CN113987184A

CN113987184A - Building operation and maintenance work order automatic filling system and method based on voice recognition

Info

Publication number: CN113987184A
Application number: CN202111273397.8A
Authority: CN
Inventors: 刘鹏飞; 曲志刚; 何晓; 张玉彬; 李彦; 饶冬东
Original assignee: Shandong Tongyuan Digital Technology Co ltd
Current assignee: Shandong Tongyuan Digital Technology Co ltd
Priority date: 2021-10-29
Filing date: 2021-10-29
Publication date: 2022-01-28

Abstract

The invention discloses a method and a system for automatically filling a building operation and maintenance work order based on voice recognition, wherein the method comprises the following steps: the method comprises the steps of constructing a mapping relation between each form item and a target keyword type in various work orders in advance, obtaining voice data, and converting the voice data into text data; performing keyword identification on the text data, and determining the type of the keyword; and matching the identified keywords with the form items according to the form mapping model of the corresponding work order, verifying the data, and automatically filling if the matching is successful and the verification is passed. According to the method, firstly, the repair voice description of the user is converted into a large text, the key field information can be extracted from the whole fault description by constructing the mapping relation between each form item and the target keyword type in various work orders and combining semantic analysis, and the structured form is automatically filled in to form structured form data, so that the method is beneficial to sorting and analyzing the subsequent data collection.

Description

Building operation and maintenance work order automatic filling system and method based on voice recognition

Technical Field

The invention belongs to the technical field of artificial intelligence, and particularly relates to a building operation and maintenance work order automatic filling method and system based on voice recognition.

Background

The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.

In the field of building operation and maintenance, when a common user initiates a work order, at least the following fields are manually filled in a work order system through a computer or a mobile phone and other terminals: the name, the position and the fault state of the fault equipment are described, and the efficiency is low. In order to enhance user experience and improve work order reporting efficiency, the existing work order systems provide a voice reporting function, but most of the systems do not have a dictation escape function, the intelligent degree of the reporting mode is low, special posts are often needed, voice texts reported by users are manually converted into structured form data, and the labor cost and the scheduling time cost are increased. Although few products have voice repair and automatically convert the voice repair into text to describe the repair fault and upload voice attachments, the voice repair is only used for the repair fault information review for the maintenance management personnel.

For special industries such as hospitals, medical staff often have no time or fill in different fields in a complex work order form in detail in a sterile environment during the repair process, so that high requirements are imposed on the efficiency and accuracy of work order filling. Although a work order automatic generation method based on voice recognition has been proposed at present, most of the existing methods provide a solution, for example, the recognition of work order contents is performed by a named entity recognition method, but specific application scenarios, such as work order content recognition in a complex field of hospital building operation and maintenance, are not combined, and the hospital building operation and maintenance field has the characteristics of more operation and maintenance objects, more types, various description modes and the like; the accuracy of work order content identification also needs to be improved.

Disclosure of Invention

In order to overcome the defects of the prior art, the invention provides a building operation and maintenance work order automatic filling method and system based on voice recognition. The method has the advantages that the repair voice description of a user is converted into a large text, the key field information can be extracted from the whole fault description by constructing the mapping relation between each form item and the target keyword type in various work orders and combining semantic analysis, and structured forms are automatically filled in to form structured form data, so that the method is beneficial to sorting and analyzing subsequent data collection.

In order to achieve the above object, one or more embodiments of the present invention provide the following technical solutions:

a building operation and maintenance work order automatic filling method based on voice recognition is provided, a mapping relation between each form item and a target keyword type in various work orders is constructed in advance, and the method comprises the following steps:

acquiring voice data and converting the voice data into text data;

performing keyword identification on the text data, and determining the type of the keyword;

and matching the identified keywords with the form items according to the form mapping model of the corresponding work order, verifying the data, and automatically filling if the matching is successful and the verification is passed.

Further, performing keyword recognition on the text data, and determining a keyword type includes:

performing word segmentation on the text data to obtain a word segmentation set;

performing keyword recognition based on a pre-constructed building operation and maintenance professional lexicon to obtain an object to be maintained and a repair description keyword;

and performing semantic analysis on the remaining participles, and determining the keywords belonging to the specified target type, wherein the specified target type is the type required in the form item.

Further, if the word segmentation which cannot be identified exists, the type of the word segmentation is determined through semantic analysis or manual work, and the word segmentation is expanded to a building operation and maintenance professional word bank.

Further, if the form item does not match the keyword, performing secondary matching on the text data according to the character type of the form item and the type of the required keyword.

Further, the method for generating the word stock of the building operation and maintenance specialty comprises the following steps:

performing word segmentation processing on the speech material based on a pre-constructed building operation and maintenance basic word bank to obtain a word segmentation set;

and performing semantic and part-of-speech analysis on the participles, judging whether the participles belong to the building operation and maintenance field, adding the newly recognized vocabularies into the building operation and maintenance basic word bank, and generating the building operation and maintenance professional word bank.

One or more embodiments provide a building operation and maintenance work order automatic filling system based on voice recognition, which comprises:

the form mapping model management module is used for managing the mapping relation between each form item in various work orders and the target keyword type;

the voice recognition module is used for acquiring voice data and converting the voice data into text data;

the keyword identification module is used for carrying out keyword identification on the text data and determining the type of the keyword; and matching the identified keywords with the form items according to the form mapping model of the corresponding work order, verifying the data, and automatically filling if the matching is successful and the verification is passed.

Further, still include: and the work order form item management module is used for managing form item structures of various work orders.

Further, still include:

the corpus collection module is used for acquiring historical work order filling information;

and the word bank building module is used for performing word segmentation and classification processing according to the linguistic data to generate a building operation and maintenance professional word bank.

Further, still include: and the database management module is used for managing the word segmentation table, the log record table, the word segmentation classification table and the message storage table.

Further, still include: and the word bank expansion module is used for acquiring the segmented words which cannot be identified by the keyword identification module, determining the type of the segmented words through semantic analysis or manual work, and expanding the segmented words to the building operation and maintenance professional word bank.

The above one or more technical solutions have the following beneficial effects:

by identifying the key word category and constructing the mapping relation between each form item in various work orders and the target key word type, the work order can be quickly filled;

in the process of carrying out keyword recognition and type recognition, firstly, the semantic analysis is carried out on the text data to match with the target type, then, the secondary matching is carried out on the text data according to the target type, and through bidirectional matching, the accuracy of keyword recognition is ensured, omission is not easy to occur, and the subsequent manual workload is saved;

by the work order content identification method, a large amount of equipment fault information can be collected, and a data basis is provided for performing big data analysis on data such as fault equipment types, equipment positions, equipment faults and maintenance means.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.

FIG. 1 is a flow chart of a method for automatically filling a building operation and maintenance work order based on speech recognition according to one or more embodiments of the present invention;

fig. 2 is a frame diagram of an automatic building operation and maintenance work order filling system based on speech recognition in one or more embodiments of the invention.

Detailed Description

It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.

The embodiments and features of the embodiments of the present invention may be combined with each other without conflict.

Example one

The embodiment discloses building operation and maintenance work order automatic filling system based on speech recognition, including:

and the corpus collection module is used for acquiring historical work order filling information, and specifically, storing the filling information to a queue every time a filled work order is received.

And the word bank building module is used for performing word segmentation and classification processing according to the linguistic data to generate a building operation and maintenance professional word bank. Specifically, the building operation and maintenance word segmentation extraction method comprises the following steps:

(1) constructing word bank classification based on professional knowledge in the field of building operation and maintenance;

(2) performing word segmentation processing on the speech material based on a pre-constructed building operation and maintenance basic word bank, and removing stop words to obtain a word segmentation set;

(3) carrying out vocabulary expansion by combining a synonym table according to the word segmentation set to obtain different expressions aiming at the same keyword;

(4) and performing semantic and part-of-speech analysis on the participles, and judging whether the participles belong to the building operation and maintenance field vocabularies. As one implementation, the determination may be made in conjunction with the semantics of the lexicon and the part-of-speech of the participle.

And the word bank expansion module is used for acquiring the words which can not be identified by the keyword identification module, analyzing the words and determining the categories of the words.

Specifically, the module comprises two parts of automatic expansion and manual expansion. When a new word is found in the word segmentation process, the system stores the entry into the database and records the frequency of the new word. The new words are processed in two ways: manual processing and automatic processing. The manual processing mode is that an administrator can manually set the classification information and synonyms of the new words; the automatic processing mode is as follows: firstly, the system can identify information such as word meaning and part of speech of the current word, and automatically classify the current word according to the information such as word meaning.

The database management module adopts a relational database management system (MySQL) and comprises a word segmentation table, a log record table, a word segmentation classification table and a message storage table. And the word segmentation table is used for storing the building operation and maintenance word segmentation. The log record table is used for recording the word segmentation condition of each time, and comprises the corpus on which the word segmentation is based, the word segmentation result, the word segmentation time, the word segments which are not successfully segmented, and the like. The word segmentation classification table is used for recording the unified words of the same keyword with different expressions, for example, expressions such as a pc, a computer and a display are unified into a computer. The message storage table is in a form of a message queue and is used for recording automatic filling and sending conditions of the form, for example, a certain form is automatically filled and sent to a certain operation and maintenance personnel, and the form cannot be automatically filled and sent to an administrator.

The work order form item management module is used for managing form item structures of various work orders, and the form item structures can comprise form items of repair equipment, repair positions, repair personnel, repair telephones, repair departments and the like, and define data types and bit numbers of the form items, such as text types, numerical types, floating point types and the like.

And the form mapping model management module is used for managing the mapping relation between each form item in various work orders and a target keyword type, wherein the target keyword type can be an address, a name, a number sequence and the like.

And the voice recognition module is used for acquiring voice data and converting the voice data into text data.

In this embodiment, third party speech to text sdk is used to convert the speech to text.

Performing keyword recognition on the text data, and determining the type of the keyword specifically comprises:

(1) performing word segmentation on the text data, and removing stop words to obtain a word segmentation set;

(2) in order to enable a computer to fully identify the text input by a user, synonym expansion is carried out on a participle set based on a corpus to obtain different expressions of each participle;

(3) performing keyword recognition based on a pre-constructed building operation and maintenance professional lexicon to obtain an object to be maintained and a repair description keyword, such as a computer and an elevator; and if the word which cannot be identified exists, performing word expansion through a word bank expansion module.

(4) And performing semantic analysis on the remaining participles, and determining the keywords belonging to a specified target type, wherein the specified target type is a type required in the form item, and comprises time, name, department, telephone and the like.

(5) And judging whether the form item does not match the keyword, if so, performing secondary matching on the text data according to the character type of the form item and the type of the required keyword. For example, by form field name, field type, or keyword type, etc. The form field name is used to define and match the form item corresponding to the form, such as: repair equipment and repair positions; form field types, which are used to define character types, such as: character type, integer type, floating point type, date type; the corresponding keyword type is used for defining the part of speech range of the matching list item content, such as: computer, elevator.

According to the technical scheme included in the step (4) and the step (5), the target type is matched through semantic analysis on the text data, then the text data is matched secondarily according to the target type, and through bidirectional matching, the accuracy of keyword identification is guaranteed, omission is not prone to occurring, and the subsequent manual workload is saved.

(6) And filling in a fault description according to the text data. Since the description of the fault may be unclear when the user inputs the voice, for example, only "lamp is out" but not what kind of situation is described, and there may be various situations such as flashing and not lighting of the fault state of the lamp, in this embodiment, after the text data is acquired, related keyword expansion is performed based on the pre-trained semantic analysis-based association prediction model, and after the user confirms, description sentences are generated, for example, the user inputs "lamp is out", and the system automatically associates the possible lamp states, such as "flashing" and "not lighting", and "may be the voltage cause".

Specifically, the associated prediction model based on semantic analysis adopts a hidden markov model, and solving the most probable hidden state sequence is one of three typical problems of the hidden markov model, and is usually solved by a viterbi algorithm. The viterbi algorithm is an algorithm for solving the shortest path (-log (prob), i.e. the maximum probability) on the hidden markov model. The associated prediction model training method based on semantic analysis comprises the following steps: acquiring equipment, fault state information and a fault reason; and training the association relation between the equipment state and the fault reason based on the hidden Markov model. Assume that a device has N states, for example: device aging, power outage, component problems, etc. (expressions are state _1, state _2,. and. state _ N), then the transition probabilities of these N states can be recorded by a matrix:

transProbMatrix＝

[tp_1_1,tp_1_2,...,tp_1_N；

tp_2_1,tp_2_2,...,tp_2_N；

...

tp_N_1,tp_N_2,...,tp_N_N

]

wherein the element in the ith row and the jth column represents the probability that the state of the device will switch from state _ i to state _ j. The probability distribution of the initial state values of the device is initProb [ ip _1, ip _ 2.,. ip _ n ], and then n elements ip _ n represent the probability that the state is state _ n when the device fails.

Let us note the observation as v, and assume that there are M possible fault state values of the device, then under state _ n, the value probability distribution of the observation value is:

vProb_n＝[vp_n_1,vp_n_2,...,vp_n_m]

and vp _ m represents the probability that an observation value is generated to be v _ m when the equipment is at state _ n.

It is assumed that the device user can enter a message into the system: "the light in xx room is continuously flashing for half an hour, but now returns to normal, and the equipment is not repaired".

Then in this case the explicit state is a lamp flashing without interruption and the implicit state is three causes of failure:

and solving the lamp tube replacement time and the fault probability.

And solving the replacement time and the fault probability of the rectifier.

The most likely voltage case is solved.

Assuming that the lamp and rectifier problem is eliminated, we proceed with the following analysis, taking voltage failure as an example:

defining T [ time ] [ voltage condition ] ═ probability, note that the device operating condition refers to the probability that the voltage condition 30 minutes before the current time determines (has the highest probability) that the current voltage condition is X, where the probability is an accumulated probability.

Since the lamp has flashed for half an hour continuously, the probability T [30 minutes ago ] [ voltage stable ] ═ initial probability [ unstable ]. emission probability [ voltage stable ] [ flashing ] ═ 0.6 × 0.1 ═ 0.06 before half an hour, and in the same way, we can get that T [30 minutes ago ] [ voltage unstable ] ═ 0.24, the strategy engine predicts: since the lamp begins to flicker before 30 minutes and the lamp flicker is caused only when the voltage is unstable, the probability of the unstable voltage before 30 minutes is relatively high, and the data is consistent with the guessed prediction result.

After 30 minutes, for each lamp state Y, there is a probability that the state before 30 minutes is X, X transitions to Y, and the probability that the voltage condition will be determined for the lamp state Y. Since there are two possibilities for the 30 minute headlight state X, there are two probabilities for Y, the larger of which is chosen as the probability for T [30 minute later ] [ lamp state Y ], while adding the 30 minute later lamp state to the resulting sequence.

And comparing the probability of the flashing light in T [30 minutes later ] with the probability of the non-flashing light in T [30 minutes later ], and finding out the larger corresponding sequence, namely the final result.

And the form filling module is used for filling forms according to the identified keywords to generate the building operation and maintenance work order.

The word library system is developed by adopting java language, a mysql database and redis adopted for caching, and the development mode is a front-end and back-end separation mode. In the development style, we choose the MVC style, i.e., model-view-controller architecture, in development. A Model represents an object accessing data and its data Model. View (View) represents the representation of the data contained in the model, and is generally expressed as a visual interface. A Controller acts on the model and the view, controls the flow of data to the model objects, and updates the view as data changes. The controller may decouple the view from the model.

Example two

Based on the system provided by the first embodiment, the first embodiment provides a building operation and maintenance work order automatic filling method based on voice recognition, a mapping relation between each form item in various work orders and a target keyword type is constructed in advance, and the method comprises the following steps:

step 1: acquiring voice data and converting the voice data into text data;

step 2: performing keyword identification on the text data, and determining the type of the keyword;

and step 3: and matching the identified keywords with the form items according to the form mapping model of the corresponding work order, verifying the data, and automatically filling if the matching is successful and the verification is passed.

The step 2 specifically comprises:

The above one or more embodiments can realize the rapid filling of the work order by identifying the keyword category and constructing the mapping relation between each form item in various work orders and the target keyword type; in addition, in the process of carrying out keyword identification and type identification, by means of various identification and matching methods, the accuracy of keyword identification is ensured, the omission of form items is not easy to occur, and the subsequent manual workload is saved.

Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using general purpose computer means, or alternatively, they can be implemented using program code that is executable by computing means, such that they are stored in memory means for execution by the computing means, or they are separately fabricated into individual integrated circuit modules, or multiple modules or steps of them are fabricated into a single integrated circuit module. The present invention is not limited to any specific combination of hardware and software.

Although the embodiments of the present invention have been described with reference to the accompanying drawings, it is not intended to limit the scope of the present invention, and it should be understood by those skilled in the art that various modifications and variations can be made without inventive efforts by those skilled in the art based on the technical solution of the present invention.

Claims

1. A building operation and maintenance work order automatic filling method based on voice recognition is characterized in that a mapping relation between each form item and a target keyword type in various work orders is constructed in advance, and the method comprises the following steps:

acquiring voice data and converting the voice data into text data;

2. The method as claimed in claim 1, wherein the step of performing keyword recognition on the text data and determining the type of the keyword comprises:

3. The method as claimed in claim 2, wherein if there is an unrecognized participle, the type of the participle is determined by semantic analysis or manual work and extended to the professional word bank of building operation and maintenance.

4. The method as claimed in claim 2, wherein if there is a form item not matching with the keyword, performing secondary matching on the text data according to the character type of the form item and the type of the required keyword.

5. The method for automatically filling in the building operation and maintenance work order based on the voice recognition as claimed in claim 1, wherein the method for generating the building operation and maintenance professional lexicon comprises:

6. A building operation and maintenance work order automatic filling system based on voice recognition is characterized by comprising:

7. The system as claimed in claim 6, further comprising: and the work order form item management module is used for managing form item structures of various work orders.

8. The system as claimed in claim 6, further comprising:

9. The system as claimed in claim 6, further comprising: and the database management module is used for managing the word segmentation table, the log record table, the word segmentation classification table and the message storage table.

10. The system as claimed in claim 6, further comprising: and the word bank expansion module is used for acquiring the segmented words which cannot be identified by the keyword identification module, determining the type of the segmented words through semantic analysis or manual work, and expanding the segmented words to the building operation and maintenance professional word bank.