscreen), a voice or sound input device, various types of sensor devices and / or imaging devices, and / or an output device such as a display device, a printer, a speaker and / or a network card. An exemplary input / output device (24) may be included inside the computing device (12) as a component constituting the computing device (12), or it may be connected to the computing device (12) as a separate device distinct from the computing device (12).

[0035] FIG. 2 is a flowchart regarding item pricing by a merchant NPC (Non-player Character) according to one embodiment of the present invention.

[0036] Referring to FIG. 2, a method for determining item prices by merchant NPCs (Non-player Characters) in a computing device (12) may include the step of the computing device (12) creating merchant NPCs (Non-player Characters) (S210), the step of the merchant NPC appraising the price of a new item that a player wishes to trade using an appraisal model (S220), the step of the merchant NPC negotiating the price of a new item with a player using a negotiator model based on the appraised price of the new item (S230), and the step of the negotiation being concluded and the price of the new item that the player wishes to trade being determined (S240).

[0037] Below, we will examine each step of Figure 2.

[0038] In step S210, the computing device (12) can generate a merchant NPC (Non-Player Character, NPC). The merchant NPC generated by the computing device (12) to trade items within the game may refer to a non-player character (Non-Player Character, NPC) that can buy and sell items within the game.

[0039] As an example, a merchant NPC may sell various items such as weapons, armor, potions, and materials to a player, or the merchant NPC may purchase items from a player. A step in which the merchant NPC actively predicts the price regarding the sale or purchase of items and negotiates with the player based on the predicted price to determine the item price will be described later.

[0040] In step S2220, the processor (14) of the computing device (12) can execute an appraisal model, and through the appraisal model, the merchant NPC can predict the transaction price for the item that the player intends to trade with the merchant NPC. The appraisal model can be designed to interpret item description information and predict the price of the item by utilizing a language model. The appraisal model can receive item description information in the form of natural language sentences as input, which includes information such as the item's required level, effect, and durability. Since this item description information may be expressed differently in each game, a natural language expression method can be selected to ensure universality.

[0041] Below, we will examine the operation of each step of FIG. 3 in relation to the price prediction of the appraisal model. FIG. 3 is a flowchart regarding item price prediction by an appraisal model according to an embodiment of the present invention.

[0042] In step S310, the merchant NPC may obtain an item dataset from the computing device (12) that includes item description information and item price information crawled from one or more other games. The item description information may include information such as the item's name, level, effect, rarity, durability, and required level.

[0043] As an example, item description information and item price information can be obtained from the World of Warcraft Classic game to obtain an item dataset consisting of a list of pairs of purchasable items and their prices. The item dataset thus obtained can be further processed through the following preprocessing steps to obtain a more useful item dataset. 3,376 basic items are extracted from the item dataset crawled from the World of Warcraft Classic game, and 106 items priced at less than 10 copper (the minimum currency unit in the World of Warcraft Classic game is 1 copper) are removed from among them. Then, all item prices are converted to copper units (conversion standards are 1 gold = 100 silver, 1 silver = 100 copper), and as a result of this preprocessing process, a total of 3,270 items are collected to obtain an item dataset.

[0044] In step S320, the appraisal model can be trained to predict the price of a new item that a player intends to trade using an item dataset obtained from at least one other game. The appraisal model receives an item dataset as input and can learn the relationship between one or more item description information and the item price information included in the item dataset.

[0045] Although the price of a new item a player intends to trade can be predicted by an appraisal model that does not require separate training, its prediction performance may be poor; this will be discussed later in the item price prediction stage.

[0046] As an example, the training of the appraisal model may be performed using a Supervised Fine-Tuning (SFT) method. The Supervised Fine-Tuning method may be a method of providing a new dataset to an existing language model to perform additional training.

[0047] While supervised fine-tuning methods involve updating all parameters of the entire existing language model, they are difficult to apply to large language models; therefore, adapter-based fine-tuning can be used. Adapter-based fine-tuning fixes the default parameters of the existing large language model and allows training only on added small adapter modules. An adapter can be a small neural network module added to enable the learning of new information without altering the structure of the existing large language model. Thus, an existing large language model can be trained on the relationship between item description information and corresponding item price information from item datasets acquired from one or more different games.

[0048] In step S330, the player can predict the price of a new item they wish to trade with a merchant NPC. The new item price prediction is achieved by obtaining descriptive information about the new item the player wishes to trade in natural language form and inputting this natural language descriptive information into a pre-trained appraisal model to predict the price of the new item.

[0049] Alternatively, using an untrained appraisal model, the price of a new item that a player intends to trade with a merchant NPC can be predicted through n-shot In-Context Learning (ICL). To predict the price of the new item, after obtaining the descriptive information of the new item the player intends to trade, one or more items with similar descriptive information to the new item are extracted from item datasets of one or more other games. The items possessing descriptive information similar to the new item are then sorted in order of highest string similarity, and the price of the new item can be predicted based on the sorted item price information.

[0050] String similarity refers to a method of numerically measuring how similar two strings (text) are to each other. This can be used when a player compares the description information of a new item they wish to trade with a merchant NPC with the item description information within the item database. As an example, string similarity can be evaluated using methods such as edit distance, cosine similarity, and Jaccard similarity.

[0051] In step S340, the player can evaluate whether the price prediction of the appraisal model for the new item they intend to trade with the merchant NPC is appropriate.

[0052] To evaluate the suitability of two methods for implementing an appraiser model, four indicators can be used. The four indicators may be the Mean Absolute Percentage Error (MAPE), Standard Deviation, Skewness, and Unexpected Output Rate.

[0053] First, regarding the average absolute percentage error, the reason for using percentage error is that if absolute error were used, the range of item prices is wide, which could lead to the absolute error of high-priced items being reflected excessively large compared to low-priced items. Therefore, percentage error (PE) can be used instead of absolute error. The average absolute percentage error is the average of the absolute values ​​of percentage errors and can show how much the predicted value of an item differs from the item's actual value on average. The average absolute percentage error is based on the following mathematical formulas.

[0054] [Mathematical Formula]

[0055]

[0056] In the above mathematical formula is a percentage error, and is the predicted price of item i, and is the actual price of item i.

[0057] [Mathematical Formula]

[0058]

[0059] In the above mathematical formula, is the mean absolute percentage error, and is the average value.

[0060] Secondly, the standard deviation of the percentage error (PE) can be measured. This indicates the volatility of the appraised item prices, and if the percentage error is distributed at similar levels across various items, the standard deviation should be low.

[0061] Thirdly, skewness can indicate whether the data distribution is skewed in a positive or negative direction. Therefore, analyzing skewness can determine whether the appraisal model underestimates (negative skewness) or overestimates (positive skewness) the price of an item. Skewness can be calculated using the following mathematical formula.

[0062] [Mathematical Formula]

[0063]

[0064] In the above mathematical formula, is the average of the percentage error, and is the standard deviation of the percentage error.

[0065] Finally, unexpected output ratios can be calculated. When the appraisal model predicts the price of an item, it may fail to extract the exact price from the output statement. As an example, there are cases where multiple prices are output even though a single prediction was requested. It is possible to determine whether the price prediction for a new item by the appraisal model is valid by the four methods described above.

[0066] In step S220 of Figure 2, the appraisal model conducted an experiment to predict the retail price of a game item using two methods: In-Context Learning (ICL) and Supervised Fine-Tuning (SFT), as an example.

[0067] First, context-based learning can be used. This method involves inference performed by providing a few examples to a pre-trained large language model without additional model training. In other words, without altering the weights of the pre-trained large language model, a few demonstration examples are provided with the prompt (input sentence) to allow the model to learn patterns.

[0068] The experimental procedure involved calling a large language model using a public API while maintaining the default hyperparameter settings. Ten randomly selected examples were included in the prompt to have the model predict prices. Performance was compared using the Llama-3.1 8b, 405b, and 3.2 1b models as the large language models.

[0069] The experimental results showed that the ICL-405b model demonstrated the best performance. The ICL-405b model successfully predicted retail prices with a probability of 94.8%, had the lowest MAPE at 1.34%, and the lowest error rate (ratio of unexpected outputs) at 5.2%. However, it showed a tendency to underestimate prices for some items (skewness of -5.06). On the other hand, the performance of the ICL-8b and ICL-1b models was relatively low. In particular, ICL-1b had the highest MAPE (14.68%), which significantly lowered prediction accuracy, and had low practicality as the ratio of unexpected outputs reached 29.05%.

[0070] Secondly, a supervised fine-tuning method can be used. Supervised fine-tuning is a method that specializes a pre-trained large language model for a specific task (predicting item prices) by further training it. Computational costs can be reduced by maintaining the model's existing weights and adding only adapters for new training.

[0071] For the experiment, an SFT model was implemented using PyTorch 2.5.1 and transformers 4.46.2. Adapter heads were added to solve the regression problem, and a single prediction was generated by applying mean pooling to the last hidden state. The model was trained using two A6000 GPUs. For large language models, the performance was compared using Llama 3.1 8b and 3.2 1b models.

[0072] The experimental results showed that the SFT-8b model demonstrated the best performance. Although the SFT-8b model had a slightly higher MAPE (2.66%) than the ICL-405b, it exhibited the lowest skewness in price prediction (-3.47), indicating a stable model with less overestimation or underestimation. However, the SFT-1b model showed relatively lower performance. Its MAPE was 3.59%, which was lower than that of the SFT-8b, and it showed a stronger tendency to underestimate in price prediction (-5.84).

[0073] In step S230, the computing device (12) can negotiate a trading price for an item that a player wishes to trade with a merchant NPC through a negotiator model running on a processor (14). The negotiator model can conduct a negotiation conversation with the player using a language model. The negotiator model can receive descriptive information about the new item the player wishes to trade, a price predicted by an appraisal model, and a previous conversation history as input, and generate an appropriate response to persuade the player.

[0074] Below, we will examine the operation of each step of FIG. 4 in relation to the item price negotiation of the negotiator model. FIG. 4 is a flowchart regarding the progress of item price negotiation by the negotiator model according to an embodiment of the present invention.

[0075] Price negotiation for a new item involves creating a virtual NPC and a virtual player, and based on the negotiator model, generating a price negotiation conversation for an item in an item dataset crawled from an arbitrary game between a virtual merchant NPC and a virtual player to obtain a negotiation dataset (S410); training the negotiator model (S420) to enable transaction negotiation for the new item based on the negotiation dataset; and, through the trained negotiator model, the merchant NPC negotiates the transaction price of the new item with the player (S430), and evaluates the suitability of the negotiated price for the new item (S440).

[0076] In step S410, the merchant NPC can obtain a negotiation dataset by generating a price negotiation dialogue for items in an item dataset crawled from an arbitrary game between a virtual merchant NPC and a virtual player based on a negotiator model.

[0077] The negotiation dataset is acquired by providing information on virtual negotiation items included in an item dataset crawled from an arbitrary game to a virtual merchant NPC and a virtual player; after predicting the item's price using an appraisal model and using it as a reference price during the negotiation process, one or more pre-set negotiation strategy prompts are input into the negotiator model; one or more previous conversation contents between the virtual merchant NPC and the virtual player are input as prompts; based on the previous conversation history, an appropriate strategy is selected, and utterances corresponding to the strategy can be generated. The negotiation dataset can be created according to a format in which strategies are injected into the NPC and the Player respectively, and a language model acting as the NPC and a language model acting as the Player interact with each other to generate dialogue.

[0078] In the negotiation process, before the negotiation begins, the virtual merchant NPC and the virtual player can receive a description of the item being negotiated as shared information. The negotiation can start when the virtual player intends to purchase the item at a discounted price of 10% to 25%. The virtual player can initiate the negotiation with the sentence, “Hello. I am looking for [Item Name].” Although actual negotiations may include greetings or small talk, these are removed to allow the model to focus on learning the negotiation. The conversation then proceeds until the player decides whether to purchase the item, during which time the virtual merchant can continue to persuade the player. The virtual player can end the negotiation by saying “conversation over.”

[0079] In Step S420, to develop a negotiator, we can adopt two approaches: zero-shot prompting (ZSP) and knowledge distillation (KD). Since it is difficult to construct a negotiation dataset using supervised learning, we can adopt two methods distinct from the appraiser model. Among these, we examine the knowledge distillation method, which requires prior training.

[0080] Knowledge Distillation (KD) is a technique that transfers knowledge learned by a large model (Teacher Model) to a small model (Student Model) to reduce computational costs while maintaining performance. In other words, it is a method of compressing and transferring knowledge from a high-performance, complex model (Teacher Model) to a lightweight model (Student Model). This method can be widely used, particularly when optimizing large language models (LLM) or neural network-based AI models. In the present invention, to distill the persuasiveness of the teacher model, an appropriate tactic is selected before generating explicit utterances, and in addition, to reduce the computational requirements during the training of the large language model, quantization and low-rank adaptation (LoRA) methods may be adopted.

[0081] The zero-shot prompting method, which does not require prior learning, will be discussed later in the section on item price negotiation between merchant NPCs and players.

[0082] In step S430, the merchant NPC can negotiate the price with the player for a new item the player wishes to trade. The merchant NPC can negotiate the price for the new item through a negotiator model that is pre-trained by one or more methods including a knowledge distillation method, based on the price predicted by the appraiser model for the new item the player wishes to trade. The step of negotiating the trade price may consist of inputting the price of the new item the player wishes to trade predicted by the appraiser model, inputting one or more previous conversation contents between the merchant NPC and the player as a prompt, inputting information about the new item the player wishes to trade, and conducting the negotiation by the learned negotiator model.

[0083] Alternatively, the merchant NPC can negotiate the price of a new item using the Zero-Shot Prompting (ZSP) method, which does not require prior training of the negotiator model. Zero-Shot Prompting is a method in which a question is given to a Large Language Model (LLM) without additional examples (demo data) and an answer is received immediately. In other words, it is a method in which the model infers using only previously learned information, and can respond immediately to general questions without a training process. However, since the Zero-Shot Prompting method may deviate significantly from the game environment, the present invention may include 10 negotiation strategies in the prompt to conduct price negotiations for a new item.

[0084] In the transaction price negotiation using the Zero-Shot Prompting method, one or more pre-set negotiation strategy prompts are input into the negotiator model, the price of a new item predicted by the appraisal model is input, if one or more previous conversations between the merchant NPC and the player exist, the conversation content is input as a prompt, item information of the new item the player wishes to trade is input, and based on the negotiation strategy prompts, the negotiator model enables the NPC merchant to proceed with price negotiation with the player regarding the new item.

[0085] Once the price of a new item the player wishes to trade is determined through negotiation, the transaction between the merchant NPC and the player can be completed at that price.

[0086] In step S440, it can be determined whether price negotiation by the negotiator model between the player and the merchant NPC is appropriate, and appropriateness can be determined using three indicators: persuasiveness, dominance, and agreement.

[0087] Persuasiveness can be evaluated by measuring the persuasiveness of the sentences generated by the negotiator. Similar to actual commercial behavior, merchant NPCs must effectively persuade players to buy and sell items at better prices. Since various tactics are used in effective negotiations, the goal was to evaluate the effectiveness of the negotiator using 10 different tactics.

[0088] Dominance in negotiations can be determined by measuring whether a merchant holds more bargaining power than a player. In the real world, merchants hold the initiative in price negotiations; as the negotiated price increases, the merchant gains more profit, while the player incurs higher costs. Since negotiation is a form of zero-sum game, dominance can be measured using the following mathematical formula.

[0089] (Mathematical formula)

[0090]

[0091] : Final agreed price of Item i

[0092] : Retail price of item i

[0093] : The price the player wants for item i

[0094] Agreement Rate: Since negotiations in the real world do not always end successfully, they can be evaluated by measuring the frequency with which an agreement is reached. Based on the results of the dominance assessment, the number of times an agreement was reached can be estimated, and the agreement rate can be calculated as the ratio of the number of agreed-upon conversations to the total number of conversations.

[0095] In this invention, three ZSP (Zero Shot Prompt) methods and two KD (Knowledge Distillation) methods were compared. Regarding persuasiveness, the KD-8b method recorded the highest persuasiveness score (3.99), followed by ZSP-405b (3.92), ZSP-8b (3.74), KD-1b (3.65), and ZSP-1b (2.95). Regarding dominance, the ZSP-405b model recorded the highest dominance score (0.47), followed by ZSP-8b (0.42) and KD-8b (0.40). Regarding agreement, the ZSP-8b model recorded the highest agreement rate at 96.94%, followed by ZSP-405b (90.83%) and ZSP-1b (90.52%). The KD-models showed a relatively low consensus rate (KD-1b: 82.57%, KD-8b: 81.04%).

[0096] FIG. 5 is a diagram illustrating examples of 10 negotiation strategies for conducting negotiations in a negotiator model according to an embodiment of the present invention. Let us examine the 10 strategies related to the negotiation process. The negotiation strategies consist of the following strategies, and are exemplary negotiation strategies that are not limited to these 10 strategies. Since the negotiation strategies are presented in the form of text, they may be applicable to other strategy configurations or methods.

[0097] Negotiation strategies can consist of the first negotiation strategy, which is a general negotiation strategy and the oldest and most widely used strategy, and the second negotiation strategy, which is a game-specific strategy that influences the motivation to purchase game items.

[0098] The first negotiation strategy prompt may consist of a prompt including at least one of the following: the Liking strategy, which builds a relationship of trust by finding common ground or compliments; the Reciprocity strategy, which utilizes the psychology that the other party will reciprocate when a favor is shown; the Social Proof strategy, which gains trust by emphasizing that others have also chosen this item; the Consistency strategy, which induces the maintenance of consistency with previous behavior; the Authority strategy, which gains trust by emphasizing expert opinions or authority; and the Scarcity strategy, which stimulates the desire to purchase by emphasizing that the product is limited in quantity or hard to obtain. The second negotiation strategy prompt may consist of a prompt including at least one of the following: the Enjoyment strategy, which emphasizes elements that enhance the game experience; the Character Competency strategy, which emphasizes elements such as ability enhancement; the Visual Authority strategy, which emphasizes elements that allow for character appearance customization or attract attention; and the Monetary Value strategy, which induces purchase by emphasizing price-to-performance ratio. It is possible.

[0099] As an example of a negotiation strategy prompt, the Social Proof strategy prompt among the first negotiation strategy prompts can be configured as, “Thank you for considering a compromise. However, another buyer has already offered to purchase these pants for 18 gold. I can still give them to you for 17 gold, but I cannot lower the price below that without losing value.” The Scarcity strategy can be configured as, “Ah, a lead mace! That is a truly rare weapon. You have a good eye for recognizing great weapons. Surely, as a skilled adventurer, you are looking to upgrade your armory.” And the Visual Authority strategy among the second negotiation strategy prompts can be configured as, “You have an excellent eye for rare items! The Bristly Whisker is a truly unique item, and it is something many people seek to enhance their character's appearance. Would you like to know more about the attributes of this item and why it is worth 733 gold?”

[0100] FIG. 6 is a diagram relating to an example of generating a negotiation dataset by at least one of a negotiation strategy according to one embodiment of the present invention.

[0101] Based on a negotiator model, the merchant NPC can obtain a negotiation dataset by generating a price negotiation dialogue for items in an item dataset crawled from an arbitrary game between a virtual merchant NPC and a virtual player.

[0102] The negotiation dataset can be acquired by providing information on virtual negotiation items included in an item dataset crawled from an arbitrary game to a virtual merchant NPC and a virtual player, predicting the item's price using an appraisal model and using it as a reference price during the negotiation process, receiving one or more pre-set negotiation strategy prompts into a negotiator model, inputting one or more previous conversation contents between the virtual merchant NPC and the virtual player as prompts, allowing them to select an appropriate strategy based on the previous conversation history, and generating utterances corresponding to that strategy. The negotiation dataset can be generated according to a format in which strategies are injected into the NPC and the Player respectively, and a language model acting as the NPC and a language model acting as the Player interact with each other to generate dialogue.

[0103] FIG. 7 is a diagram illustrating examples of conventional passive interaction, active price prediction, and active negotiation according to one embodiment of the present invention.

[0104] FIG. 710 illustrates a conventional trading method based on passive item prices between a merchant NPC and a player within a game, FIG. 720 illustrates how a merchant NPC predicts the price of a new item that a player intends to trade using an appraisal model based on item information obtained from an item dataset crawled from one or more other games, and FIG. 730 illustrates the negotiation process between the merchant NPC and the player regarding the new item using a negotiator model.

[0105] The method for determining the price of in-game items described through FIGS. 1 to 7 may also be implemented in the form of a computer program stored on a computer-readable recording medium executed by a computer, or in the form of a recording medium containing instructions executable by a computer. Additionally, the method for determining the price of in-game items described through FIGS. 1 to 7 may also be implemented in the form of a computer program stored on a computer-readable recording medium executed by a computer.

[0106] A computer-readable recording medium may be any available medium accessible by a computer and includes both volatile and non-volatile media, and both removable and inremovable media. Additionally, a computer-readable recording medium may include a computer storage medium. A computer storage medium includes both volatile and non-volatile, removable and inremovable media implemented by any method or technique for storing information such as computer-readable instructions, data structures, program modules, or other data.

[0107] Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware device may be configured to operate as one or more software modules to perform processing according to the present invention, and vice versa.

[0108] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical spirit or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0109] Although representative embodiments of the present invention have been described in detail above, those skilled in the art will understand that various modifications can be made to the above-described embodiments without departing from the scope of the present invention. Therefore, the scope of the present invention should not be limited to the described embodiments, but should be defined by the claims set forth below as well as equivalents thereof.

Claims

Claim 1 A method for determining the price of an item in a game, comprising: a step of creating a merchant NPC (Non-player Character) for trading items in a game; and a step in which the merchant NPC predicts the price of a new item based on information of the new item that the player intends to trade, using an Appraiser Model. Claim 2 A method for determining the price of an in-game item according to claim 1, wherein the step of predicting the price of the new item comprises: a step of obtaining an item dataset including item description information and item price information crawled from one or more other games; a step of training the appraisal model to predict the price of the new item based on the item dataset; and a step of predicting the price of the new item that the player intends to trade with the NPC using the trained appraisal model. Claim 3 A method for determining the price of an in-game item, wherein, in claim 2, the step of training the appraisal model comprises: a step of inputting the item dataset into the appraisal model; and a step of causing the appraisal model to learn the relationship between one or more item description information included in the item dataset and the price information of the item. Claim 4 A method for pricing an in-game item, wherein the step of predicting the price of the new item comprises: a step of obtaining descriptive information of the new item that the player intends to trade; and a step of the learned appraisal model predicting the price of the new item based thereon. Claim 5 A method for determining the price of an in-game item according to claim 1, wherein the step of predicting the price of the new item comprises: obtaining descriptive information of the new item that the player intends to trade; extracting one or more items similar to the item descriptive information of the new item from item datasets of one or more other games; sorting items containing item descriptive information similar to the new item in order of high similarity of strings based on the extracted items; and predicting the price of the new item based on the sorted item price information in the appraisal model. Claim 6 A method for pricing an in-game item, wherein, in claim 1, the pricing step further comprises the step of having the merchant NPC negotiate the price of the new item with the player using the negotiator model based on the price of the new item predicted by the appraiser model. Claim 7 In claim 6, the step of negotiating the price of the new item comprises: a step of creating a virtual NPC and a virtual player; a step of obtaining a negotiation dataset by generating a price negotiation dialogue for an item in an item dataset crawled from any game between a virtual merchant NPC and a virtual player based on the negotiator model; a step of training the negotiator model to enable transaction negotiation of the new item based on the negotiation dataset; and a step in which the merchant NPC negotiates the transaction price of the new item with the player by means of the trained negotiator model; a method for determining the price of an item within a game. Claim 8 In claim 7, the step of acquiring the negotiation dataset comprises: providing information on any negotiation item included in an item dataset crawled from any game to a virtual merchant NPC and a virtual player; inputting one or more negotiation strategy prompts pre-set in the negotiator model; inputting the price of the virtual negotiation item predicted by the appraisal model; inputting one or more previous conversation contents between the virtual merchant NPC and the virtual player as prompts; generating a conversation between the virtual merchant NPC and the virtual player through a negotiation simulation based on the negotiator model; and acquiring the generated conversation as the negotiation dataset for training the negotiator model; a method for determining the price of an in-game item. Claim 9 In claim 7, the step of negotiating the transaction price comprises: a step of inputting the price of the new item predicted by the appraisal model; a step of inputting one or more previous conversation contents between the merchant NPC and the player as a prompt; a step of inputting item information of the new item that the player intends to trade; and a step of conducting the negotiation by the learned negotiator model; a method for determining the price of an in-game item. Claim 10 In claim 6, the step of negotiating the transaction price comprises: a step of inputting one or more negotiation strategy prompts pre-set in the negotiator model; a step of inputting the price of the new item predicted by the appraisal model; a step of inputting one or more previous conversation contents between the merchant NPC and the player as prompts; a step of inputting information of the new item that the player intends to trade; and a step of the negotiator model causing the NPC merchant to proceed with price negotiation with the player regarding the new item based on the negotiation strategy prompts; a method for determining the price of an in-game item. Claim 11 A method for pricing in-game items, wherein, in claim 8 or 10, the negotiation strategy prompt comprises a first negotiation strategy prompt and a second negotiation strategy prompt, and the first negotiation strategy prompt comprises at least one of the following: a Liking strategy that builds a trust relationship by finding common ground or praising; a Reciprocity strategy that utilizes the psychology of the other party to reciprocate when a favor is shown; a Social Proof strategy that gains trust by emphasizing the fact that others have also chosen this item; a Consistency strategy that induces consistency with previous behavior; an Authority strategy that gains trust by emphasizing expert opinions or authority; and a Scarcity strategy that stimulates the desire to purchase by emphasizing that the item is limited in quantity or difficult to obtain. Claim 12 In Paragraph 11, A method for pricing in-game items, wherein the second negotiation strategy prompt described above is a prompt that includes at least one of the following: an Enjoyment strategy that emphasizes elements that enhance the game experience; a Character Competency strategy that emphasizes elements such as ability enhancement; a Visual Authority strategy that emphasizes elements that can decorate the character's appearance or attract attention; and a Monetary Value strategy that induces purchase by emphasizing price-to-performance ratio. Claim 13 In claim 6, the pricing method further includes a step of evaluating whether the negotiation result of a new item through the negotiator model is appropriate, and the evaluation of appropriateness evaluates the performance of the negotiator model by at least one of the following indicators: an evaluation of persuasiveness, which is the ability of a merchant NPC to persuade a player; an evaluation of dominance, which is the degree of power held; and an evaluation of agreement, which is the rate at which an agreement is successfully concluded. Claim 14 In Clause 13, the above-mentioned dominance is a method for pricing in-game items calculated by the following mathematical formula. (Mathematical Formula) : The final agreed price of the new item i traded by the player : Retail price of new item i traded by the player : The price the player wants for the new item i traded by the player Claim 15 A computing device comprising: a processor; and a memory for storing one or more programs executed by the processor, wherein the processor performs the operation of generating a merchant NPC (Non-player Character) for trading items within a game; the operation of causing the merchant NPC to predict the price of a new item based on information of the new item that a player intends to trade using an Appraiser Model; and the operation of causing the merchant NPC to conduct a negotiation with the player regarding the price of the new item using a Negotiator Model. Claim 16 A computer program stored on a non-transitory computer-readable storage medium, wherein the computer program comprises one or more instructions, and when the instructions are executed by a computing device having one or more processors, the computing device performs the steps of: generating a merchant NPC (Non-player Character) for trading items within a game; causing the merchant NPC to predict the price of a new item based on information of the new item that a player intends to trade using an Appraiser Model; and causing the merchant NPC to conduct a negotiation with the player regarding the price of the new item using a Negotiator Model.

Images