Agent search system

Abstract

We provide an agent search system that objectively and comprehensively matches the user's desired conditions with the agent's past performance and achievements to suggest the most suitable agent. [Solution] The agent search system comprises a performance storage unit 10 that stores past case information for each agent, a results storage unit 12 that stores result information corresponding to each case, an extraction means 20 that extracts past cases of the agent from the performance storage unit 10 based on conditions entered by the user, an evaluation calculation means 30 that analyzes the result information of the extracted cases stored in the results storage unit 12 and generates an evaluation value for the conditions, and a presentation means 40 that presents an agent based on the evaluation values ​​generated for multiple agents. The evaluation calculation means 30 may have a function to correct the evaluation value according to the difficulty index based on the information contained in either the performance storage unit 10 or the results storage unit 12.

Description

Technical Field

[0001] The present invention relates to an agent search system that searches for agents such as lawyers and patent attorneys and presents the most suitable agent to the client. In particular, it relates to a technique for calculating an evaluation value based on the past case information and achievement information of each agent and recommending an agent in consideration of a difficulty index.

Background Art

[0002] Conventionally, as a system for matching a client and an expert, a system that automatically selects an expert for the client's problem based on the occupation, field of expertise, attribute information, etc. registered in the expert database is known (for example, see Japanese Patent Application Laid-Open No. 2022-70187). In such a conventional system, an expert corresponding to the problem content input by the client can be extracted based on the registration information and the degree of keyword matching, and a plurality of experts can be presented. However, this type of system depends on the registration information and self-declared attributes of the expert and does not consider objective achievement data such as past achievements and case results. Therefore, it has not been possible to comprehensively collate the user's desired conditions and the actual business results of the agent, calculate a highly reliable evaluation value based on the results, and present the most suitable agent.

[0003] Conventional posted matching depends on the self-declared information of the agent and keyword matching, and it has been difficult to select using objective achievement data of past cases. In addition, the confirmation of conflict of interest depends on the hearing on the applicant's side and the self-declaration of the agent, and it has not been possible to cover the involvement of related company names or joint applicant names, name notation fluctuations, etc., which has required time and labor. The system of the present application resolves the above bottlenecks by using evaluation calculation based on the performance storage unit and the achievement storage unit and automatic collation by the conflict-of-interest detection means.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

[0005] Conventional agent searches typically involved listing agent information and requiring users to manually select an agent. This made it difficult to objectively and comprehensively match the user's desired conditions with the agent's past performance and results to present the most suitable agent. To solve this problem, the present invention aims to provide an agent search system that analyzes the user's desired conditions with the agent's performance and results to calculate an evaluation value, and then presents the user with the most suitable agent based on that evaluation value. [Means for solving the problem]

[0006] To solve the above problems, the present invention provides an agent search system for searching for agents, comprising: a performance storage unit that stores past case information of each agent; an outcome storage unit that stores outcome information corresponding to each case; an extraction means for extracting past cases of the agent from the performance storage unit based on conditions entered by the user; an evaluation calculation means that analyzes the outcome information of the extracted cases stored in the outcome storage unit and generates an evaluation value for the conditions; and a presentation means that presents an agent based on the evaluation values ​​generated for a plurality of agents.

[0007] Furthermore, the evaluation calculation means may include a function to correct the evaluation value according to the difficulty index based on information contained in at least one of the performance memory unit or the results memory unit.

[0008] Furthermore, the presentation means includes a recommendation generation means, which can select and present agents as recommendation candidates that meet the user's requirements based on the conditions and evaluation values ​​entered by the user.

[0009] Furthermore, the presentation means can visualize and present the degree of contribution of each evaluation element to the evaluation value of each agent.

[0010] Furthermore, the conflict of interest detection means can compare information on entities involved in the agent's past cases with information on companies and other entities related to the user, and if a match or relationship is detected, it can generate a warning information regarding the conflict of interest and present it using the presentation means. [Effects of the Invention]

[0011] According to the present invention, objective and quantitative evaluation based on the agent's past case performance and results can be achieved, and the agent best suited to the user's desired conditions can be automatically presented.

[0012] Furthermore, by adjusting evaluation values ​​to take difficulty indicators into account, it becomes possible to appropriately evaluate agents who are capable of handling challenging cases.

[0013] Furthermore, by using a recommendation generation system to prioritize recommending agents that meet the user's requirements, user satisfaction can be improved.

[0014] In addition, by visualizing the contribution of each evaluation element, the transparency and explainability of the evaluation can be improved, supporting users in making appropriate decisions.

[0015] Furthermore, the conflict of interest detection mechanism allows for prior confirmation of potential conflicts of interest during the agent selection stage, thereby preventing ethical and legal risks. [Brief explanation of the drawing]

[0016] [Figure 1] This is a block diagram showing the overall configuration of an agent search system according to one embodiment of the present invention. [Figure 2] This is a table diagram showing an example of agent information stored in the performance memory unit. [Figure 3] This is a table diagram showing an example of project outcome information stored in the outcome storage unit. [Figure 4] This is a table diagram showing an example of a difficulty level index master. [Figure 5] This is a flowchart showing the operation flow of the agent search system. [Figure 6] It is a diagram showing an example of a display screen for agent search results. [Figure 7] It is a schematic diagram showing an example of a contribution degree visualization screen of evaluation elements. [Figure 8] It is a block diagram showing the configuration of an agent search system equipped with an interest conflict detection means. [Figure 9] It is a flowchart of the interest conflict detection process.

Mode for Carrying Out the Invention

[0017] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. The following embodiments are examples for explaining the present invention and do not limit the scope of the present invention. Also, in each drawing, the same elements are denoted by the same reference numerals, and redundant explanations are omitted.

[0018] <First Embodiment> In the first embodiment, a basic agent search system that calculates an evaluation value based on the past case information and achievement information of agents and presents a plurality of agents will be described. Overview of the System

[0019] The agent search system of this embodiment is a system that searches for the agent most suitable for the needs of the client from among agents such as lawyers and patent attorneys. This system calculates an evaluation value based on the past case performance and achievement information of agents, and automatically searches for and presents the optimal agent from among a plurality of agents.

[0020] (Overall Configuration of the System) As shown in FIG. 1, the agent search system of this embodiment is composed of a performance storage unit 10, an achievement storage unit 12, a difficulty master 14, an extraction means 20, an evaluation calculation means 30, a presentation means 40, and a user interface unit 50.

[0021] This system is configured, for example, as a cloud server or a standalone information processing device. The system's hardware configuration includes a control unit (CPU, MPU, etc.), main memory (RAM), auxiliary storage (HDD, SSD), communication interface, input / output interface, etc. The auxiliary storage unit stores a system program for executing the processing according to the present invention and various databases. The control unit reads the system program into the main memory and executes it, thereby realizing each functional block such as the extraction means 20, evaluation calculation means 30, and presentation means 40.

[0022] The user interface unit 50 displays the search results in a list format, including rank, agent name, area of ​​expertise, evaluation value, success rate, average satisfaction, average resolution time, and difficulty level. It also features contribution visualization, presenting the contribution of each evaluation element to the evaluation value in a graph to enhance the transparency of the evaluation basis. These screen configurations correspond to Figures 6 and 7.

[0023] (Performance Record Department) The performance storage unit 10 is a database that stores past case information for each agent. The performance storage unit 10 is a storage device that records the agent's past work performance and is composed of a storage medium such as a hard disk, SSD, or cloud storage. As shown in Figure 2, the performance storage unit 10 stores agent ID, agent name, area of ​​expertise, number of past cases, details of past cases, and case difficulty information.

[0024] The data stored in the performance memory unit 10 can be configured to be periodically acquired and accumulated from a publicly known patent information database (e.g., J-PlatPat, commercial patent databases, etc.) through batch processing or API integration. This allows the system to always maintain the latest agent performance information, providing a basis for objective evaluation.

[0025] Agent ID is an identifier used to uniquely identify an agent, such as a code consisting of a combination of numbers or letters (e.g., A001, B002). Agent Name is the agent's name or the name of their office, such as "Taro Yamada" or "Patent Office ABC".

[0026] The "specialized field" refers to the agent's area of ​​expertise, such as patents, trademarks, designs, litigation, or appeals. The "number of past cases" refers to the total number of cases the agent has handled in the past, such as 100 or 200 cases.

[0027] Past case details refer to detailed information about each case, including case ID, technical field, case size, and case complexity. The technical field is the technical area to which the case belongs, such as electrical, mechanical, chemical, bio, and AI-related classifications. Case size is an indicator of the size of the case, such as single application, multi-country application, and large-scale litigation. Case complexity refers to factors that affect the difficulty of the case, such as the amount of prior art, the complexity of the claims, and the difficulty of the legal issues.

[0028] The details of past cases also include information on the parties involved. This information includes the names of the applicant companies, co-applicants, related companies, assignees, or client companies in past cases, and indicates which companies the agent has worked for in the past, or which parties they have dealt with as opposing parties.

[0029] Specific examples of past case details include implementations that store application number, publication number, registration number, filing date, technology classification (IPC, FI, F-term, etc.), information on involved parties (applicant name, co-applicant name, assignee name, etc.), name of the patent attorney in charge, and case type (application, rejection response, appeal, litigation, etc.). This allows for centralized management of key information necessary for searching, difficulty calculation, and conflict of interest detection.

[0030] Project difficulty information refers to a difficulty index or difficulty level (high, medium, low) calculated from the technical field, project size, complexity, etc., and is an indicator of how difficult each project was. Project difficulty information may be stored in the performance storage unit 10 or in the results storage unit 12, which will be described later.

[0031] (Achievement storage) The results storage unit 12 is a database that stores results information corresponding to each case. The results storage unit 12 is a storage device that records the results and outcomes of each case handled by the agent, and like the performance storage unit 10, it is composed of a storage medium such as a hard disk, SSD, or cloud storage. As shown in Figure 3, the results storage unit 12 stores the case ID, agent ID, success or failure status, resolution period, client satisfaction, and case difficulty.

[0032] The case ID is an identifier used to uniquely identify a case, and it corresponds to the case ID in the past case details of the performance storage unit 10. The agent ID is an identifier used to identify the agent who handled the case.

[0033] The success / failure status is a flag indicating whether the case was successful or unsuccessful. For example, it is a binary data value where success (1) means the patent was registered, and failure (0) means the application was rejected. The resolution period is the number of days required from the start to the completion of the case. For example, this could be the number of days from patent application to registration, or the number of days from filing a lawsuit to the final judgment.

[0034] Client satisfaction is an evaluation by the client, for example, on a 5-point scale from 1 to 5 (1: dissatisfied, 2: somewhat dissatisfied, 3: average, 4: somewhat satisfied, 5: satisfied). Case difficulty is the difficulty level (high, medium, low) of the case or the overall difficulty value, which is either the difficulty level entered when registering the case or the overall difficulty value calculated by referring to Difficulty Master 14.

[0035] The results storage unit 12 can also store information such as the number of rejection notices issued, the number of intermediate processing responses, and the examiner's name. The number of rejection notices is the number of rejection notices issued by the Japan Patent Office for the case in question, and serves as an indicator of the difficulty level. The examiner's name is the name of the examiner who examined the case in question, and can be used for difficulty level adjustments that take into account the differences in statistical assessment rates for each examiner.

[0036] Examples of fields in the results storage unit 12 include the latest status (pending, examination requested, rejection decision, patent granted, registered), patenting flag, number of rejection notices, number of intermediate responses, time to patenting, examiner name (for difficulty adjustment), appeal information (type, number, parties, result), and litigation information (case number, court, parties, result).

[0037] (Difficulty Master) The difficulty level master 14 stores master data for determining the difficulty level of a project. The difficulty level master 14 is a database that stores standard data for calculating the objective difficulty level of a project. As shown in Figure 4, the difficulty level master 14 stores the technical field code, basic difficulty level, project size coefficient, and complexity coefficient.

[0038] A technology field code is a code used to classify technology fields, such as E01 (electrical), M01 (mechanical), and C01 (chemical). The basic difficulty level is a value for each technology field, such as a number in the range of 0.8 to 1.5. Advanced technology fields (e.g., AI-related, biotechnology) have a high basic difficulty level (1.3 to 1.5), while mature technology fields (e.g., mechanical structures) have a low basic difficulty level (0.8 to 1.0).

[0039] The case size coefficient is a difficulty adjustment coefficient that corresponds to the size of the case, and is a value in the range of 0.9 to 1.3. For example, a coefficient of 0.9 is set for a single application, 1.2 for a multi-country application, and 1.3 for a large-scale lawsuit. The complexity coefficient is a difficulty adjustment coefficient that corresponds to the complexity of the case, and is a value in the range of 0.8 to 1.4. For example, a coefficient of 0.8 is set when there is little prior art, and 1.4 is set when there is a lot of prior art and a complex search is required.

[0040] (Method for calculating difficulty level) The evaluation calculation means 30 calculates the overall difficulty level of the case using the following formula. Overall difficulty = Basic difficulty × Project size coefficient × Complexity coefficient

[0041] Based on the calculated overall difficulty level, the difficulty level is determined by the following threshold criteria. • Overall difficulty ≥ 1.2 → High difficulty (correction factor 1.3) • 0.9 ≤ Overall difficulty < 1.2 → Medium difficulty (correction factor 1.0) • Overall difficulty < 0.9 → Low difficulty (correction factor 0.8)

[0042] In addition to the calculation method for the overall difficulty described above, the difficulty index can be adjusted based on the number of rejection notices and examiner information stored in the results storage unit 12. For example, if there are three or more rejection notices, +0.5 points are added to the difficulty. Also, if the examiner's name is statistically low (strict) in terms of the examination rate, +0.3 points are added to the difficulty. If it is a specific technical field (e.g., business model patent), +0.2 points are added to the difficulty. This makes it possible to perform a more precise difficulty assessment that is in line with the actual situation of the case.

[0043] The calculated difficulty level and overall difficulty value are stored in the project difficulty information field of the performance storage unit 10 or the project difficulty field of the results storage unit 12. This allows the evaluation calculation means 30 to obtain difficulty information from either the performance storage unit 10 or the results storage unit 12.

[0044] (extraction means) The extraction means 20 extracts past cases of the target agent from the performance storage unit 10 based on conditions entered by the user. The extraction means 20 is a software module that realizes the function of searching for and retrieving data that matches specific conditions from the database, and is realized by the CPU executing a program.

[0045] Specifically, the extraction means 20 performs the following step-by-step processing steps. In the first step, it compares the desired area of ​​expertise (e.g., patents, trademarks, designs, etc.) entered from the user interface unit 50 with the area of ​​expertise information of each agent stored in the performance storage unit 10, and extracts agents whose area of ​​expertise matches as primary candidates. In the second step, it filters by the type of case specified by the user (e.g., application representation, rejection response, opposition response, etc.) and extracts agents with experience in the relevant case type as secondary candidates. In the third step, if a lower limit condition for the number of past cases handled is specified, it extracts only agents that meet this condition. In the fourth step, if the user specifies a difficulty level (high, medium, low), it refers to the difficulty information of each case stored in the performance storage unit 10 and extracts agents with experience in cases corresponding to the specified difficulty level as final candidates. By performing this step-by-step narrowing, the past cases of agents that match the user's conditions are efficiently extracted and handed over to the evaluation calculation means 30.

[0046] (Method for calculating evaluation) The evaluation calculation means 30 analyzes the performance information of the extracted cases stored in the performance storage unit 12 and generates an evaluation value for the given conditions. The evaluation calculation means 30 is a software module that implements the function of quantifying and evaluating the performance of agents, and like the extraction means 20, it is implemented by the CPU executing a program.

[0047] (Calculation of basic evaluation value) The evaluation calculation means 30 first obtains information on "success or failure," "resolution period," and "client satisfaction" from the results storage unit 12 for each case of the extracted agent, and calculates a basic evaluation value. The basic evaluation value is calculated using the following formula. Basic evaluation value = (Success rate × 50) + (Resolution time evaluation value × 30) + (Satisfaction evaluation value × 20)

[0048] The success rate is the percentage of the target projects that were successful (success rate = 1), and is calculated using the following formula. Success rate = (Number of successful cases ÷ Number of target cases) × 100 For example, if 85 out of 100 target cases are successful, the success rate is 85%.

[0049] The resolution period evaluation value is a value that evaluates the shortness of the resolution period, and is calculated using the following formula. Resolution time evaluation value = 100 - (Average resolution time ÷ Industry average time × 100) This formula shows that a shorter average resolution period results in a higher evaluation score. For example, if the average resolution period is 300 days and the industry average is 400 days, the resolution period evaluation score would be 100 - (300 ÷ 400 × 100) = 25.

[0050] The satisfaction rating is a value obtained by converting the average client satisfaction score to a 100-point scale, and is calculated using the following formula. Satisfaction rating = (Average satisfaction ÷ 5) × 100 For example, if the average satisfaction level is 4.5, the satisfaction rating would be (4.5 ÷ 5) × 100 = 90.

[0051] The weight values ​​for each element in the basic evaluation formula (success rate × 50, resolution period evaluation value × 30, satisfaction evaluation value × 20) can be dynamically adjusted according to the user's requests and the characteristics of the case. For example, the weight of the resolution period evaluation value can be increased (e.g., 50) for urgent cases, and the weight of the satisfaction evaluation value can be increased (e.g., 40) for quality-focused cases.

[0052] (Determination of difficulty correction coefficient) The evaluation calculation means 30 determines a difficulty correction coefficient based on the difficulty information obtained from the performance storage unit 10 or the results storage unit 12. If the difficulty level is "High", the difficulty correction coefficient will be set to 1.3. If the difficulty level is "medium," the difficulty correction coefficient will be set to 1.0. If the difficulty level is "Low," the difficulty correction coefficient will be set to 0.8.

[0053] (Calculation of corrected evaluation value) The evaluation calculation means 30 calculates a corrected evaluation value by multiplying the basic evaluation value by a difficulty correction coefficient. Corrected evaluation value = Base evaluation value × Difficulty correction coefficient For example, if the base score is 80 points and the difficulty correction coefficient is 1.3, the corrected score would be 80 × 1.3 = 104 points.

[0054] (Normalization of evaluation values) The evaluation calculation means 30 can normalize the corrected evaluation value to a range of 0 to 100 as needed. The normalized evaluation value is then passed to the presentation means 40 as the final evaluation value.

[0055] (Presentation means) The presentation means 40 presents a suitable agent to the user based on the evaluation values ​​generated for each agent. The presentation means 40 is a software module that implements a function to display search results in a format that the user can see, and outputs the results through the user interface unit 50.

[0056] The presentation method 40 displays agents in a ranking format, ordered from highest to lowest evaluation score. Display items include agent name, evaluation score, success rate, average satisfaction, average resolution time, and the difficulty level they can handle.

[0057] Furthermore, presentation means 40 provides links to detailed information about each agent, allowing users to view more detailed information. This detailed information includes a list of past cases, track record in each technical field, client reviews, etc.

[0058] (User Interface Section) The user interface unit 50 provides an interface for users to input search criteria and view search results. The user interface unit 50 is implemented as a web application that runs on a web browser, or as a dedicated desktop application or smartphone application.

[0059] On the input screen of the user interface unit 50, the user can enter the following items. • Desired area of ​​expertise (e.g., patents, trademarks, designs, litigation, appeals, etc.) • Case type (e.g., filing an application, responding to rejections, responding to oppositions, filing appeals) • Desired difficulty level (high, medium, low) • Minimum number of past cases handled Other conditions (region, language proficiency, price range, etc.)

[0060] (Operation Flow) As shown in Figure 5, the operation flow of the agent search system in this embodiment is as follows.

[0061] Step S1: The user enters search criteria through the user interface unit 50. Step S2: The extraction means 20 extracts past cases of the relevant agent from the performance storage unit 10 based on the input conditions. Step S3: The evaluation calculation means 30 obtains result information from the result storage unit 12 for the extracted cases. Step S4: The evaluation calculation means 30 calculates a basic evaluation value based on the performance information. Step S5: The evaluation calculation means 30 determines a difficulty correction coefficient based on the difficulty information. Step S6: The evaluation calculation means 30 calculates a corrected evaluation value by applying a difficulty correction coefficient to the basic evaluation value. Step S7: Presentation means 40 presents agents in a ranked format based on their evaluation scores. Step S8: The user views the search results and selects an appropriate agent.

[0062] (Example of search results display) As shown in Figure 6, the search results screen displays a list of agent names, ratings, success rates, average satisfaction levels, average resolution times, and more. Users can view and compare detailed information about each agent.

[0063] (Effects of the first embodiment) According to this embodiment, objective and quantitative evaluation based on the agent's past case performance and results can be achieved, and the agent best suited to the user's desired conditions can be automatically presented. Furthermore, by adjusting the evaluation value considering the difficulty index, agents capable of handling difficult cases can be appropriately evaluated.

[0064] (Modified version of the first embodiment) The first embodiment may be modified as follows. The evaluation calculation means 30 can add evaluation elements other than success rate, resolution period, and satisfaction level. For example, cost performance (ratio of results to fees), communication ability (feedback from clients), depth of expertise (degree of concentration of experience in a specific technical field), etc., can be added as evaluation elements.

[0065] Furthermore, the evaluation calculation means 30 can apply temporal weighting. For example, more recent cases can be given higher weight, reducing the influence of older cases. This makes it possible to perform an evaluation that more accurately reflects the agent's current capabilities.

[0066] Furthermore, the presentation means 40 can learn the user's past selection and evaluation history and provide personalized recommendations. This makes it possible to prioritize the presentation of agents that match the user's preferences.

[0067] <Second Embodiment> In the second embodiment, the difficulty correction function will be described in more detail.

[0068] (Evaluation of track record in handling highly complex projects) The evaluation calculation method 30 analyzes the number and success rate of high-difficulty cases that the agent has handled in the past, and evaluates their ability to handle high-difficulty cases. A high-difficulty case refers to a case with an overall difficulty level of 1.2 or higher.

[0069] The score for the ability to handle high-difficulty cases is calculated using the following formula. High-difficulty case handling ability score = (Number of high-difficulty cases ÷ Total number of cases) × (Success rate of high-difficulty cases ÷ 100) × 100 For example, if out of 100 total cases, 30 are high-difficulty cases, and the success rate for those cases is 80%, then the high-difficulty case handling ability score would be (30 ÷ 100) × (80 ÷ 100) × 100 = 24 points.

[0070] The evaluation calculation means 30 further adjusts the difficulty correction coefficient based on this score. If the high-difficulty case handling ability score is 20 points or higher, +0.2 is added to the difficulty correction coefficient. This ensures that agents who can handle difficult cases are evaluated more highly.

[0071] (Difficulty level evaluation by technical field) The evaluation calculation means 30 can evaluate the difficulty level for each technology field. In certain technology fields (e.g., AI, bio, business models), projects generally tend to be more difficult.

[0072] The difficulty level master for each technical field records the basic difficulty level for each technical field, and the evaluation calculation means 30 uses this to perform evaluations for each technical field. If a user specifies a particular technical field, agents with extensive experience in that technical field are recommended on a priority basis.

[0073] (Evaluation of ability to handle multiple difficulty levels) The evaluation calculation means 30 assesses whether the agent can handle a wide range of cases, from low difficulty to high difficulty. An agent who can handle a wide range of difficulty levels is evaluated as highly versatile.

[0074] The difficulty level coverage score is calculated using the following formula: Difficulty level coverage score = (Number of covered difficulty levels ÷ 3) × 100 For example, if a candidate has experience teaching at all three levels—low, medium, and high—their score will be 100 points. This score is used to determine the recommendation priority.

[0075] <Third Embodiment> In the third embodiment, the recommendation generation means will be described.

[0076] (Overview of recommendation generation methods) The presentation means 40 includes a recommendation generation means. Based on the conditions and evaluation values ​​entered by the user, the recommendation generation means selects and presents agents that meet the user's requirements as recommendation candidates.

[0077] (Recommendation logic) The recommendation generation method executes the following recommendation logic. First, agents whose evaluation score is above a certain threshold (e.g., 80 points) are selected as candidates for recommendation. Secondly, agents who perfectly match the conditions specified by the user (specialized field, difficulty level, region, etc.) will be given priority as candidates. Thirdly, agents with a high level of past user feedback will be placed at the top of the list. Fourthly, considering their current availability (level of busyness), we will prioritize agents who are most likely to accept the case.

[0078] (Application of collaborative filtering) The recommendation generation method uses collaborative filtering technology to recommend agents highly rated by other users who have searched using similar criteria. This enables recommendations tailored to the individual user's preferences.

[0079] (Ensuring diversity) The recommendation generation method can include not only top-ranking agents but also agents with different characteristics in a balanced manner to add diversity to the recommendation list. For example, agents prioritizing success rate, agents prioritizing resolution time, and agents prioritizing satisfaction can each be included in the recommendation list.

[0080] (Presentation of reasons for recommendation) The recommendation generation method explicitly presents the reasons for recommending each agent. For example, it might display a message such as, "This agent has extensive experience in the AI ​​field you specified and has an 85% success rate for high-difficulty projects."

[0081] (Effects of the third embodiment) According to this embodiment, user satisfaction can be improved not only by displaying rankings, but also by prioritizing the recommendation of agents that meet the user's requirements. Furthermore, by clearly stating the reasons for the recommendation, it is possible to support the user's decision-making.

[0082] <Fourth Embodiment> In the fourth embodiment, the visualization of the contribution of evaluation elements will be described.

[0083] (Overview of contribution visualization) The presentation means 40 visualizes and presents the contribution of each evaluation element to the evaluation value of each agent. By visualizing the contributions, users can understand why a particular agent is highly (or low) rated.

[0084] (Calculation of contribution) The evaluation calculation means 30 records the extent to which each evaluation element (success rate, resolution period, satisfaction level) contributes to the evaluation value during the calculation process of the basic evaluation value. For example, if the success rate is 85%, the contribution from the success rate would be 85 × 0.5 = 42.5 points. If the resolution period evaluation score is 25 points, the contribution from the resolution period is 25 × 0.3 = 7.5 points. If the satisfaction rating is 90 points, the contribution from satisfaction is 90 × 0.2 = 18 points. The sum of these values ​​constitutes the base evaluation score.

[0085] (Method for visualizing contribution) As shown in Figure 7, the presentation means 40 displays the contribution of each evaluation element using a pie chart, bar graph, or radar chart. For example, a pie chart might show that the success rate contributes 60%, the resolution time contributes 15%, and satisfaction contributes 25%.

[0086] (Comparison display function) The presentation means 40 has a function to display and compare the contributions of multiple agents side by side. This allows the user to visually understand the strengths and weaknesses of each agent. For example, you can quickly grasp characteristics such as Agent A having a high success rate but a long resolution period, while Agent B having a shorter resolution period but slightly lower satisfaction levels.

[0087] (Detailed analysis function) The presentation means 40 can further subdivide and display each evaluation element. For example, the success rate can be displayed separately for each technical field and project type to show which areas the company is particularly strong in.

[0088] (Sensitivity analysis) The presentation means 40 includes a function to simulate how the evaluation value changes when the weight of the evaluation element is changed. If the user sets "I want to prioritize the success rate," the weight is changed, the evaluation is re-evaluated, and the results are displayed.

[0089] (Improved transparency) Visualizing contributions improves the transparency of the evaluation process. Users can select a representative not just based on high evaluation scores, but also after understanding the reasoning behind them.

[0090] (Improved reliability) Having clear evaluation criteria makes it easier for users to trust the system's evaluation results. This, in turn, improves system usage and user satisfaction.

[0091] (Effects of the fourth embodiment) According to this embodiment, by visualizing the contribution of each evaluation element, the transparency and explainability of the evaluation can be improved, thereby supporting appropriate decision-making by users.

[0092] <Fifth Embodiment> In the fifth embodiment, a conflict of interest detection function will be described.

[0093] (The necessity of detecting conflicts of interest) When selecting an agent, checking for conflicts of interest is extremely important. If an agent has previously acted as an agent for a competitor, there is a possibility that their interests may conflict with those of the current client. This embodiment provides a function to automatically detect such conflicts of interest.

[0094] (Configuration of the conflict of interest detection means) The agent search system includes a conflict of interest detection means 16. The conflict of interest detection means 16 compares the information of the entities involved in the agent's past cases with the information of the companies, etc., related to the user, and if a match or relationship is detected, it generates a warning information regarding the conflict of interest and presents it using the presentation means 40.

[0095] (Operation of the conflict of interest detection mechanism) The conflict of interest detection means 16 performs the following process. First, we obtain company information (company name, group company names, competitor company names) entered by users. Secondly, the system retrieves information on the entities involved in cases that each agent targeted for the search has been involved in in the past from the performance memory unit 10. Thirdly, user company information and involved entity information are compared to detect matches or relationships. Fourth, if a match or relevance is detected, a conflict of interest warning will be generated. Fifth, warning information is presented to the user through the presentation means 40.

[0096] (System Configuration) As shown in Figure 8, the agent search system equipped with the conflict of interest detection means 16 includes a subject information storage unit, a user information acquisition unit, a matching unit, and a conflict determination unit.

[0097] (Subject information storage unit) The Subject Information Storage Unit is a database that stores the information of the entities involved in each agent's past cases, which is stored in the Performance Storage Unit 10. The information of the entities involved includes information such as the name of the applicant company, the name of the co-applicant, the name of the related company, or the name of the client company in past cases, and indicates which companies the agent has worked for in the past, or which parties the agent dealt with as an opposing party.

[0098] The information on the participating entities includes the names of the client, applicant, co-applicant, and opposing company, as well as their identifying information (corporate number, company code, industry classification, etc.). The involvement category of the agent for each participating entity (client, opposing party, auxiliary participant, etc.) is also stored.

[0099] (User Information Acquisition Department) The User Information Acquisition Unit is responsible for acquiring identifying information about users (companies or individuals) who perform agent searches. Users enter their company name or corporate group name on the search screen, and that information is acquired as user information.

[0100] Subject information refers to information including the name of the company, group company, or competitor to which the user (company or individual) conducting the search belongs. Furthermore, by linking with external company databases, it is also possible to obtain information such as the company group to which the user belongs, affiliated companies, partner companies, joint research partners, and licensing agreements.

[0101] (Verification section) The matching unit is responsible for comparing the information of each agent involved, stored in the subject information storage unit, with the user information acquired by the user information acquisition unit. Matching is the process of confirming whether there is a match or relationship between the companies that the agent has been involved with in the past and the company or related company to which the user belongs.

[0102] The matching process can perform name similarity matching, taking into account abbreviations, English spellings, etc., in addition to exact matches. The conflict of interest detection means 16 may be configured to normalize variations in company name spellings (e.g., Co., Ltd. / (Co., Ltd.)) before matching, and further perform a matching determination considering parent company, subsidiary, and group relationships by referring to an affiliate company information database.

[0103] (Conflict judgment section) The conflict of interest determination unit determines, based on the matching results, whether it would be inappropriate for the agent to accept a new case if a relationship exists between the parties involved in past cases and the user. For example, if the agent has previously handled patent application work as an agent for a competitor of the user, a "conflict of interest" will be determined.

[0104] The conflict of interest determination unit has the function of determining the risk level of a conflict of interest based on the detected relationship. The risk level is an indicator of the severity of the conflict of interest and is classified into three stages, for example, high, medium, and low. High risk is when the company is currently representing a company that is a direct competitor, or when it is representing the opposing party in a disputed case. Medium risk is when the company has previously represented a competitor in a field that overlaps with its technology, or when it has a relationship with a joint research partner or licensing agreement partner. Low risk is when the company is only involved in cases of indirectly related companies (such as different business divisions of group companies) or in old cases (more than 5 years ago).

[0105] (Generating warning information) If a match or relevance is found, the conflict of interest detection means 16 generates warning information indicating a potential conflict of interest. The warning information is a message and related data to notify the user that a conflict of interest has been detected, and includes the type of conflict of interest, the risk level, information on the underlying past case, and a specific warning message.

[0106] Warning information includes the following elements: • Types of conflicts of interest (competitors, group companies, joint research partners, licensing partners, opposing parties in lawsuits, etc.) • Risk level (high, medium, low) • Information on past cases that serve as the basis (application number, applicant name, technology classification, agency date, etc.) • Specific warning messages (e.g., "This agent has a history of handling cases for a competitor designated by your company.")

[0107] One example of a warning information display format is a message card that includes the detection type (exact match / related company match / name variation match), basis (application number, publication number, etc.), number of cases, period of involvement, relevance score, and recommended action (exclusion of the agent, whether additional verification is necessary, etc.). This allows users to immediately check the basis of the warning and the recommended action on the same screen.

[0108] (Warning) The presentation means 40 displays warning information corresponding to the agent when displaying the agent list. Preferably, the presentation format is to display the agent search results list with categories such as "Conflict of Interest," "Confirmation Required," and "No Problem."

[0109] The presentation method 40 can change its display format according to the severity of the conflict of interest. In the case of high risk, a red warning icon and a detailed message are displayed, recommending exclusion from the selection candidates. In the case of medium risk, a yellow caution icon is displayed, prompting confirmation. In the case of low risk, it is presented in a faint display as informational material.

[0110] The information used for matching the parties involved includes not only the name of the agent (firm name) but also the name of the patent attorney in charge. When issuing a warning, not only the name of the firm (A Patent Firm) but also the name of the patent attorney in charge of the competing case (Taro Kono) will be listed. This will allow users to make more detailed risk assessments, such as "the firm is the same, but the patent attorney in charge is different."

[0111] (Customizing warnings) When registering subject information, users can set a "warning level (e.g., high, medium, low)" and a "monitoring period (e.g., within the last 3 years, within the last 5 years, all period)" for each competitor. The conflict of interest detection means 16 performs matching based on these custom settings. For example, if "monitoring period: within the last 3 years" is set, cases older than 3 years will be excluded from matching. The warning display color and urgency are changed and presented according to the set warning level (high, medium, low).

[0112] (Specific reason for warning) When issuing a warning about a conflict of interest, the presentation method 40 does not simply display "Warning," but clearly indicates the specific grounds for the detection. For example, it provides the following detailed information: • Warning level: High (direct competition and subject matter matching) • Target of detection: X Corporation (direct competitor) • Basis for this case: Patent Application No. 2024-XXXXXX ·Responsibility period: 2024 (1 year ago) • Technology field: G06N(AI) - (※Matches the search field this time) • Patent attorney in charge: Taro Kono (A Patent Office) This allows users to understand the basis of warnings in detail and make more appropriate risk assessments.

[0113] (Technical implementation) This section describes the technical implementation of the conflict of interest detection method. The corporate normalization dictionary is a dictionary that uniformly manages variations in company name spelling, aliases, former trade names, mergers and acquisitions, and trade name changes. This dictionary allows for the recognition of variations in spelling such as "ABC Corporation," "ABC Corporation," "ABC (Co., Ltd.)," and "ABC Co., Ltd." as belonging to the same company.

[0114] The data sources used will be an integrated collection of patent publications (applicant information, co-applicant information), examination progress information (history of agent changes, etc.), litigation databases (litigation parties, agent information), appeal databases (appellant, respondent, agent information), firm publication records (patent firm websites, etc.), corporate information databases (capital relationships, group structure), and transactional relationship information (license agreements, joint research agreements, etc.).

[0115] In terms of classifying the types of involvement, the types of involvement by agents are classified and recorded as follows: Agent = Applicant side (representing the client), Opposing side (representing the opposing party in litigation or appeal), Third party (intervenor, etc.), Co-applicant side (representing one party in a joint application).

[0116] In terms of managing the time axis, we manage the temporal information of past projects and consider factors such as the simultaneity of past projects (high risk if multiple competing companies were represented simultaneously), the determination of ongoing engagements (high risk for ongoing projects), and the reduction of weight given to older projects (risk assessment is lowered for projects older than 5 years). effect

[0117] According to this embodiment, conflicts of interest can be automatically detected during the agent search stage, thus avoiding the risk of users selecting an ethically or legally inappropriate agent. By automating the conflict of interest check, which previously required manual verification, oversights and omissions can be prevented. Furthermore, automating the matching process significantly reduces the time required for conflict of interest verification, preventing human error. In addition, by referencing information on similar names and related companies, it becomes possible to detect potential conflicts of interest early, which was previously difficult.

[0118] <Combination of Embodiments> The first to fifth embodiments described above can be implemented independently, or multiple embodiments can be combined and implemented in combination.

[0119] (Preferred combination 1: Difficulty adjustment and recommendation generation) By combining the difficulty correction function of the second embodiment with the recommendation generation means of the third embodiment, it is possible to prioritize the recommendation of agents capable of handling difficult cases. This combination has the effect of allowing users to efficiently find experienced agents when requesting highly difficult cases.

[0120] (Preferred combination 2: Contribution visualization and conflict of interest detection) By combining the contribution visualization function of the fourth embodiment with the conflict of interest detection means of the fifth embodiment, it is possible to understand the basis for evaluating the agent while also confirming the risk of a conflict of interest. This combination has the effect of enabling users to make more appropriate decisions by considering both the agent's strengths and potential risks.

[0121] (Preferred combination 3: Integration of all embodiments) By combining all of the first through fifth embodiments, the most comprehensive agent search system can be realized. In this configuration, all functions such as objective evaluation based on past performance, precise evaluation through difficulty correction, recommendations tailored to user needs, visualization of evaluation basis, and automatic detection of conflicts of interest are integrated, resulting in the most user-friendly and reliable system.

[0122] In particular, the difficulty adjustment function appropriately evaluates the track record of handling highly difficult cases, the recommendation generation method prioritizes the presentation of agents that meet the user's needs, the visualization of contributions ensures transparency in evaluations, and the detection of conflicts of interest avoids ethical risks, resulting in a combination of benefits. This combination represents the most desirable embodiment for an agent search system.

[0123] <Variation> The embodiments described above may be modified as follows.

[0124] (Variation 1: Weight adjustment of multiple evaluation metrics) The weights used to calculate the basic evaluation score (success rate x 50, resolution time evaluation score x 30, satisfaction evaluation score x 20) can be dynamically adjusted according to the client's requests. For example, the weight of the resolution time evaluation score can be increased for urgent cases, and the weight of the satisfaction evaluation score can be increased for quality-focused cases.

[0125] (Example 2: Optimization of evaluation models using machine learning) By building a machine learning model using past selection results and client feedback as training data, the calculation method for evaluation values ​​and weight values ​​can be automatically optimized. This enables more accurate recommendations.

[0126] (Modification example 3: Integration of real-time information) Integrating real-time information such as the agent's current workload, availability, and fee structure will enable more practical recommendations. For example, an agent who is highly rated but currently too busy to take on a case will be excluded from the list of recommended candidates.

[0127] (Variation 4: Handling international cases) By adding information such as the agent's foreign language proficiency, network with overseas offices, and experience with international applications, we can handle international cases. For example, we can incorporate experience with PCT applications and Paris Convention applications into our evaluation.

[0128] (Variation 5: Application of blockchain technology) By applying blockchain technology to the performance and results storage sections, data tampering prevention and transparency can be ensured. Agent performance data is stored without tampering, allowing users to select agents based on highly reliable information.

[0129] (Modification 6: Evaluation adjustment using confidence coefficient) For agents with a limited number of past cases, applying a statistical confidence interval to the evaluation value allows for a more stable assessment. For example, for agents with fewer than 10 past cases, a discount factor is applied to the evaluation value. The confidence-adjusted evaluation value is calculated by multiplying the adjusted evaluation value by the confidence coefficient. The confidence coefficient is set to 1.0 for 30 or more past cases, 0.9 for 20 to less than 30 cases, 0.8 for 10 to less than 20 cases, and 0.7 for fewer than 10 cases. This method prevents agents without sufficient experience from being overvalued.

[0130] As described above, the agent search system of the present invention can perform objective and quantitative evaluations based on the agent's past case information and performance data, and present the user with the most suitable agent. Furthermore, functions such as difficulty correction, recommendation generation, contribution visualization, and conflict of interest detection can improve user satisfaction and avoid ethical and legal risks in agent selection. [Examples]

[0131] The following describes embodiments of the present invention.

[0132] (Example 1: Searching for a patent application agent) Users use the agent search system to find an agent to handle patent applications in the AI ​​technology field. Users enter the following search criteria: Specialty field "Patents (AI field)", Case type "Patent application agency", Desired difficulty level "High", Number of past cases handled "50 or more".

[0133] The extraction means 20 extracts Agent A, Agent B, and Agent C from the performance storage unit 10 as agents with experience in handling patent applications in the field of AI. Next, the evaluation calculation means 30 analyzes the performance information of each agent's past cases stored in the results storage unit 12.

[0134] Agent A's past performance is as follows: He has handled 100 patent applications in the AI ​​field, of which 85 were granted registration (85% success rate). He has handled 30 highly difficult cases (difficulty coefficient of 1.2 or higher), of which 24 were successful (80% success rate for highly difficult cases). The average client satisfaction rate is 4.5 / 5.0. The average processing time is 300 days.

[0135] The evaluation calculation method 30 first calculates a basic evaluation score. 40 points are allocated for a success rate of 85%, 25 points for a satisfaction level of 4.5, and 15 points for a processing period of 300 days, resulting in a total basic evaluation score of 80 points. Next, a difficulty adjustment is made considering the agent's experience with high-difficulty cases. Agent A has experience with 30 high-difficulty cases with an 80% success rate, so a difficulty adjustment coefficient of 1.3 is applied. The adjusted evaluation score is calculated as 80 points × 1.3 = 104 points.

[0136] Similarly, for Agent B, who has handled 80 cases in the past with an 82% success rate, a satisfaction rating of 4.2, an average processing time of 350 days, a base evaluation score of 75 points, and a difficulty adjustment coefficient of 1.23 applied to 20 high-difficulty cases, the adjusted evaluation score is calculated to be 92 points. For Agent C, who has handled 120 cases in the past with an 88% success rate, a satisfaction rating of 4.6, an average processing time of 280 days, a base evaluation score of 85 points, and a difficulty adjustment coefficient of 1.15 applied to 15 high-difficulty cases, the adjusted evaluation score is calculated to be 98 points.

[0137] The presentation means 40 presents the results to the user in a ranking format based on these evaluation values, with Agent A ranked 1st (104 points), Agent C 2nd (98 points), and Agent B 3rd (92 points). At the same time, the breakdown of each agent's evaluation value (success rate: 40 / 40 points, satisfaction: 25 / 30 points, processing time: 15 / 20 points, difficulty adjustment: ×1.3) is also displayed so that the user can understand the basis for the evaluation.

[0138] (Example 2: Searching for a trademark registration agent) Users search for an agent to handle their trademark registration. The search criteria include: Specialty field "Trademarks," Case type "Trademark registration application," Desired difficulty level "Medium," and Number of past cases handled "100 or more." Trademark cases tend to be less difficult than patent cases, hence the difficulty level is set to "Medium."

[0139] Extraction method 20 extracts agents D, E, and F who have experience in trademark registration. Agent D has handled 300 trademark registration cases in the past, with a success rate of 90%, a satisfaction rating of 4.2, an average processing time of 120 days, and a basic evaluation score of 85 points. Since many of the cases are of moderate difficulty, a difficulty correction factor of 1.0 is applied, and the adjusted evaluation score is also 85 points.

[0140] Agent E has handled 200 cases in the past, with a success rate of 92%, a satisfaction rating of 4.8, an average processing time of 100 days, a base evaluation score of 90 points, and an adjusted evaluation score of 90 points after applying a difficulty adjustment factor of 1.0. Agent F has handled 150 cases in the past, with a success rate of 88%, a satisfaction rating of 4.0, an average processing time of 140 days, a base evaluation score of 80 points, and an adjusted evaluation score of 80 points after applying a difficulty adjustment factor of 1.0.

[0141] Presentation method 40 presents agent E as first place (90 points), agent D as second place (85 points), and agent F as third place (80 points). In this way, appropriate evaluation values ​​can be calculated according to the type and difficulty of the case, and the most suitable agent can be presented to the user. In cases with relatively low difficulty, such as trademark cases, the difficulty correction coefficient tends to be close to 1.0, and the basic evaluation value tends to become the final evaluation value.

[0142] (Example 3: Search using conflict of interest detection function) User X searches for an agent regarding a patent dispute with competitor Y. In addition to the search criteria, the user enters their own company information (Company name: X, Related companies: X Group A, B, C) and information about competitor Y.

[0143] The extraction means 20 extracts agents G, H, and I as candidates through a normal search process. Next, the conflict of interest detection means 16 refers to the information of entities involved in each agent's past cases from the performance storage unit 10 to check whether there is any history of transactions with competitor company Y or its group companies.

[0144] As a result, it is discovered that agent G has previously handled five patent applications for company Y. The conflict of interest detection means 16 generates this information as a warning. The warning includes specific details such as, "Agent G has a track record of handling five patent applications for company Y (a competitor) in the past. Case details: Patent applications for AI-related technologies (2020-2022). Please be aware that there may be a conflict of interest."

[0145] The presentation means 40 displays the above warning information along with the evaluation information of agent G (evaluation value, area of ​​expertise, number of past cases, etc.). The user can use this warning information to select agent H or I, who do not have a risk of conflict of interest. In this way, the conflict of interest detection function allows the user to understand potential risks in advance and select an appropriate agent.

[0146] Furthermore, it is desirable that the conflict of interest detection means 16 be configured to detect not only exact matches, but also affiliated companies within a corporate group and variations in the spelling of company names (e.g., "Y Corporation" and "Y Corporation"). This will allow for more reliable detection of conflict of interest risks. [Explanation of Symbols]

[0147] 10. Performance Memory Unit 12 Results storage section 14 Difficulty Master 16. Conflict of Interest Detection Method 20 Extraction means 30 Evaluation calculation method 40 Presentation means 50 User Interface Section

Claims

1. An agent search system for searching for agents, A performance memory unit that stores past case information for each agent, A results storage unit that stores results information corresponding to each project, An extraction means for extracting the agent's past cases from the performance storage unit based on conditions entered by the user, An evaluation calculation means that analyzes the results information of past projects stored in the results storage unit and generates an evaluation value for the conditions, The system includes a presentation means for presenting a multiple agent based on the evaluation values ​​generated for each agent, The evaluation calculation means generates the evaluation value based at least on the depth of expertise and the weighting of time. Agent search system.

2. An agent search system for searching for an agent, A performance memory unit that stores past case information for each agent, A results storage unit that stores results information corresponding to each project, An extraction means for extracting the agent's past cases from the performance storage unit based on conditions entered by the user, An evaluation calculation means that analyzes the results information of past projects stored in the results storage unit and generates an evaluation value for the conditions, The system includes a presentation means for presenting a multiple agent based on the evaluation values ​​generated for each agent, The aforementioned performance storage unit stores information from each agent's past case information, including at least the technical classification. Agent search system.

3. An agent search system according to claim 1 or 2, The evaluation calculation means includes a function to correct the evaluation value according to the difficulty index, based on at least one of the past case information contained in the performance storage unit or the result information contained in the result storage unit. Agent search system.

4. An agent search system according to claim 1 or 2, The aforementioned presentation means includes a recommendation generation means, The recommendation generation means selects and presents agents as recommendation candidates that meet the user's requirements, based on the conditions entered by the user and the evaluation values. Agent search system.

5. An agent search system according to claim 1 or 2, The aforementioned presentation means visualizes and presents the contribution of each evaluation element to the evaluation value of each agent. Agent search system.

6. An agent search system according to claim 1 or 2, Equipped with means for detecting conflicts of interest, The conflict of interest detection means compares the information of the parties involved in the agent's past cases with the information of the user, and if a match or relationship is detected, generates a warning information regarding the conflict of interest and presents it using the presentation means. Agent search system.

Images