Computer-implemented procedure, computer program and system for mediating between applicant and service provider in a company

Abstract

A computer-implemented method for mediating between applicants and service providers in a business environment, wherein, by means of an input device of a data processing system (10), the applicant specifies, requests, or commissions expert knowledge and / or the service provider offers the expert knowledge (V1); a machine learning model (20) trained on descriptions of requests for expert knowledge and on descriptions of offered services is executed on a data processing device of the data processing system (10) (V2); the machine learning model (20) processes the specified, requested, or commissioned expert knowledge and / or the offered expert knowledge and calculates a match count for the mediation, wherein a higher match count indicates a greater match between the commissioned expert knowledge and the offered service (V3);the machine learning model (20) determines an explanation for a quality of match (V4); by means of an output device of the data processing system (10) the number of hits and the explanation are output to the applicant and / or the service provider (V5).;

Description

[0001] The invention relates to a computer-implemented method, a computer program and a system for mediating between applicants and service providers in a company.

[0002] The following definitions, descriptions and statements retain their respective meaning for and apply to the entire disclosed subject matter of the invention.

[0003] Systems and procedures for personnel recruitment, also known as recruitment, are known from the state of the art. Recruitment refers to the entire process of identifying, searching, selecting, pre-screening, and interviewing applicants for permanent or temporary positions within a company. Staffing also includes the process associated with selecting individuals for unpaid tasks. Managers, human resources professionals, and recruitment specialists may be entrusted with the implementation of staffing, but in some cases, parts of the process are also carried out by public bodies, commercial recruitment agencies, or specialized human resources consulting firms. Internet-based technologies that improve all aspects of staffing are now widespread, including the use of artificial intelligence.

[0004] For example, US 10 832 219 B2 discloses techniques for dynamically changing weighting factors of an artificial neural network to reweight attributes of an ordering model based on feedback in a streaming environment for candidate selection.

[0005] DE10 2013 008 860 A1 discloses a system that assesses the cultural fit of new employees—regardless of whether they are permanent employees or freelancers—with companies or individual organizational units down to the team level. A variety of corporate and organizational cultures are mapped into a homogeneous system. The determination, evaluation, and presentation of the fit are automated in real time.

[0006] Existing solutions for finding and hiring experts, especially for short-term projects within companies, are based on personal networks, job boards, recruitment agencies, or internal talent pools. These existing solutions often suffer from limitations in terms of scope, efficiency, and accuracy, leading to missed opportunities, lengthy hiring processes, and suboptimal project outcomes. Personal networks can be based on recommendations from acquaintances; for example, employees often rely on their own networks and recommendations from colleagues to find suitable experts. This disadvantage comes with limited reach and potential bias that can exclude qualified candidates from the search. On job boards and platforms, such as freelance platforms, companies can post project requests and find experts who have offered their services.Disadvantages include intense competition, the difficulty of assessing quality and suitability, and time-consuming candidate screening processes. Furthermore, companies may outsource the search for skilled workers to external agencies, known as staffing agencies, that specialize in sourcing and screening talent. Disadvantages include high costs, potential mismatch with the company culture and project requirements, and dependence on the agency's expertise and network. Companies may maintain internal talent pools and databases of previous contractors and consultants, which they can draw on when new projects arise. Disadvantages include outdated information, limited scalability, and the need for constant maintenance and updates.

[0007] The object of the invention was to provide an assessment of a need for experts and a received expert offer for this need, in order to obtain an improved matching model between demand and supply in the field of personnel placement. As described above, personnel placement also includes the search for or placement of experts for short-term and / or relatively small-volume assignments. The object of the invention was, in particular, how such an assessment can be generated using artificial intelligence.

[0008] The subject matter of the independent or subordinate claims each solves this problem. The disclosed methods, systems, and computer programs each offer their own functionalities. Advantageous embodiments of the invention emerge from the definitions, the dependent claims, the drawings, and the description of preferred embodiments.

[0009] According to one aspect, the invention provides a computer-implemented method for mediating between a requester and a service provider in a corporate environment. The mediation also includes the search for experts for short-term and / or relatively small-volume orders. For example, the requester has a specific need and requires support. Using the method, they can commission proven experts and have the request met quickly. The service provider is an expert in their field and offers their know-how as a service. The service provider can be an internal or external service provider. Using the method, they find and receive requests that match their range of services.

[0010] One step of the process involves the requestor inviting tenders for, requesting, or commissioning expert knowledge and / or the service provider offering the corresponding expert knowledge using an input device of a data processing system. According to one aspect, the requestor invites tenders for, requests, or commissions services whose execution requires corresponding expert knowledge. The service provider offers the services accordingly. The data processing system can be a company's IT platform, for example, a cloud computing platform. The input device can be, for example, a terminal, a prompt input of application software, or a mobile device, such as a notebook or a smartphone. For example, the requestor has the option of selecting a tender scope appropriate to their needs.For example, the service provider has the option of entering a meaningful description of their services, expertise, and / or references. This makes the service provider visible within the company and more relevant to previously unknown customers.

[0011] A further step of the method involves executing a machine learning model trained on descriptions of requests for expert knowledge and on descriptions of offered services on a data processing device of the data processing system. Execution may include loading and / or processing. The data processing device may include one or more processor units and memory units. The expert knowledge may include services related to the expert knowledge or problem-solving activities.

[0012] In a further step, the machine learning model processes the advertised, requested, or commissioned expert knowledge and / or the offered expert knowledge. It calculates a match score for the placement. This match score is also called a matching score. A larger match score indicates a greater match between the commissioned expert knowledge and the offered service. The match score calculated by the machine learning model enables a precise match between project requirements and expert offers. This achieves high-quality results and increases the project success rate. By tapping into a larger pool of experts using the machine learning model, the limitations of traditional personal networks are overcome and the visibility of the services of individual specialists is increased for potential customers.

[0013] In a further step, the machine learning model determines an explanation for the quality or goodness of the match. This means that an explanation can be provided even if one would not speak of a match in everyday language. For example, if the applicant is looking for training for high-voltage systems, the provider offers training for soft skills, this would not be a match; the only match would be training. However, the type of training shows that it is not a match. This explanation provides users with valuable insights and allows them to refine their project requirements or service offerings.In contrast to the prior art, the subject matter disclosed here differs, among other things, in the following ways: A coherent explanation is provided as to why a specific pair consisting of a demand request and a supply from an expert / service provider is a very good, fairly good, neutral, fairly poor, or very poor match. For example, a service provider might be an expert in a specific domain that is in the context of a broader domain. If the model now identifies a potential match, it may be that the specific request is a good fit only for the broader domain, but not for the specific domain. In the field of gear design, for example, there are metal and plastic gears. The state of the art in machine learning methods would recommend a possible match even if the service provider is an expert in plastic gears but not in metal gears.The subject matter disclosed here is capable of providing a rationale for each match, helping users in two ways: The service provider can choose only very specific matches that also consider expertise in plastic gears, thus avoiding wasting time on matches that don't fit their expertise. The service provider can also choose specific matches and clearly explore the opportunities for other expertise they could develop. Requesting users with a specific need could directly see which service provider has unique expertise in a particular area and derive actions accordingly.For the plastic gears example, the inquiring user can see that the suggested service provider only has expertise in metal gears, not plastic gears, and then decide whether or not to hire the service provider. When entering the query, the user could get an initial sense of how many potential service providers were found and also receive recommendations on how to fine-tune the query text to obtain more precise matches.

[0014] The number of hits and the explanation are output to the applicant and / or the service provider using an output device of the data processing system.

[0015] This process can be used to advantageously identify an expert for short-term or small-scale problems within a large corporation, especially when it comes to smaller tasks that arise unexpectedly during projects. The process is particularly focused on customized services that can be used individually by anyone and that are not already offered through another structured and established channel within the company.

[0016] According to a further embodiment of the method, the machine learning model determines a suggestion for the applicant as to how the applicant can modify the advertised, requested, or commissioned expertise in order to address an even better selection of available service providers. For example, when searching for a gear cutting service provider for plastic gears, it could suggest that the selection of service providers for specific plastic gears is very limited, and the applicant could also omit the specialization in plastic to find general gear cutting specialists. Conversely, it would also be possible for the applicant to seek a gear cutting specialist without a specific specialization, but the machine learning model recognizes that service providers with various specializations are available.Accordingly, the model could suggest to the requester that they refine the query and simultaneously indicate which refinement options would lead to good results. The machine learning model's suggestion can be output.

[0017] According to another embodiment of the method, the machine learning model determines a suggestion for the service provider as to how the service provider can reformulate the expertise / service offered. For example, the service provider could respond to a request for gearing for plastic gears. The machine learning model would recognize that the service provider currently has no plastic specialization stored in their service description and thus suggest that they could add it if they want to make their specialization in plastic gears visible. It is also possible that the service provider has stored a specialization in their profile that they (unknowingly) have not removed.If the service provider then encounters such a request and indicates with their feedback that this request is not a good fit for them, the machine learning model could recognize this and point out that they still have a specialization in their profile, but perhaps no longer want it to be visible. Accordingly, the machine learning model would suggest to the service provider how to reformulate their profile to remove this aspect in the future. The machine learning model's suggestion can be output.

[0018] According to a further embodiment of the method, the number of hits calculated by the machine learning model and / or the explanation determined by the machine learning model are assessed by a human. The assessments flow back into the training of the machine learning model. The assessment by a human is an interaction with the machine learning model through which the prediction of the machine learning model is improved. This interaction, which provides feedback to the machine learning model, is also referred to as human in the loop. As a result of this interaction, the machine learning model continually adapts its view of the world. By flowing the assessment back into the training process of the machine learning model, the machine learning model can continuously learn and adapt to improve the matching process and the recommendations. The training process is, for example, reinforcement learning.

[0019] According to another aspect, the machine learning model is a generative machine learning model. Generative machine learning models learn the patterns and structure of the input training data and then generate new data with similar characteristics. For example, the generative machine learning model is based on transformer machine learning models. For example, the machine learning model is a generative pre-trained transformer, such as GPT-n. By generating match scores between project requirements and expert offers using the generative machine learning model, a relatively accurate and efficient identification of suitable candidates is possible. By using match scores generated by the machine learning model between demand and supply, the invention effectively overcomes the search for experts in personal networks, streamlines the talent acquisition process, and increases the visibility of individual specialists' achievements.

[0020] Another aspect involves commissioning and / or offering expert knowledge using prompt engineering, for example, using a generative machine learning model. Prompt engineering. In prompt engineering, the description of the task the machine learning model is supposed to perform is embedded in the input, for example, in the form of a question, rather than being explicitly specified. In prompt engineering, one or more tasks are typically converted into a prompt-based dataset, and a language model is trained using prompt-based learning, or simply prompt learning. An example of a question that can be embedded in the machine learning model using prompt engineering is: Is there an expert for plastic gears?

[0021] Another aspect involves commissioning and / or offering expert knowledge related to support with process optimization in a production environment, technical consulting during a product development process, and system engineering. This allows, for example, experts to be found for short-term or small-scale problems within a large corporation.

[0022] According to another aspect, calculating the number of hits includes a semantic comparison of application details with the description of the services offered, a frequency comparison of keywords based on the number of keywords in the application and in the description of the services offered, or a combination of semantic comparison and frequency comparison. The semantic comparison can be performed, for example, using a bidirectional encoder representations from transformers language model. The BERT model determines how similar these descriptions are from a semantic perspective.

[0023] According to a further aspect, the invention provides a computer program for mediating between the requester and the service provider in a company. The computer program comprises program instructions that cause a computer to execute the steps of the method disclosed here according to one of the preceding claims when the computer program is loaded or executed by the computer.

[0024] The computer program can be provided as interpretable code that is loaded onto the computer, for example directly into the computer's internal memory, or as compiled code that is executed by the computer. During interpretation, an interpreter carries out operations on the program to be interpreted during runtime. The interpreter is, for example, another computer program that can enforce compliance with higher-level security policies during runtime of the computer program disclosed here. During compilation, a program written in one source language is translated into a program in another language that is equivalent in many respects. Provision can be made, for example, on a machine-readable medium on which the computer program is stored, for example on a non-volatile storage medium, or via a data carrier signal.The program instructions can be machine instructions, source code or object code, written in assembly language, an object-oriented programming language such as C++, or in a procedural programming language such as C.

[0025] According to a further aspect, the invention provides a system for mediating between a requester and a service provider in a company. The system is configured to determine, based on a result of a comparison between a request from a requester for expert knowledge and a description of a service from a service provider for the expert knowledge, wherein the request and the description are received in a first message from an input device, a hit count for the mediation, wherein a larger hit count indicates a higher match between the requested expert knowledge and the offered service, and an explanation for the match. Furthermore, the system is configured to initiate transmission of a second message to the requester and / or the service provider, the second message comprising the hit count and the explanation. The system can be a network entity.According to one aspect, the system is a cloud infrastructure, for example, a cloud computing platform. The cloud infrastructure, for example, the cloud computing platform, may comprise one or more computers, computer networks, one or more storage units, and one or more transmitting and receiving devices. For example, the receiving device receives the first message. The transmitting device, for example, sends the second message.

[0026] According to a further embodiment, the system comprises a hardware environment. The hardware environment is configured to execute a machine learning model trained on descriptions of requests for expert knowledge and on descriptions of offered services to determine the number of hits and the explanation.

[0027] According to a further aspect, the system comprises a hardware environment configured to execute a machine learning model trained on descriptions of requests for expert knowledge and on descriptions of offered services to determine the number of hits and the explanation, wherein the machine learning model is a generative machine learning model and the commissioning of the expert knowledge and / or the offering of the expert knowledge is carried out by means of prompt engineering.

[0028] According to a further aspect, the calculation of the number of hits by the system comprises a semantic comparison of application details with the description of the services offered, a frequency comparison of keywords based on a number of keywords in the application and in the description of the services offered, or a combination of semantic comparison and frequency comparison.

[0029] In one aspect, the system is configured to perform the method disclosed herein.

[0030] Instead of using general job boards or freelance platforms, companies can use the content disclosed here to access specialized platforms specific to their industry or field, offering a more targeted selection of qualified experts. The content disclosed here also allows companies to develop programs to rotate internal employees between different departments or projects, allowing them to share their expertise and acquire new skills. Invention.

[0031] The invention is illustrated in the following exemplary embodiments. They show: Fig. 1 an embodiment of a method disclosed here, Fig. 2 shows a further embodiment of the method disclosed here and Fig. 3 an embodiment of a system disclosed here.

[0032] In the figures, identical reference symbols designate identical or functionally similar reference parts. For clarity, only the relevant reference parts are highlighted in the individual figures.

[0033] Fig. Figure 1 shows the computer-implemented method for mediating between a requester and a service provider in a company. In a step V1, the requester requests expert knowledge and / or the service provider offers the expert knowledge. The request and offer are made via an input device of a data processing system 10. In a step V2, a machine learning model 20 trained on descriptions of requests for expert knowledge and on descriptions of offered services is executed on a data processing device of the data processing system 10. In a step V3, the machine learning model 20 processes the requested expert knowledge and / or the offered expert knowledge and calculates a hit count for the mediation, with a larger hit count indicating a higher match between the requested expert knowledge and the offered service.In a step V4, the machine learning model 20 determines an explanation for the match. In a step V5, the number of hits and the explanation are output to the applicant and / or the service provider via an output device of the data processing system 10.

[0034] The Fig. The method shown in Figure 2 corresponds to the method shown in Figure 1 except for step V4. Fig. In the embodiment shown in Figure 2, in a step V6, the machine learning model 20 determines, based on the explanation from step V4, a suggestion for the applicant as to how the applicant can modify the advertised, requested, or commissioned expert knowledge in order to address an even better selection of available service providers. In a step V7, the machine learning model 20 determines, based on the explanation from step V4, a suggestion for the service provider as to how the service provider can reformulate its offered expert knowledge / service. The suggestion for the applicant and the suggestion for the service provider are output or provided in step V5, along with the number of hits and the explanation.

[0035] Fig.Figure 3 shows the system 30 for mediating between the requester and the service provider within a company. The system 30 is, for example, a cloud computing platform. The data processing system 10 is integrated into the cloud. For example, computing and storage capacities and the data processing device of the data processing system 10 are provided as cloud services, for example, as infrastructure-as-a-service. Reference symbol 10 Data processing system 20 Machine learning model 30 systems V1-V7 procedural steps QUOTES CONTAINED IN THE DESCRIPTION

[0000] This list of documents submitted by the applicant was generated automatically and is included solely for the convenience of the reader. This list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions. Cited patent literature

[0000] US 10 832 219 B2

[0004] DE 10 2013 008 860 A1

[0005]

Claims

[1] Computer-implemented procedure for mediating between applicant and service provider in a business environment, whereby • by means of an input device of a data processing system (10) the applicant advertises, requests or orders expert knowledge and / or the service provider offers the expert knowledge (V1); • a machine learning model (20) trained on descriptions of requests for expert knowledge and on descriptions of services offered is executed on a data processing device of the data processing system (10) (V2); • the machine learning model (20) processes the advertised, requested or commissioned expert knowledge and / or the offered expert knowledge and calculates a number of hits for the mediation, whereby a larger number of hits indicates a higher match between the commissioned expert knowledge and the offered service (V3); • the machine learning model (20) determines an explanation for a quality of agreement (V4); • the number of hits and the explanation for the applicant and / or the service provider are output by means of an output device of the data processing system (10) (V5). [2] The method of claim 1, wherein the machine learning model (20) determines a suggestion for explaining how the requester can modify the advertised, requested or commissioned expert knowledge in order to address an improved selection of available service providers (V6). [3] Method according to one of the preceding claims, wherein the machine learning model (20) determines a suggestion for explaining how the service provider can reformulate its offered expert knowledge (V7) in order to address an improved selection of potential applicants. [4] Method according to one of the preceding claims, wherein the number of hits calculated by the machine learning model (20) and / or the explanation determined by the machine learning model (20) are assessed by a human and the assessments are fed back into the training of the machine learning model (20). [5] Method according to one of the preceding claims, wherein the machine learning model (20) is a generative machine learning model (20). [6] Method according to one of the preceding claims, wherein the commissioning of the expert knowledge and / or the offering of the expert knowledge is carried out by means of prompt engineering. [7] Computer program for mediating between applicant and service provider in a company, the computer program comprising program instructions which cause a computer to carry out the steps of the method according to one of the preceding claims when the computer program is loaded or executed by the computer. [8] System (30) for mediating between applicant and service provider in a company, the system is configured to: • Determining, based on a result of a comparison between a request from a requester for expert knowledge and a description of a service from a service provider for the expert knowledge, wherein the request and the description are received in a first message from an input device, a hit number for the mediation, wherein a larger hit number indicates a higher match between the requested expert knowledge and the offered service, and an explanation for the match; • Initiate transmission of a second message to the applicant and / or the service provider, the second message comprising the hit count and the explanation. [9] System (30) according to claim 8, comprising a hardware environment configured to execute a machine learning model (20) trained on descriptions of requests for expert knowledge and on descriptions of services offered to determine the number of hits and the explanation. [10] System (30) according to claim 8, comprising a hardware environment which is designed to execute a machine learning model (20) trained on descriptions of requests for expert knowledge and on descriptions of offered services in order to determine the number of hits and the explanation, wherein the machine learning model (20) is a generative machine learning model (20) and the commissioning of the expert knowledge and / or the offering of the expert knowledge is carried out by means of prompt engineering. [11] System (30) according to one of claims 8 to 10, wherein the calculation of the number of hits comprises a semantic comparison of application details with the description of the services offered, a frequency comparison of keywords based on a number of keywords in the application and in the description of the services offered, or a combination of semantic comparison and frequency comparison.

Images