CN113657091A - Government affair hot line work order allocation method based on event extraction and authority list - Google Patents
Government affair hot line work order allocation method based on event extraction and authority list Download PDFInfo
- Publication number
- CN113657091A CN113657091A CN202111025600.XA CN202111025600A CN113657091A CN 113657091 A CN113657091 A CN 113657091A CN 202111025600 A CN202111025600 A CN 202111025600A CN 113657091 A CN113657091 A CN 113657091A
- Authority
- CN
- China
- Prior art keywords
- work order
- government affair
- vector
- authority list
- hotline
- 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.)
- Withdrawn
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 30
- 230000000694 effects Effects 0.000 claims abstract description 9
- 239000013598 vector Substances 0.000 claims description 47
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000013075 data extraction Methods 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 239000004576 sand Substances 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 description 10
- 238000005516 engineering process Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012790 confirmation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 230000001681 protective effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biophysics (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a government affair hotline work order allocation method based on event extraction and a authority list, which comprises the following steps: step 1, constructing a semantic coding layer; step 2, constructing an event extraction layer; step 3, constructing a authority list embedding layer; and 4, allocating the work orders. The method can realize the rapid positioning of the hot events and hot areas in the government affair hot line work order, thereby providing reasonable suggestion suggestions for government services, optimizing the government service effect and improving the satisfaction of people.
Description
Technical Field
The invention relates to a government affair hotline work order allocation method based on event extraction and a authority list.
Background
In recent years, with the rapid increase of the traffic volume of the government affair hotline, governments try to embed IT technology into the internal management and service flow of the government affair hotline, and digital transformation of the government affair hotline is gradually promoted. Although the application of IT technology improves the quality of service of government hot line traffic, there are still some problems in the work order processing. For example, the operator needs to find the responsibility unit to be distributed for each work order, and the alternative responsibility units are usually as many as several tens of responsibility units, so that the process of confirming the responsibility units is time-consuming and labor-consuming. In addition, the manual confirmation also depends on the accurate judgment of the operators on the classification of citizens' appeal and the deep understanding of the functions of each responsibility unit, so that the accuracy of work order allocation needs to be further improved. With the increasingly numerous and complicated types and the increasingly large number of urban events accepted by the government affair hotline, it is of great significance to research an intelligent allocation method capable of quickly and accurately positioning work order event handling departments.
In order to realize automatic classification of government affair hotline worksheets, a patent with publication number CN112541351A discloses a method and a system for dispatching government affair hotline worksheets in the field of civil construction, and the method and the system finish dispatching the government affair hotline worksheets in the field of civil construction by using textRank and SVM. The method based on machine learning needs a great amount of feature engineering, the selection and analysis mode of the features is complex, and the features which are designed by spending a great deal of effort are probably irrelevant to the designated task. Patent publication No. CN112308251A discloses a work order assignment method and system based on machine learning, which uses TF-IDF and KNN to extract and classify key information of IT operation and maintenance work order fault data. However, work orders belonging to the same general class but different subclasses have many similarities in the textual information description. For example, the descriptions of "parking on the road" and "repairing the vehicle on the road" belonging to the category of "street order" are very similar, but belong to different units of responsibility (police and city management bureaus). It is difficult to find such a small difference by using a neural network method, and it is difficult to make a correct judgment.
Disclosure of Invention
The invention aims to provide a government affair hot line work order allocation method based on event extraction and a authority list.
In order to achieve the purpose, the invention provides a government affair hot line work order allocation method based on event extraction and authority list, which comprises the following steps:
step 3, constructing a authority list embedding layer;
and 4, allocating the work orders.
Preferably, step 1 comprises:
step 1.1, performing semantic coding on the work order text content by using a RoBERTA pre-training language model, expressing a token sequence into a vector form after the text input is converted into the token sequence, wherein the token sequence consists of token embedding, sentence embedding and position embedding, and sending the token sequence into a coding layer stacked by 12 layers of transform coder modules to extract semantic features;
step 1.2, after being coded by a RoBERTA model, splicing and fusing the coded output of any layer in a 12-layer transform coder to obtain a context embedded matrix, and inputting the context embedded matrix into a subsequent layer to obtain the best effect;
and step 1.3, after the semantic representation vector with the context feature information is obtained, inputting the fused code into an event extraction layer.
Preferably, step 2 comprises:
step 2.1, extracting word level characteristics through CNN;
step 2.2, obtaining text context semantic information by using a BiGRU network to obtain sentence level characteristics;
step 2.3, splicing sentence level characteristics and word level characteristics after applying Self-Attention;
and 2.4, optimizing the event extraction network by using the conditional random field.
Preferably, step 2.1 comprises:
step 2.1.1, using the work order text encoding vector output by Roberta as an input sequence of a CNN network, extracting corresponding local features, and outputting semantic features of the whole sentence text in the form of feature vectors;
step 2.1.2, after the convolution operation is completed, sending each obtained feature vector to a pooling layer to generate potential local features, pooling the output result of the convolution layer by adopting a maximum pooling strategy, capturing the most important features after convolution, and processing a variable-length sentence into a fixed length;
step 2.1.3, scanning each word in the sentence by using a convolution kernel to obtain a feature vector of the whole sentence;
and 2.1.4, acquiring text features on different scales through a convolution set, repeating the convolution process, and splicing results of convolution kernels to obtain a feature vector of the whole government affair service hotline text sentence.
Preferably, step 2.2 comprises:
step 2.2.1, respectively inputting elements in the semantic representation vector L into a forward GRU and a reverse GRU to obtain a forward hidden state
And reverse hidden state
Step 2.2.2, splicing the forward and reverse hidden states to obtain the hidden state of the BiGRU
Preferably, step 2.3 comprises:
step 2.3.1, encoding each input part to form semantic code Iembedding；
Step 2.3.2, establishing a parameter matrix WQ、WKAnd WVEncoding semantics into IembeddingLinear mapping is carried out in a feature space to form Q, K, V three vectors;
step 2.3.3, similarity calculation is carried out on Q and K to obtain a calculated attention weight, normalization operation is carried out on the obtained weight, and finally the weight and Value are subjected to weighted summation to obtain a final attention score Vout。
Preferably, step 2.4 comprises:
2.4.1, selecting a BIO label system to label the sequence, wherein all words marked by event locations LOC and event trigger words TRG in the government affair hot line work order text data set are marked by words with event locations LOC and event trigger words TRG, the first word in the words is marked by B-LOC and B-TRG again, the rest words in the words are marked by I-LOC and I-TRG again, more than two continuous numbers are marked by combination, and other words are marked by O uniformly;
step 2.4.2, inputting the splicing vector of Self-orientation and CNN into CRF after passing through a full-connection network to obtain a label sequence corresponding to the work order text;
step 2.4.3: using CRF to assign marks to each word and calculate the score of the whole sequence, wherein the final result is the labeling sequence with the highest score;
and 2.4.4, comparing the output sequence with the actual label sequence, calculating network loss, and updating and optimizing Self-orientation, BiGRU and CNN according to the network loss so as to enhance the extraction effect of the work order event vector.
Preferably, step 3 comprises:
step 3.1, collecting authority and responsibility list webpage data of all departments of a plurality of cities to the local, and after data cleaning and data extraction are carried out on the webpage, classifying and fusing according to the names of the departments, the names of the authority and responsibility lists and the contents of the authority and responsibility lists to form an authority and responsibility list knowledge base;
and 3.2, processing the authority list of each implementation department L in the authority list knowledge base into a key-value list form:
L＝[(k1,v1),(k2,v2),...,(kn,vn)]
wherein k isiDenotes "department name-title", viThe concrete content of the authority list is represented;
step 3.3, Embedding is carried out on L to obtain a vector L ═ I1,I2,…,In]Wherein, Ii＝(ki,vi)，kiAnd viOther is kiAnd viThe Embellding result of (1);
step 3.4, using the hidden state h of the BiGRU networkiAnd a key-value pair vector L as input, in the context direction of the key-value pairL 'quantity'sAnd L'vIs an output;
step 3.5, mixing L'kAnd L'vAnd splicing, performing dimension transformation by using a fully-connected network, and finally completing the government affair hot line work order distribution task in combination with the event extraction layer.
Preferably, step 4 comprises: output vector V after splicing Self-orientation and CNNoutOutput L 'embedded with authority list in network'kAnd L'vSplicing to obtain VconcatWill VconcatAnd inputting the data into a classifier to finish the allocation of the government affair hotline work order.
According to the technical scheme, the semantic representation vector with the context feature information is obtained from the government affair hotline work order text by constructing the semantic coding layer; constructing an event extraction layer based on a combined neural network to obtain a label sequence corresponding to a work order text, so as to enhance the extraction effect of the work order event vector; constructing a key value pair embedded network based on a knowledge base to further mine the effect of promoting the work order allocation of the authority and liability list of the government department; and combining the enhanced work order event vector with the output vector of the authority list embedding layer to finish the allocation of the government affair hot line work orders.
Additional features and advantages of the invention will be set forth in the detailed description which follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
FIG. 1 is a flow chart of the method for government affair hot line work order allocation based on event extraction and authority list in the invention;
fig. 2 is a model structure diagram of the government affair hot line work order allocation method based on event extraction and authority list in the invention.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings. It should be understood that the detailed description and specific examples, while indicating the present invention, are given by way of illustration and explanation only, not limitation.
Referring to fig. 1 and 2, the present invention provides a method for allocating government affair hotline work orders based on event extraction and authority list, including:
step 3, constructing a authority list embedding layer;
and 4, allocating the work orders.
Specifically, the step 1 comprises:
step 1.1, performing semantic coding on the work order text content by using a RoBERTA pre-training language model, expressing a token sequence into a vector form after the text input is converted into the token sequence, wherein the token sequence consists of token embedding, sentence embedding and position embedding, and sending the token sequence into a coding layer stacked by 12 layers of transform coder modules to extract semantic features;
step 1.2, after being coded by a RoBERTA model, splicing and fusing the coded output of any layer in a 12-layer transform coder to obtain a context embedded matrix, and inputting the context embedded matrix into a subsequent layer to obtain the best effect;
and step 1.3, after the semantic representation vector with the context feature information is obtained, inputting the fused code into an event extraction layer.
The step 2 comprises the following steps:
step 2.1, extracting word level characteristics through CNN; further, step 2.1 specifically includes:
step 2.1.1, using the work order text encoding vector output by Roberta as an input sequence of a CNN network, extracting corresponding local features, and outputting semantic features of the whole sentence text in the form of feature vectors;
step 2.1.2, after the convolution operation is completed, sending each obtained feature vector to a pooling layer to generate potential local features, pooling the output result of the convolution layer by adopting a maximum pooling strategy, capturing the most important features after convolution, and processing a variable-length sentence into a fixed length;
step 2.1.3, scanning each word in the sentence by using a convolution kernel to obtain a feature vector of the whole sentence;
step 2.1.4, acquiring text features on different scales through a convolution set, repeating the convolution process, and splicing results of convolution kernels to obtain a feature vector of a whole government affair service hotline text sentence;
step 2.2, obtaining text context semantic information by using a BiGRU network to obtain sentence level characteristics; further, step 2.2 specifically includes:
step 2.2.1, respectively inputting elements in the semantic representation vector L into a forward GRU and a reverse GRU to obtain a forward hidden state
And reverse hidden state
Step 2.2.2, splicing the forward and reverse hidden states to obtain the hidden state of the BiGRU
Step 2.3, splicing sentence level characteristics and word level characteristics after applying Self-Attention; further, step 2.3 specifically includes:
step 2.3.1, encoding each input part to form semantic code Iembedding；
Step 2.3.2, establishing a parameter matrix WQ、WKAnd WVEncoding semantics into IembeddingLinear mapping is carried out in a feature space to form Q, K, V three vectors;
step 2.3.3, similarity calculation is carried out on Q and K to obtain a calculated attention weight, normalization operation is carried out on the obtained weight, and finally the weight and Value are subjected to weighted summation to obtain a final attention score Vout；
Step 2.4, optimizing an event extraction network by using the conditional random field; further, step 2.4 specifically includes:
2.4.1, selecting a BIO label system to label the sequence, wherein all words marked by event locations LOC and event trigger words TRG in the government affair hot line work order text data set are marked by words with event locations LOC and event trigger words TRG, the first word in the words is marked by B-LOC and B-TRG again, the rest words in the words are marked by I-LOC and I-TRG again, more than two continuous numbers are marked by combination, and other words are marked by O uniformly;
step 2.4.2, inputting the splicing vector of Self-orientation and CNN into CRF after passing through a full-connection network to obtain a label sequence corresponding to the work order text;
step 2.4.3: using CRF to assign marks to each word and calculate the score of the whole sequence, wherein the final result is the labeling sequence with the highest score;
and 2.4.4, comparing the output sequence with the actual label sequence, calculating network loss, and updating and optimizing Self-orientation, BiGRU and CNN according to the network loss so as to enhance the extraction effect of the work order event vector.
The step 3 comprises the following steps:
step 3.1, collecting authority and responsibility list webpage data of all departments of a plurality of cities to the local, and after data cleaning and data extraction are carried out on the webpage, classifying and fusing according to the names of the departments, the names of the authority and responsibility lists and the contents of the authority and responsibility lists to form an authority and responsibility list knowledge base;
and 3.2, processing the authority list of each implementation department L in the authority list knowledge base into a key-value list form:
L＝[(k1,v1),(k2,v2),...,(kn,vn)]
wherein k isiDenotes "department name-title", viThe concrete content of the authority list is represented;
step 3.3, Embedding is carried out on L to obtain a vector L ═ I1,I2,…,In]Wherein, Ii＝(ki,vi)，kiAnd viOther is kiAnd viThe Embellding result of (1);
step 3.4, using the hidden state h of the BiGRU networkiSum-key-value pair vectorL is input, and is a context vector L 'of key-value pairs'sAnd L'vIs an output;
step 3.5, mixing L'kAnd L'vAnd splicing, performing dimension transformation by using a fully-connected network, and finally completing the government affair hot line work order distribution task in combination with the event extraction layer.
Step 4 comprises the following steps: output vector V after splicing Self-orientation and CNNoutOutput L 'embedded with authority list in network'kAnd L'vSplicing to obtain VconcatWill VconcatAnd inputting the data into a classifier to finish the allocation of the government affair hotline work order.
By the technical scheme, the method can efficiently assist government affair hotline operators to automatically complete work order allocation, and overcomes the defects that manual allocation by operators is time-consuming and labor-consuming. The technology of the invention is utilized to carry out text analysis processing on the government hot line work order, and the time, the event, the address and the key information of the user complaint in the original data are extracted. Meanwhile, core keywords are extracted from government hot line data by the technology, and the topic model training, the space-time model training and the data optimization are carried out on the basis of a machine learning algorithm, so that hot events and hot areas can be quickly positioned, and reasonable suggestion suggestions are provided for government services.
The preferred embodiments of the present invention have been described in detail with reference to the accompanying drawings, however, the present invention is not limited to the specific details of the above embodiments, and various simple modifications can be made to the technical solution of the present invention within the technical idea of the present invention, and these simple modifications are within the protective scope of the present invention.
It should be noted that the various technical features described in the above embodiments can be combined in any suitable manner without contradiction, and the invention is not described in any way for the possible combinations in order to avoid unnecessary repetition.
In addition, any combination of the various embodiments of the present invention is also possible, and the same should be considered as the disclosure of the present invention as long as it does not depart from the spirit of the present invention.
Claims (9)
1. A government affair hotline work order allocation method based on event extraction and authority list is characterized by comprising the following steps:
step 1, constructing a semantic coding layer;
step 2, constructing an event extraction layer;
step 3, constructing a authority list embedding layer;
and 4, allocating the work orders.
2. The method for government affair hotline work order allocation based on event extraction and authority list according to claim 1, wherein the step 1 comprises:
step 1.1, performing semantic coding on the work order text content by using a RoBERTA pre-training language model, expressing a token sequence into a vector form after the text input is converted into the token sequence, wherein the token sequence consists of token embedding, sentence embedding and position embedding, and sending the token sequence into a coding layer stacked by 12 layers of transform coder modules to extract semantic features;
step 1.2, after being coded by a RoBERTA model, splicing and fusing the coded output of any layer in a 12-layer transform coder to obtain a context embedded matrix, and inputting the context embedded matrix into a subsequent layer to obtain the best effect;
and step 1.3, after the semantic representation vector with the context feature information is obtained, inputting the fused code into an event extraction layer.
3. The method for government affair hotline work order allocation based on event extraction and authority list according to claim 1, wherein the step 2 comprises:
step 2.1, extracting word level characteristics through CNN;
step 2.2, obtaining text context semantic information by using a BiGRU network to obtain sentence level characteristics;
step 2.3, splicing sentence level characteristics and word level characteristics after applying Self-Attention;
and 2.4, optimizing the event extraction network by using the conditional random field.
4. The event extraction and authority list based government affair hotline work order allocation method according to claim 3, wherein the step 2.1 comprises:
step 2.1.1, using the work order text encoding vector output by Roberta as an input sequence of a CNN network, extracting corresponding local features, and outputting semantic features of the whole sentence text in the form of feature vectors;
step 2.1.2, after the convolution operation is completed, sending each obtained feature vector to a pooling layer to generate potential local features, pooling the output result of the convolution layer by adopting a maximum pooling strategy, capturing the most important features after convolution, and processing a variable-length sentence into a fixed length;
step 2.1.3, scanning each word in the sentence by using a convolution kernel to obtain a feature vector of the whole sentence;
and 2.1.4, acquiring text features on different scales through a convolution set, repeating the convolution process, and splicing results of convolution kernels to obtain a feature vector of the whole government affair service hotline text sentence.
5. The event extraction and authority list based government affair hotline work order allocation method according to claim 3, wherein the step 2.2 comprises:
step 2.2.1, respectively inputting elements in the semantic representation vector L into a forward GRU and a reverse GRU to obtain a forward hidden state
And reverse hidden state
Step 2.2.2, splicing the forward and reverse hidden states to obtain the hidden state of the BiGRU
6. The event extraction and authority list based government affair hotline work order allocation method according to claim 3, wherein the step 2.3 comprises:
step 2.3.1, encoding each input part to form semantic code Iembedding；
Step 2.3.2, establishing a parameter matrix WQ、WKAnd WVEncoding semantics into IembeddingLinear mapping is carried out in a feature space to form Q, K, V three vectors;
step 2.3.3, similarity calculation is carried out on Q and K to obtain a calculated attention weight, normalization operation is carried out on the obtained weight, and finally the weight and Value are subjected to weighted summation to obtain a final attention score Vout。
7. The event extraction and authority list based government affair hotline work order allocation method according to claim 3, wherein the step 2.4 comprises:
2.4.1, selecting a BIO label system to label the sequence, wherein all words marked by event locations LOC and event trigger words TRG in the government affair hot line work order text data set are marked by words with event locations LOC and event trigger words TRG, the first word in the words is marked by B-LOC and B-TRG again, the rest words in the words are marked by I-LOC and I-TRG again, more than two continuous numbers are marked by combination, and other words are marked by O uniformly;
step 2.4.2, inputting the splicing vector of Self-orientation and CNN into CRF after passing through a full-connection network to obtain a label sequence corresponding to the work order text;
step 2.4.3: using CRF to assign marks to each word and calculate the score of the whole sequence, wherein the final result is the labeling sequence with the highest score;
and 2.4.4, comparing the output sequence with the actual label sequence, calculating network loss, and updating and optimizing Self-orientation, BiGRU and CNN according to the network loss so as to enhance the extraction effect of the work order event vector.
8. The method for government affair hotline work order allocation based on event extraction and authority list according to claim 1, wherein the step 3 comprises:
step 3.1, collecting authority and responsibility list webpage data of all departments of a plurality of cities to the local, and after data cleaning and data extraction are carried out on the webpage, classifying and fusing according to the names of the departments, the names of the authority and responsibility lists and the contents of the authority and responsibility lists to form an authority and responsibility list knowledge base;
and 3.2, processing the authority list of each implementation department L in the authority list knowledge base into a key-value list form:
L＝[(k1,v1),(k2,v2),...,(kn,vn)]
wherein k isiDenotes "department name-title", viThe concrete content of the authority list is represented;
step 3.3, Embedding is carried out on L to obtain a vector L ═ I1,I2,…,In]Wherein, Ii＝(ki,vi)，kiAnd viOther is kiAnd viThe Embellding result of (1);
step 3.4, using the hidden state h of the BiGRU networkiAnd a key-value pair vector L ' as input, with a context vector L ' of key-value pairs 'sAnd L'vIs an output;
step 3.5, mixing L'kAnd L'vAnd splicing, performing dimension transformation by using a fully-connected network, and finally completing the government affair hot line work order distribution task in combination with the event extraction layer.
9. The method for government affair hotline work order allocation based on event extraction and authority list according to claim 1, wherein the step 4 comprises: output vector V after splicing Self-orientation and CNNoutOutput L 'embedded with authority list in network'kAnd L'vSplicing to obtain VconcatWill VconcatAnd inputting the data into a classifier to finish the allocation of the government affair hotline work order.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111025600.XA CN113657091A (en) | 2021-09-02 | 2021-09-02 | Government affair hot line work order allocation method based on event extraction and authority list |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111025600.XA CN113657091A (en) | 2021-09-02 | 2021-09-02 | Government affair hot line work order allocation method based on event extraction and authority list |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113657091A true CN113657091A (en) | 2021-11-16 |
Family
ID=78493485
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111025600.XA Withdrawn CN113657091A (en) | 2021-09-02 | 2021-09-02 | Government affair hot line work order allocation method based on event extraction and authority list |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113657091A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116108169A (en) * | 2022-12-12 | 2023-05-12 | 长三角信息智能创新研究院 | Hot wire work order intelligent dispatching method based on knowledge graph |
-
2021
- 2021-09-02 CN CN202111025600.XA patent/CN113657091A/en not_active Withdrawn
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116108169A (en) * | 2022-12-12 | 2023-05-12 | 长三角信息智能创新研究院 | Hot wire work order intelligent dispatching method based on knowledge graph |
| CN116108169B (en) * | 2022-12-12 | 2024-02-20 | 长三角信息智能创新研究院 | Hot wire work order intelligent dispatching method based on knowledge graph |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111694924B (en) | Event extraction method and system | |
| CN111177326B (en) | Key information extraction method and device based on fine labeling text and storage medium | |
| CN109325116B (en) | Urban event automatic classification dispatching method and device based on deep learning | |
| CN111324717B (en) | Open domain question-answering-oriented answer acquisition method and system based on machine reading understanding | |
| CN112581106B (en) | Government affair event automatic order dispatching method fusing grid semantics of handling organization | |
| CN106991374A (en) | Handwritten Digit Recognition method based on convolutional neural networks and random forest | |
| CN109885824A (en) | A kind of Chinese name entity recognition method, device and the readable storage medium storing program for executing of level | |
| CN110175857B (en) | Method and device for determining optimal service | |
| CN116108169B (en) | Hot wire work order intelligent dispatching method based on knowledge graph | |
| CN109919175A (en) | A kind of more classification methods of entity of combination attribute information | |
| CN115935245B (en) | Automatic classification and allocation method for government affair hot line cases | |
| CN115145551A (en) | Intelligent auxiliary system for machine learning application low-code development | |
| CN116737922A (en) | Tourist online comment fine granularity emotion analysis method and system | |
| CN115905538A (en) | Event multi-label classification method, device, equipment and medium based on knowledge graph | |
| CN113657091A (en) | Government affair hot line work order allocation method based on event extraction and authority list | |
| CN115965789A (en) | Scene perception attention-based remote sensing image semantic segmentation method | |
| CN111815485A (en) | Sentencing prediction method and device based on deep learning BERT model | |
| CN117151222B (en) | Domain knowledge guided emergency case entity attribute and relation extraction method thereof, electronic equipment and storage medium | |
| CN113920379A (en) | Zero sample image classification method based on knowledge assistance | |
| CN112632406B (en) | Query method, query device, electronic equipment and storage medium | |
| CN115797795B (en) | Remote sensing image question-answer type retrieval system and method based on reinforcement learning | |
| CN115658942A (en) | Financial scene-oriented joint credit investigation intelligent data retrieval method | |
| CN114969511A (en) | Content recommendation method, device and medium based on fragments | |
| CN113901781A (en) | Similar case matching method for fusing segmented coding and affine mechanism | |
| CN113076425A (en) | Event related viewpoint sentence classification method for microblog comments |
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 | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20211116 |
|
| WW01 | Invention patent application withdrawn after publication |