CN109684473A - A kind of automatic bulletin generation method and system - Google Patents
A kind of automatic bulletin generation method and system Download PDFInfo
- Publication number
- CN109684473A CN109684473A CN201811616494.0A CN201811616494A CN109684473A CN 109684473 A CN109684473 A CN 109684473A CN 201811616494 A CN201811616494 A CN 201811616494A CN 109684473 A CN109684473 A CN 109684473A
- Authority
- CN
- China
- Prior art keywords
- bulletin
- corpus
- feature
- template
- data library
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/10—Text processing
- G06F40/166—Editing, e.g. inserting or deleting
- G06F40/186—Templates
Abstract
It include: the data that obtain information present invention relates particularly to a kind of automatic bulletin generation method and system, method；Feature extraction is carried out to the information data of acquisition；The corpus in matching knowledge corpus data library and logic corpus data library is removed using the feature of extraction；The corpus transferring default dynamic template, and will match to is added in default dynamic template, generates the bulletin template of sentence level；It organizes the bulletin template of generation to form bulletin.Compared to the prior art, the feature that the application passes through extraction structural data, according to knowledge corpus data library and logic corpus data library dynamic generation template, bulletin is generated according to template filling data, it can reduce the workload for manually writing bulletin, it can be improved and write efficiency, guarantee the timeliness of bulletin.
Description
Technical field
The present invention relates to information process field more particularly to a kind of automatic bulletin generation methods.
Background technique
Bulletin is the brief information for transmitting certain aspect information, also known as " news in brief ", in some fields, such as finance neck
Domain, the data faced are many and diverse, and very high to timeliness requirement, lead to the intricate operation of bulletin author, pressure is big.Therefore,
Lack a kind of method and system that can automatically generate bulletin in the prior art.
Summary of the invention
The purpose of the present invention is to provide the method and system that one kind can automatically generate bulletin according to information data, with solution
The certainly big technical problem of bulletin author intricate operation, pressure.
To achieve the above object, the present invention provides a kind of automatic bulletin generation method, comprising:
Obtain information data；
Feature extraction is carried out to the information data of acquisition；
The corpus in matching knowledge corpus data library and logic corpus data library is removed using the feature of extraction；Wherein, described
Knowledge corpus data library stores the corpus for the relevant knowledge that bulletin generates, and the storage of logic corpus data library carries out logical description
Corpus；
The corpus transferring default dynamic template, and will match to is added in default dynamic template, generates sentence level
Bulletin template；
The bulletin template of generation is organized, and combines the information data obtained, forms bulletin.
In the above-mentioned technical solutions, further, feature extraction is carried out to the information data of acquisition to specifically include:
Using natural language processing feature extraction algorithm, word frequency analysis or semantic analysis are carried out to information data,
To extract feature；
Alternatively, extracting feature using default feature extraction algorithm.
In the above-mentioned technical solutions, further, feature is extracted using natural language processing feature extraction algorithm, it is specific to wrap
It includes:
Information data are split into several document feature sets；
Every document feature sets are assessed using valuation functions, obtain the weight of every document feature sets；
Document feature sets and its weight, the feature as extracted.
In the above-mentioned technical solutions, further, every document feature sets are assessed using valuation functions, obtains every
The weight of document feature sets, refers to: the weight of every document feature sets is indicated using word frequency and the product of inverse document frequency；
Wherein,
In the above-mentioned technical solutions, further, matching knowledge corpus data library and logic language are gone to using the feature of extraction
The content of text for expecting lane database, refers to:
Using document feature sets and its weight by content of text vectorization；
According to cosine similarity algorithm, immediate language is matched from knowledge corpus data library and logic corpus data library
Material.
In the above-mentioned technical solutions, further, default dynamic template is transferred, comprising:
Labeling processing is carried out to all dynamic templates, extracts associated dynamic template using specific query statement, and
By query result with matching degree sequence in case subsequent replacement demand.
In the above-mentioned technical solutions, further, before carrying out feature extraction to the information data of acquisition, to information data
It is pre-processed；The pretreatment includes: to filter out participle and stop words.
The application also provides a kind of automatic bulletin generation system, comprising:
Information data acquisition module, the information data acquisition module is for the data that obtain information；
Data processing module, the data processing module are used to analyze the information data of acquisition to extract the spy of information data
Sign；
Knowledge corpus data library, knowledge corpus data library are used to store the corpus of the relevant knowledge of bulletin generation；
Logic corpus data library, logic corpus data library are used to store the related corpus for carrying out logical description；
Logic processing module, the logic processing module be used for according to the feature of extraction go matching knowledge corpus data library and
Corpus in logic corpus data library；
Template generation module, the template generation module is for transferring default dynamic template, and the corpus that will match to adds
It adds in default dynamic template, generates the bulletin template of sentence level；
Bulletin generation module organizes the bulletin template of generation to form bulletin.
In the above-mentioned technical solutions, further, the bulletin generation module includes:
Coupling analysis submodule, the coupling analysis submodule are used to analyze the degree of coupling between the content of bulletin template；
Framework tissue submodule, the framework tissue submodule is used to select the Rational of demonstration briefing template, with shape
At bulletin.
In the above-mentioned technical solutions, further, the bulletin generation module further include:
Correlation analysis and back forecasting submodule, the correlation analysis and back forecasting submodule for pair: framework
The bulletin for organizing submodule to generate carries out correlation analysis, to guarantee the accuracy of bulletin；
Risk control submodule, the risk control submodule are used to carry out compliance inspection to the bulletin of generation.
Compared to the prior art, automatic bulletin provided by the present application generates system, by extracting the feature of structural data,
According to knowledge corpus data library and logic corpus data library dynamic generation template, bulletin is generated according to template filling data, it can
Mitigate the workload for manually writing bulletin, can be improved and write efficiency, guarantee the timeliness of bulletin.
Detailed description of the invention
It, below will be to specific in order to illustrate more clearly of the application specific embodiment or technical solution in the prior art
Embodiment or attached drawing needed to be used in the description of the prior art be briefly described, it should be apparent that, it is described below
Attached drawing is some embodiments of the application, for those of ordinary skill in the art, before not making the creative labor
It puts, is also possible to obtain other drawings based on these drawings.
Fig. 1 is the schematic diagram that system is generated according to the automatic bulletin that some embodiments of the present application provide；
Fig. 2 is the schematic diagram for the bulletin generation module that automatic bulletin shown in FIG. 1 generates system；
Fig. 3 is the exemplary process diagram of the automatic bulletin generation method provided according to some embodiments of the present application.
Specific embodiment
It is described as the application defined in requirement and its equivalent that has the right convenient for Integrated Understanding below with reference to attached drawing
Various embodiments.These embodiments include various specific details in order to understand, but these are considered only as illustratively.Cause
This, it will be appreciated by those skilled in the art that carrying out variations and modifications without departing from this to various embodiments described herein
The scope and spirit of application.In addition, briefly and to be explicitly described the application, the application will be omitted to known function and structure
Description.
The term used in following description and claims and phrase are not limited to literal meaning, and being merely can
Understand and consistently understands the application.Therefore, for those skilled in the art, it will be understood that provide to the various implementations of the application
The description of example is only the purpose to illustrate, rather than limits the application of appended claims and its Equivalent definitions.
Below in conjunction with the attached drawing in some embodiments of the application, technical solutions in the embodiments of the present application carries out clear
Chu is fully described by, it is clear that described embodiments are only a part of embodiments of the present application, instead of all the embodiments.
Based on the embodiment in the application, obtained by those of ordinary skill in the art without making creative efforts all
Other embodiments shall fall in the protection scope of this application.
It should be noted that the term used in the embodiment of the present application is the mesh for being only in description specific embodiment
, it is not intended to be limiting the application." one " of the embodiment of the present application and singular used in the attached claims,
"one", "an", " described " and "the" be also intended to including most forms, unless the context clearly indicates other meaning.Also
It should be appreciated that term "and/or" used herein refer to and include one or more mutually bindings list any of project or
All possible combinations.Express " first ", " second ", " first " and " second " is for modifying respective element without examining
Worry sequence or importance are used only for distinguishing a kind of element and another element, without limiting respective element.
It is illustrated below that concrete example is generated as with financial and economic news bulletin, but automatic bulletin generates system and method not
It is limited to financial and economic news, can is any field and the bulletin of form.
Fig. 1 is the schematic diagram that system is generated according to the automatic bulletin that some embodiments of the present application provide.
As shown in Figure 1, automatic bulletin provided by the present application generates system, including information data acquisition module 110, data
Processing module 120, knowledge corpus data library 130, logic corpus data library 140, logic processing module 150, template generation module
160 and bulletin generation module 170.
Information data acquisition module 110
For information data acquisition module 110 for the data that obtain information, advisory data is the Data entries that bulletin generates system,
In some embodiments, data source can be the big data laboratory structuring number collected of user (such as certain bank)
According to.News data acquisition module can read data according to the requirement of client from database or data server, then will count
According to transferring to bulletin to generate system, bulletin is automatically generated.
Data processing module 120
Data processing module 120 mainly does signature analysis to data and carries out feature extraction.The analysis and extraction point of feature
It is two kinds, one is using, natural language processing carries out word frequency analysis to data or semantic analysis, another kind are default
Feature extraction.
The first feature extracting method, it is different according to the usage scenario of data, different feature analysis als can be customized.With
Under divide three levels, to introduce natural language processing feature extraction algorithm:
One, text vector
Classical vector space model (VSM:Vector Space Model) is applied successfully to famous SMART text
This searching system.VSM concept is simple, and the processing to content of text is reduced to the vector operation in vector space, and it with
The semantic similarity of similarity expression spatially, it is intuitive and easy to understand.When document is represented as the vector of document space, so that it may logical
It crosses the similitude calculated between vector and measures the similitude between document.Most common similarity measurement mode is in text-processing
COS distance.Text Mining System uses vector space model, represents mesh with document feature sets (T1, T2 ... Tn) and its weight Wi
Information is marked, when carrying out information matches, uses the degree of correlation of these characteristic items evaluation unknown text and target sample.
The selection of document feature sets and its weight is known as the feature extraction of target sample, and the superiority and inferiority of feature extraction algorithm will be direct
Influence the operational effect of system.
Feature extraction algorithm of two, based on statistics
This type algorithm is assessed each feature in characteristic set, and by construction valuation functions to each spy
Sign marking, word each in this way obtain an assessed value, also known as weight.Then all features are sorted by weight size,
The optimal characteristics of predetermined number are extracted as the character subset for extracting result.Clearly for this type algorithm, text feature is determined
The principal element of extraction effect is the quality of valuation functions.
Three .TF-IDF
The maximally efficient implementation method of word weight is exactly TF-IDF.Wherein TF is known as word frequency, for calculating word description
The ability of document content, IDF are known as inverse document frequency, the ability for distinguishing document for calculating the word.The guiding theory of TF-IDF
Establish on such basic assumption: occur word many times in a text, another in class text go out
Occurrence number also can be very much, and vice versa.So estimating if feature space coordinate system takes TF word frequency to be used as, so that it may embody similar
The characteristics of text.In addition it is also contemplated that word distinguishes different classes of ability, TF-IDF method thinks the text that a word occurs
Frequency is smaller, it is distinguished, and different classes of ability is bigger, so the concept of inverse text frequency IDF is introduced, with TF and IDF
Product estimate as the value of feature space coordinate system.The weighted value that Feature Words are calculated with TF-IDF algorithm is to indicate to work as
The frequency that one word occurs in this document is higher, while the number occurred in other documents is fewer, then shows the word pair
It is stronger in the separating capacity for indicating this document, so its weighted value just should be bigger.
Specifically:
TF-IDF=TF × IDF.
Use case:
After carrying out TF-IDF feature extraction to text data, using in its TF-IDF feature and knowledge corpus data library 130
Data carry out similarity mode, and most similar knowledge data should be just in the case where some other restrictive condition all meets
The maximally related background knowledge data of this article notebook data.
Second of default feature extraction refers to the matching that according to specific logic data are carried out with Feature Words, to will count
It is associated according to specific knowledge corpus and logic corpus.After feature extraction, two kinds of algorithms require will be important in data,
Significant data are extracted and are saved, and are supplied to the use of bulletin generation module 170.
In some embodiments, system is classified and being handled data using default feature using decision-tree model.
Decision tree is the model established by some attributes of text.Here attribute refers to such as comprising some specific
Keyword.Then decision tree uses top-down mode, and each node in tree represents an attribute, and every of the node
Subtree all represents a determining value of this attribute, puts decision value on leaf node.Objective attribute target attribute is sentenced with decision tree
When disconnected, subtree will be entered along known attribute from tree root, until finding leaf node, the result also just judged.
So the path from root node to each leaf node is a decision rule.Here judging result is referred to textual data
According to successful characteristic matching has been carried out, the generation template for being suitble to expression text data is had found, in conjunction with next modules
Carry out the generation of article.
Use case:
Such as when system is using default feature " limit-up ", " opening quotation ", etc. features determine that text data is to express one
The news of stock market's limit-up, to carry out the integration and generation of article using relevant data template.
In above two default feature extracting method, it is required to pre-process text data.The main packet of pretreatment
Containing two parts, participle and stop words filtering.
For participle, very mature participle dictionary has been had accumulated, accurate word segmentation can have been carried out to article.
Stop words refers to that those cannot reflect the function word of theme, they cannot not only reflect the theme of document, but also
The extraction of keyword can be interfered, it is necessary to be filtered out.Such as: " ", " ground ", " obtaining " etc auxiliary word, it is necessary to
It is filtered out.Stop words is determined as all function words and punctuation mark under normal circumstances.
Logic processing module 150
Logic processing module 150 mainly goes matching 130 He of knowledge corpus data library using specific algorithm with information feature
Content of text in logic corpus data library 140.Such as when needing to generate article background, logic processing module 150 can root
According to the feature of input information, similarity mode is carried out in knowledge corpus data library 130, maximally related background information is found, returns
Back to logic processing module 150 (will be described in detail in next module knowledge corpus data library 130 this search with it is matched
Process).For another example, the corpus data of input needs the relationship between Expressive Features, and such as " limit-up ", " limit down ", " rising sharply " is " big
Fall " when, logic processing module 150 can match in logic corpus data library 140 according to the feature and demand of input corpus
Maximally related logic corpus.
By access knowledge corpus data library 130 and logic corpus data library 140, logic processing module 150 has obtained phase
The corpus data of pass transfers to template generation module to carry out template generation.
Knowledge corpus data library 130
The relevant background knowledge that mainly storage bulletin generates of knowledge corpus data library 130, and pass through and input corpus
It compares to find most similar corpus data.Specifically, it is necessary first to the corpus of input be segmented and be calculated word frequency, so
An article is indicated using vector according to word frequency afterwards, the article and knowledge corpus number of input can be thus indicated using vector
According to all articles in library 130.
So, this problem that " finds maximally related background information ", be changed into find how in the database one with it is defeated
Most classic cosine similarity algorithm can be used again in the maximally related vector of incoming vector.More specifically, can be input
A certain data in corpus and knowledge corpus data library 130 is imagined as the two lines section in hyperspace, is all from origin
([0,0 ...]) it sets out, it is directed toward different directions.Between two lines section shape in an angle, if angle is 0 degree, it is meant that
Direction is identical, line segment is overlapped；If angle is 90 degree, it is meant that form right angle, direction is completely dissimilar；If angle is 180
Degree, it is meant that direction is exactly the opposite.It therefore, can be by the size of angle, to judge the similarity degree of vector.Angle is smaller,
It is more similar with regard to representing.
By taking two-dimensional space as an example, a (x1, y1) and b (x2, y2) it is two vectors in coordinate system, it is fixed according to cosine
Reason, can acquire their angle theta with following formula:
This calculation method of cosine also sets up n-dimensional vector.It is assumed that A and B are two n-dimensional vectors, A be [A1,
A2 ..., An], B is [B1, B2 ..., Bn], then the cosine of the angle theta of A and B is equal to:
Wherein cosine value is closer to 1, indicates that angle closer to 0 degree, that is, two vectors are more similar, i.e. " cosine phase
Like property ".The most similar article in knowledge corpus data library 130 therewith can be found according to the corpus of input as a result,.
Logic corpus data library 140
Logic corpus data library 140 mainly stores the related corpus for carrying out logical description, is described during generating bulletin
The important quarry of property phrase and clause.Although logic processing module 150 is to match corpus according to the feature of trigger event
And random call, but synonymous and different attribute description can be labeled and segment in the database.Such as: " pull out big sun
Line ", " blowout rises sharply ", " skip limit-up " although etc. can be employed as describe stock excellent situation, " pulling out big land "
The presentation that share price sharp rises is only described, without " blowout rises sharply " to the metaphor of trading volume, " skip limit-up " is then even more
The difference of rise form is presented.Corpus can also be enriched constantly in actual use by artificial and machine learning.
Wherein, manual intervention is that more descriptive materials are manually entered into database；Machine learning is then by algorithm from big
Description that is similar to a certain event but being not quite similar is found out in amount news report, is then marked and is put in storage.Here it is used to
The algorithm for extracting corpus material is similar to the aforementioned feature extraction algorithm based on statistics, such as TF-IDF and decision-tree model.
Template generation module
Template generation module 160 is responsible for transferring default dynamic template and adds matching corpus, and bulletin template, this template are generated
Generation granularity be sentence level.This module can obviously be divided into two word steps:
The inquiry of dynamic template: dynamic template is defined herein as the sentence of a fixed clause, (i.e. including template part
Known clause part) and dynamic fill part (i.e. unknown message part).
Such as: " X1 " stock code " X2 " " X4 " " X5 " in " X3 "；This template can be used for describing single stock specific
Market summary info in period.
Meanwhile in order to improve the dynamic and feasibility of module generation, it is dynamic that a template can be created multiple sons by succession
Morphotype version, such as:
" X1 " stock code " X2 " slumps " X5 " in morning quotation
" X1 " stock code " X2 " rises slightly " X5 " in morning quotation
In order to improve the accuracy of template generation, this module carries out labeling processing to all templates, and using specific
Query statement high efficiency extraction associated dynamic template, and query result is sorted with matching degree in case of subsequent replacement demand.Language
Material matching addition: due to that may include a large amount of unknown messages in dynamic template, corpus matching adding module needs first will dynamic
Unknown message in template is matched with data type and description information, for example, the matching of time type matching, degree adverb,
Value type matching etc..Fills can be carried out to it after unknown domain is matched, need simultaneously during filling internal
Appearance is handled, including numerical precision adjustment (such as 123,450,000 yuan be adjusted to 1.23 hundred million yuan), numerical value it is regular (such as
5.0002% be adjusted to 5%), numerical value replacement (such as go up 10% be adjusted to limit-up) etc..
By two above step, system produces the template comprising a certain the details complete information, in order to guarantee this
The accuracy of template final news in brief, also needs the template to all generations to be reconstructed and organize, bulletin generation module 170 is
It can reach this purpose.
Bulletin generation module 170
Bulletin generation module 170 is responsible for that previous step template integrative organization generated gets up to form the money finally shown
News, and the information of generation is analyzed and verified, guarantee that this information is qualified and closes rule.
The specific structure and function of bulletin generation module 170 are refering to Fig. 2.
Fig. 2 is the schematic diagram for the bulletin generation module that automatic bulletin shown in FIG. 1 generates system.
As shown in Fig. 2, bulletin generation module includes coupling analysis submodule 210, framework tissue submodule 220, correlation
Analysis and back forecasting submodule 230 and risk control submodule 240.
Coupling analysis submodule 210: a module (i.e. template generation module) exports template content on this module Main Analysis
Between the degree of coupling.The degree of coupling of template is defined as between different template contents with the presence or absence of dependence and degree of dependence.
The degree of coupling between template can be loose couplings (or for lower coupling), be also possible to close-coupled (or for high coupling
Property).The degree of coupling for analyzing the template generated is conducive to the information structure that more efficient tissue ultimately generates, and makes the money generated
News have higher readability.
Framework tissue submodule 220: this submodule is responsible for the Rational for selecting finally to show, generates including previous step
The tissue of template module be associated with, the segmentation of sentence and paragraph and continuity authentication function.Wherein, according to coupling analysis submodule
The content of 210 output of block, this submodule can establish framework organization chart in memory, and generated by optimization graph traversal algorithm and relatively closed
Suitable institutional framework, this institutional framework be correspond to previous step template result this how to be connected.Meanwhile different mould
It needs to connect by smooth conjunction or divide by modes such as paragraphs between version, with the integrative organization's knot being optimal
Structure.
Correlation analysis and back forecasting submodule 230: it needs to carry out phase by the information that organizational structure submodule generates
The analysis of closing property, that is, the information generated will correspond to relevant initial data to the full extent, to guarantee the accuracy of information.Meanwhile
Backtracking analyze the information corresponding information daughter element can be corresponded into corresponding initial data and retain the correspondence between data,
Sequentially and causality.It is regarded as accurate agonic by the information of this submodule.
Risk control submodule 240: in view of the ambiguity of Chinese text, the text that certain machines generate may include sensitivity
Information, after the different submodules especially generated are readjusted sequence and add conjunction or be divided, these texts can
It can produce ambiguity.The information of generation is carried out compliance inspection by the risk control submodule 240 again, guarantees that system is most lifelong
At module accurately close rule.
Fig. 3 is the exemplary process diagram of the automatic bulletin generation method provided according to some embodiments of the present application.
As shown in figure 3, the application also provides a kind of automatic bulletin generation method, comprising:
S310, obtain information data；
S320 pre-processes information data；
S330 carries out feature extraction to information data；
In some embodiments, using natural language processing feature extraction algorithm, word frequency statistics point are carried out to information data
Analysis or semantic analysis, to extract feature.Specifically, comprising:
Information data are split into several document feature sets；
Every document feature sets are assessed using valuation functions, obtain the weight of every document feature sets；
Document feature sets and its weight, the feature as extracted.
In some embodiments, feature is extracted using default feature extraction algorithm.
S340 removes the corpus in matching knowledge corpus data library and logic corpus data library using the feature of extraction；
The corpus that S350 transfers default dynamic template, and will match to is added in default dynamic template, generates Sentence-level
Other bulletin template；
S360 organizes the bulletin template of generation to form bulletin.
The above is only the preferred embodiment of the present invention, it is noted that those skilled in the art are come
It says, without departing from the technical principles of the invention, several improvements and modifications can also be made, these improvements and modifications are also answered
It is considered as protection scope of the present invention.
Claims (10)
1. a kind of automatic bulletin generation method characterized by comprising
Obtain information data；
Feature extraction is carried out to the information data of acquisition；
The corpus in matching knowledge corpus data library and logic corpus data library is removed using the feature of extraction；Wherein, the knowledge
Corpus data library stores the corpus for the relevant knowledge that bulletin generates, and the storage of logic corpus data library carries out the language of logical description
Material；
The corpus transferring default dynamic template, and will match to is added in default dynamic template, generates the bulletin of sentence level
Template；
It organizes the bulletin template of generation to form bulletin.
2. automatic bulletin generation method according to claim 1, which is characterized in that carry out feature to the information data of acquisition
Extraction specifically includes:
Using natural language processing feature extraction algorithm, word frequency analysis or semantic analysis are carried out to information data, to mention
Take feature；
Alternatively, extracting feature using default feature extraction algorithm.
3. automatic bulletin generation method according to claim 2, which is characterized in that utilize natural language processing feature extraction
Algorithm extracts feature, specifically includes:
Information data are split into several document feature sets；
Every document feature sets are assessed using valuation functions, obtain the weight of every document feature sets；
Document feature sets and its weight, the feature as extracted.
4. automatic bulletin generation method according to claim 3, which is characterized in that using valuation functions to every Feature Words
Item is assessed, and is obtained the weight of every document feature sets, is referred to: every feature is indicated using word frequency and the product of inverse document frequency
The weight of entry；
Wherein,
5. automatic bulletin generation method according to claim 3 or 4, which is characterized in that go to match using the feature of extraction
Content of text in knowledge corpus data library and logic corpus data library, refers to:
Using document feature sets and its weight by content of text vectorization；
According to cosine similarity algorithm, immediate corpus is matched from knowledge corpus data library and logic corpus data library.
6. automatic bulletin generation method according to claim 1, which is characterized in that transfer default dynamic template, comprising:
Labeling processing is carried out to all dynamic templates, extracts associated dynamic template using specific query statement, and will look into
Result is ask with matching degree sequence in case subsequent replacement demand.
7. automatic bulletin generation method according to claim 1, which is characterized in that carry out feature to the information data of acquisition
Before extraction, information data are pre-processed；The pretreatment includes: to filter out participle and stop words.
8. a kind of automatic bulletin generates system characterized by comprising
Information data acquisition module, the information data acquisition module is for the data that obtain information；
Data processing module, the data processing module are used to analyze the information data of acquisition to extract the feature of information data；
Knowledge corpus data library, knowledge corpus data library are used to store the corpus of the relevant knowledge of bulletin generation；
Logic corpus data library, logic corpus data library are used to store the related corpus for carrying out logical description；
Logic processing module, the logic processing module are used to go matching knowledge corpus data library and logic according to the feature of extraction
Corpus in corpus data library；
Template generation module, the template generation module is for transferring default dynamic template, and the corpus that will match to is added to
In default dynamic template, the bulletin template of sentence level is generated；
Bulletin generation module organizes the bulletin template of generation to form bulletin.
9. automatic bulletin according to claim 8 generates system, which is characterized in that the bulletin generation module includes:
Coupling analysis submodule, the coupling analysis submodule are used to analyze the degree of coupling between the content of bulletin template；
Framework tissue submodule, the framework tissue submodule is used to select the Rational of demonstration briefing template, to form letter
Report.
10. automatic bulletin according to claim 9 generates system, which is characterized in that the bulletin generation module further include:
Correlation analysis and back forecasting submodule, the correlation analysis and back forecasting submodule for pair: framework tissue
The bulletin that submodule generates carries out correlation analysis, to guarantee the accuracy of bulletin；
Risk control submodule, the risk control submodule are used to carry out compliance inspection to the bulletin of generation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811616494.0A CN109684473A (en) | 2018-12-28 | 2018-12-28 | A kind of automatic bulletin generation method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811616494.0A CN109684473A (en) | 2018-12-28 | 2018-12-28 | A kind of automatic bulletin generation method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109684473A true CN109684473A (en) | 2019-04-26 |
Family
ID=66190616
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811616494.0A Pending CN109684473A (en) | 2018-12-28 | 2018-12-28 | A kind of automatic bulletin generation method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109684473A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414736A (en) * | 2020-03-23 | 2020-07-14 | 腾讯科技（深圳）有限公司 | Story generation model training method, device, equipment and storage medium |
| CN112289301A (en) * | 2020-11-03 | 2021-01-29 | 苏州思必驰信息科技有限公司 | TTS pre-synthesis method and system |
| CN112765950A (en) * | 2021-01-08 | 2021-05-07 | 首都师范大学 | Template library generation method and system based on cosine similarity and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101667194A (en) * | 2009-09-29 | 2010-03-10 | 北京大学 | Automatic abstracting method and system based on user comment text feature |
| US20110087671A1 (en) * | 2009-10-14 | 2011-04-14 | National Chiao Tung University | Document Processing System and Method Thereof |
| CN103324622A (en) * | 2012-03-21 | 2013-09-25 | 北京百度网讯科技有限公司 | Method and device for automatic generating of front page abstract |
| CN109063147A (en) * | 2018-08-06 | 2018-12-21 | 北京航空航天大学 | Online course forum content recommendation method and system based on text similarity |
-
2018
- 2018-12-28 CN CN201811616494.0A patent/CN109684473A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101667194A (en) * | 2009-09-29 | 2010-03-10 | 北京大学 | Automatic abstracting method and system based on user comment text feature |
| US20110087671A1 (en) * | 2009-10-14 | 2011-04-14 | National Chiao Tung University | Document Processing System and Method Thereof |
| CN103324622A (en) * | 2012-03-21 | 2013-09-25 | 北京百度网讯科技有限公司 | Method and device for automatic generating of front page abstract |
| CN109063147A (en) * | 2018-08-06 | 2018-12-21 | 北京航空航天大学 | Online course forum content recommendation method and system based on text similarity |
Non-Patent Citations (3)
| Title |
|---|
| 仇培元等: ""互联网文本蕴含道路交通信息抽取的模式匹配方法"", 《地球信息科学学报》 * |
| 沈焕生等: ""基于信息抽取的自动摘要生成技术"", 《2009年中国信息技术应用学术研讨会论文集》 * |
| 邹剑章等: ""基于事件框架的移动摘要方法研究"", 《微计算机信息》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111414736A (en) * | 2020-03-23 | 2020-07-14 | 腾讯科技（深圳）有限公司 | Story generation model training method, device, equipment and storage medium |
| CN112289301A (en) * | 2020-11-03 | 2021-01-29 | 苏州思必驰信息科技有限公司 | TTS pre-synthesis method and system |
| CN112765950A (en) * | 2021-01-08 | 2021-05-07 | 首都师范大学 | Template library generation method and system based on cosine similarity and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2022022045A1 (en) | Knowledge graph-based text comparison method and apparatus, device, and storage medium | |
| US9542477B2 (en) | Method of automated discovery of topics relatedness | |
| CN109918506B (en) | Text classification method and device | |
| Sunilkumar et al. | A survey on semantic similarity | |
| US20190392035A1 (en) | Information object extraction using combination of classifiers analyzing local and non-local features | |
| Li et al. | A policy-based process mining framework: mining business policy texts for discovering process models | |
| Lydia et al. | Correlative study and analysis for hidden patterns in text analytics unstructured data using supervised and unsupervised learning techniques | |
| CN109684473A (en) | A kind of automatic bulletin generation method and system | |
| CN111737997A (en) | Text similarity determination method, text similarity determination equipment and storage medium | |
| CN116628229B (en) | Method and device for generating text corpus by using knowledge graph | |
| Hossari et al. | TEST: A terminology extraction system for technology related terms | |
| Storopoli | Topic modeling: how and why to use in management research | |
| Repke et al. | Extraction and representation of financial entities from text | |
| Jeon et al. | Making a graph database from unstructured text | |
| Chen et al. | Using latent Dirichlet allocation to improve text classification performance of support vector machine | |
| Rodriguez et al. | An efficient wikipedia-based approach for better understanding of natural language text related to user requirements | |
| Panthum et al. | Generating functional requirements based on classification of mobile application user reviews | |
| Picco et al. | Zshot: An open-source framework for zero-shot named entity recognition and relation extraction | |
| Duque et al. | A multiview clustering approach for mining authorial affinities in literary texts | |
| Franciscus et al. | Beyond word-cloud: A graph model derived from beliefs | |
| Gomes et al. | A multi-document summarization system for news articles in Portuguese using integer linear programming | |
| US20230325606A1 (en) | Method for extracting information from an unstructured data source | |
| Piryani et al. | An algorithmic formulation for extracting learning concepts and their relatedness in ebook texts | |
| Balahur et al. | A semantic relatedness approach to classifying opinion from Web reviews | |
| Kang et al. | Banking System Incidents Analysis Using Knowledge Graph |
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 | ||
| CB02 | Change of applicant information |
Address after: Room 6013, building 14, 785 Hutai Road, Jing'an District, Shanghai Applicant after: DANHAN INTELLIGENT TECHNOLOGY (SHANGHAI) Co.,Ltd. Address before: Room 118, Building 20, No. 83 Lane 1-42, Hongxiang North Road, Wanxiang Town, Pudong New Area, Shanghai Applicant before: DANHAN INTELLIGENT TECHNOLOGY (SHANGHAI) Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190426 |
|
| WD01 | Invention patent application deemed withdrawn after publication |