CN114730443B

CN114730443B - Building information processing device

Info

Publication number: CN114730443B
Application number: CN201980101771.0A
Authority: CN
Inventors: 远藤嘉人; 庄司圭一
Original assignee: Mitsubishi Electric Building Solutions Corp
Current assignee: Mitsubishi Electric Building Solutions Corp
Priority date: 2019-11-06
Filing date: 2019-11-06
Publication date: 2023-05-16
Anticipated expiration: 2039-11-06
Also published as: CN114730443A; JPWO2021090389A1; JP7160503B2; WO2021090389A1

Abstract

A building information processing device (70) is provided with a machine learning processing unit (86), an extraction unit (82), and an AI (84). A machine learning processing unit (86) performs machine learning using a feature quantity dictionary (74) on an AI (84), wherein the feature quantity dictionary (74) is a sentence relating to a failure of a device or a component of a building. Thus, in AI (84), learning of the characteristic sentences set in the selected characteristic sentence dictionary (78) is performed, wherein the characteristic sentences represent maintenance management characteristics related to the equipment or the components. An extraction unit (82) extracts, for each character string, a sentence relating to a failure of a device or component of a building, with respect to a plurality of character strings in which a contact or report relating to the failure of the device or component is described. Then, the character string and the characteristic sentence corresponding to the extracted sentence related to the failure of the device or the component are selected using the AI (84).

Description

Building information processing device

Technical Field

The present invention relates to a building information processing apparatus that handles operation of a building or failure of equipment or components of the building.

Background

In a building, in addition to the operation of the building, maintenance management services including coping with a failure are generally implemented for various devices or components such as an elevator, an air conditioner, and an electrical device. In the case where a failure occurs or a failure is handled, generally, a document for making a contact or report is generated.

Patent document 1 describes an information processing apparatus including: in the case where a user of an elevator owner, manager, boarding or the like encounters a trouble, an operation for communicating with a monitoring center or the like is assisted. In the information processing apparatus, the user can freely record the fault state in addition to selecting the fault state from the items registered in advance.

Patent document 2 describes a report generation system in the case where an operator who performs maintenance inspection of an elevator performs work on a work object. The operator sequentially inputs the work item determination information, the work result and the advice for each work item through voice. The management server documents the voice, extracts keywords, and classifies the keywords for each job object item to generate a report.

When a trouble contact is received, a contact is made to the worker.

Patent document 3 describes the following: the voice call sent from the manager of the elevator or escalator, the rider, etc. to the monitoring center is documented and information processing is performed, thereby generating a message sent to the operator who performs maintenance inspection. In the information processing, when a predetermined keyword is extracted from a document, a format message corresponding to the keyword is prepared. In addition to inputting keywords extracted from a document to a blank section set in a format message, the transmission of the message is completed by the operator making an input.

The handling of faults may be performed in areas other than buildings.

Patent document 4 describes the following system: based on the similarity, documents in which computer failures are described are hierarchically classified, whereby similar documents in the past can be accessed.

Prior art literature

Patent literature

Patent document 1: japanese patent laid-open publication No. 2019-1618

Patent document 2: japanese patent application laid-open No. 2015-75792

Patent document 3: japanese patent laid-open No. 2018-156134

Patent document 4: japanese patent laid-open publication No. 2005-267351

Disclosure of Invention

Problems to be solved by the invention

As described in patent documents 1 and 2, when a device or a component of a building fails, a user such as an owner, tenant, visitor, or the like of the building or a maintenance manager of the building makes a contact or report. Such a document (or may be referred to as a character string) of the contact or report contains information useful for maintenance management of the building. Therefore, it is convenient if information corresponding to a failure can be extracted from such a document or character string of a contact or report. However, in general, since devices or parts of a building involve a plurality of aspects, extraction of information corresponding to a failure is complicated, and time and cost are required.

In patent document 3, since a voice is documented, a document or a character string which is freely described in proximity is processed. However, in patent document 3, since the content of the maintenance management work is limited for only the elevator or only the escalator, it is considered that the selection of the format message by the keyword can be relatively easily performed. Further, in patent document 3, since the operator directly confirms and generates a message, even if the accuracy of selecting a format message is poor, an accurate message is finally generated. Therefore, it is difficult to apply the technology of patent document 3 to various devices or components of a building.

Patent document 4 deals with limited contents such as a computer failure, and sets documents or character strings having a fixed format as classification targets. Therefore, it is difficult to apply the technique of patent document 4 to failure of equipment or parts of a building.

The purpose of the present invention is to obtain a sentence indicating a maintenance management characteristic from a character string in which a contact or report relating to the operation of a building or the failure of a device or a component of the building is described.

Means for solving the problems

The building information processing device of the present invention comprises: AI that learns, by machine learning using a sentence related to the operation of a building or the failure of a device or a component of the building, the selection of a characteristic sentence representing a maintenance management characteristic related to the building, the device or the component; an extraction unit that extracts, from a character string in which a contact or report relating to the operation of a building or a failure of a device or a component of the building is recorded, a sentence relating to the operation of the building or the failure of the device or the component; and a selection unit that uses the AI to select the characteristic sentence corresponding to the extracted sentence related to the operation of the building or the malfunction of the device or the component.

In one aspect of the present invention, the system further includes a preprocessing unit that obtains a plurality of character strings from a plurality of documents in different formats, each of the documents being generated by a user or a maintenance manager of the building and describing a contact or report related to operation of the building or failure of a device or a component of the building, wherein the extracting unit extracts a sentence related to the failure of the device or the component for each of the character strings, and the selecting unit selects the characteristic sentence for each of the character strings using the AI.

In one embodiment of the present invention, the system further includes a setting unit that sets at least a part of a sentence related to the operation of the building or the failure of the device or the component, according to an instruction from an administrator of the building information processing apparatus.

In one embodiment of the present invention, the term relating to the operation of the building or the failure of the device or the component includes a name of the building, the device or the component and a description term describing the operation or the failure occurring in the building, the device or the component.

In one embodiment of the present invention, the description statement is a statement describing a specification of an operation or a failure, a request for handling the operation or the failure, or an implementation of handling the operation or the failure.

In one embodiment of the present invention, the characteristic statement refers to a statement that is summarized for use or malfunction of the building, device, or component.

In one embodiment of the present invention, the characteristic sentence refers to a sentence for specifying, requesting, or predicting the execution of the operation or the failure of the building, the device, or the component.

In one embodiment of the present invention, the character string or the document from which the character string is obtained includes a predetermined sentence, and the output unit outputs the sentence or a sentence corresponding to the sentence together with the characteristic sentence for the character string.

Effects of the invention

According to the present invention, a characteristic term indicating maintenance management characteristics related to a building, equipment, or component can be selected from a character string described in relation to the operation of the building or the failure of the equipment or component of the building. Thus, for example, a document can be generated using the characteristic sentence. Further, for example, character strings related to a failure can be classified based on a characteristic sentence.

Drawings

Fig. 1 is a schematic configuration diagram of a building system according to an embodiment.

Fig. 2 is a diagram showing a schematic structure of a building.

Fig. 3 is a diagram showing a schematic configuration of the building information processing apparatus.

Fig. 4 is a diagram showing an example of a daily report as a report document in which a failure is recorded.

Fig. 5 is a diagram showing an example of a contact document in which a failure is described.

FIG. 6 is an example of an inquiry/response string generated from a document.

Fig. 7 is a diagram showing a relationship between feature amounts and characteristic sentences.

Fig. 8 is an example of selecting a characteristic sentence based on a feature amount extracted from an inquiry/response character string.

Fig. 9 is an example of a statement of a rule base.

FIG. 10 is a diagram of appending a sentence of a rule base extracted from a document to FIG. 8.

Detailed Description

Fig. 1 is a diagram showing a schematic configuration of a building system 10 of the embodiment. In the figure, 3

buildings

20a, 20b, and 20c (hereinafter, simply referred to as building 20 without distinction) are shown as examples of a plurality of buildings. Maintenance management of the building 20, including handling and fault handling, is performed by a maintenance management company of the building. The building 20 is connected to be able to communicate with a building information processing apparatus 70 via a communication network 60. A PC 100 that performs data processing is connected to the building information processing apparatus 70. The building information processing apparatus 70 and the PC 100 are provided in, for example, a maintenance management company of a building.

Fig. 2 is a diagram illustrating the building 20 shown in fig. 1 in detail. An air conditioning device 22, a water supply and drainage device 24, an electric device 26, an elevator 28, and the like are provided in the building 20.

In addition, a central monitoring room 30 is provided in the basement of the building 20. A central monitor panel 32 and a PC 34 are provided in the central monitor room 30. The central monitoring board 32 is a computer connected to the air conditioning equipment 22, the water supply and drainage equipment 24, the electric equipment 26, and the like so as to be capable of communication and control and monitoring of operation.

The PC 34 is connected to a remote monitoring device of the elevator 28 to control and monitor the elevator 28, and also to analyze operation data of the elevator. The PC 34 is also connected to the central monitoring board 32 to analyze data output from the central monitoring board 32. The PC 34 is used for generating service reports such as daily reports by

maintenance personnel

36 and 38 who perform maintenance management work using the central monitoring room 30 as a site. The PC 34 is also configured to receive a contact document related to the operation of the building or a contact document related to the failure of the equipment and components from the user of the building 20 by mail or the like.

In fig. 2, it is assumed that the component 24a of the water supply and drainage apparatus 24 has failed. Then, the tenant 40, which is a user of the building 20, discovers the failure of the component 24a and communicates with the central monitoring room 30 via the mobile phone 42. Maintenance personnel 36 of central monitoring room 30 receive contact from tenant 40 via telephone 44. Thus, the maintenance person 36 communicates with itself or other maintenance persons 38 to repair the component 24a, and the like.

In fig. 2, maintenance personnel 38 is confirming that a fault in electrical equipment 26 that is otherwise in communication is being received. Based on the result of the confirmation, the maintenance person 38 performs necessary repairs or the like.

Fig. 3 is a diagram showing a detailed structure of the building information processing apparatus 70 shown in fig. 1. The building information processing apparatus 70 is constructed by controlling computer hardware including a memory and a processor with software including an operating system and application programs. The building information processing apparatus 70 may be constructed using a single computer hardware, or may be constructed using a plurality of computer hardware connected to a network.

In the building information processing apparatus 70, a feature dictionary setting section 72, a feature dictionary 74, a feature sentence setting section 76, a feature sentence dictionary 78, a document input section 80, a preprocessing section 81, an extraction section 82, an AI 84, a machine learning processing section 86, a feature sentence acquisition section 88, an output section 90, a rule base dictionary setting section 92, a rule base dictionary 94, and a rule base sentence acquisition section 96 are constructed by application programs.

The feature dictionary setting unit 72 is an example of a setting means, and the manager of the building information processing device 70 inputs the feature dictionary 74. The feature quantity dictionary 74 is a dictionary in which a plurality of sentences are stored for a failure of a device or a component of a building. The failure refers to a state in which abnormality, malfunction, obstacle, degradation, or the like of the device or the component occurs or is estimated to occur. In the case of a failure, although there is no problem in the equipment or the component itself, the operation may include a situation in which a trouble occurs if the setting is not changed.

The term concerning the failure includes names of devices and components of the building, and description terms describing failures occurring in the devices and components. This is to include the object in which the fault has occurred and the sentence describing what the fault is, and thus the fault can be easily grasped.

The names of the devices and the components include general names, manufacturer product numbers, and the like, and names described in terms of the degree to which the

maintenance personnel

36 and 38 can determine the devices and the components. The description statement describing the fault includes at least one of determination of describing the fault, request for handling the fault, and implementation of handling the fault. The term for specifying a failure refers to a term describing a state or cause of a failure such as abnormal noise or bulb breakage. The term "request for coping with a failure" refers to a request description for confirmation of a failure, repair, replacement, etc., which indicates that temperature adjustment is desired and repair is desired. The term "means a description of handling a fault such as confirmation, repair, replacement, or the like. The term includes a term in which 2 or more words are connected in addition to 1 word (the distinction of word class such as noun, verb, adjective is not limited). Concise sentences may also be included in the sentence.

In the feature quantity dictionary 74, other feature sentences may be also included in relation to the failure of the equipment or parts of the building. Specifically, a term indicating a time associated with the fault, such as a date and time of inquiry of the fault or a date and time of response of the fault, or a term indicating details of the fault, such as a floor where the fault has occurred, a room where the fault has occurred, a name of equipment where the fault has occurred, and a name of a component, may be used. As another example, a sentence indicating a failure-correspondent person such as a company name or a person name that has been subjected to the failure may be given.

The sentence registered in the feature amount dictionary 74 is set to have a large correlation with the accuracy of the sentence selected by the AI 84. Therefore, a part of the feature dictionary 74 can be generated by automatically performing the feature selection process, but in this case, the administrator also searches for the adequacy of the feature via the feature dictionary setting unit 72 to set the feature dictionary 74. In particular, by setting the feature quantity dictionary 74 by a person who is skilled in maintenance management of the building, improvement of the accuracy of the selection of the AI 84 can be expected.

In general, since synonyms or paraphraseology having similar meanings exist in sentences, the feature dictionary 74 is also provided with synonyms or paraphraseology.

The characteristic sentence setting unit 76 sets a characteristic sentence dictionary 78 by being input by an administrator of the building information processing apparatus 70. A part of the characteristic sentence dictionary 78 may be automatically generated using the feature quantity dictionary 74, for example. However, in this case, the administrator also searches the validity of the characteristic sentence through the characteristic sentence setting unit 76, and sets the characteristic sentence dictionary 78.

In the characteristic sentence dictionary 78, a plurality of characteristic sentences which are sentences representing maintenance management characteristics concerning devices or components are stored. The characteristic statement may also be a statement that is summarized for a failure of a device or component. For example, a sentence that compactly combines the names of devices or components and the states thereof is a sentence that summarizes a failure. The characteristic term may include a term describing a specification of a failure of the device or the component, a request for handling the failure, or an implementation of handling the failure. The sentences set in the characteristic sentence dictionary 78 correspond to the sentences set in the feature quantity dictionary 74. The correspondence relationship is used in machine learning described later.

In general, since synonyms or paraphraseology having similar meanings exist in sentences, it is also possible to set and register synonyms or paraphraseology in the characteristic sentence dictionary 78. However, as will be described later, the characteristic sentence dictionary 78 defines sentences selected by the AI 84. In order to avoid ambiguities in the sentences selected by the AI 84 due to synonyms or paraphraseology, a manner may be considered in which synonyms or paraphraseology are not set in the characteristic sentence dictionary 78.

The document input unit 80 obtains a document in which a contact or report concerning a failure of the device or the component is recorded from the building 20, the storage server, or the like shown in fig. 1. The contact document or report document related to the failure refers to a document generated by a user of the building (refer to a person who uses the building, such as an owner, tenant, visitor, etc.) or a maintenance manager such as

maintenance personnel

36, 38 of the building, for notifying that a device or component has a failure. In general, the characteristics of the report document generated under certain responsibilities or rights are stronger than the contact document. However, since the contact document and the report document have similar features, sometimes it is not necessarily clearly distinguishable.

The contact document or report document may also be generated by a user and sent to a maintenance manager of the building, for example. In addition, contact documents or report documents may also be generated and sent to users of the building or other maintenance managers, for example, by

maintenance personnel

36, 38. The contact document or report document may be created in consideration of various purposes such as a purpose for which a fault is known to occur, a purpose for recording a fault occurrence, a purpose for requesting to cope with a fault, and a purpose for notifying that a fault has been dealt with.

Contact documents and report documents are typically electronically generated using word processing software. However, the contact document and the report document may be electronic documents in which a voice is converted into characters by a voice recognition process, or documents in which handwritten characters are converted into characters by OCR (Optical Character Recognition: optical character recognition).

The preprocessing unit 81 is an example of preprocessing means, and extracts character strings in which records concerning malfunctions of devices or components are made from each document acquired by the document input unit 80, and collects them into a list. In general, the contact document or report document input from the document input unit 80 is described in various formats. Therefore, when a plurality of failures are listed in one document, the preprocessing unit 81 performs processing of selecting a portion in which each failure is described. The feature of the document format can be identified (e.g., based on the line boundaries) and selected automatically or in accordance with a pre-made setting. Further, for example, the meaning of each region is interpreted by natural language processing, and thus selection can be performed.

Next, when a trouble is described in a plurality of columns, the preprocessing unit 81 performs a process of extracting a character string in which the columns are described in a combined manner. For example, in a plurality of contact documents or report documents, an item corresponding to "inquiry contents" and an item corresponding to "response to inquiry" are prepared. Therefore, a string is extracted in which the recorded strings of the two items (or items corresponding to synonyms and paraphrases thereof) are integrated.

In the present embodiment, the term "character string" refers to a character, a number, a symbol, a space, or the like connected to each other. The term such as a character string is a concept similar to the term such as a sentence, but may not necessarily be constituted by a sentence. For example, since a plurality of entry fields are combined, a term that is formed by connecting sentences that do not form a sentence pattern is also included in the character string. The term document (sometimes also referred to as text) is used to represent a larger collection of electronic files. In contrast, in the present embodiment, terms such as character strings are used for the purpose of being able to be a part of elements extracted from a document. For example, in the case of forming an electronic file for each character string, the character string and the document can be regarded as substantially identical.

The extraction unit 82 is an example of extraction means, and extracts the above-described sentence concerning the failure of the device or the component from the character string generated by the preprocessing unit 81. The extraction unit 82 extracts the sentence related to the failure by the coincidence determination with the sentence set in the feature quantity dictionary 74. That is, the extracting unit 82 extracts the sentences stored in the feature quantity dictionary 74 from the target character string.

AI 84 is an abbreviation for artificial intelligence (Artificial Intelligence) that enables intelligent actions of a person in a software manner. In the embodiment, the AI 84 is assumed to be a model capable of machine learning. For example, an algorithm using deep learning as a model.

The AI 84 is also an example of the selection means, and after performing machine learning, the sentences extracted by the extraction unit 82 are input, and the characteristic sentences corresponding thereto are selected. The number of the characteristic words may be 1 or a plurality of the characteristic words. In addition, when a plurality of sentences are selected, the sentences may be output as a connected characteristic sentence. The selection of the characteristic statement depends on machine learning as described below.

Note that, the AI 84 may be configured to input sentences or sentences other than the sentences extracted by the extracting unit 82. For example, in the case where the AI 84 has a language processing function in advance, the entire character string output from the input preprocessing unit 81 can be considered, and the accuracy of selecting the characteristic sentence can be improved.

The machine learning processing unit 86 performs machine learning of the AI 84. Given an appropriate combination of the sentences contained in the feature quantity dictionary 74, the machine learning processing section 86 performs machine learning to select a corresponding characteristic sentence. Thus, in the case where a combination of sentences relating to the failure of the device or the component is given, the AI 84 can select the most similar characteristic sentence even without directly learning the combination.

The characteristic term acquisition unit 88 acquires the characteristic term selected by the AI 84 and transmits the acquired characteristic term to the output unit 90.

The rule base dictionary setting unit 92 sets the rule base dictionary 94 by inputting the rule base dictionary setting unit by the manager of the building information processing apparatus 70. When the sentence to be obtained is clear, the sentence is set in the rule base dictionary 94. In particular, when information common to various devices and components is acquired, it is conceivable that the information is set in the rule base dictionary 94 independently of the AI 84. In addition, the rule base dictionary 94 can be effectively used also when it is desired to directly acquire information from a document input to the document input section 80 without going through the preprocessing section 81. In addition, a part of the rule base dictionary 94 can be generated by an automatically performed process, but in this case, the manager also searches for and sets the validity of the rule base dictionary 94 via the rule base dictionary setting unit 92.

Various sentences can be set in the rule base dictionary 94. Examples of the setting include setting related to an attribute of a building, such as a building name, a total building area, a prefecture where the building is located, and a building application. Further, a term indicating a time related to the fault, such as a date and time of inquiry of the fault or a date and time of handling the fault, a term indicating details of the fault, such as a floor where the fault has occurred or a room where the fault has occurred, may be used. As another example, a sentence indicating a failure-correspondent person such as a company name or a person name that has been subjected to the failure may be given. In general, since synonyms or paraphraseology having similar meanings exist in sentences, synonyms or paraphraseology are also set in the rule base dictionary 94.

The rule base sentence acquisition unit 96 acquires a sentence set in the rule base dictionary or a sentence corresponding to the sentence from the document acquired by the document input unit 80. That is, the rule base sentence acquisition unit 96 can acquire a sentence that matches a sentence set in the rule base dictionary, or a sentence that has a value corresponding to a case where the sentence set in the rule base dictionary is an item name or the like. The obtained sentence is sent to the output unit 90. The output unit 90 is an example of an output means, and outputs the characteristic sentence acquired from the characteristic sentence acquisition unit 88 together with the sentence acquired by the rule base sentence acquisition unit 96.

Next, a document input to the document input section 80 of the building information processing apparatus 70 will be described with reference to fig. 4 and 5. Fig. 4 is an example of a business day report used in a certain building, and fig. 5 is an example of a contact document managed by a building received by telephone.

The business daily report shown in fig. 4 is a business report generated by an operator of a company performing maintenance and management business of a building having a name "good, good building". The business daily report is submitted to the good, generated in 2019, 10, 15 days of the building. The names of the shift man and night man, the work items carried out on the day and the inspection results thereof are recorded. The 2 boxes indicated by

reference numerals

110, 112 in the center of the daily newspaper represent the contents of the fault occurring on the day and the disposition thereof.

The column of reference numeral 110 is provided with each item of the date and time of occurrence, the date and time of end, the person to be handled, the content, and the treatment. The occurrence date and time is 9 hours 52 minutes of 10 months 15 days of 2019, the ending date and time is 11 hours 13 minutes of the same day, and the counter is good. The content item is an item corresponding to "inquiry content", and a request item for receiving a contact (i.e., a free format description) is freely recorded in a corresponding description field. Specifically, "good company good" states that the 14-layer sink is not discharged from the innermost faucet of the sink in the bathroom, and thus the order should be dealt with. The item to be processed corresponds to "response to inquiry", and the corresponding entry field freely records a response to the person and a response to the person. Specifically, the "arrival time status is recorded as a status that the automatic faucet of the corresponding basin is not discharging water. Reinsertion of the power plug of the automatic faucet carries out reset operation and confirms normal operation. Good company good. "such handling records".

Similarly, in the column 112, items relating to other failures that can be handled by the failure are described in the same day from 13:35 to 15:21. Specifically, in the column corresponding to the content item, there is described that the door handle of the "ΔΔ company ΔΔ" contact name 19-layer ΔΔ zone exclusive portion exit door is loose. In addition, in the column corresponding to the treatment item, "the door handle of the arrival condition confirmation 19-layer ΔΔ zone exclusive use portion exit door is loose" is recorded. Tightening the door screw and confirming that the door handle is not loose. Delta Co-confirmation of Delta is completed by Delta company.

Fig. 5 is a contact document showing a contact received via a call center of a maintenance management company and showing a corresponding item. In this contact document, only one fault is noted. Specifically, at 13 hours and 30 minutes, the recipient receives a call from the request object and receives a contact related to an x building located in the central area of tokyo. The received content item corresponds to "inquiry content", and in the corresponding column, the very good company is recorded in a freely recorded manner, and the 6-layer very good meeting request changes the set temperature of the meeting room to 25 degrees to 18 points. Please respond. ".

The term of cause handling corresponds to the term of "response to inquiry", and "confirm that the measured temperature of the corresponding FCU is 26.7 ℃ is recorded in a freely recorded manner in the corresponding column, and the following changes are made. The set temperature of the corresponding FCU was changed from 27.0 ℃ to 25.0 ℃. At 18 h 00 min, the set temperature of the corresponding VAV was restored as it was. This was done by attendance x at 10 months 18 of 2019. The items of the company names are the same company, and are therefore omitted.

As illustrated in fig. 4 and 5, in maintenance management of a building, a document based on a uniform format is not necessarily generated in connection with or reporting of a failure. Even the buildings for which the same maintenance management company is responsible sometimes use different formats of contacts or reports depending on the preference of the owners or on the histories of the buildings. In addition, the format of the contact or report may also vary depending on the delivery tool used by the person making the contact or report. In particular, when the cross-over process is performed for a plurality of buildings, it is necessary to perform a process of acquiring character strings from documents having a plurality of formats.

Fig. 6 is a diagram showing character string generation performed in the preprocessing unit 81 of the building information processing apparatus 70. When the document shown in fig. 4 and 5 is input, the preprocessing unit 81 combines the description of the item corresponding to the "query content" and the "response to the query" to generate the "query/response character string". Items 1 and 2 in fig. 6 are character strings extracted from columns of

reference numerals

110 and 112 in fig. 4, respectively. Further, item 3 of fig. 6 is a character string generated by extraction from the document of fig. 5. It can be said that the information from delegation to coping with one failure is added to the character string of each item in fig. 6.

Fig. 7 is a diagram showing an example of the feature quantity dictionary 74 and the characteristic sentence dictionary 78 set in the building information processing apparatus 70. In the example shown in fig. 7, the feature quantity set in the feature quantity dictionary 74, that is, the sentence related to the failure of the equipment or the component of the building is described as being divided into items of feature quantities 1, 2, and 3. The feature quantity 1 is a device name, the feature quantity 2 is a component name, and a description statement for describing a fault is set as the feature quantity 3, and a statement indicating the determination of the fault and the request for coping with the fault is set. Further, a characteristic sentence related to the request set in the characteristic sentence dictionary 78 is described.

Items a to F in the left column in fig. 7 are items set for each failure. For example, the feature amounts 1, 2, and 3 and the characteristic statement described in item a are all described for any one failure. Specifically, "air conditioner", "FCU (fan coil unit)", as the device name of the feature quantity 1, and "VAV (variable air volume mode)", and "set temperature" as the component name of the feature quantity 2 are set. The set temperature is not necessarily the component name, but is a main element of the air conditioner, and is described in the feature quantity 2 and is subjected to the processing as the component name. As descriptions of faults in the feature quantity 3, "operation request", "change request", "hot" and "cold" are described. As characteristic sentences closely related to these feature amounts 1, 2, and 3, "temperature change request" is set. The characteristic term is a summary term that simply represents a request for requesting a change in a temperature regarded as a component name.

In item B of fig. 7, as in item a, as the feature quantity 1, "air conditioner", "FCU" are set. However, in the failure of item B, there is no content corresponding to feature quantity 2, and feature quantity 2 is a blank space. Regarding the feature amounts 1 and 2, if at least one of them is described, the device or the component to be the failure target can be specified, and therefore, one can be set as a blank. In the feature quantity 3, "abnormal noise", "odor", "inspection request", and "repair request" are set as faults. The feature quantity 3 indicates a failure, and therefore must be set. As a characteristic term, "air conditioner inspection request" corresponding to these feature amounts 1 to 3 is set. The characteristic sentence is a summary sentence that briefly represents a request for checking a device such as an air conditioner.

Similarly, in the example shown in fig. 7, characteristic sentences such as "toilet trouble", "bulb replacement request", "door trouble", "inquiry (alarm)" are set. The phrase "except for inquiry (alarm)" means that an alarm indicating an abnormality is notified to the central monitoring board. In the alarm, since there is no person who requests a response, a description other than the inquiry (alarm) is set. Further, feature amounts 1 to 3 corresponding to these feature words are set. In the feature quantity 3, for example, the phrases such as "deal with order" and "repair order" are set in a plurality of items, but since the feature quantity 3 represents a failure in groups with the feature quantities 1 and 2, there is no particular problem.

Fig. 8 is a diagram in which the results obtained by processing the extraction unit 82 and AI 84 of the building information processing apparatus 70 for each character string shown in fig. 6 are added to the table of fig. 6. The extraction unit 82 extracts sentences included in the entire feature quantity dictionary 74 from the query/response character strings indicating the faults of the items 1 to 3. As a result, in item 1, "automatic faucet", "wash basin" is extracted as feature 1, and "power plug", "water" is extracted as feature 2, and "no outlet", "response request" is extracted as feature 3. Similarly, in item 2, "door", "door handle", "screw", "loose" is extracted, and in item 3 "FCU", "VAV", "set temperature", "change request" is extracted. In the example of fig. 8, although mutually different feature words are extracted for items 1 to 3, feature words common to a plurality of items may be extracted depending on the content of the query/response string.

In AI 84, machine learning is performed using the relationship between the feature quantity dictionary 74 and the characteristic sentence dictionary 78 shown in fig. 7. That is, when the feature amounts 1 to 3 of any one of the items a to F in all fig. 7 are extracted and no other feature amount is extracted, the AI 84 selects the corresponding feature sentence. In addition, in the AI 84, even in a case where only a part of the feature amounts 1 to 3 of any one of the items a to F in fig. 7 is extracted and a part of the feature amounts 1 to 3 included in the other items is also extracted, learning is performed so that a proper characteristic sentence can be selected according to machine learning.

In the example shown in fig. 8, the feature amounts 1 to 3 extracted in item 1 are similar to item C shown in fig. 7, and therefore "toilet trouble" is selected as a characteristic sentence. Since the feature amounts 1 to 3 extracted in item 2 are similar to item E of fig. 7, the "gate failure" is selected as a characteristic sentence. Since the feature amounts 1 to 3 extracted in item 3 are similar to item a of fig. 7, a "temperature change request" is selected as a characteristic term. The selected characteristic sentence may be a classification item for classifying the fault.

In the example shown above, the query/response character string listed in fig. 6 is classified into 6 types of failures of items a to F illustrated in fig. 7. In the example of fig. 7, the feature amounts 1 to 3 are also very simple examples, and in practice, the number of setting sentences including synonyms and paraphraseology increases. The failure occurring in an actual building is related to various aspects and depends on the degree of division, but may be 100 or more or 1000 or more. The inquiry/response character string also depends on the number of buildings and the total period of failures, but it is also conceivable that the number of inquiry/response character strings can be easily 1000 or more, or 1 ten thousand or more, or 10 ten thousand or more. The building information processing device 70 can classify a plurality of character strings at high speed and with high accuracy.

Next, a process based on a rule base will be described with reference to fig. 9 and 10. Fig. 9 shows an example of sentences set in the rule base dictionary 94 of the building information processing apparatus 70. Here, the building name is set as item I, and the prefecture name where the building is located is set as item II.

The rule base sentence acquisition unit 96 of the building information processing apparatus 70 acquires a sentence or a corresponding value set in the rule base dictionary 94 from the document acquired by the document input unit 80. For example, in the business daily report shown in fig. 4, the term "good" is extracted based on the character "good, building public start". However, since there is no description about the prefecture name, the prefecture name is not extracted. In the contact document shown in fig. 5, since the item of the building name and the item of the address are provided, the sentence of "××building" and "tokyo" are acquired from the corresponding entry field. The obtained sentence is outputted from the output unit 90 of the building information processing apparatus 70.

Fig. 10 is a table of fig. 8 with building names and prefecture names added thereto. In fig. 10, in addition to the query/response character strings and feature amounts 1 to 3, the characteristic sentences and sentences of the rule base (building names, prefecture) output by the output unit 90 are described. By using not only characteristic sentences but also sentences of a rule base, multi-angle information can be extracted for building faults.

The extracted information is variously utilized by the PC 100. For example, by giving an output for each document as a summary of the document, it is expected that the document can be sorted or intuitively grasped. Further, the present invention is useful for summarizing business month report, business year report, and the like, for example, in performing business summary of a building.

Alternatively, for a plurality of buildings, statistical studies can be performed by counting failures for a period of time (for example, 1 year) long. In the case of comparing the statistics with a specific building, it is also possible to analyze the fault characteristics in the specific building.

In the above description, an example is shown in which a characteristic sentence centering on the request shown in fig. 7 is selected as the characteristic sentence concerning the failure. However, as the characteristic term, a term for specifying a failure or a term indicating the content of the response may be selected. In particular, when only the request is not included in the feature amounts 1 to 3, information useful for, for example, instructing an operator to perform a task is provided when a characteristic term such as a content to be predicted is selected.

In the above description, the handling of faults of devices or components of a building is exemplified. However, the present embodiment can be similarly applied to the operation of a building. Here, the operation of a building means a service required for using the building or performing maintenance management. The operation of a building includes the operation of devices or components of the building (operation herein refers to a service for continuing use regardless of the presence or absence of a failure). The operation of the building includes operation related to the building itself (for continuing use of business of the building) such as a building or a room of the building, and trouble handling related to the building itself.

The above-described manner is merely illustrative of embodiments, and various embodiments may be employed.

Description of the reference numerals

10: a building system; 20. 20a, 20b, 20c: building; 22: an air conditioning apparatus; 24: a water supply and drainage device; 24a: a component; 26: an electrical device; 28: an elevator; 30: a central monitoring room; 32: a central monitoring board; 36: maintenance personnel; 38: maintenance personnel; 40: tenant; 42: a mobile telephone; 44: a telephone; 60: a communication network; 70: building information processing means; 72: a feature dictionary setting unit; 74: a feature quantity dictionary; 76: a characteristic sentence setting unit; 78: a characteristic sentence dictionary; 80: a document input section; 81: a preprocessing section; 82: an extraction unit; 84: AI;86: a machine learning processing unit; 88: a characteristic sentence acquisition unit; 90: an output unit; 92: a rule base dictionary setting unit; 94: a rule base dictionary; 96: and a rule base sentence acquisition unit.

Claims

1. A building information processing apparatus, characterized in that the building information processing apparatus has:

an AI that learns, by machine learning using a sentence related to the operation of a building or a failure of a device or a component of the building, selection of a characteristic sentence indicating maintenance management characteristics related to the building, the device or the component, wherein the AI learns based on a sentence correspondence relationship set in a characteristic sentence dictionary;

an extraction unit that extracts, from a character string in which a contact or report relating to the operation of a building or a failure of a device or a component of the building is recorded, a sentence relating to the operation of the building or the failure of the device or the component; and

and a selection unit that selects, using the AI, the characteristic sentence corresponding to the extracted sentence related to the operation of the building or the malfunction of the device or the component.

2. The building information processing apparatus according to claim 1, wherein,

the building information processing apparatus further includes a preprocessing unit that obtains a plurality of character strings from a plurality of documents in different formats generated by a user or a maintenance manager of the building and describing contacts or reports related to operation of the building or failure of devices or components of the building,

the extraction unit extracts, for each of the character strings, a sentence relating to a failure of the device or the component,

the selection unit selects the characteristic sentence for each character string using the AI.

3. The building information processing apparatus according to claim 1, wherein,

the building information processing apparatus further includes a setting unit that sets at least a part of a sentence related to the operation of the building or the failure of the device or the component, in accordance with an instruction from an administrator of the building information processing apparatus.

4. The building information processing apparatus according to claim 1, wherein,

the term relating to the operation of the building or the failure of the device or the component includes the name of the building, the device or the component and a description term describing the operation or the failure occurring in the building, the device or the component.

5. The building information processing apparatus according to claim 4, wherein,

the description statement is a statement describing the specification of an operation or a failure, the request for handling the operation or the failure, or the implementation of handling the operation or the failure.

6. The building information processing apparatus according to claim 1, wherein,

the characteristic statement refers to a statement that is summarized for the operation or failure of the building, device or component.

7. The building information processing apparatus according to claim 1, wherein,

the characteristic sentence refers to a sentence for specifying an operation or a failure of the building, the device, or the component, requesting an operation or a failure to be handled, or predicting an operation or a failure to be handled.

8. The building information processing apparatus according to claim 1, wherein,

the building information processing apparatus further includes an output unit that outputs, when a predetermined sentence is included in the character string or a document in which the character string is acquired, the predetermined sentence or a sentence having a value corresponding to a case where the predetermined sentence is included in the character string together with the characteristic sentence.