CN110489749B - Business process optimization method of intelligent office automation system - Google Patents
Business process optimization method of intelligent office automation system Download PDFInfo
- Publication number
- CN110489749B CN110489749B CN201910723500.0A CN201910723500A CN110489749B CN 110489749 B CN110489749 B CN 110489749B CN 201910723500 A CN201910723500 A CN 201910723500A CN 110489749 B CN110489749 B CN 110489749B
- Authority
- CN
- China
- Prior art keywords
- layer
- word vector
- word
- official document
- business process
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
Abstract
The invention discloses an intelligent office automation system business process optimization method, wherein the intelligent office automation system business process mainly comprises the steps of building an office system environment based on a BPM workflow engine, constructing a official Word2vec Word vector library, carrying out regularization processing on the Word vector library, extracting an official document feature set and identifying official document information based on a convolutional neural network. The business system environment based on the BPM workflow engine realizes information transmission, data synchronization, business monitoring and continuous upgrading optimization of enterprise business processes. The environment supports the system mapping of the actual model management business and the workflow, comprehensively defines the operation content and the executive personnel of each task accurately, and converts the business process management into uniform, automatic and standard computer system operation. The invention further realizes the business process optimization based on the convolutional neural network by constructing the Word2vec Word vector library of the official document, coordinates various system resources in an enterprise, reduces the time cost and improves the response efficiency of a business process system.
Description
Technical Field
The invention relates to a document information intelligent identification technology in an office automation system, realizes automatic recommendation of a service process, and mainly relates to construction of a document Word2vec Word vector library and document information identification based on a convolutional neural network.
Background
In recent years, with the temperature rise of big data and artificial intelligence concepts, more and more enterprises begin to attach importance to the commercial value in the data, and with the increasing maturity of data mining technology, machine learning is more and more widely and deeply applied in multiple industries such as government, finance, communication, medical treatment and retail, a data analysis model is built based on specific business target problems and historical data, and model-based management behaviors such as model storage, model verification, model deployment application and model performance monitoring are implemented, so that the method becomes an important link in the data mining process of the enterprises. In order to meet the changing market environment and the changing customer requirements, how to establish a high-efficiency, repeatable, fast-iterating and high-flexibility model management process is one of the key topics for enterprises to improve the data analysis capability and convert the capability and resources into values, and has very strong research and practice significance. The patent explores an application mode of data analysis model management and workflow integration based on the direction.
Conventional message-driven or event-driven workflow systems can better support the operation of office systems, but have not kept pace with the rapid development of artificial intelligence. To combine the official document information and the historical data analysis and provide intelligent services, a data-driven technology must be incorporated into the original workflow system. In addition, the government of China mainly bases on documents in office work, so the importance of intelligent services based on documents is particularly prominent. How to apply the document cooperative processing technology and the semantic processing technology to an office system to realize intelligent identification of official document information and provide intelligent services for users is an important problem to be solved.
Although the existing system changes the original document processing mode, the document processing flow is not changed, that is, an advanced management method is not added. In spite of the products which are commonly used at present, the functions which can be provided by the products are basically processing around official documents, and the additional functions are far from satisfactory in terms of depth and breadth. The fundamental reason is that the systems only process the official documents and do not utilize the processed official documents, that is, the knowledge is not accumulated, and the decision service for the enterprise needs cannot be made according to the existing information. The official document system does not have the automatic official document sorting function when the official documents are distributed, and the requirement on users is high, which is a demand to be solved urgently. In addition, with the rapid expansion of information technology, the official documents generated by enterprises are increased rapidly, the data volume in the process logs is huge, and how to effectively utilize the data to effectively work becomes a demand to be solved by each company. Therefore, for the characteristics of massive, dispersed and dynamic change of information data in the private cloud environment, a key technology research for data-driven business process optimization in the private cloud environment needs to be performed.
Disclosure of Invention
The invention provides a business process optimization method of an intelligent office automation system, which is mainly based on intelligent identification of official document information by a data-driven method and mainly has the following functions: the method comprises the steps of building an office system environment based on a BPM (Business process management) workflow engine, building a Word vector library of a document Word2vec, performing regularization processing on the Word vector library, extracting a document feature set and identifying document information based on a convolutional neural network, wherein the document information is shown in FIG. 1.
The intelligent office automation system adopts an office system environment based on a BPM workflow engine, and realizes information transmission, data synchronization, service monitoring and continuous upgrading optimization of enterprise service processes. The environment supports the system mapping of the actual model management business and the workflow, comprehensively defines the operation content and the executive personnel of each task accurately, and converts the business process management into uniform, automatic and standard computer system operation. Meanwhile, a deep learning frame is incorporated in the environment as an auxiliary support, the business process optimization based on the convolutional neural network is further realized through the construction of a Word2vec Word vector library, various system resources in an enterprise are coordinated, the time cost is reduced, and the response efficiency of a business process system is improved.
The invention has the advantages that:
1. the business process recommendation system adopts an Office Automation (OA) system based on a business process engine (BPM), compared with the traditional workflow engine, the system has strong information integration, the circulating data among documents is transmitted in real time, various IT systems and resources can be integrated, uniform process services are provided for other applications, and the business process recommendation system is a business process approval management platform integrating multiple information services. In addition, the high configurability of the process can greatly reduce the development amount, and various simple or complex business process forms can be managed quickly at low cost, and meanwhile, the unified management of manual processes and automatic processes can adapt to the requirement of quick change of the process.
2. The technique of the Word2vec Word vector library models the context semantics and the relationship between the context and the target through a neural network. The method provides a form of converting words into vectors aiming at complex official document information, effectively solves the problem, can reduce dimensionality, and can capture position associated information of a certain word in the official document.
3. The business process optimization system introduces TF-IDF (Word frequency-inverse text frequency) algorithm to calculate the weight of each Word vector in the text on the basis of Word2vec Word vector library technology, and is used for evaluating the importance degree of a Word or a document in a corpus. This provides the BPM workflow engine with a measure or rating as to how relevant the document is to the user query.
4. The business process optimization system realizes intelligent identification of official document information based on a convolutional neural network, different convolutional cores are used for carrying out convolution on an official document matrix, the width of a convolutional kernel is equal to the length of a word vector, then max-posing (maximum value sampling) is used for operating a vector extracted by each convolutional kernel, finally each convolutional kernel corresponds to a number, the convolutional kernels are spliced to obtain a sentence vector representing the official document, and finally a recommended result of a business process is obtained through weighting of weights of all parts, so that the efficiency of office staff and the accuracy of the business process are improved.
Drawings
FIG. 1 is a functional framework of a business process optimization system based on a BPM workflow engine;
FIG. 2 is a BPM workflow engine framework;
FIG. 3 is an exemplary diagram of a business process message flow;
FIG. 4 is a Word2 vec-based document to Word vector mapping;
FIG. 5 is a flowchart of the document feature extraction and weight assignment;
FIG. 6 is a document classification model of a convolutional neural network;
FIG. 7 is a business process intelligent recommendation system interface.
Detailed Description
The following describes in detail a method for optimizing a business process of an intelligent office automation system according to the present invention with reference to the accompanying drawings:
1. OA office system based on BPM workflow engine
The BPM is based on the management of a business process, and fully supports the full-automatic cooperative operation of all aspects of process design, execution, management, association and optimization. In order to integrate all service flows and manual flows into BPMN (BPM network), a complex architecture design is required, as shown in fig. 2, a BPM engine integrating BPEL4 peoples is a real core of the whole system, a domain expert can design a test deployment flow into the BPMN engine by using BPMN2, a management console of the BPM engine can perform various kinds of management of flow operation, and the BPM engine is not only responsible for sending tasks to a manual task system, but also responsible for interactive communication to each rest (representational State transfer) endpoint service. Where the BPM message flow acts on a BPMN collaboration diagram showing how two or more flows without centralized control interact with each other in a synchronized manner. Message flow is a way to express how two separately controlled flows communicate and cooperate with each other, an activity or event in one pool may initiate a message to the other pool, the message flow is depicted as a dashed line, where an empty circle represents the source of a message, an empty arrow represents the termination of a message, the message flow representation is for example as shown in fig. 3. Each flow is contained in its own Pool, and the pools are usually labeled with participant names. The process oriented to business management can drive various resources and services, and realize various queries and statistical analysis. The office system unifies information exchange, knowledge, notice announcement and the like on the process portal, improves organization and team cooperation in the process of process execution, and can analyze and trace the process and the result.
2. Construction of official Word2vec Word vector library and extraction of official document feature set
Word2vec is a group of correlation models used to generate Word vectors. These models are two-layer neural networks, which reconstruct the word vector library by training. After training is completed, the word2vec model can be used to map each word to a vector, which can be used to represent the word-to-word relationship, and the specific mapping process is shown in fig. 4.
The training process for Word2vec is as follows:
1) and expressing the words subjected to word segmentation in an independent hot type coding mode.
2) And multiplying the word vector obtained in the last step by a projection matrix to obtain the input of the hidden layer, and then obtaining the output of the hidden layer through an activation function. But for simple calculations it is common to sum the outputs of the input layers directly to obtain the input of the hidden layer.
3) The output layer is a huffman tree in which leaf nodes correspond to words in the vocabulary and non-leaf nodes are equivalent to the parameters from the hidden layer to the output layer. The weights between the input layer and the hidden layer can be obtained by training the model.
4) The word vector representation of the word is obtained by multiplying the initial one-hot type coded word representation by the weight between the input layer and the hidden layer.
3. Document feature extraction and weight assignment
The book module firstly checks the document diPerforming word segmentation processing Wi＝[w1，w2，…，wn]N is the number of words, and then the text after Word segmentation is replaced by a low-dimensional numerical value vector according to a Word2vec Word vector library
Is wiThe word vector of (a) is,
k is the dimension of the word vector, so that the official document representation is changed from high-latitude high-sparsity traditional data which are difficult to process by a neural network into continuous dense matrix data representation similar to an image. The official document representation method avoids the tedious work of manual feature selection in the traditional machine learning text classification algorithm, and enables official document information to obtain the maximum degreeAnd the patent adopts an improved TF-IDF algorithm to carry out word vector weight calculation, and the specific processing flow is shown in FIG. 5.
4. Official document information identification based on convolutional neural network
The convolutional neural network generally consists of an input layer, a plurality of convolutional layers, a pooling layer, a full link layer and a softmax layer, and a convolutional neural network model for official document information identification is shown in fig. 6. One convolutional layer has a plurality of different convolutional kernels, and omega belongs to RhkH is the height of a convolution kernel, k is the space dimension of a word vector, omega is a convolution kernel matrix weight parameter, the convolution kernel slides downwards by step length 1, and convolution operation is carried out when a window of a text vector h x k passes through to generate a new characteristic value; wi：i+hIs a word sequence (W) with the length of h +1i，Wi+1，…，Wi+h) B is a bias term, b is equal to R, the operator (. cndot.) is convolution calculation, and f is an activation function. Processing the text vector by a convolution kernel to obtain a feature map c ═ c (c)1，c2，…，cn-h+1) And n is the number of words in the official document. The pooling layer is used for extracting the characteristics of the characteristic diagram by 1-max-posing, cmMax c, the different length texts become the same length features after being processed by the pooling layer. The input of the full connection layer is the characteristic output of the pooling layer, and the input is
p is the type of the convolution kernel, q is the number of each convolution kernel, and the output layer uses the softmax function to perform category judgment, so that the intelligent recommendation of the business process is achieved, and the specific implementation interface is shown in fig. 7.
Claims (4)
1. A business process optimization method of an intelligent office automation system is characterized by comprising the following steps: the method comprises the steps of establishing an office system environment based on a BPM workflow engine, constructing a official Word2vec Word vector library, carrying out regularization processing on the Word vector library, extracting an official document feature set and identifying official document information based on a convolutional neural network; the construction of the Word2vec Word vector library maps historical official document data to Word vectors; carrying out convolution weighted classification on a matrix formed by the official document word vectors based on official document information identification of the convolution neural network to obtain a recommendation result of the service process.
2. The business process optimization method of an intelligent office automation system of claim 1, wherein: the construction of the Word2vec Word vector library of the official document models the context semantics and the relation between the context and the target through a convolutional neural network; the words of the official document after word segmentation are expressed in a single hot coding mode; multiplying the word vector obtained in the last step by a projection matrix to obtain the input of a hidden layer, and directly summing the output of the input layer to obtain the input of the hidden layer; the output layer is a Hoffman tree, wherein leaf nodes are words in a corresponding vocabulary table, non-leaf nodes are equivalent to parameters from the hidden layer to the output layer, and the weight between the input layer and the hidden layer is obtained through a training model; multiplying the initial one-hot type coded word representation by the weight between the input layer and the hidden layer to obtain a word vector representation of the word; aiming at complex official document information, a form of converting words into vectors is provided.
3. The business process optimization method of an intelligent office automation system of claim 1, wherein: regularization processing of a word vector library and extraction of a document feature set are realized, the TF-IDF algorithm is utilized to distribute word vector weights, and a document d is firstly subjected toiPerforming word segmentation processing Wi＝[w1，w2，…，wn]N is the number of words, and then the text after Word segmentation is replaced by a low-dimensional numerical value vector according to a Word2vec Word vector library
Is wiThe word vector of (a) is,
k is the dimension of the word vector, so that the official document representation is changed from high-latitude high-sparsity traditional data which are difficult to process by a neural network into continuous dense matrix data representation similar to an image.
4. The business process optimization method of an intelligent office automation system of claim 1, wherein: identifying official document information based on a convolutional neural network, wherein the identification comprises information processing of an input layer, a plurality of convolutional layers, a pooling layer, a full-link layer and a softmax layer; one convolutional layer has a plurality of different convolutional kernels, and omega belongs to RhkH is the height of a convolution kernel, k is the space dimension of a word vector, omega is a convolution kernel matrix weight parameter, the convolution kernel slides downwards by step length 1, and convolution operation is carried out when a window of a text vector h x k passes through to generate a new characteristic value; wi：i+hIs a word sequence (W) with the length of h +1i，Wi+1，…，Wi+h) B is a bias term, b belongs to R, the operator (-) is convolution calculation, and f is an activation function; processing the text vector by a convolution kernel to obtain a feature map c ═ c (c)1，c2，…，cn-h+1) N is the number of words in the official document; the pooling layer is used for extracting the characteristics of the characteristic diagram by 1-max-posing, cmMax { c }, after being processed by the pooling layer, texts with different lengths all become the features with the same length; the input of the full connection layer is the characteristic output of the pooling layer, and the input is
p is the type of convolution kernel, q is the number of each convolution kernel, and the output layer uses the softmax function to judge the type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910723500.0A CN110489749B (en) | 2019-08-07 | 2019-08-07 | Business process optimization method of intelligent office automation system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910723500.0A CN110489749B (en) | 2019-08-07 | 2019-08-07 | Business process optimization method of intelligent office automation system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110489749A CN110489749A (en) | 2019-11-22 |
| CN110489749B true CN110489749B (en) | 2021-05-18 |
Family
ID=68550041
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910723500.0A Active CN110489749B (en) | 2019-08-07 | 2019-08-07 | Business process optimization method of intelligent office automation system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110489749B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111782811A (en) * | 2020-07-03 | 2020-10-16 | 湖南大学 | E-government affair sensitive text detection method based on convolutional neural network and support vector machine |
| CN114331226B (en) * | 2022-03-08 | 2022-06-24 | 天津联创科技发展有限公司 | Intelligent enterprise demand diagnosis method and system and storage medium |
| CN115907674B (en) * | 2022-12-13 | 2023-12-01 | 广州明动软件股份有限公司 | Intelligent efficiency analysis method and system based on AI algorithm and neural engine |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102904885A (en) * | 2012-09-26 | 2013-01-30 | 北京工业大学 | Compound authentication method of multi-identity authentication information feature |
| CN107133785A (en) * | 2017-06-12 | 2017-09-05 | 山东浪潮云服务信息科技有限公司 | A kind of pending business prompting method of office system based on domestic CPU |
| CN108629550A (en) * | 2017-03-17 | 2018-10-09 | 武汉唐科技有限公司 | It is a kind of can backstage automatic running the office automation system implementation method |
| CN109635116A (en) * | 2018-12-17 | 2019-04-16 | 腾讯科技（深圳）有限公司 | Training method, electronic equipment and the computer storage medium of text term vector model |
-
2019
- 2019-08-07 CN CN201910723500.0A patent/CN110489749B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102904885A (en) * | 2012-09-26 | 2013-01-30 | 北京工业大学 | Compound authentication method of multi-identity authentication information feature |
| CN108629550A (en) * | 2017-03-17 | 2018-10-09 | 武汉唐科技有限公司 | It is a kind of can backstage automatic running the office automation system implementation method |
| CN107133785A (en) * | 2017-06-12 | 2017-09-05 | 山东浪潮云服务信息科技有限公司 | A kind of pending business prompting method of office system based on domestic CPU |
| CN109635116A (en) * | 2018-12-17 | 2019-04-16 | 腾讯科技（深圳）有限公司 | Training method, electronic equipment and the computer storage medium of text term vector model |
Non-Patent Citations (2)
| Title |
|---|
| Richard Yu-Yuan Hung."Business process management as competitive advantage: a review and empirical study".《Total Quality Management & Business Excellence》.2006,第17卷(第1期), * |
| 曾少宁."基于jBPM的工作流驱动办公系统的设计与实现".《计算机与现代化》.2010,(第6期),159-165. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110489749A (en) | 2019-11-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109767255B (en) | Method for realizing intelligent operation and accurate marketing through big data modeling | |
| CN110489749B (en) | Business process optimization method of intelligent office automation system | |
| US20150134401A1 (en) | In-memory end-to-end process of predictive analytics | |
| CN110196711B (en) | Plug-in type work order assistant decision-making method and system based on artificial intelligence | |
| CN111813958A (en) | Intelligent service method and system based on innovation and entrepreneurship platform | |
| CN104281615A (en) | Complaint handling method and system | |
| CN116737129B (en) | Supply chain control tower generation type large language model and construction method thereof | |
| WO2020162943A1 (en) | Holistic intelligence and autonomous information system and method thereof | |
| CN113806560A (en) | Power data knowledge graph generation method and system | |
| CN114238655A (en) | Enterprise association relation identification method, device, equipment and medium | |
| CN113705242B (en) | Intelligent semantic matching method and device for education consultation service | |
| CN114493535A (en) | Asset model-based data center system for data driving | |
| CN110490333A (en) | The professional content intelligent generation method write based on AI | |
| CN112241785B (en) | Book interview method based on deep reinforcement learning | |
| CN113902394A (en) | Service platform for realizing artificial intelligence industry ecology | |
| Shi et al. | Human resources balanced allocation method based on deep learning algorithm | |
| Wang et al. | A Web-based CBR knowledge management system for PC troubleshooting | |
| CN112818229A (en) | Ornament recommendation method, system, device and medium based on home space | |
| CN107505852B (en) | Method for constructing cloud manufacturing service description file containing simulation information | |
| CN115035044A (en) | Be applied to intelligent AI platform of industry quality inspection | |
| Kopp et al. | Enterprise Architecture Tools Assessment for Smart Cities Governance using Fuzzy Logic Techniques | |
| Cusumano et al. | Intelligent building contract tendering-potential and exploration | |
| Weber | AIoT Chances in Resource-Constrained Environments via Manual Pruning | |
| CN113868322B (en) | Semantic structure analysis method, device and equipment, virtualization system and medium | |
| Zhao et al. | Visual Analysis and Mining of Knowledge Graph for Power Network Data Based on Natural Language Processing |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |