Information processing device, information processing method, and program

Abstract

The goal is to comprehensively estimate a person's skills. [Solution] The information processing device includes a skill estimation unit. The skill estimation unit derives the similarity between input text data 30 of a predetermined person and skills corresponding to a predefined skill list, and outputs the estimated skills 50 related to the person based on the similarity. The skill estimation unit performs skill estimation on the input text data 30 using multiple different skill estimation methods, namely techniques M1, M2, M3, and M4.

Description

Technical Field

[0001] The present invention relates to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] Conventionally, a system for estimating a user's skills from a carrier summary (job summary) text freely described by the user is known. For example, a skill estimation device that estimates the skills of a person when related terms defined for each skill are used in data such as emails or the frequency of appearance is above a predetermined level is known (see Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, since the above-described conventional skill estimation device estimates skills using one type of similarity algorithm, there is a possibility that skills cannot be comprehensively estimated. Therefore, there is a demand for estimating also skills that have been overlooked by the conventional skill estimation device.

[0005] An object of the present invention is to comprehensively estimate a person's skills.

Means for Solving the Problems

[0006] To solve the above problems, the information processing device of the present invention includes a skill estimation unit that derives the similarity between textual information of a predetermined person and skills corresponding to a predefined skill list, and outputs estimated skills possessed by the person based on the similarity, wherein the skill estimation unit derives the similarity for the textual information using a plurality of different skill estimation methods. [Effects of the Invention]

[0007] This invention makes it possible to comprehensively estimate a person's skills. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the functional configuration of an information processing device in an embodiment of the present invention. [Figure 2] This diagram shows an overview of the skill estimation process. [Figure 3] This is a flowchart showing the skill estimation process. [Figure 4] This figure shows an example of the processing method of the first method. [Figure 5] This figure shows an example of training data for BERT in the second method. [Figure 6] This figure shows an example of the processing method of the second method. [Figure 7] This figure shows an example of training data for BERT in the third method. [Figure 8] This figure shows an example of the processing method of the third method. [Figure 9] This figure shows another example of the processing of the third method. [Figure 10] This figure shows an example of the processing method of the fourth method. [Figure 11] This figure shows an example of the final similarity score for each skill in the first to fourth methods. [Figure 12] This figure shows an example of an embodiment and a processing result of a conventional method for input text data. [Modes for carrying out the invention]

[0009] Embodiments of the present invention will be described in detail below with reference to the drawings. However, the scope of the present invention is not limited to the illustrated examples.

[0010] Referring to Figure 1, the device configuration of this embodiment will be described. The information processing device 1 is an information processing device such as a PC (Personal Computer) installed in a department such as the human resources department of a company and used by administrators such as human resources staff. In particular, the information processing device 1 estimates and displays the skills that an employee is thought to possess, based on text data such as a career summary entered by the employee as a person belonging to the company, as estimated skills of the employee. However, the information processing device 1 is not limited to estimating the skills of employees of a company. For example, the information processing device 1 may be configured to estimate the skills of people in other groups, such as school teachers and students, or members of various organizations.

[0011] The internal functional configuration of the information processing device 1 is described below. PC1 comprises a CPU (Central Processing Unit) 11, an operation unit 12, a RAM (Random Access Memory) 13, a display unit 14, a storage unit 15, and a communication unit 16. Each part of PC1 is connected via a bus 17. The CPU 11 functions as a skill estimation unit.

[0012] The CPU 11 controls various parts of the PC 11. The CPU 11 reads a specified program from the system programs and various application programs stored in the memory unit 15 and loads it into the RAM 13. The CPU 11 executes various processes in cooperation with the loaded program. The operation unit 12 has a keyboard with various keys and a pointing device such as a mouse. The operation unit 12 receives key operations and position operations from the user and outputs the operation information to the CPU 11. The operation unit 12 may also include a touch panel provided on the display unit 14. The RAM 13 is a volatile semiconductor memory and has a work area where various data and programs are temporarily stored. The display unit 14 has an LCD (Liquid Crystal Display) or ELD (Electro-Luminescent Display) display panel. The display unit 14 displays various display information on the display panel according to the control of the CPU 11. The memory unit 15 is composed of an HDD (Hard Disk Drive), SSD (Solid State Drive), etc., and stores information in a readable and writable manner. The memory unit 15 stores various programs and information. In particular, the memory unit 15 stores the skill estimation program 151, the skill list 20, and the like. The skill estimation program 151 is a program for executing the skill estimation process described later. The skill list 20 is a list of pre-stored information containing multiple skills that can be used as estimation candidates. As shown in Figure 4, the skill list 20 stores terms that indicate skills such as "language processing," "image processing," and "speech analysis." The communication unit 16 is a wired or wireless communication module connected to a communication network such as the Internet. The CPU 11 sends and receives information with external devices such as PCs owned by employees via the communication unit 16.

[0013] Next, referring to FIGS. 2 to 12, the operation of the information processing apparatus 1 will be described. As shown in FIG. 2, the skill estimation process executed in the information processing apparatus 1 is a process of estimating the skills that are considered to be possessed by an arbitrary employee from the input text data 30 of the career summary input by the employee, and outputting (displaying) the estimated possessed skills 50 related to the employee. In skill estimation, the processes of four types of skill estimation methods M1, M2, M3, and M4, each having different processing contents, are all executed. The methods M1 to M4 are methods for independently estimating skills.

[0014] Method M1 is a method of finding words that match the skills in the skill list 20 from the input text data 30, determining the matching skills as the highest similarity, and outputting them as the estimated possessed skills. Method M2 is a method of analyzing the meaning of the entire input text data 30, deriving the similarity between the meaning of the text and the skills in the skill list 20, and outputting the estimated possessed skills based on the similarity. Method M3 is a method of analyzing the sentence structure of the input text data 30, extracting the words at locations where skills are considered to be described, deriving the similarity with each skill in the skill list 20, and outputting the estimated possessed skills based on the similarity. Method M4 is a method of extracting words (nouns) in the input text data 30, deriving the similarity for each skill in the skill list 20, and outputting the estimated possessed skills based on the similarity. Here, the similarity is the degree indicating the similarity between the skill candidate words or skill candidate sentences in the input text data 30 and each skill in the skill list 20. Also, in the present embodiment, each method only derives the similarity, and the estimated possessed skills are output based on the similarity derived by all the methods.

[0015] The processing for methods M2 to M4 is performed using a pre-trained machine learning model. The pre-trained models for methods M2 to M4 are, for example, included in the skill estimation program 151, but they may also be stored separately in the memory unit 15. Each pre-trained model is machine-trained using training data (teaching data) to improve the accuracy of deriving the similarity of each skill in the skill list 20 to the input text data 30. The training data includes, for example, combinations of employee input text data and the skills that the employee actually possesses. Method M1 does not use a pre-trained model. In method M1, the similarity of a skill in the skill list 20 that matches a term in the input text data 30 is set to 1, and the similarity of a skill that does not match is set to 0. In this way, in each of methods M1 to M4, the similarity of each skill is a common parameter that represents the degree of similarity with a numerical value between 0 and 1.

[0016] In the information processing device 1, pre-trained models of methods M2 to M4 are incorporated into the skill estimation program 151. Furthermore, the information processing device 1 receives input text data for career summaries, pre-entered by any employee, from the employee's PC via, for example, the communication unit 16, and stores it in the storage unit 15. For example, a user acting as an administrator inputs an execution instruction for the skill estimation process via the operation unit 12. Triggered by this execution instruction, the CPU 11 executes the skill estimation process according to the skill estimation program 151. As shown in Figure 3, the CPU 11 receives a specified input of input text data for the target employee (referred to as the target employee) from the user via the operation unit 12 (step S1). In step S1, the CPU 11 reads the specified input text data from the storage unit 15.

[0017] The CPU 11 executes the processing of method M1 (step S2). As shown in FIG. 4, in step S2, when the same character string (carrier candidate word) as each skill in the skill list 20 is included (completely matches) in the input sentence data, the CPU 11 derives the similarity of the skill as 1. The CPU 11 derives the similarity of the skills in the skill list 20 that are not included in the input sentence data as 0. For the processing of method M1, for example, the named entity extraction function of the morphological analyzer Sudachi is used. Assume that the morphological analyzer (not shown) is stored in the storage unit 15 in advance. All the skills in the skill list 20 prepared in advance are registered in the named entity dictionary of the named entity extraction function of the morphological analyzer. The CPU 11 inputs the input sentence data to the morphological analyzer to output the estimated possessed skills. For example, corresponding to the input sentence data "I have been researching image processing for many years", the CPU 11 derives the similarity of "image processing" that completely matches in the skill list 20 as 1. The skills that do not completely match in the skill list 20 are derived with a similarity of 0.

[0018] Returning to FIG. 3, the CPU 11 executes the processing of method M2 (step S3). In step S3, the CPU 11 extracts each sentence (skill candidate sentence) of the input sentence data, analyzes the meaning of each extracted sentence, and derives the similarity with the skills in the skill list 20 corresponding to the meaning of each sentence. For example, the pre-trained model of the large language model BERT is fine-tuned for analyzing the meaning of each sentence as a whole. Fine-tuning is to re-train a pre-trained model pre-trained using predetermined training data using another training data and finely adjust the parameters of the pre-trained model.

[0019] As shown in Figure 5, the trained model is fine-tuned using a dataset of, for example, 300 pairs of skills and their descriptive sentences as training data. As shown in Figure 6, this fine-tuning generates a trained BERT model 60 for analyzing the meaning of sentences, which is pre-stored in the memory unit 15. When the trained model 60 receives input sentence data consisting of, for example, one input sentence, it analyzes the meaning of the input sentence. The CPU 11 inputs the input sentence data of one input sentence into the trained model 60 and has it analyze the meaning of the input sentence. The CPU 11 derives the likelihood of each skill in the skill list 20 with respect to the analyzed meaning of the input sentence. The sum of the likelihoods of each skill in the skill list 20 corresponding to one input sentence is set to 1.0. The CPU 11 derives the likelihood of the skills corresponding to one input sentence as similarity. For example, it derives the similarity of each skill in the skill list 20 corresponding to the input sentence data "I have been researching image processing for many years". The similarity scores for "Programming," "Image Processing," "Speech Analysis," etc., are derived as 0.18, 0.7, 0.09, etc., for all 20 skills on the skill list, so that their sum equals 1.0.

[0020] Furthermore, when the trained model 60 receives input sentence data consisting of multiple input sentences, it extracts multiple input sentences and analyzes the meaning of each input sentence. The CPU 11 inputs the input sentence data of multiple input sentences into the trained model 60 and has it analyze the meaning of each input sentence. The CPU 11 derives the likelihood of each skill in the skill list 20 being related to the analyzed meaning of each input sentence. For example, the CPU 11 outputs the likelihood of the skill corresponding to each input sentence as a similarity. At this time, similarity to all skills in the skill list 20 is derived for each input sentence. The similarities corresponding to each input sentence are compared, and if different similarities are derived for the same skill depending on the input sentence, the highest similarity is adopted as the similarity for that skill and is derived as the similarity in method M2.

[0021] Returning to Figure 3, CPU 11 executes the processing of method M3 (step S4). In step S4, CPU 11 analyzes the structure of the input sentence, extracts words from the sections where skills are likely to be described, and derives similarity scores corresponding to the extracted words (skill candidate words). For example, a trained model of the large-scale language model BERT is fine-tuned to identify the sections where skills are likely to be described.

[0022] As shown in Figure 7, the trained model is fine-tuned using a dataset of approximately 3000 pairs of sentences using skills and the skills themselves as training data. As shown in Figure 8, this fine-tuning generates a trained BERT model 70 for analyzing sentence structure, which is pre-stored in the memory unit 15. When the trained model 70 receives input sentence data, it estimates and extracts words that appear to be skills in the input sentence. The CPU 11 inputs the input sentence data into the trained model 70 and has it extract words that appear to be skills in the input sentence. The CPU 11 creates latent vectors from each extracted word and each skill in the skill list 20, derives the similarity using cosine similarity, and outputs it. One method for latent vectorization is to use ChiVe. ChiVe is a Japanese word vector based on a large corpus and multiple granularity segmentation.

[0023] For example, corresponding to the input sentence data "I have been conducting research on image processing for many years," the extracted words "image processing" and "research" are extracted. The similarity between the extracted word "image processing" and each skill in Skill List 20 is derived, and the similarity between the extracted word "research" and each skill in Skill List 20 is derived. For example, the similarity for "image processing" in Skill List 20 is... "language processing": 0.1, "image analysis": 0.8, "speech analysis": 0.1..., and the similarity for "research" in Skill List 20 is... "programming": 0.25, "basic research": 0.7, "experiment": 0.1..., and so on. The similarity for each extracted word is derived for all skills in Skill List 20.

[0024] As shown in Figure 9, when the CPU 11 obtains multiple similarity scores for the same skill in the skill list 20 for each extracted word from the input sentence data, it uses the highest similarity score as the similarity score for that skill. For example, the words "experiment" and "development" are extracted for the input sentence data "I have conducted many experiments to carry out development." The similarity scores between the extracted word "experiment" and each skill in the skill list 20 are derived, and the similarity scores between the extracted word "development" and each skill in the skill list 20 are derived. If the similarity score for the skill "basic research" corresponding to the extracted word "experiment" is 0.8, and the similarity score for the skill "basic research" corresponding to the extracted word "development" is 0.7, the higher similarity score for the skill "basic research" = 0.8 is adopted as the similarity score for that skill. The similarity between the extracted word "experiment" and the skill "programming" is 0.1, and the similarity between the extracted word "development" and the skill "programming" is 0.25. The higher similarity score for the skill "programming" (0.25) is adopted as the similarity score for that skill and is derived as the similarity score in method M3.

[0025] Returning to Figure 3, CPU 11 executes the processing of method M4 (step S5). In step S5, CPU 11 uses a morphological analysis tool (not shown) to extract words (nouns) from the input text data and derives the similarity between the extracted words (skill candidate words) and each skill in the skill list 20. The morphological analysis tool extracts only nouns from the input text data as extracted words. The morphological analysis tool is stored in the memory unit 15 in advance. CPU 11 creates latent vectors of all extracted words and each skill in the skill list 20, calculates the cosine similarity, and derives the similarity between each extracted word and all skills in the skill list 20. However, if nouns are consecutive in the input text data, such as "image processing", CPU 11 adds the latent vector of "image" and the latent vector of "processing" to obtain the latent vector of "image processing".

[0026] For example, corresponding to the input sentence data "I have been conducting research on image processing for many years," the nouns "for many years," "image," "processing," and "research" are extracted and converted into latent vectors. At this time, the latent vectors of consecutive extracted words "image" and "processing" are added together to form the latent vector of the extracted words "for many years," "image processing," and "research." The similarity between the extracted word "for many years" and each skill in skill list 20 is calculated, the similarity between the extracted word "image processing" and each skill in skill list 20 is derived, and the similarity between the extracted word "research" and each skill in skill list 20 is derived. Also, similar to method M3, if the CPU 11 obtains multiple different similarity values ​​for the same skill in skill list 20 corresponding to each extracted word from the input sentence data, it adopts the highest similarity value as the similarity value for that skill and derives it as the similarity value in method M4.

[0027] Returning to Figure 3, the CPU 11 displays the estimated skills held by the final employee corresponding to the input text data on the display unit 14 based on the similarity derived in steps S2 to S5 (step S6). In step S6, the CPU 11 displays a predetermined number of skills as estimated held skills, sorted by the highest final similarity, based on all the similarity values ​​of methods M1 to M4 corresponding to each skill in the skill list 20, for example, as shown in the similarity list 80 in Figure 11. In this case, if different similarity values ​​are derived for the same skill by each method, the highest similarity value is adopted as the final similarity value. If the same similarity value is derived for the same skill by each method, that same similarity value is adopted as the final similarity value. For example, if the similarity value of method M1 for the estimated held skill "Basic Research" is 1, and the similarity values ​​of methods M2, M3, and M4 are all 0.02, then the highest similarity value of method M1, which is 1, becomes the final similarity value. Note that the skills and similarities in similarity lists 80 and 90 in Figure 11 are different examples from those in Figures 4, 6, 8, and 9. Furthermore, CPU 11 may be configured to display similarity along with each estimated skill. Additionally, if, for the same estimated skill across methods M1 to M4, the similarity in at least one (or more) methods exceeds a predetermined similarity, CPU 11 may change the display method of the estimated skill according to the number of methods that exceed the limit. The display method may include font boldness (for example, bolder font the more methods exceed the limit), size, and color.

[0028] Furthermore, as shown in the similarity list 90 in Figure 11, the CPU 11 may be configured to derive the final similarity by adding the similarities for each skill in the same skill list 20 for methods M1 to M4. For example, if the similarity of the estimated possessed skill "Basic Research" for method M1 is 1, and the similarities for methods M2, M3, and M4 are each 0.02, then the sum of these values, 1.06, becomes the final similarity. In addition, the CPU 11 compares the final similarities of each skill and selects a predetermined number of skills in descending order of final similarity to display them as estimated possessed skills.

[0029] Returning to Figure 3, the CPU 11 receives input from the user via the operation unit 12 to select skills related to the target employee from the estimated skills displayed in step S6 (step S7). In step S7, the CPU 11 stores the selected skills in the memory unit 15 in association with the target employee's identification information, and terminates the skill estimation process. For example, the assignment location of each employee is determined based on the information such as the target employee's skills stored in the memory unit 15.

[0030] As shown in Figure 12, we compare the estimated skills obtained by performing the skill estimation process of this embodiment with the processing of only Method 4 as a conventional method, using an example of input text data. However, the predetermined number of estimated skills obtained by the skill estimation process is greater than the predetermined number of estimated skills obtained by the conventional method. The estimated skills obtained by the skill estimation process include all the estimated skills obtained by the conventional method, and other estimated skills are also derived. In Figure 12, the estimated skills with a gray background are misestimated estimated skills (skills that the target employee does not actually possess) and correspond to estimated skills that were not selected in step S7.

[0031] Furthermore, among methods M1 to M4, the priority of the method that derives the similarity that is thought to have high accuracy in skill estimation may be increased. Specifically, the CPU 11 stores and learns the estimated possessed skills selected in step S7, derives the accuracy of skill estimation corresponding to the similarity derived for each method, and sets the priority of the method with high accuracy to increase. In step S6 of the skill estimation process after this setting, the CPU 11 calculates the final similarity by adding the similarity of each skill after weighting (for example, ×100) the similarity derived for the high-priority method. The CPU 11 compares the final similarities and displays a predetermined number of estimated possessed skills in descending order of final similarity.

[0032] As described above, according to this embodiment, the information processing device 1 includes a CPU 11. The CPU 11 derives the similarity between input text data of a predetermined employee and skills corresponding to a predefined skill list 20, and displays the estimated skills held by the employee based on the similarity. The CPU 11 derives similarity from the input text data using multiple different methods M1 to M4. Therefore, the skills of employees can be comprehensively estimated as estimated skills held.

[0033] Furthermore, the CPU 11 derives the similarity between candidate skill words or candidate skill sentences in the input text data and the skills corresponding to the skill list 20. The CPU 11 displays a predetermined number of skills corresponding to the skill list 20 as estimated skills possessed by the employee, in descending order of similarity. Therefore, estimated skills possessed can be quantitatively determined easily and accurately using similarity.

[0034] Furthermore, the multiple methods include methods M1 to M4. Method M1 derives a similarity of a skill as 1 based on whether or not the input text data contains words that match the skills corresponding to the skills in the skill list 20. Method M2 extracts each sentence in the input text data as a skill candidate sentence, analyzes the meaning of the skill candidate sentence, and derives the similarity between each of the skills corresponding to the skills in the skill list 20 and each of the skill candidate sentences. Method M4 extracts words in the input text data as skill candidate words and derives the similarity between each of the skills corresponding to the skills in the skill list 20 and each of the words. Method M3 analyzes the sentence structure of the input text data, extracts words in sections where skills are likely to be described as skill candidate words, derives the similarity between each of the skills corresponding to the skills in the skill list 20 and each of the words, and estimates them as estimated possessed skills. Therefore, methods M1 to M4 can estimate employees' skills more comprehensively.

[0035] Furthermore, if the CPU 11 derives multiple different similarity values ​​for skills corresponding to the same skill list 20 using a predetermined method (e.g., M3), it adopts the highest similarity value. Therefore, similarity values ​​can be easily and appropriately derived using a predetermined method.

[0036] Furthermore, if the similarity of the same estimated skill exceeds a predetermined level in two or more of the multiple different methods M1 to M4, the CPU 11 displays the estimated skill related to the employee in a manner corresponding to the number of methods that exceeded the predetermined similarity level. This allows users to easily and visually recognize estimated skills that are highly likely to be possessed by an employee, based on similarity levels that exceed a predetermined level in multiple methods.

[0037] Furthermore, after deriving similarity scores using all different methods M1 to M4, if multiple different similarity scores are derived for the same skill list 20 using multiple different methods, the CPU 11 adopts the highest similarity score as the final similarity score. The CPU 11 then displays a predetermined number of skills corresponding to the skill list 20 as estimated skills possessed by the individual, in descending order of final similarity scores. This increases the accuracy of the estimated skills possessed by the employee.

[0038] Furthermore, the CPU 11 derives the sum of similarity scores from multiple different methods M1 to M4 for skills corresponding to the same skill list 20, and displays a predetermined number of skills corresponding to the skill list 20 as estimated skills possessed by the person, in descending order of the sum of these similarity scores. Therefore, the estimated skills possessed by an employee with a high degree of certainty can be easily determined by the sum of similarity scores from multiple methods.

[0039] Furthermore, CPU11 derives similarity by assigning weights according to priority to the highest-priority method among several different methods M1 to M4. This makes it possible to increase the accuracy of the estimated skills possessed by employees.

[0040] The above description discloses an example in which the storage unit 15 is used as a computer-readable medium for the program according to the present invention, but the invention is not limited to this example. Other computer-readable mediums that can be used include ROM (Read Only Memory) and portable recording media such as CD-ROM. Furthermore, a carrier wave can also be used as a medium for providing the program data according to the present invention via a communication line.

[0041] The above-described embodiments are merely examples of the information processing apparatus, information processing method, and program according to the present invention, and are not limited thereto.

[0042] For example, in the above embodiment, the skill estimation process in Figure 3 is configured so that the processes of methods M1 to M4 are executed one by one in that order in series, but it is not limited to this configuration. The serial execution order of the processes of methods M1 to M4 may be changed. Alternatively, the processes of methods M1 to M4 may be executed in parallel.

[0043] Furthermore, while Figure 3 shows a configuration in which four different methods M1 to M4 are executed in the skill estimation process, it is not limited to this configuration. The skill estimation process may be configured to execute at least two of the methods M1 to M4. Alternatively, the skill estimation process may be configured to execute at least one of the methods M1 to M4 along with at least one method (skill estimation method) different from methods M1 to M4. Alternatively, the skill estimation process may be configured to execute at least two methods (skill estimation methods) different from methods M1 to M4 instead of the methods M1 to M4.

[0044] Furthermore, in the skill estimation process shown in Figure 3, the CPU 11 is configured to display the estimated skill on the display unit 14, but this is not the only configuration. The output of the estimated skill may also be printed using a printing device (not shown), output as audio using a speaker, transmitted to an external device via the communication unit 16, or stored in the storage unit 15 or a storage medium (not shown).

[0045] In the skill estimation method described above, only similarity is derived for each of the methods M1 to M4, and the estimated skills possessed are output after the similarity has been derived for all methods. However, it is also possible to output the estimated skills possessed based on similarity for each method. In that case, the estimated skills possessed output for each method are considered estimated skill candidates, and after the output of estimated skill candidates for all methods is complete, a predetermined number of these candidates, ordered by similarity, may be displayed as the estimated skills possessed by the employee. This skill estimation method also allows for the comprehensive estimation of an employee's skills as estimated skills possessed.

[0046] Furthermore, when displaying estimated owned skills based on final similarity, the number of skills displayed may not be limited, and all skills whose final similarity exceeds a predetermined similarity may be displayed as estimated owned skills in descending order of final similarity.

[0047] While embodiments of the present invention have been described, the scope of the present invention is not limited to the embodiments described above, but includes the scope of the invention as described in the claims and its equivalents. [Explanation of Symbols]

[0048] 1 information processing unit, 11 CPUs

Claims

1. The system includes a skill estimation unit that derives the similarity between the textual information of a given person and the skills corresponding to a predefined skill list, and outputs the estimated skills possessed by the person based on the similarity. The information processing device is characterized in that the skill estimation unit derives the similarity score from the text information using a plurality of different skill estimation methods.

2. The information processing apparatus according to claim 1, characterized in that the skill estimation unit derives the similarity between a candidate skill word or candidate skill sentence in the text information and a skill corresponding to the skill list, and outputs the skills corresponding to the skill list in descending order of similarity as estimated skills possessed by the person.

3. The information processing apparatus according to claim 2, characterized in that the plurality of different skill estimation methods include at least two of the following methods: a method for deriving the similarity of a skill based on whether or not the text information contains words that match the skills corresponding to the skill list; a method for extracting each sentence of the text information as a candidate skill sentence, analyzing the meaning of the candidate skill sentences, and deriving the similarity between the skills corresponding to the skill list and each of the candidate skill sentences; a method for extracting words from the text information as candidate skill words, and deriving the similarity between the skills corresponding to the skill list and each of the words; and a method for analyzing the structure of the sentences in the text information, extracting words in places where skills are likely to be described as candidate skill words, and deriving the similarity between the skills corresponding to the skill list and each of the words.

4. The information processing apparatus according to claim 3, wherein the skill estimation unit adopts the highest similarity when multiple different similarity values ​​are derived for skills corresponding to the same skill list in a predetermined skill estimation method.

5. The information processing apparatus according to any one of claims 2 to 4, characterized in that, with respect to the similarity of the same estimated possessed skill, if two or more of the plurality of different skill estimation methods exceed a predetermined similarity, the skill estimation unit outputs the estimated possessed skill relating to the person in a manner corresponding to the number of skill estimation methods that exceeded the predetermined similarity.

6. The information processing apparatus according to any one of claims 2 to 4, characterized in that, after deriving the similarity in all of the multiple different skill estimation methods, if multiple different similarity values ​​have been derived for a skill corresponding to the same skill list by the multiple different skill estimation methods, the skill estimation unit adopts the highest similarity as the final similarity, and outputs the skills corresponding to the skill list as estimated skills possessed by the person in descending order of the final similarity.

7. The information processing apparatus according to any one of claims 2 to 4, characterized in that the skill estimation unit derives the sum of the similarity scores from the plurality of different skill estimation methods for skills corresponding to the same skill list, and outputs the skills corresponding to the skill list as estimated skills possessed by the person, in descending order of the sum of the sums.

8. The information processing apparatus according to any one of claims 2 to 4, characterized in that the skill estimation unit derives the similarity by applying weights according to the priority of the skill estimation method with the highest priority among the plurality of different skill estimation methods.

9. The information processing apparatus according to claim 3, characterized in that, each time the derivation of the similarity by each of the plurality of different skill estimation methods is completed, the skill estimation unit outputs a predetermined number of skills corresponding to the skill list in descending order of similarity as estimated possessed skill candidates, and outputs a predetermined number of estimated possessed skill candidates in all skill estimation methods in descending order of similarity as estimated possessed skills relating to the person.

10. The process includes a skill estimation step which derives the similarity between the written information of a given person and the skills corresponding to a predefined skill list, and outputs the estimated skills possessed by the person based on the similarity. An information processing method characterized in that, in the skill estimation step, the similarity score is derived from the textual information using a plurality of different skill estimation methods.

11. Computers, This system functions as a skill estimation unit that derives the similarity between the textual information of a given person and the skills corresponding to a predefined skill list, and outputs the estimated skills possessed by the person based on the similarity. The skill estimation unit is a program characterized by deriving the similarity score from the text information using a plurality of different skill estimation methods.

Images