CN109933791A - Material recommended method, device, computer equipment and computer readable storage medium - Google Patents
Material recommended method, device, computer equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN109933791A CN109933791A CN201910166860.5A CN201910166860A CN109933791A CN 109933791 A CN109933791 A CN 109933791A CN 201910166860 A CN201910166860 A CN 201910166860A CN 109933791 A CN109933791 A CN 109933791A
- Authority
- CN
- China
- Prior art keywords
- writing
- sentence
- viewpoint
- sample
- type
- 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.)
- Granted
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Machine Translation (AREA)
Abstract
The embodiment of the invention provides a kind of material recommended method, device, computer equipment and computer readable storage mediums, wherein this method comprises: obtaining theme to be checked；The theme to be checked is matched with the viewpoint deposited；According to the corresponding relationship of the viewpoint and writing material deposited, recommend the corresponding writing material of the viewpoint of successful match.The technical solution is conducive to avoid the problem that recall or screen available material from the indefinite resource of viewpoint in the prior art, or avoid needing author to read the process for screening identical of views material one by one in the prior art, be conducive to that user is time saving, efficiently obtains effective writing material.
Description
Technical field
The present invention relates to natural language processing technique field, in particular to a kind of material recommended method, device, computer are set
Standby and computer readable storage medium.
Background technique
Currently, the material of argumentative writing writing is mainly passed through author's accumulation at ordinary times or is searched using keyword in search engine
The mode of rope obtains.First way the problem is that: to obtain effective writing material, not only need long-term accumulation and
And need therefrom to remember useful material, but the amount of reading of people is limited with memory, which has limited the optional of material
Range, while the burden of memory or memory is increased, energy is especially very expended for beginner；Existing for the second way
Problem is: many true resources only carry out objective statement to story on internet, have plenty of subjective sight in author's brains
Point, and it is frequently present of semantic gap between subjective viewpoint and objective statement, for example, a story can illustrate some reason
Or viewpoint, but in the statement of story there is no in the direct description of reason or viewpoint or the statement of story include key
Word, but the reason that illustrates of the story or viewpoint may not be inconsistent with the viewpoint of author, just author be needed to read the fact one by one at this time
Resource, and the potential reason of each true resource, viewpoint, meaning are rule of thumb inferred to common sense, then judge the fact resource
Whether potential reason, viewpoint, meaning are consistent with the viewpoint of author's subjectivity, and then just can determine which true resource for author
Subjective viewpoint is effective, available material, so that author expends energy, reduces the efficiency for obtaining material.
Summary of the invention
The embodiment of the invention provides a kind of material recommended methods, to exist when solving and obtaining writing material in the prior art
Optional material it is limited, expend the technical issues of energy, low efficiency.This method comprises:
Obtain theme to be checked；
The theme to be checked is matched with the viewpoint deposited；
According to the corresponding relationship of the viewpoint and writing material deposited, recommend the corresponding writing material of the viewpoint of successful match.
The embodiment of the invention also provides a kind of material recommendation apparatus, to deposit when solving and obtaining writing material in the prior art
Optional material it is limited, expend the technical issues of energy, low efficiency.The device includes:
Theme obtains module, for obtaining theme to be checked；
Matching module, for matching the theme to be checked with the viewpoint deposited；
Recommending module recommends the viewpoint pair of successful match for the corresponding relationship according to the viewpoint and writing material deposited
The writing material answered.
The embodiment of the invention also provides a kind of computer equipments, including memory, processor and storage are on a memory
And the computer program that can be run on a processor, the processor realize above-mentioned arbitrary element when executing the computer program
Material recommended method, existing optional material is limited when solving to obtain writing material in the prior art, expend energy, low efficiency
Technical problem.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer-readable recording medium storage
There is the computer program for executing above-mentioned arbitrary material recommended method, it is existing when solving to obtain writing material in the prior art
Optional material is limited, expends the technical issues of energy, low efficiency.
In embodiments of the present invention, by matching theme to be checked with the viewpoint deposited, and then according to having deposited
Viewpoint and writing material corresponding relationship, recommend the corresponding writing material of the viewpoint of successful match, realize can be directed to it is to be checked
Subject recommending writing material corresponding with the matched viewpoint of theme to be checked is ask, so that the writing material and theme to be checked recommended
It is corresponding.Be conducive to avoid the problem that recall or screen available material from the indefinite resource of viewpoint in the prior art, or
It avoids needing author to read the process for screening identical of views material one by one in the prior art, it is time saving, efficiently to be conducive to user
Obtain effective writing material.
Detailed description of the invention
The drawings described herein are used to provide a further understanding of the present invention, constitutes part of this application, not
Constitute limitation of the invention.In the accompanying drawings:
Fig. 1 is a kind of flow chart of material recommended method provided in an embodiment of the present invention；
Fig. 2 is a kind of functional block diagram of material recommended method provided in an embodiment of the present invention；
Fig. 3 is a kind of schematic diagram of coding and decoding model provided in an embodiment of the present invention；
Fig. 4 is a kind of structural block diagram of computer equipment provided in an embodiment of the present invention；
Fig. 5 is a kind of structural block diagram of material recommendation apparatus provided in an embodiment of the present invention.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention clearer, right below with reference to embodiment and attached drawing
The present invention is described in further details.Here, exemplary embodiment and its explanation of the invention is used to explain the present invention, but simultaneously
It is not as a limitation of the invention.
In embodiments of the present invention, a kind of material recommended method is provided, as shown in Figure 1, this method comprises:
Step 102: obtaining theme to be checked；
Step 104: the theme to be checked is matched with the viewpoint deposited；
Step 106: according to the corresponding relationship of the viewpoint and writing material deposited, recommending the viewpoint of successful match is corresponding to write
Make material.
Process as shown in Figure 1 is it is found that in embodiments of the present invention, by carrying out theme to be checked with the viewpoint deposited
Matching, and then according to the corresponding relationship of the viewpoint and writing material deposited, recommend the corresponding writing material of the viewpoint of successful match,
Realizing can be for subject recommending to be checked writing material corresponding with the matched viewpoint of theme to be checked, so that recommends writes
It is corresponding with theme to be checked to make material.Be conducive to avoid to recall from the indefinite resource of viewpoint in the prior art or screening can
The problem of with material, or avoid that author is needed to read the process for screening identical of views material one by one in the prior art, favorably
In time saving, efficiently material is effectively write in acquisition.
When it is implemented, above-mentioned material recommended method can read magnanimity material by machinery equipment, and then acquire writing
Material simultaneously determines the corresponding viewpoint of writing material, is equivalent to the amount of reading and memory capability for expanding people, writer only needs to input
Theme to be checked, it can directly acquire a large amount of writing materials corresponding with theme to be checked, available, allow to efficiently,
Accurately recommend effective writing material for writer.Further, it is also possible to targetedly improve the amount of reading of reader and know
Storage level is known, for example, reader can obtain largely writing element corresponding with theme to be checked by inputting theme to be checked
Material, allow reader for a certain theme expand amount of reading, and by viewpoint with writing material be mapped memory, lay in.
When it is implemented, can recommend and the matched sight pair of theme to be checked for theme to be checked to user to realize
The writing material answered, so that the writing material recommended is corresponding with theme to be checked, in the present embodiment, as shown in Fig. 2, pushing away
Need to establish the corresponding relationship of writing material and viewpoint (viewpoint can be theme, meaning or position) before recommending, for example, can be with
Realize that writing material is corresponding with viewpoint by following steps:
Acquisition writing material；
Using the writing material of acquisition as the input of machine learning component, machine learning component output writing material is corresponding
Viewpoint.
When it is implemented, the process of above-mentioned acquisition writing material can use search engine, field website carries out information pumping
It takes, to obtain a large amount of writing material, writing material may include true type writing material and/or theoretical type writing material, thing
Full mold writing material may include the narrative contents such as story, the description of account property, and theoretical type writing material may include famous person
The contents such as well-known saying, ancient poetry, poem.
Specifically, for example, acquisition for true type writing material, it can be used that " celebrity story ", " famous person writes element
The keywords such as material " are scanned in search engine, obtain search result web page；It extracts the network address of webpage and is polymerize, obtained
The domain list of websites of story V-neck V；Page analysis is carried out to field website and text extracts, the content of story is obtained and is write for true type
Make material；The story of all acquisitions is deposited to the acquisition that true type writing material is realized to database.
For the acquisition of theoretical type writing material, keywords such as " famous sayings of famous figures, ancient poetry, poems " can be used and draw in search
It holds up and scans for, obtain search result web page；It extracts the network address of webpage and is polymerize, obtain famous sayings of famous figures, ancient poetry, poem neck
Domain list of websites；Page analysis is carried out to field website and text extracts, obtaining famous sayings of famous figures text, ancient poetry, poem is theory
Type writes material；The theoretical type writing material of all acquisitions is deposited to database, the i.e. acquisition of realization theory type writing material.
It, can also be with when it is implemented, true type writing material and theoretical type writing material be there may be in a database
It is respectively stored in two databases, i.e., establishes true type material database and theoretical type material database respectively.
When it is implemented, after acquisition writing material, it is thus necessary to determine that the corresponding viewpoint of writing material, in order to can be efficiently
Determine the unseen new corresponding viewpoint of writing material in the past, it is true present applicant proposes being come by way of machine learning component
Surely the corresponding viewpoint of writing material, i.e., using the writing material of acquisition as the input of machine learning component, machine learning component is
The corresponding viewpoint of exportable writing material.Specifically, the machine learning component can be arbitrary can learning training nerve net
Network, machine learning component can have different titles according to particular condition in use, for example, in this application, machine learning component
For determining the corresponding viewpoint of writing material, i.e. machine learning component is properly termed as viewpoint annotator；It can also be made according to difference
It is named as model, module, learning tool, learning framework etc. with situation, i.e. machine learning component is substantially that can instruct by study
Experienced component or structure, the application are not specifically limited the title of machine learning component.
When it is implemented, the machine learning component of the viewpoint of writing material can be determined in order to obtain, in the present embodiment,
As shown in Fig. 2, machine learning component can be trained based on following steps:
Sample writing material and corresponding sample viewpoint are obtained, the sample writing material that will acquire and corresponding sample viewpoint
As sample data, inputs machine learning component and be trained.
When it is implemented, existing sample writing element can be directly acquired in order to improve the accuracy of machine learning component
Material and corresponding sample viewpoint as sample data, can also through this embodiment in following steps obtain sample writing element
Material and corresponding sample viewpoint are as sample data, for example, the process for obtaining sample writing material is as follows:
A large amount of data resource is obtained, according to the feature of sentence in data resource, sample writing is extracted from data resource
Material and corresponding sample viewpoint.Specifically, the data resource can be arbitrary in the verbal descriptions such as article, record, comment
Hold, for example, the data resource may include words type, narrate type etc. different types of data resource.
When it is implemented, can be used existing for the other kinds of data resource except the data resource of words type
Method extract sample writing material and corresponding sample viewpoint, for example, manually extract sample writing material and corresponding sample
This viewpoint, the application are not specifically limited.For the data resource of words type, the present embodiment proposes a kind of extraction sample writing
The method of material, for example, can sentence in the data resource by way of analyzing argumentative writing debate structure according to argumentative writing type
Feature, from the data resource of argumentative writing type extract sample write material:
Classify according to argumentative writing debate structure to the sentence in the data resource of argumentative writing type；For example, being divided into introduction
The sentence types such as sentence, topic sentence, sub- argument sentence, argument sentence and concluding sentence specifically can carry out sentence using existing classification method
Subclassification, the application are not specifically limited.The sentence of each type has different expressions respectively, for example, introduction sentence, is introduced related
Material introduces theme；Topic sentence provides article central idea；Never Tongfang is described sub- argument sentence in face of theme；Argument sentence
Support purport and argument；Concluding sentence summarizes full text.
Story is searched in the data resource of argumentative writing type and triggers sentence, by the sentence of story triggering sentence and adjacent specified type
Merge into true this writing of pattern material；
Sentence in the data resource of argumentative writing type is matched with known theoretical type sentence, by the sentence of successful match
Son merges into theoretical this writing of pattern material.
Specifically, it is narrative that story triggering sentence can be characterization during extracting true pattern this writing material
Sentence, for example, it may be the sentence of element is described including time, place, personage generation title, name etc., due in an argumentative writing
Middle argument is for supporting paper viewpoint, in turn it is considered that the meaning of story and the argument of argumentative writing and its phase in argument
It closes, therefore, participle and part-of-speech tagging can be carried out to argument sentence, trigger sentence to search story in argument sentence, find story touching
After sending out sentence, story triggering sentence is merged to obtain narrative content with the sentence of adjacent specified type, as true this writing of pattern element
Material.Complete in order to obtain, effective narrative content, story can be triggered sentence and adjacent specified type, comprising narration element
Sentence merge, for example, story can be triggered sentence and adjacent continuous argument sentence merge into true this writing of pattern material.
Specifically, during extracting theoretical pattern this writing material, by the sentence in the data resource of argumentative writing type
It is matched with known theoretical type sentence (for example, well-known saying, ancient poetry etc.), it is theoretical in the data resource to find argumentative writing type
Type sentence, in order to it is more acurrate, effectively obtain theoretical pattern this writing material, can be by argument sentence in the data resource of argumentative writing type
It is matched with known theoretical type sentence, the sentence of successful match is merged into theoretical this writing of pattern material.
When it is implemented, extracting sample writing material (theoretical this writing of pattern material and this writing of true pattern here
Material is referred to as sample writing material) after, sample writing element can also be obtained by way of analyzing argumentative writing debate structure
The sample viewpoint of material, to obtain the corresponding data of sample writing material and sample viewpoint as sample.For example, in the present embodiment
In, the sentence write according to argumentative writing debate structure to sample in the data resource of the argumentative writing type where material carries out classifying it
Afterwards, the classification results obtained include topic sentence and sub- argument sentence, and classification results can also include introduction sentence, argument sentence and concluding sentence
Etc. types sentence；The amalgamation result of one of any or any combination item in item following in the data resource of argumentative writing type is made
The corresponding sample viewpoint of material: topic, topic sentence and the sub- argument sentence nearest with sample writing material distance is write for sample,
Such a sample writing material can correspond to one or more sample viewpoints, and sample writing material is seen with each sample respectively
Point corresponds to one<sample writing material, and sample viewpoint>to storing, without merging to multiple sample viewpoints, i.e., correspondence is more
The sample writing material of a sample viewpoint may generate multiple<sample writing material, sample viewpoint>right, because of different people pair
The understanding of same material may be different.
Specifically, sub- argument sentence passes through the section sequence of section where sub- argument sentence and sub- argument at a distance from sample writing material
Sentence sentence sort to indicate, for example, sample writing material where section sub- argument sentence than other sections sub- argument sentence distance more
Closely；If section where sample writing material does not have sub- argument sentence, there is sub- argument sentence in other multistages, then the sub- argument sentence ratio in leading portion
Sub- argument sentence in back segment is apart from closer；If there is multiple sub- argument sentences in other sections, son of the sub- argument sentence in front than back
Argument sentence is apart from closer.
For example, extracting true this writing of pattern material by taking true this writing of pattern material as an example and determining sample viewpoint
Process is as follows:
Step 1: the sentence in data resource by way of analyzing argumentative writing debate structure to argumentative writing type divides
The analysis result of class is as follows:
[topic sentence]: the life of people is limited, we should make great efforts to create oneself significant life.
[sub- argument sentence]: the meaning of life is to change the world.
[discussion]: human development is the process of a continuous reforming world.
[argument sentence 1]: Qiao Busi changes electronic world, is dedicated to digital revolution throughout one's life.
[argument sentence 2]: his contribution revolutionizes people's lives mode.
Position name from argument sentence first: " Qiao Busi ", i.e. argument sentence 1 are that story triggers sentence；To the generation in argument sentence
Word carries out reference resolution, and therefore " he " also refers to Qiao Busi.Argument sentence 1 is closed with the argument sentence 2 for continuously including same personage
It and is true pattern this writing material.
Step 2: searching out the nearest sub- argument sentence of distance according to the relationship with true this writing of pattern material distance.
Correspond to " meaning of life is to change the world " in this example.
If the sub- argument sentence nearest with true this writing of pattern material distance is considered as fact pattern this writing material
Sample viewpoint, the sample writing material indicated in available example 1 as shown in table 1 below.When it is implemented, due to topic sentence
It is to summarize full text, topic sentence can also be merged to the viewpoint collectively as sample writing material with apart from nearest sub- argument sentence
(or meaning).
By repeating step 1 and step 2, extensive argumentative writing is handled, fairly large < sample can be obtained and write
Make material, sample viewpoint > right, the example of as shown in table 1 below 2 true this writing of pattern materials and sample viewpoint.
Table 1
When it is implemented, after obtaining sample writing material and corresponding sample viewpoint as sample data, it can be by sample
This writing material and corresponding sample viewpoint input machine learning component are trained, which can be arbitrary
Can learning training neural network, for example, in the present embodiment, machine learning component is with the generation model of a sequence to sequence
For, the specific can be that classical coding and decoding model as shown in Figure 3, the coding layer of coding and decoding model is using circulation mind
Through network, LSTM is may be selected in the basic unit (box shown in Fig. 3 is a basic unit) of Recognition with Recurrent Neural Network
(Long Short-Term Memory, shot and long term memory network), two-way LSTM or stacking LSTM (Stacking LSTM) etc. become
Kind.
The process of specific training coding and decoding model can be, and write material and corresponding sample viewpoint as sample using sample
Notebook data writes material as input, using corresponding sample viewpoint as output, for example, firstly, writing element to sample using sample
Each sentence in material is segmented, and by the word vectors after participle, obtains the term vector of each sentence, and by each sentence
Input of the term vector as coding layer, the output of coding layer is as the semantic expressiveness for inputting sentence.By the semantic expressiveness of sentence
Decoding layer is inputted, decoding layer then generates object statement by word according to the semantic expressiveness of input sentence, which is to input
Sample writes the corresponding sample viewpoint of material.
When it is implemented, can be jointly using true this writing of pattern material and corresponding sample viewpoint, theoretical pattern sheet
It writes material and corresponding sample viewpoint is one machine learning component of sample training, i.e., write material with different types of sample
It comes together to train a machine learning component.
When it is implemented, in order to enable the viewpoint of machine learning component mark is more acurrate, in the present embodiment, for difference
Machine learning component can be respectively trained in the sample writing material of type, for example, with true this writing of pattern material and corresponding
Sample viewpoint is sample data, and input machine learning component is trained, obtains true type machine learning component, be referred to as
True type viewpoint annotator, it is subsequent that true type viewpoint annotator can be used to true type writing material progress viewpoint mark；With
Theoretical this writing of pattern material and corresponding sample viewpoint are sample data, and input machine learning component is trained, is managed
By type machine learning component, it is referred to as theoretical type viewpoint annotator, it is subsequent that theoretical type viewpoint annotator can be used to reason
Viewpoint mark is carried out by type writing material.
When it is implemented, in the present embodiment, sample is write material for the generalization ability of reinforcement machine learning component
In proper noun replace with the character of no concrete meaning, the sample writing material and corresponding sample of proper noun is substituted
Viewpoint is sample data, and input machine learning component is trained.
Specifically, the proper noun in above-mentioned sample writing material refers to specific or unique people or object, for example,
It can be place name, name, building name, country name, unit name, organization name, time etc..It is with true this writing of pattern material
Sample can be write specific name in material and be numbered according to appearance sequence, and replace with the word of no concrete meaning by example
Symbol, such as PER1, PER2 ....
Such as: Qiao Busi changes electronic world, is that PER1 changes electronic world after replacement name.
PER1 or PER2 ... is considered as a new word and vocabulary is added, also can be by training machine learning object
Distributing a vector indicates.
The characteristics of training will make machine learning component more pay attention to event in this way is ignored or weakens proprietary by name etc.
Noun bring is different.Such as: it encounters new story and teaches " search that Li Yanhong changes China ", then being based on " PER1
Change " this mode, the characteristics of machine learning component is based only on event marks viewpoint, it is possible to will the story of " Li Yanhong " and
The story of " Qiao Busi " connects, and the story of " Li Yanhong " is recommended with the story of " Qiao Busi " as similar writing material
To user.
Specific implementation, as shown in Fig. 2, machine learning component training after the completion of, so that it may by machine learning component come
Viewpoint is marked to the writing material in database, establishes the corresponding relationship of writing material and viewpoint, for example, can be first by writing element
The proper nouns such as the name for including in material replace with PER1, PER2 ... wait characters, then the writing element of proper noun will be substituted
Material inputs machine learning component, and the output of machine learning component is the viewpoint of the writing material.At this point, the writing in database
Material is no longer simple narration, but establish writing material and viewpoint corresponding relationship, specifically, writing material with it is corresponding
Viewpoint can by<writing material, viewpoint>pair mode store.
When it is implemented, the viewpoint vectorization of obtained writing material, the semantic vector for obtaining viewpoint can also be indicated,
In database each record can be including<writing material word sequence, viewpoint word sequence, viewpoint semantic vector indicate>etc.
Content.Specifically, the process of vectorization can using existing vectorization method realize, as word2vec method,
Method or the method for deep neural network of paragraph2vec etc., the application is not specifically limited.
When it is implemented, can be by the corresponding relationship storage of viewpoint and writing material in the database, viewpoint and writing element
The corresponding relationship of material can be stored by different modes, for example, with<viewpoint, writing material>pair mode by viewpoint and
The specific data of writing material carry out corresponding storage, alternatively, being recorded in the form of table etc. corresponding between viewpoint and writing material
The specific data of relationship or incidence relation, viewpoint and writing material can not correspond to storage, etc., as long as can be shown that sight
Corresponding relationship between point and writing material, the application do not do viewpoint and the form of expression of the corresponding relationship of writing material
It is specific to limit.
When it is implemented, recommending writing material that can also adopt in the present embodiment for the ease of inquiry, quickly to user
Viewpoint is indicated with the mode of inverted index and writes the corresponding relationship of material, includes index terms in each record of inverted index
And writing material corresponding with index terms, index terms include viewpoint.Specifically, including index terms and element in the list of inverted index
Material list two parts, inverted index each record in may include all writing materials corresponding with index terms, when multiple
When the corresponding viewpoint of writing material is identical, then the same index terms corresponds to multiple writing materials in a record, for example, writing
Make material 1 and write the viewpoint of material 8 to be all " struggle ", as shown in table 2 below, index terms " struggle " corresponds to writing 1 He of material
Material 8 is write, similarly, index terms " gratitude " corresponds to writing material 2 and writing material 6.
Specifically, index terms other than including viewpoint, can also include the keyword in the corresponding writing material of viewpoint,
It is more with the keyword of theme successful match to be checked when viewpoint and theme successful match to be checked, then it represents that writing element
The matching degree of material and theme to be checked is higher.
When it is implemented, when being indexed using inverted index to writing material and corresponding viewpoint, it can also be to writing
Material and corresponding viewpoint are segmented, remove stop words, and therefore, each writing material record becomes the sequence of a word,
It is equivalent to a document.All writing material records are equivalent to a collection of document.Any inverted index building work can be used
Tool (such as Lucene) establishes word to the corresponding relationship of document (writing material) list, as shown in table 2 below.
Table 2
| Index terms | Material list |
| …… | …… |
| Struggle | Material 1 is write, material 8 ... is write |
| Feel grateful | Material 2 is write, material 6 ... is write |
| …… | …… |
When it is implemented, as shown in Fig. 2, line can be carried out after establishing the corresponding relationship of writing material and viewpoint
Upper material is recommended.For example, the theme to be checked (or viewpoint character string) of user's input is received, it will be in theme to be checked and database
Viewpoint matched (for example, can to user input theme to be checked segment, remove stop words, use remaining word
Matched with the viewpoint (or index terms) in database), the corresponding writing material of the viewpoint of successful match is candidate writing
Writing material in candidate writing material is recommended user by material.
When it is implemented, recommending the validity of material to recommend successful match in the present embodiment to further increase
The corresponding writing material of viewpoint, comprising:
The similarity between theme to be checked and the viewpoint of successful match is calculated, in the corresponding writing of the viewpoint of successful match
In material, writing material is recommended according to similarity.
Specifically, vectorization first can be carried out to theme to be checked, the semantic vector for obtaining theme to be checked is indicated, then counts
The similarity between the semantic vector expression of theme to be checked and the semantic vector expression of the viewpoint of successful match is calculated, for example, can
To calculate similarity using the methods of cosine similarity or Euclidean distance, the application is not specifically limited.Calculate similarity it
Afterwards, which can be used as the score of candidate writing material (i.e. the corresponding writing material of the viewpoint of successful match), Jin Ergen
Candidate's writing material is ranked up according to the sequence of score from high to low, default writing material for sorting forward can be recommended
To user, the writing material after sequence can also all be recommended into user according to sequence.
When it is implemented, may be used also in the present embodiment to meet user to the inquiry needs of different type writing material
With the specified type of determination writing material to be recommended；Then in the corresponding writing material of the viewpoint of successful match, recommend to meet
The writing material of specified type.
Specifically, the information for the specified type that can be inputted according to user determines the specified type of writing material to be recommended,
The writing material for recommending which kind of specified type to the user is determined, for example, the type information of user A input is true type, then
In the corresponding writing material of viewpoint of successful match, true type is recommended to write material to user A；It can also be recommended by material
Itself setting information of method recommends specified type to designated user to determine the specified type to be recommended for writing material
Material is write, for example, the equipment of operation material recommended method or plateform system can input setting information, setting information is for fixed
Justice recommends the writing material of specified type to some or certain type of user, and then is assured that be recommended write based on setting information
Make the specified type of material, recommend true type writing material for example, realizing to user A, recommends theoretical type writing element to user B
Material.
When it is implemented, in the present embodiment, obtaining master to be checked to provide different form, flexible inquiry mode
Topic, comprising:
When input content is article (for example, article of argumentative writing type or narrative type), according to argumentative writing debate structure
Classify to the sentence in article, for example, classification results include topic sentence and sub- argument sentence, classification results can also include drawing
It, will be one of any or any in the theme of article, topic sentence, sub- argument sentence by sentence types such as sentence, argument sentence and concluding sentences
Combination is used as theme to be checked.
For example, user can input entire chapter composition after composition writing is completed, operation material recommended method
System automatically classifies to the sentence in composition using chapter debate structure analyzer, analyzes theme, the topic sentence of article
With sub- argument sentence etc., one of any or any combination in theme, topic sentence and sub- argument sentence that analysis is obtained as to
Theme is inquired, it is subsequent that writing material is recommended to user based on theme to be checked.
In the present embodiment, additionally provide a kind of calculation machine equipment, as shown in figure 4, include memory 402, processor 404 and
The computer program that can be run on a memory and on a processor is stored, the processor executes real when the computer program
Existing above-mentioned arbitrary material recommended method.
Specifically, the computer equipment can be terminal, server or similar arithmetic unit.
In the present embodiment, a kind of computer readable storage medium, the computer-readable recording medium storage are provided
There is the computer program for executing above-mentioned arbitrary material recommended method.
Specifically, computer readable storage medium includes permanently and non-permanent, removable and non-removable media can
To realize that information is stored by any method or technique.Information can be computer readable instructions, data structure, the module of program
Or other data.The example of computer readable storage medium includes, but are not limited to phase change memory (PRAM), static random-access is deposited
Reservoir (SRAM), dynamic random access memory (DRAM), other kinds of random access memory (RAM), read-only memory
(ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory techniques, CD-ROM are read-only
Memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape magnetic disk storage or
Other magnetic storage devices or any other non-transmission medium, can be used for storage can be accessed by a computing device information.According to
Herein defines, and computer readable storage medium does not include temporary computer readable media (transitory media), such as adjusts
The data-signal and carrier wave of system.
Material recommended method provided in an embodiment of the present invention realizes following technical effect: will by the material recommended method
Theme to be checked is matched with the viewpoint deposited, and then according to the corresponding relationship of the viewpoint and writing material deposited, to user
The corresponding writing material of viewpoint for recommending successful match, realizing can match for subject recommending to be checked with theme to be checked
The corresponding writing material of viewpoint so that recommend writing material it is corresponding with theme to be checked.The material recommended method is advantageous
In avoiding the problem that writer recalls from the indefinite resource of viewpoint or screens available material in the prior art, or avoid existing
Have and author needed to read the process for screening identical of views material one by one in technology, be conducive to it is time saving, efficiently obtain it is effective
Write material.
Based on the same inventive concept, a kind of material recommendation apparatus is additionally provided in the embodiment of the present invention, such as following implementation
Described in example.Since the principle that material recommendation apparatus solves the problems, such as is similar to material recommended method, the reality of material recommendation apparatus
The implementation that may refer to material recommended method is applied, overlaps will not be repeated.It is used below, term " unit " or " mould
The combination of the software and/or hardware of predetermined function may be implemented in block ".Although device described in following embodiment is preferably with soft
Part is realized, but the realization of the combination of hardware or software and hardware is also that may and be contemplated.
Fig. 5 is a kind of structural block diagram of the material recommendation apparatus of the embodiment of the present invention, as shown in figure 5, the device includes:
Theme obtains module 502, for obtaining theme to be checked；
Matching module 504, for matching the theme to be checked with the viewpoint deposited；
Recommending module 506 recommends the viewpoint of successful match for the corresponding relationship according to the viewpoint and writing material deposited
Corresponding writing material.
In one embodiment, above-mentioned apparatus further include:
Material collection module, for acquiring writing material；
Viewpoint labeling module, input of the writing material as machine learning component for that will acquire, machine learning component
The corresponding viewpoint of output writing material.
In one embodiment, above-mentioned apparatus further include:
Training module, for being trained based on following steps to machine learning component:
Sample writing material and corresponding sample viewpoint are obtained, the sample writing material that will acquire and corresponding sample viewpoint
As sample data, inputs machine learning component and be trained.
In one embodiment, the training module, comprising:
Data capture unit, for obtaining data resource；
Sample acquisition unit extracts sample writing element for the feature according to sentence in data resource from data resource
Material and corresponding sample viewpoint.
In one embodiment, the sample acquisition unit, specifically for being directed to the data resource of argumentative writing type, according to view
Paper debate structure classifies to the sentence in the data resource of argumentative writing type；Event is searched in the data resource of argumentative writing type
Thing triggers sentence, and the sentence of story triggering sentence and adjacent specified type is merged into true this writing of pattern material；By argumentative writing type
Data resource in sentence matched with known theoretical type sentence, the sentence of successful match is merged into theoretical pattern sheet
Write material.
In one embodiment, the sample acquisition unit, also particularly useful for for the words where sample writing material
The data resource of literary type carries out classification classification to the sentence in the data resource of argumentative writing type according to argumentative writing debate structure, point
Class result includes topic sentence and sub- argument sentence；By one of any or any combination in item following in the data resource of argumentative writing type
Amalgamation result as the corresponding sample viewpoint of sample writing material: topic, topic sentence and with sample writing material distance
Nearest sub- argument sentence.
In one embodiment, the training module, further includes:
Training unit, for inputting machine using true this writing of pattern material and corresponding sample viewpoint as sample data
Learning object is trained, and obtains true type machine learning component；
Using theoretical this writing of pattern material and corresponding sample viewpoint as sample data, input machine learning component is instructed
Practice, obtains theoretical type machine learning component.
In one embodiment, training unit, the proper noun for being also used to write sample in material are replaced with without specific
The character of meaning, the sample proper noun is substituted write material and corresponding sample viewpoint as sample data, input machine
Learning object is trained.
In one embodiment, theme obtains module, is also used to when input content is article, according to argumentative writing debate knot
Structure classifies to the sentence in article, and classification results include topic sentence and sub- argument sentence, by the theme of article, topic sentence and
One of any or any combination in sub- argument sentence is as theme to be checked.
In one embodiment, the recommending module, for calculating between theme to be checked and the viewpoint of successful match
Similarity recommends writing material according to similarity in the corresponding writing material of viewpoint of successful match.
In one embodiment, further includes:
Memory module, for indicating viewpoint by the way of inverted index and writing the corresponding relationship of material, inverted index
Each record in include index terms and writing material corresponding with index terms, index terms include viewpoint.
In one embodiment, the recommending module, for determining the specified type of writing material to be recommended；Matching at
In the corresponding writing material of the viewpoint of function, recommend the writing material for meeting specified type.
Material recommendation apparatus provided in an embodiment of the present invention realizes following technical effect: will by the material recommendation apparatus
Theme to be checked is matched with the viewpoint deposited, and then according to the corresponding relationship of the viewpoint and writing material deposited, to user
The corresponding writing material of viewpoint for recommending successful match, realizing can match for subject recommending to be checked with theme to be checked
The corresponding writing material of viewpoint so that recommend writing material it is corresponding with theme to be checked.The material recommendation apparatus is advantageous
In avoiding the problem that writer recalls from the indefinite resource of viewpoint or screens available material in the prior art, or avoid existing
Have and author needed to read the process for screening identical of views material one by one in technology, be conducive to it is time saving, efficiently obtain it is effective
Write material.
Obviously, those skilled in the art should be understood that each module of the above-mentioned embodiment of the present invention or each step can be with
It is realized with general computing device, they can be concentrated on a single computing device, or be distributed in multiple computing devices
On composed network, optionally, they can be realized with the program code that computing device can perform, it is thus possible to by it
Store and be performed by computing device in the storage device, and in some cases, can be held with the sequence for being different from herein
The shown or described step of row, perhaps they are fabricated to each integrated circuit modules or will be multiple in them
Module or step are fabricated to single integrated circuit module to realize.In this way, the embodiment of the present invention be not limited to it is any specific hard
Part and software combine.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the embodiment of the present invention can have various modifications and variations.All within the spirits and principles of the present invention, made
Any modification, equivalent substitution, improvement and etc. should all be included in the protection scope of the present invention.
Claims (17)
1. a kind of material recommended method characterized by comprising
Obtain theme to be checked；
The theme to be checked is matched with the viewpoint deposited；
According to the corresponding relationship of the viewpoint and writing material deposited, recommend the corresponding writing material of the viewpoint of successful match.
2. material recommended method as described in claim 1, which is characterized in that further include:
Acquisition writing material；
Using the writing material of acquisition as the input of machine learning component, the corresponding sight of machine learning component output writing material
Point.
3. material recommended method as claimed in claim 2, which is characterized in that further include:
Machine learning component is trained based on following steps:
Sample writing material and corresponding sample viewpoint are obtained, the sample writing material that will acquire and corresponding sample viewpoint conduct
Sample data, input machine learning component are trained.
4. material recommended method as claimed in claim 3, which is characterized in that obtain sample writing material and corresponding sample is seen
Point, comprising:
Obtain data resource；
According to the feature of sentence in data resource, sample writing material and corresponding sample viewpoint are extracted from data resource.
5. material recommended method as claimed in claim 4, which is characterized in that according to the feature of sentence in data resource, from number
Material is write according to sample is extracted in resource, comprising:
For the data resource of argumentative writing type, the sentence in the data resource of argumentative writing type is carried out according to argumentative writing debate structure
Classification；
Story is searched in the data resource of argumentative writing type and triggers sentence, and story triggering sentence is merged with the sentence of adjacent specified type
For true this writing of pattern material；
Sentence in the data resource of argumentative writing type is matched with known theoretical type sentence, the sentence of successful match is closed
It and is theoretical pattern this writing material.
6. material recommended method as claimed in claim 4, which is characterized in that obtain the corresponding sample of sample writing material and see
Point, comprising:
The data resource that the argumentative writing type where material is write for sample, according to argumentative writing debate structure to the number of argumentative writing type
Classify according to the sentence in resource, classification results include topic sentence and sub- argument sentence；
The amalgamation result of one of any or any combination item in item following in the data resource of argumentative writing type is write as sample
Make the corresponding sample viewpoint of material: topic, topic sentence and the sub- argument sentence nearest with sample writing material distance.
7. material recommended method as claimed in claim 3, which is characterized in that the sample writing material that will acquire and corresponding sample
This viewpoint is used as sample data, and input machine learning component is trained, comprising:
Using true this writing of pattern material and corresponding sample viewpoint as sample data, input machine learning component is trained,
Obtain true type machine learning component；
Using theoretical this writing of pattern material and corresponding sample viewpoint as sample data, input machine learning component is trained,
Obtain theoretical type machine learning component.
8. the material recommended method as described in any one of claims 1 to 7, which is characterized in that obtain theme to be checked, wrap
It includes:
When input content is article, classify according to argumentative writing debate structure to the sentence in article, classification results include
Topic sentence and sub- argument sentence, using one of any or any combination in the theme of article, topic sentence and sub- argument sentence as to
Inquire theme.
9. the material recommended method as described in any one of claims 1 to 7, which is characterized in that further include:
Determine the specified type of writing material to be recommended；
In the corresponding writing material of viewpoint of successful match, recommend the writing material for meeting specified type.
10. a kind of material recommendation apparatus characterized by comprising
Theme obtains module, for obtaining theme to be checked；
Matching module, for matching the theme to be checked with the viewpoint deposited；
Recommending module recommends the viewpoint of successful match corresponding for the corresponding relationship according to the viewpoint and writing material deposited
Write material.
11. material recommendation apparatus as claimed in claim 10, which is characterized in that further include:
Material collection module, for acquiring writing material；
Viewpoint labeling module, input of the writing material as machine learning component for that will acquire, machine learning component output
Write the corresponding viewpoint of material.
12. material recommendation apparatus as claimed in claim 11, which is characterized in that further include:
Training module, for being trained based on following steps to machine learning component:
Sample writing material and corresponding sample viewpoint are obtained, the sample writing material that will acquire and corresponding sample viewpoint conduct
Sample data, input machine learning component are trained.
13. material recommendation apparatus as claimed in claim 12, which is characterized in that the training module, comprising:
Data capture unit, for obtaining data resource；
Sample acquisition unit, for the feature according to sentence in data resource, extracted from data resource sample writing material and
Corresponding sample viewpoint.
14. material recommendation apparatus as claimed in claim 13, which is characterized in that the sample acquisition unit is specifically used for, needle
To the data resource of argumentative writing type, classify according to argumentative writing debate structure to the sentence in the data resource of argumentative writing type；
Story is searched in the data resource of argumentative writing type and triggers sentence, and story triggering sentence is merged with the sentence of adjacent specified type
For true this writing of pattern material；
Sentence in the data resource of argumentative writing type is matched with known theoretical type sentence, the sentence of successful match is closed
It and is theoretical pattern this writing material.
15. material recommendation apparatus as claimed in claim 13, which is characterized in that the sample acquisition unit, also particularly useful for,
The data resource that the argumentative writing type where material is write for sample is provided according to data of the argumentative writing debate structure to argumentative writing type
Sentence in source is classified, and classification results include topic sentence and sub- argument sentence；By item following in the data resource of argumentative writing type
In one of any or any combination item amalgamation result as the corresponding viewpoint of writing material: topic, topic sentence and with write
Make the nearest sub- argument sentence of material distance.
16. a kind of computer equipment including memory, processor and stores the meter that can be run on a memory and on a processor
Calculation machine program, which is characterized in that the processor realizes any one of claims 1 to 9 institute when executing the computer program
The material recommended method stated.
17. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has perform claim
It is required that the computer program of material recommended method described in any one of 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910166860.5A CN109933791B (en) | 2019-03-06 | 2019-03-06 | Material recommendation method, device, computer equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910166860.5A CN109933791B (en) | 2019-03-06 | 2019-03-06 | Material recommendation method, device, computer equipment and computer readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109933791A true CN109933791A (en) | 2019-06-25 |
| CN109933791B CN109933791B (en) | 2023-08-01 |
Family
ID=66986580
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910166860.5A Active CN109933791B (en) | 2019-03-06 | 2019-03-06 | Material recommendation method, device, computer equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109933791B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112182210A (en) * | 2020-09-25 | 2021-01-05 | 四川华空天行科技有限公司 | Language generation model based on composition data feature classifier and writing support method |
| CN116090473A (en) * | 2023-04-06 | 2023-05-09 | 北京大学深圳研究生院 | Intelligent writing assisting method, device and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140188866A1 (en) * | 2012-12-31 | 2014-07-03 | Microsoft Corporation | Recommendation engine based on conditioned profiles |
| KR20180036460A (en) * | 2016-09-30 | 2018-04-09 | 주식회사 솔트룩스 | System and method for recommending knowledge actively to write document |
| CN108763516A (en) * | 2018-05-31 | 2018-11-06 | 悦未来科技（深圳）有限公司 | Write exchange method, device and computer readable storage medium |
-
2019
- 2019-03-06 CN CN201910166860.5A patent/CN109933791B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140188866A1 (en) * | 2012-12-31 | 2014-07-03 | Microsoft Corporation | Recommendation engine based on conditioned profiles |
| KR20180036460A (en) * | 2016-09-30 | 2018-04-09 | 주식회사 솔트룩스 | System and method for recommending knowledge actively to write document |
| CN108763516A (en) * | 2018-05-31 | 2018-11-06 | 悦未来科技（深圳）有限公司 | Write exchange method, device and computer readable storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 何少华: "新闻主题的提炼与深化", 《新闻爱好者》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112182210A (en) * | 2020-09-25 | 2021-01-05 | 四川华空天行科技有限公司 | Language generation model based on composition data feature classifier and writing support method |
| CN112182210B (en) * | 2020-09-25 | 2023-11-24 | 四川华空天行科技有限公司 | Language generation model based on composition and theory data feature classifier and composition supporting method |
| CN116090473A (en) * | 2023-04-06 | 2023-05-09 | 北京大学深圳研究生院 | Intelligent writing assisting method, device and system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109933791B (en) | 2023-08-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10896214B2 (en) | Artificial intelligence based-document processing | |
| Lui et al. | Accurate language identification of twitter messages | |
| CN104933164B (en) | In internet mass data name entity between relationship extracting method and its system | |
| Argyrou et al. | Topic modelling on Instagram hashtags: An alternative way to Automatic Image Annotation? | |
| CN103853834B (en) | Text structure analysis-based Web document abstract generation method | |
| CN107590219A (en) | Webpage personage subject correlation message extracting method | |
| Shen et al. | A hybrid model for quality assessment of Wikipedia articles | |
| US20200004792A1 (en) | Automated website data collection method | |
| CN103744981A (en) | System for automatic classification analysis for website based on website content | |
| Bisandu et al. | Data clustering using efficient similarity measures | |
| CN113569050B (en) | Method and device for automatically constructing government affair field knowledge map based on deep learning | |
| WO2016114790A1 (en) | Reading difficulty level based resource recommendation | |
| CN111563384A (en) | Evaluation object identification method and device for E-commerce products and storage medium | |
| CN110019642A (en) | A kind of Similar Text detection method and device | |
| CN108228612B (en) | Method and device for extracting network event keywords and emotional tendency | |
| Outsios et al. | Word embeddings from large-scale greek web content | |
| Baisa et al. | Benchmark dataset for propaganda detection in Czech newspaper texts | |
| KR102185733B1 (en) | Server and method for automatically generating profile | |
| CN112015907A (en) | Method and device for quickly constructing discipline knowledge graph and storage medium | |
| Antunes et al. | Readability of web content | |
| CN109933791A (en) | Material recommended method, device, computer equipment and computer readable storage medium | |
| CN104133913B (en) | A kind of city retail shop information bank automatic build system being polymerized with search based on video analysis and method | |
| Shrestha | Detecting Fake News with Sentiment Analysis and Network Metadata | |
| KR102126911B1 (en) | Key player detection method in social media using KeyplayerRank | |
| CN116976321A (en) | Text processing method, apparatus, computer device, storage medium, and program product |
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 |