Information provision device, information provision method, and program

Abstract

This invention provides an information provision device, an information provision method, and a program that can appropriately select the information to be provided. [Solution] The information providing device comprises: an acquisition unit that acquires input information which is input information; a word extraction unit that randomly extracts a portion of a plurality of words from a plurality of words based on the input information at a predetermined rate; a document vector calculation unit that calculates a first document vector which is a document vector based on the extracted portion of a plurality of words; a similarity calculation unit that calculates the similarity of a plurality of second document vectors which are a plurality of document vectors based on a plurality of provided information which is information related to the input information and provided to the input information, and the first document vector; and a presentation unit that presents a portion of the plurality of provided information based on each similarity.

Description

Technical Field

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

Background Art

[0002] Patent Document 1 describes a safety information providing system aimed at providing appropriate safety information for an input. In the safety information providing system described in Patent Document 1, a search information input means acquires and inputs a set keyword according to time from a plurality of set keywords that are set and stored in advance for each time and type of disaster and correspond to the type of disaster, and a search means performs a search using the set keyword input by the search information input means. According to this configuration, it is said that it is possible to prevent the search results from becoming routine. In this safety information providing system, for example, a plurality of days' worth of set keywords are input in advance by an administrator, and the set keyword selected on a daily basis is changed.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] As described above, in the safety information providing system described in Patent Document 1, by including a set keyword that changes daily in the keyword used for searching safety information, the search results are prevented from becoming routine. However, since the set keyword is set by, for example, an administrator, there is a problem that the effect of preventing routine may not always be sufficient depending on the setting method of the set keyword.

[0005] This disclosure was made to solve the above-mentioned problems and aims to provide an information provision device, an information provision method, and a program that can appropriately select the information to be provided. [Means for solving the problem]

[0006] To solve the above problems, the information providing device according to this disclosure comprises: an acquisition unit that acquires input information which is input information; a word extraction unit that randomly extracts a portion of a plurality of words from a plurality of words based on the input information at a predetermined rate; a document vector calculation unit that calculates a first document vector which is a document vector based on the extracted portion of a plurality of words; a similarity calculation unit that calculates the similarity of a plurality of second document vectors which are a plurality of document vectors based on a plurality of provided information which is information related to the input information and provided to the input information, and the first document vector; and a presentation unit that presents a portion of the plurality of provided information based on the similarity of each.

[0007] The information provision method relating to this disclosure includes the steps of: acquiring input information which is information that has been entered; randomly selecting a portion of a plurality of words from a plurality of words based on the input information at a predetermined rate; calculating a first document vector which is a document vector based on the extracted portion of a plurality of words; calculating the similarity between a plurality of second document vectors which are a plurality of document vectors based on a plurality of provided information which is information related to the input information and provided to the input information, and the first document vector; and presenting a portion of the plurality of provided information based on the similarity.

[0008] The program relating to this disclosure causes a computer to perform the following steps: acquire input information which is information that has been entered; randomly extract a portion of a plurality of words from a plurality of words based on the input information at a predetermined rate; calculate a first document vector which is a document vector based on the extracted portion of a plurality of words; calculate the similarity between a plurality of second document vectors which are a plurality of document vectors based on a plurality of provided information which is information related to the input information and provided to the input information, and the first document vector; and present a portion of the plurality of provided information based on the similarity. [Effects of the Invention]

[0009] According to the information provision device, information provision method, and program disclosed herein, it is possible to appropriately select the information to be provided. [Brief explanation of the drawing]

[0010] [Figure 1] This is a schematic diagram of the information provision system according to the embodiment of this disclosure. [Figure 2] This is a schematic diagram showing an example of a work instruction sheet according to an embodiment of this disclosure. [Figure 3] This is a schematic diagram illustrating an example of the work outline according to the embodiments of this disclosure. [Figure 4] This is a schematic diagram illustrating an example of the work content according to the embodiments of this disclosure. [Figure 5] This is a schematic diagram illustrating an example of the work requiring qualification according to the embodiments of this disclosure. [Figure 6] This is a schematic diagram illustrating an example of work requiring a work permit according to the embodiments of this disclosure. [Figure 7] This is a schematic diagram illustrating an example of the risks associated with the embodiments of this disclosure. [Figure 8] This is a schematic diagram showing an example of safety measures instructions according to the embodiments of this disclosure. [Figure 9] This is a schematic diagram showing an example of an output screen according to the embodiment of this disclosure. [Figure 10]It is a schematic diagram showing examples of laws and standards to be complied with according to an embodiment of the present disclosure. [Figure 11] It is a schematic diagram showing an example of risk assessment according to an embodiment of the present disclosure. [Figure 12] It is a schematic diagram showing an example of past disaster cases according to an embodiment of the present disclosure. [Figure 13] It is a schematic diagram showing an example of a disaster case report according to an embodiment of the present disclosure. [Figure 14] It is a schematic diagram for explaining an example of calculating a score regarding the relevance between a work instruction and laws and standards to be complied with according to an embodiment of the present disclosure. [Figure 15] It is a schematic diagram for explaining an example of calculating the similarity between a work instruction and a risk assessment according to an embodiment of the present disclosure. [Figure 16] It is a schematic diagram for explaining an example of calculating the similarity between a work instruction and a disaster case report according to an embodiment of the present disclosure. [Figure 17] It is a schematic diagram showing an example of an output screen according to an embodiment of the present disclosure. [Figure 18] It is a flowchart showing an operation example of an information providing device according to an embodiment of the present disclosure. [Figure 19] It is a schematic block diagram showing the configuration of a computer according to an embodiment of the present disclosure.

Mode for Carrying Out the Invention

[0011] Hereinafter, an information providing device, an information providing method, and a program according to an embodiment of the present disclosure will be described with reference to FIGS. 1 to 19. In each figure, the same or corresponding components are denoted by the same reference numerals, and the description thereof will be omitted as appropriate.

[0012] FIG. 1 is a schematic configuration diagram of an information providing system 100 according to an embodiment of the present disclosure. The information providing system 100 shown in FIG. 1 includes an information providing device 1 and a user terminal 2. The information providing system 100 is a system for performing natural language processing on work information 31, which is information related to the work to be performed, and providing safety information 37, which is information for reducing the risk of work and has a high relevance to the work information 31. In the present embodiment, the work information 31 and the work information 32 described later are examples of the "input information" according to the present disclosure. Also, in the present embodiment, the safety information 37 and the safety information 33 described later are examples of the "provided information" according to the present disclosure.

[0013] The information providing device 1 can be configured using a computer such as a personal computer or a server, and includes the following components as functional blocks composed of a combination of hardware such as a computer and peripheral devices, and software such as a program executed by the computer. That is, as a functional block, the information providing device 1 includes an acquisition unit 11, a word extraction unit 12, a database 13, a document vector calculation unit 14, a second document vector calculation unit 15, a similarity calculation unit 16, a presentation unit 17, a reception unit 18, a modification unit 19, a score calculation unit 20, and a target document preprocessing unit 21.

[0014] Also, the user terminal 2 is a terminal such as a personal computer, a tablet terminal, or a smartphone, and inputs the work information 31 and transmits it to the information providing device 1, or receives the safety information 37 and outputs it to an output unit such as a display unit.

[0015] Here, with reference to Figures 2 to 8, an example of the structure of the work information 31 will be described. Figure 2 is a schematic diagram showing an example of a work instruction sheet 31 (shown as work instruction sheet 31A in Figure 2) according to the embodiment of this disclosure. The work instruction sheet 31A shown in Figure 2 is an example of the structure of the work information 31, and is data including text information, etc., and includes work overview (information representing the work) 31A1, work content (information representing the work (hereinafter omitted)) 31A2, work requiring qualifications 31A3, work requiring a work permit 31A4, what kind of hazards there are 31A5, and safety measures instructions 31A6.

[0016] Figure 3 is a schematic diagram showing an example of a work summary 31A1 according to the embodiment of this disclosure. The work summary 31A1 shown in Figure 3 includes information such as "Work Order ID (Identification Code)", "Project Name", "Work Date", "Company Name", "Work Group Name", "Group Leader", "Number of Personnel", "Work Location", "Work Hours", "Person to be Confirmed", and "Work Title". In the work summary 31A1, the string before the symbol ":" is fixed information, and the string after the symbol ":" is entered (or selected) by, for example, the operator of user terminal 2. Note that multiple symbols "*" represent any character or symbol. Also, one or more symbols "〇 (circle)" described later also represent any character or symbol.

[0017] Figure 4 is a schematic diagram showing an example of work content 31A2 according to the embodiment of this disclosure. Work content 31A2 shown in Figure 4 includes tasks numbered 1 to 4: "Work preparation," "Char supply hopper char vacuuming, pulverized coal supply hopper pulverized coal removal," "Flexible container bag transport and temporary storage," and "Cleanup and tidying up." In work content 31A2 shown in Figure 4, "Work preparation," "Char supply hopper char vacuuming, pulverized coal supply hopper pulverized coal removal," "Flexible container bag transport and temporary storage," and "Cleanup and tidying up" are input (or selected) by the operator of the user terminal 2.

[0018] Figure 5 is a schematic diagram showing an example of a qualified task 31A3 according to the embodiment of this disclosure. The qualified task 31A3 shown in Figure 5 includes information on the name of the holder of the qualifications (license, completion of skills training, etc.) required for the task content 31A2 shown in Figure 4. In the qualified task 31A3 shown in Figure 5, the "name" is entered (or selected) by the operator of the user terminal 2.

[0019] Figure 6 is a schematic diagram showing an example of a task 31A4 that requires a work permit according to an embodiment of this disclosure. In this embodiment, the work permit is a permit for a task that requires special permission in addition to approval in this work order. The task 31A4 that requires a work permit shown in Figure 6 includes a checkbox 31A41 to indicate whether or not the task is required. In the task 31A4 that requires a work permit shown in Figure 6, the presence or absence of a check 31A42 for checkbox 31A41 is set by the operator of the user terminal 2.

[0020] Figure 7 is a schematic diagram showing an example of a potential hazard 31A5 according to the embodiment of this disclosure. The potential hazard 31A5 shown in Figure 7 includes information indicating the presence or absence of hazards such as "falling," "tipping," "collision," "being hit," "falling objects," "traffic accident," "contact with harmful objects," and "being caught or entangled." The potential hazard 31A5 shown in Figure 7 includes a checkbox 31A51 indicating the presence or absence of a hazard, and whether or not the checkbox 31A51 is checked is set by the operator of the user terminal 2.

[0021] Figure 8 is a schematic diagram showing an example of safety instructions 31A6 according to the embodiment of this disclosure. It includes information on matters requiring safety instructions for each type of work, such as "general work" and "work at height," "transportation work," and "work with oxygen deficiency hazards." The safety instructions 31A6 shown in Figure 8 include checkboxes 31A61 to indicate whether or not instructions are required. In the safety instructions 31A6 shown in Figure 8, the presence or absence of a checkbox 31A61 is set by the operator of the user terminal 2.

[0022] Returning to Figure 1, the acquisition unit 11 acquires work information 31, which is information related to the work to be performed, and stores it as work information 32 in a predetermined memory area.

[0023] The word extraction unit 12 extracts multiple words based on the work information 32, and further extracts a predetermined percentage (e.g., 50%) of some of the words extracted from the work information 32. Here, the multiple words based on the work information 32 are, for example, multiple words extracted from multiple sentences contained in the work information 32 by word segmentation processing. If, for example, 100 words are extracted from the work information 32, the word extraction unit 12 will randomly select, for example, 50 words. The word extraction unit 12 extracts words using, for example, strings entered or selected by the operator of the user terminal 2 contained in the work information 32, strings representing item names corresponding to the entered or selected strings, strings representing the content of items with check marks, etc. By deliberately using, for example, 50% of the words randomly by the word extraction unit 12, it is possible to prevent the results from diverging in the search for related documents compared to using 100%. Also, by randomly selecting, for example, 50% of the words, it is possible to prevent the search results for related information from becoming monotonous. This configuration ensures that, for example, the same results are not obtained even if the same task continues for several days to several weeks. However, as will be discussed later, for information regarding laws and standards, it may not be appropriate for the results to change randomly. Therefore, in this embodiment, all multiple words based on the work information 32 are used as input information. The word extraction unit 12 does not delete duplicate words even if the same word is extracted multiple times. In this case, the fact that a word has been extracted multiple times can be utilized as a characteristic.

[0024] Database 13 stores multiple pieces of safety information 33, which are information for reducing work risks. In this embodiment, safety information 33 includes rule information 33-1, which is information representing rules to be followed in the work; risk assessment information 33-2, which is information about past risk assessments; and accident case information 33-3, which is information about accident cases that have occurred in the past. Here, risk assessment means a series of procedures for identifying hazards and harmful elements in the workplace, estimating risks, setting priorities, and deciding on risk reduction measures. Note that safety information 33 may not include, for example, rule information 33-1. Also, safety information 33 may not include either risk assessment information 33-2 or accident case information 33-3.

[0025] Refer to Figures 9 to 13 to explain the outlines of rule information 33-1, risk assessment information 33-2, and disaster case information 33-3. Figure 9 is a schematic diagram showing an example of the output screen in user terminal 2 according to the embodiment of this disclosure. Figure 10 is a schematic diagram showing an example of the laws and standards 385 to be complied with as shown in Figure 9 according to the embodiment of this disclosure. Figure 11 is a schematic diagram showing an example of risk assessment 386 as shown in Figure 9 according to the embodiment of this disclosure. Figure 12 is a schematic diagram showing an example of past disaster case 387 according to the embodiment of this disclosure. Figure 13 is a schematic diagram showing an example of disaster case report 33-3A according to the embodiment of this disclosure.

[0026] The output screen 38 shown in Figure 9 is a screen for selecting candidate safety information 37 that have been extracted as highly relevant. The output screen 38 includes a button 381 for confirming and outputting the content of safety measures / past disaster information, the content of the work instruction 382, ​​the work overview 383, the work details 384, the laws and standards to be complied with 385, the risk assessment 386, and past disaster cases 387. The work overview 383 contains the same content as the work overview 31A1 shown in Figure 3. The work details 384 contains the same content as the work details 31A2 shown in Figure 4.

[0027] As shown in Figure 10, the laws and standards to be complied with (385) include an adoption (selection) checkbox 3851, a "law / law number," "safety measures," and an unsuitable checkbox 3852. In this embodiment, for example, 10 candidates are presented, and some are selected by the operator of the user terminal 2 by checking the adoption checkbox 3851. The unsuitable checkbox 3852 is checked by the operator of the user terminal 2 if the information is deemed unsuitable for the task, and processing is performed to make it less likely that such information will be selected (described later). Each row of the laws and standards to be complied with (385) corresponds to one rule information 33-1.

[0028] As shown in Figure 11, the risk assessment 386 includes corresponding checkboxes for "accepted," "type of disaster," "work content," "what happens, how, and how," "safety measures," and "unsuitable." In this embodiment, for example, 10 candidates are presented, and some are selected by the operator of the user terminal 2 by checking the "accepted" checkboxes. Each row of this risk assessment 386 corresponds to one piece of risk assessment information 33-2.

[0029] As shown in Figure 12, past disaster case 387 includes a checkbox for adoption, a "Disaster Case Report," a "Type of Disaster," a "Severity of Injury," a "Date of Occurrence," a "Work Details," a "Summary of Disaster," a "Location of Occurrence," and a checkbox for rejection. In this embodiment, for example, 10 candidates are presented, and some are selected by the operator of user terminal 2 by checking the "Adopt" checkbox. Each row of this past disaster case 387 corresponds to one disaster case information 33-3. By clicking the "Disaster Case Report" button 3871 within the "Disaster Case Report" section, a disaster case report 33-3A is displayed as shown in Figure 13. Each disaster case report 33-3A corresponds to one disaster case information 33-3.

[0030] Furthermore, the document vector calculation unit 14 calculates a first document vector 34, which is a document vector based on some of the words unintentionally extracted by the word extraction unit 12. Here, a document vector (or sentence vector) is a vector that represents the semantic features of a document as multidimensional numerical values ​​through natural language processing.

[0031] The second document vector calculation unit 15 calculates multiple second document vectors 36, which are each document vector based on the risk assessment information 33-2 and the disaster case information 33-3 from each safety information 33 stored in the database 13. In this case, the second document vector 36 will include the vectorized risk assessment information 36-2 and the vectorized disaster case information 36-3. Note that the first document vector calculation unit 14 and the second document vector calculation unit 15 may use a common configuration.

[0032] The similarity calculation unit 16 calculates the similarity between each of the multiple second document vectors 36 and the first document vector 34. The similarity can be expressed as cosine similarity. Cosine similarity is an index that represents how similar two vectors are, taking a value from -1 to 1, where 1 means they are perfectly similar (angle difference is 0 degrees and they are vectors pointing in the same direction), and -1 means they are not perfectly similar (angle difference is 180 degrees and they are vectors pointing in opposite directions). The similarity calculation unit 16 calculates the similarity between the second document vector 36 and the first document vector 34 based on the risk assessment information 33-2, and the similarity between the second document vector 36 and the first document vector 34 based on the disaster case information 33-3.

[0033] The target document preprocessing unit 21, as a process before the score calculation unit 20 calculates the score related to relevance, uses natural language processing to extract or calculate word groups 22 for each document and the IDF (inverse document frequency) 23 for each word from the safety information 33 stored in the database 13, based on the rule information 33-1. The word group 22 for each document consists of multiple words extracted for each document by performing word segmentation processing on each document included in the rule information 33-1. For example, the word group 22 for each document is information that associates the identification number of the document included in the rule information 33-1 with information representing the multiple words contained in that document. In addition, the IDF 23 for each word includes the result of calculating the IDF (logarithm of the reciprocal of the probability that each word is included in all documents) for all words. The IDF value increases as the frequency of inclusion of the word in a document decreases.

[0034] The score calculation unit 20 calculates a score regarding the relationship between the work information 32 and each rule information 33-1 using natural language processing. In this embodiment, the relationship score is an indicator that a higher score indicates a stronger relationship. In this embodiment, the LPS (Lucene's Practical Score) obtained by LPSF (Lucene's Practical Scoring Function) is used as the relationship score (also called relationship score, relevance score, etc.). However, this is just one example, and other relationship scores may be used. The score calculation unit 20 uses multiple words contained in the work information 32 as the search text (query), and the search target is all rule information 33-1, and calculates a score (LPS) for each rule information 33-1.

[0035] Here, an overview of natural language processing in this embodiment will be described with reference to Figures 14 to 16. Figure 14 is a schematic diagram illustrating an example of calculating a score regarding the relevance between a work instruction and applicable laws and standards according to the embodiment of this disclosure. Figure 15 is a schematic diagram illustrating an example of calculating the similarity between a work instruction and a risk assessment according to the embodiment of this disclosure. Figure 16 is a schematic diagram illustrating an example of calculating the similarity between a work instruction and an accident case report according to the embodiment of this disclosure.

[0036] As shown in Figure 14, in calculating the LPS score for the relationship between work instructions and the laws and standards to be complied with, first, a word group 22A (corresponding to the word group 22 for each document shown in Figure 1) is extracted from multiple documents 41 that have undergone word segmentation processing on multiple laws and business standard documents 33-1A (an example of the configuration of rule information 33-1), representing multiple words contained in each document 41. Next, a trained model (trained machine learning model) 21A-2 using IDFs obtained by machine learning with a population of multiple law / document word groups 22A as a corpus is used to calculate the IDF 23A (corresponding to the IDF 23 for each word) for each word contained in the word group 22A for each document. Here, the process of extracting words by performing word segmentation processing (21A-1) and the process of calculating the IDF for each word (21A-2) are processes performed by the target document preprocessing unit 21. Furthermore, for each work instruction sheet 32A (corresponding to work information 32), a relevance score 43 is calculated for each document (rule information 33-1) using multiple words extracted from the word segmentation process document 42, the word group 22A for each document, and the idf 23A for each word (20A (corresponding to processing by the score calculation unit 20)). The score can be calculated using the formula and variables shown in Figure 14.

[0037] As shown in Figure 15, in calculating the similarity between work instructions and risk assessments, first, a trained model 53 is prepared using an IDF that has been machine-trained using a population of multiple risk assessments 33-2B (an example of risk assessment information 33-2) as a corpus. Next, a trained model 54 is prepared using Fasttext that has been machine-trained using a publicly available corpus. Then, by inputting a document 51, which has undergone word segmentation processing for multiple risk assessments 33-2B, into the trained models 53 and 54, a word score table 55 and a word vector 56 are calculated. The word score table 55 is a table that represents an index for each word that indicates how important each word in the document 51 is within that document 51. The word vector 56 represents each word in the document 51 as a multidimensional vector. The document vector 36B (corresponding to the second document vector 36) can be calculated by, for example, taking the average of the vectors for each word over all words in the document 51. In this process, the document vector calculation process (57) weights the multidimensional vectors of each word in the word vector 56 with the values ​​of each word in the word score table 55, and then calculates the average. This allows the document vector 36B (corresponding to the second document vector 36) to be a feature that reflects the position of the risk assessment 33-2B within the population.

[0038] On the other hand, the document 52, which has undergone word segmentation processing on the work instruction sheet 32A (corresponding to work information 32), is subjected to a random sampling process at a predetermined ratio (12B (corresponding to processing by the word extraction unit 12)), and then input into trained models 58 and 59 to calculate a word score table 60 and word vectors 61. Note that trained models 53 and 58 are the same trained model, and trained models 54 and 59 are the same trained model. In the document vector calculation process (62), the multidimensional vectors of each word in the word vectors 61 are weighted by the values ​​of each word in the word score table 60, and the average is calculated to calculate the document vector 34B (corresponding to the first document vector 34). Then, the similarity 63 between the document vector 34B and each document vector 36B is calculated (16B (corresponding to processing by the similarity calculation unit 16)). In Figure 15, the configuration combining the trained model 53, the trained model 54, and the document vector calculation process (57) corresponds to configuration 15B (corresponding to the second document vector calculation unit 15). Furthermore, the configuration combining the trained model 58, the trained model 59, and the document vector calculation process (62) corresponds to configuration 14B (corresponding to the document vector calculation unit 14).

[0039] As shown in Figure 16, in calculating the similarity between work instructions and disaster incident reports, first, a trained model 73 using an IDF machine-trained with a population of multiple disaster incident reports 33-3C (an example of disaster incident information 33-3) as a corpus is prepared. Next, a trained model 74 using Fasttext machine-trained with a publicly available corpus is prepared. Then, by inputting a document 71, which has undergone word segmentation processing on multiple disaster incident reports 33-3C, into the trained models 73 and 74, a word score table 75 and a word vector 76 are calculated. In the document vector calculation process (77), the multidimensional vectors of each word in the word vector 76 are weighted by the values ​​of each word in the word score table 75, and the average is calculated to calculate the document vector 36C (corresponding to the second document vector 36).

[0040] On the other hand, the document 72, which has undergone word segmentation processing on the work instruction sheet 32A (corresponding to work information 32), is subjected to a random sampling process at a predetermined ratio (12C (corresponding to processing by the word extraction unit 12)), and then input into trained models 78 and 79 to calculate a word score table 80 and word vectors 81. Note that trained models 73 and 78 are the same trained model, and trained models 74 and 79 are the same trained model. In the document vector calculation process (82), the multidimensional vectors of each word in the word vectors 81 are weighted by the values ​​of each word in the word score table 80, and the average is calculated to calculate the document vector 34C (corresponding to the first document vector 34). Then, the similarity 83 between the document vector 34C and each document vector 36C is calculated (16C (corresponding to processing by the similarity calculation unit 16)). In Figure 16, the configuration combining the trained model 73, the trained model 74, and the document vector calculation process (77) corresponds to configuration 15C (corresponding to the second document vector calculation unit 15). Furthermore, the configuration combining the trained model 78, the trained model 79, and the document vector calculation process (82) corresponds to configuration 14C (corresponding to the document vector calculation unit 14).

[0041] Furthermore, the presentation unit 17 presents a portion of the multiple safety information 33 to the user terminal 2 based on their similarity. First, the presentation unit 17 selects, for example, the top 10 rule information 33-1, risk assessment information 33-2, and disaster case information 33-3 based on their respective scores (LPS) or similarity with the work information 32. Next, the presentation unit 17 refers to the history of the selection results, sorts the 10 pieces of information in descending order of the number of times they have been selected, and then presents (provides) them as candidate information for output, as explained with reference to Figure 9, etc. Next, when button 381 is clicked, the presentation unit 17 presents (provides) the output screen 37A (corresponding to the safety information 37) shown in Figure 17, which includes only the information for which a checkbox (such as checkbox 3851 in Figure 10) has been selected in the output screen 38 shown in Figure 9. The output screen 37A shown in Figure 17 includes the contents of the work instruction 37A2, the work overview 37A3, the work details 37A4, the laws and standards to be complied with 37A5, the risk assessment 37A6, and past disaster cases 37A7.

[0042] Furthermore, the reception unit 18 receives information (for example, a check of the "inappropriate" checkbox 3852) indicating that some of the safety information presented by the presentation unit 17 is inappropriate.

[0043] Furthermore, the modification unit 19 changes the values ​​of one or more elements (components) of the second document vector 36 based on the safety information deemed inappropriate. By changing the values ​​of one or more elements of the second document vector 36, the similarity can be changed, and the probability of the information being selected can be changed.

[0044] Next, an example of the operation of the information providing device 1 will be described with reference to Figure 18. The process shown in Figure 18 is started, for example, in response to a request from the user terminal 2. In the process shown in Figure 18, first, the acquisition unit 11 acquires the work instruction sheet (work information 31) (step S11). Next, the word extraction unit 12 extracts words (strings and strings of checked items) from the work instruction sheet (let's call this word group A) (step S12). Next, the word extraction unit 12 randomly selects a predetermined percentage of words from the extracted words (word group A) (let's call this word group B) (step S13). Next, the document vector calculation unit 14 calculates the document vector (first document vector 34) of word group B (step S14). Next, the score calculation unit 20 calculates the score (LPS) for word group A for each word group 22 based on laws and regulations and business standards (rule information 33-1) (step S15). Next, the similarity calculation unit 16 calculates the similarity between the document vector (second document vector 36) based on the risk assessment (risk assessment information 33-2) and the first document vector 34 (step S16). Next, the similarity calculation unit 16 calculates the similarity between the document vector (second document vector 36) based on the disaster incident report (disaster incident information 33-3) and the first document vector 34 (step S17). Next, the presentation unit 17 presents a predetermined number of laws / business standards and information based on the risk assessment and disaster incident report based on each score or similarity (step S18). Next, the presentation unit 17 selects the information to provide based on the user's input (step S19). Next, the reception unit 18 determines whether or not it has been set as inappropriate (step S20). If it has been set as inappropriate (step S20: YES), the modification unit 19 changes the values ​​of one or more elements of the second document vector 36 that were deemed inappropriate (step S21). If no setting indicating inappropriateness has been made (step S20: NO), or after step S21, the presentation unit 17 presents the selected information (step S22) and terminates the process shown in Figure 18.

[0045] (Effects and Benefits) The information providing device 1 with the above configuration includes: an acquisition unit 11 that acquires work information, which is information related to the work to be performed; a word extraction unit 12 that randomly extracts some of several words from a plurality of words based on the work information 32 at a predetermined ratio; a document vector calculation unit 14 that calculates a first document vector 34, which is a document vector based on the extracted plurality of several words; a similarity calculation unit 16 that calculates the similarity between a plurality of second document vectors 36, which are each document vector based on a plurality of safety information 33, which is information for reducing the risks of the work, and the first document vector 34; and a presentation unit 17 that presents a portion of the plurality of safety information 33 (safety information 37) based on each similarity. With this configuration, the information to be provided can be appropriately selected.

[0046] As described above, in this embodiment, work information 31 and work information 32 are examples of "input information" related to this disclosure, and safety information 33 and safety information 37 are examples of "provided information" related to this disclosure. Therefore, the acquisition unit 11 can be understood as a configuration that acquires input information, which is the information that has been input. The word extraction unit 12 can be understood as a configuration that randomly extracts a portion of a plurality of words from a plurality of words based on the input information at a predetermined rate. The similarity calculation unit 16 can be understood as a configuration that calculates the similarity between a plurality of second document vectors, which are a plurality of document vectors based on a plurality of provided information that is information related to the input information and provided to the input information, and the first document vector.

[0047] Furthermore, safety information 33 includes at least one of risk assessment information 33-2, which is information on past risk assessments, and disaster case information 33-3, which is information on past disaster cases.

[0048] Safety information 33 includes rule information 33-1, which is information representing rules to be followed during work. The system further includes a score calculation unit 20 that calculates a score (LPS) for each relationship between multiple word groups (word groups 22 for each document) based on multiple rule information 33-1 and multiple words based on work information 32. The presentation unit 17 presents a portion of the safety information 33, including multiple rule information 33-1, based on each score (LPS). With this configuration, it is possible to calculate a score for the relationship between rule information 33-1 and work information 32, for which words have not been randomly selected.

[0049] Furthermore, the information providing device 1 includes a receiving unit 18 that receives information to the effect that some of the presented safety information (provided information) is inappropriate, and a modification unit 19 that changes the values ​​of one or more elements of the second document vector 36 based on the safety information 33 (provided information) deemed inappropriate. With this configuration, user feedback can be reflected in future selections.

[0050] (Other embodiments) Although embodiments of this disclosure have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments and may include design changes and the like that do not depart from the gist of this disclosure.

[0051] <Computer Configuration> Figure 19 is a schematic block diagram showing the configuration of a computer according to at least one embodiment. The computer 90 includes a processor 91, main memory 92, storage 93, and an interface 94. The information providing device 1 described above is implemented in the computer 90. The operation of each processing unit described above is stored in storage 93 in the form of a program. The processor 91 reads the program from storage 93, loads it into main memory 92, and executes the above processing according to the program. The processor 91 also allocates memory areas in main memory 92 corresponding to each of the storage units described above, according to the program.

[0052] The program may be for implementing some of the functions that the computer 90 is to perform. For example, the program may perform functions in combination with other programs already stored in storage, or in combination with other programs implemented in other devices. In other embodiments, the computer may include a custom LSI (Large Scale Integrated Circuit) such as a PLD (Programmable Logic Device) in addition to, or instead of, the above configuration. Examples of PLDs include PAL (Programmable Array Logic), GAL (Generic Array Logic), CPLD (Complex Programmable Logic Device), FPGA (Field Programmable Gate Array), etc. In this case, some or all of the functions implemented by the processor may be implemented by the integrated circuit.

[0053] Examples of storage 93 include HDDs (Hard Disk Drives), SSDs (Solid State Drives), magnetic disks, magneto-optical disks, CD-ROMs (Compact Disc Read Only Memory), DVD-ROMs (Digital Versatile Disc Read Only Memory), and semiconductor memory. Storage 93 may be an internal medium directly connected to the bus of the computer 90, or an external medium connected to the computer 90 via an interface 94 or a communication line. Furthermore, if this program is distributed to the computer 90 via a communication line, the computer 90 that receives the program may expand it into main memory 92 and execute the above processing. In at least one embodiment, storage 93 is a tangible storage medium that is not temporary.

[0054] <Note> The information providing device 1 described in each embodiment can be understood, for example, as follows:

[0055] (1) The information providing device 1 according to the first embodiment includes: an acquisition unit 11 that acquires work information which is information relating to the work to be performed; a word extraction unit 12 that randomly extracts some of several words from a plurality of words based on the work information at a predetermined rate; a document vector calculation unit 14 that calculates a first document vector which is a document vector based on the extracted plurality of several words; a similarity calculation unit 16 that calculates the similarity between a plurality of second document vectors which are a plurality of document vectors based on a plurality of safety information which is information for reducing the risks of the work, and the first document vector; and a presentation unit 17 that presents a portion of the plurality of safety information based on the similarity. According to this embodiment and the following embodiments, the information to be provided can be appropriately selected.

[0056] (2) The information providing device 1 according to the second embodiment is the information providing device 1 of (1), wherein the input information is work information which is information relating to the work to be performed, and the provided information is safety information which is information for reducing the risks of the work.

[0057] (3) The information providing device 1 according to the third embodiment is the information providing device 1 according to (1) or (2), wherein the safety information includes at least one of the following: risk assessment information, which is information relating to past risk assessments, and disaster case information, which is information relating to disaster cases that have occurred in the past.

[0058] (4) The information providing device 1 according to the fourth embodiment is the information providing device 1 according to (1) to (3), wherein the safety information includes rule information which is information that represents rules to be observed in the work, and further comprises a score calculation unit 20 which calculates a score for each relationship between a plurality of word groups based on a plurality of rule information and a plurality of words based on the work information, and the presentation unit presents a portion of the safety information including a plurality of rule information based on each score.

[0059] (5) The information providing device 1 according to the fifth embodiment is the information providing device 1 according to (1) to (4), and includes a receiving unit 18 that receives information to that effect when some of the provided information presented is inappropriate, and a changing unit 19 that changes the values ​​of one or more elements of the second document vector based on the provided information deemed inappropriate. [Explanation of Symbols]

[0060] 100... Information provision system 1…Information provision device 2…User terminal 11…Acquisition part 12...Word extraction section 13…Database 14…Document vector calculation unit 15...Second Document Vector Calculation Unit 16…Similarity calculation unit 17…Presentation part 18…Reception Department 19…Changes 20... Score calculation section

Claims

1. An acquisition unit that acquires input information, which is the information that has been entered. A word extraction unit that randomly selects a subset of multiple words from a plurality of words based on the aforementioned input information in a predetermined proportion, A document vector calculation unit calculates a first document vector, which is a document vector based on a subset of the extracted words, A similarity calculation unit calculates the similarity between a plurality of second document vectors, which are a plurality of document vectors based on a plurality of provided information that is related to the input information and provided to the input information, and the first document vector. A presentation unit that presents a portion of the multiple pieces of information based on each of the aforementioned similarity levels, An information-providing device equipped with the following features.

2. The aforementioned input information is work information, which is information related to the work to be performed. The information provided is safety information, which is information for reducing the risks of the aforementioned work. The information providing device according to claim 1.

3. The aforementioned safety information includes at least one of the following: risk assessment information, which is information regarding past risk assessments, and disaster case information, which is information regarding past disaster cases. The information providing device according to claim 2.

4. The aforementioned safety information includes rule information, which is information representing the rules to be followed in the aforementioned work. The system further includes a score calculation unit that calculates scores for the relationships between multiple word groups based on multiple rule information and multiple words based on the work information, The display unit presents a portion of the safety information, including a plurality of rule information, based on each of the scores. The information providing device according to claim 3.

5. A reception desk that receives information to the effect that some of the information provided above is inappropriate, A modification unit that modifies the values ​​of one or more elements of the second document vector based on the provided information deemed inappropriate. An information providing device according to any one of claims 1 to 4, comprising:

6. The steps include obtaining input information, which is the information that has been entered, A step of randomly selecting a portion of several words from a group of words based on the aforementioned input information in a predetermined proportion, A step of calculating a first document vector which is a document vector based on the extracted subset of multiple words, A step of calculating the similarity between a plurality of second document vectors, which are information related to the input information and are based on a plurality of provided pieces of information provided for the input information, and the first document vector, A step of presenting a portion of the multiple pieces of information based on each of the aforementioned similarity levels, Information provision methods including those mentioned above.

7. The steps include obtaining input information, which is the information that has been entered, A step of randomly selecting a portion of several words from a group of words based on the aforementioned input information in a predetermined proportion, A step of calculating a first document vector which is a document vector based on the extracted subset of multiple words, A step of calculating the similarity between a plurality of second document vectors, which are information related to the input information and are based on a plurality of provided pieces of information provided for the input information, and the first document vector, A step of presenting a portion of the multiple pieces of information based on each of the aforementioned similarity levels, A program that causes a computer to execute something.

Images