CN110413737A - A kind of determination method, apparatus, server and the readable storage medium storing program for executing of synonym - Google Patents
A kind of determination method, apparatus, server and the readable storage medium storing program for executing of synonym Download PDFInfo
- Publication number
- CN110413737A CN110413737A CN201910699704.5A CN201910699704A CN110413737A CN 110413737 A CN110413737 A CN 110413737A CN 201910699704 A CN201910699704 A CN 201910699704A CN 110413737 A CN110413737 A CN 110413737A
- Authority
- CN
- China
- Prior art keywords
- text sequence
- search text
- field
- search
- synonym
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention discloses determination method, apparatus, server and the readable storage medium storing program for executing of a kind of synonym, wherein this method can be applied to the machine learning techniques of artificial intelligence field, this method comprises: obtaining the concurrent for hitting search text sequence including multiple concurrents hits search text sequence set, each concurrent hits each search text sequence in search text sequence, and there are associated search results；The Automobile driving probability of each field in each search text sequence is determined based on attention model；Synonym discrimination model is trained according to the Automobile driving probability of each field, obtains the synonym discrimination model for introducing attention mechanism；Concurrent to be discriminated is hit into the synonym discrimination model that search text sequence input introduces attention mechanism, determines the synonym pair that concurrent to be discriminated is hit in search text sequence.By this embodiment, the accuracy of determining synonym pair is improved, expands the semantic coverage of synonym pair.
Description
Technical field
This application involves field of computer technology more particularly to a kind of determination method, apparatus of synonym, server and can
Read storage medium.
Background technique
Currently, classical synonymicon constructing technology includes at least two kinds, one is linguist to pass through Modern Chinese
Dictionary definition carrys out manual sorting, and another kind can hit search text sequence set by concurrent and count by modern search engines
Calculation machine automatic aligning technology, and then the synonym pair of potential candidate is obtained, finally manually deleted by various statistical language features again
Form slection at more magnanimity synonymicon.
However, this mode can only realize concurrent hit order of the field in search text sequence after word cutting it is perfectly aligned when,
It determines the synonym pair that concurrent is hit in search text sequence, leads to the accuracy rate angle for determining synonym pair, and determine
The synonym disadvantage not wide enough to semantic coverage.
Therefore, the accuracy rate of determining synonym pair how is improved, expands the semantic coverage of synonym pair, becomes one urgently
It solves the problems, such as.
Summary of the invention
The embodiment of the invention provides determination method, apparatus, server and the readable storage medium storing program for executing of a kind of synonym, are based on
The synonym discrimination model for being introduced into attention mechanism determines the synonym pair that concurrent is hit in search text sequence, can determine not
With the text sequence of order of the field, corresponding synonym pair, improves the accuracy of determining synonym pair, expands synonym pair
Semantic coverage.
In a first aspect, the embodiment of the invention provides a kind of determination methods of synonym, comprising:
It obtains concurrent and hits search text sequence set, it includes that multiple concurrents are hit that the concurrent, which is hit in search text sequence set,
Text sequence is searched for, each concurrent hits each search text sequence correspondence in search text sequence, and there are associated search results；
The Automobile driving probability for each field for including in each search text sequence is determined based on attention model；
According to the Automobile driving probability for each field for including in each search text sequence to synonym discrimination model
It is trained, obtains the synonym discrimination model for introducing attention mechanism；
Concurrent to be discriminated is hit into search text sequence and inputs the synonym discrimination model for introducing attention mechanism, with
Determine the synonym pair that the concurrent to be discriminated is hit in search text sequence.
Second aspect, the embodiment of the invention provides a kind of determining devices of synonym, comprising:
Module is obtained, hits search text sequence set for obtaining concurrent, the concurrent is hit in search text sequence set
Search text sequence is hit including multiple concurrents, each search text sequence that each concurrent is hit in search text sequence is corresponding to have pass
The search result of connection；
Distribution module, for determining the attention for each field for including in each search text sequence based on attention model
Power allocation probability；
Training module, for according to it is described it is each search text sequence in include each field Automobile driving probability to same
Adopted word discrimination model is trained, and obtains the synonym discrimination model for introducing attention mechanism；
Determining module inputs the synonymous of the introducing attention mechanism for concurrent to be discriminated to be hit search text sequence
Word discrimination model, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.
The third aspect, the embodiment of the invention also provides a kind of servers, comprising: processor and storage device；It is described to deposit
Storage device, for storing program instruction；The processor calls described program instruction, for executing: obtaining concurrent and hits search text
This arrangement set, it includes that multiple concurrents hit search text sequence that the concurrent, which is hit in search text sequence set, and each concurrent is hit
There are associated search results for each search text sequence correspondence in search text sequence；It is determined based on attention model described each
The Automobile driving probability for each field for including in search text sequence；According to each word for including in each search text sequence
The Automobile driving probability of section is trained synonym discrimination model, and the synonym for obtaining introducing attention mechanism differentiates mould
Type；Concurrent to be discriminated is hit into search text sequence and inputs the synonym discrimination model for introducing attention mechanism, with determination
The concurrent to be discriminated hits the synonym pair in search text sequence out.
Fourth aspect, the embodiment of the invention also provides a kind of computer readable storage medium, the computer-readable storages
Program instruction is stored in medium, which is performed, for realizing method described in above-mentioned first aspect.
Available concurrent hits search text sequence set in the embodiment of the present invention, is determined based on attention model described total
The Automobile driving probability that each field for including in text sequence is respectively searched in search text sequence set is clicked, and according to described
The Automobile driving probability for each field for including in each search text sequence is trained synonym discrimination model, is introduced
The synonym discrimination model of attention mechanism, and concurrent to be discriminated is hit into search text sequence and inputs the introducing attention
The synonym discrimination model of mechanism, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.Pass through
This embodiment improves the accuracy of determining synonym pair, expands the semantic coverage of synonym pair.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of structural representation of synonym discrimination model for introducing attention mechanism provided in an embodiment of the present invention
Figure；
Fig. 2 is a kind of frame signal of synonym discrimination model for introducing attention mechanism provided in an embodiment of the present invention
Figure；
Fig. 3 is a kind of schematic diagram of the thermodynamic chart of Automobile driving probability provided in an embodiment of the present invention；
Fig. 4 is a kind of flow diagram of the determination method of synonym provided in an embodiment of the present invention；
Fig. 5 is a kind of structural schematic diagram of the determining device of synonym provided in an embodiment of the present invention；
Fig. 6 is a kind of structural schematic diagram of server provided in an embodiment of the present invention.
Specific embodiment
With artificial intelligence technology research and progress, research and application is unfolded in multiple fields in artificial intelligence technology, such as
Common smart home, intelligent wearable device, virtual assistant, intelligent sound box, intelligent marketing, unmanned, automatic Pilot, nobody
Machine, robot, intelligent medical, intelligent customer service etc., it is believed that with the development of technology, artificial intelligence technology will obtain in more fields
To application, and play more and more important value.
Artificial intelligence (Artificial Intelligence, AI) is to utilize digital computer or digital computer control
Machine simulation, extension and the intelligence for extending people of system, perception environment obtain knowledge and the reason using Knowledge Acquirement optimum
By, method, technology and application system.In other words, artificial intelligence is a complex art of computer science, it attempts to understand
The essence of intelligence, and produce a kind of new intelligence machine that can be made a response in such a way that human intelligence is similar.Artificial intelligence
The design principle and implementation method for namely studying various intelligence machines make machine have the function of perception, reasoning and decision.
Artificial intelligence technology is an interdisciplinary study, is related to that field is extensive, and the technology of existing hardware view also has software layer
The technology in face.Artificial intelligence basic technology generally comprise as sensor, Special artificial intelligent chip, cloud computing, distributed storage,
The technologies such as big data processing technique, operation/interactive system, electromechanical integration.Artificial intelligence software's technology mainly includes computer
Several general orientation such as vision technique, voice processing technology, natural language processing technique and machine learning/deep learning.
Wherein, machine learning (Machine Learning, ML) is a multi-field cross discipline, is related to probability theory, system
Count the multiple subjects such as, Approximation Theory, convextiry analysis, algorithm complexity theory.Specialize in the mankind are simulated or realized to computer how
Learning behavior reorganize the existing structure of knowledge to obtain new knowledge or skills and be allowed to constantly improve the performance of itself.
Machine learning is the core of artificial intelligence, is the fundamental way for making computer have intelligence, and application is each throughout artificial intelligence
A field.Machine learning and deep learning generally include artificial neural network, confidence network, intensified learning, transfer learning, conclusion
The technologies such as study, formula teaching habit.Wherein, the determination technology of synonym is an important technology of machine learning application.
The determination scheme of synonym provided in an embodiment of the present invention is related to the machine learning techniques of artificial intelligence, can pass through
It obtains concurrent and hits search text sequence set, hit in each search text sequence in search text sequence set according to the concurrent
The Automobile driving probability for each field for including is trained synonym discrimination model, obtains introducing the synonymous of attention mechanism
Word discrimination model, and concurrent to be discriminated is hit into search text sequence and inputs the synonym differentiation mould for introducing attention mechanism
Type, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence, to improve determining synonym pair
Accuracy, expand the semantic coverage of synonym pair.It is illustrated especially by following examples:
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention is clearly retouched
It states, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.Based on the present invention
In embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall within the protection scope of the present invention.
With reference to the accompanying drawing, it elaborates to some embodiments of the present invention.In the absence of conflict, following
Feature in embodiment and embodiment can be combined with each other.
The determination method of the synonym provided in the embodiment of the present invention can be executed by a kind of server, specifically, can be with
It is executed by the determining device of the synonym in server.
Current synonym to the synonym determined of the method for determination to there are synonym semantic coverage is small, and can not and
When include the exemplary shortcomings of neologisms；And search text sequence is hit currently based on concurrent and full word alignment techniques based on sequence obtain
Method to synonym pair can only can just be determined together in the case where concurrent hits the sequence alignment of each word in search text sequence
Adopted word pair, so as to cause the synonym disadvantage not wide enough to semantic coverage.
For example, it is assumed that have " how grilled chicken wing is nice " in search engine logs, " how grilled chicken wing is nice " the two search
Text sequence, after carrying out word cutting to the two search text sequences, obtain " how/roasting/chicken wings/nice " and " how/roasting/chicken
Wing/nice ", by sequence full word alignment techniques, finally obtaining " how " and " how " the two words is synonym.But for
Such as the search text sequence that the sequence of " how folding arms nice " and " it is nice how chaos is done " both words is not aligned, it cuts
The sequence for searching for potential synonym context in text sequence due to two after complete word is disturbed, then is not available based on sequence
Full word alignment techniques obtain synonym pair.
In addition to this, when " won ton " this Chinese language words are translated in current alignment techniques, each text sequence
In word for special translating purpose word " folding arms " contribution be identical, it is clear that " won ton " is more important for translating into " folding arms ", does not have
There is the model for introducing attention less problematic when input sentence comparison is short, but if input sentence comparison is long, at this time
All semantemes indicate that the information of word itself has disappeared by an intermediate semantic vector completely, so as to cause losing very
More detailed information.
To solve the above-mentioned problems, the embodiment of the present invention proposes a kind of synonym differentiation by introducing attention mechanism
The method that model determines synonym pair hits search text sequence set by obtaining concurrent, and the concurrent hits search text sequence
It include that multiple concurrents hit search text sequence in set, each search text sequence that each concurrent is hit in search text sequence is corresponding
There are associated search results；The attention for each field for including in each search text sequence is determined based on attention model
Allocation probability, and mould is differentiated to synonym according to the Automobile driving probability for each field for including in each search text sequence
Type is trained, and obtains the synonym discrimination model for introducing attention mechanism, and concurrent to be discriminated is hit search text sequence
The column input synonym discrimination model for introducing attention mechanism, to determine that the concurrent to be discriminated hits search text sequence
Synonym pair in column.By this embodiment, may be implemented from the entirely different search text sequence of order of the field really
Synonym pair is made, the semantic coverage of synonym is expanded, improves the accuracy rate of determining synonym pair.
In one embodiment, the synonym discrimination model proposed by the present invention for introducing attention mechanism may include but not
It is limited to the sequence based on attention mechanism to sequence Seq2Seq model；In certain embodiments, the seq2seq full name
Sequence to Sequence。
In some embodiments, the structure of the synonym discrimination model for introducing attention mechanism is as shown in Figure 1, Fig. 1
It is a kind of structural schematic diagram of synonym discrimination model for introducing attention mechanism provided in an embodiment of the present invention.As shown in Figure 1,
It is described introduce attention mechanism synonym discrimination model include coding side 11, decoding end 12, attention model 13 i.e.
Attention, the synonym discrimination model of the introducing attention mechanism are a kind of general encoder --- decoder chassis, can
For scenes such as machine translation, text snippet, conversation modeling, image subtitles.In one embodiment, it is inputted by coding side 11
First text sequence exports the second text sequence corresponding with the text sequence by decoding end 12.
In one embodiment, in coding side 11 by the first text sequence of a variable-length become regular length to
The vector of this regular length is become the second text sequence of variable-length by amount expression, decoding end 12.In certain embodiments,
It is equivalent to and the first text sequence is translated as semantic identical second text sequence, and find out synonym pair therein.For example, false
If the first text sequence is " how folding arms nice ", the second text sequence is " it is nice how chaos is done ", if coding side 11
Middle the first text sequence " how folding arms nice " by a variable-length becomes the vector expression of regular length, decoding end 12
The second text sequence " it is nice how chaos is done " that the vector of this regular length is become to variable-length can then determine
Synonym in one text sequence and the second text sequence is to fold arms and chaos.
In one embodiment, it inputs and " how to fold arms in the synonym discrimination model for being introduced into attention mechanism
It is nice " text sequence after, how text sequence " can be done by the synonym discrimination model of the introducing attention mechanism
Fold arms nice " split, obtain " how ", " doing ", " folding arms ", " nice ", and determine " how ", " doing ", " folding arms ",
The Automobile driving probability of " nice " each field (i.e. word), and determine that the Automobile driving probability is greater than predetermined probabilities threshold value
Field be synonym pair.
For example, it is assumed that determining how text sequence " is done by the synonym discrimination model of the introducing attention mechanism
Fold arms nice " in each field Automobile driving probability be (how, 0.1), (doing, 0.1), (folding arms, 0.7), (nice, 0.1),
In output sequence " it is nice how chaos is done " the Automobile driving probability of each field be (chaos, 0.7), (how, 0.1), (do,
0.1), (nice, 0.1), it is assumed that predetermined probabilities threshold value is 0.5, then can determine that " folding arms " and " chaos " is synonym pair.
In one embodiment, the synonym discrimination model for introducing attention mechanism can be illustrated by Fig. 2
Illustrate, Fig. 2 is a kind of block schematic illustration of synonym discrimination model for introducing attention mechanism provided in an embodiment of the present invention.Such as
Shown in Fig. 2, by the synonym discrimination model of the introducing attention mechanism, text sequence is inputted from coding side 21
" X1X2X3X4 " encodes text sequence " X1X2X3X4 " by coding side 21 to obtain vector " C1C2C3 ", passes through decoding
22 pairs of vectors " C1C2C3 " are held to be decoded output text sequence " Y1Y2Y3 ".Pass through the attention point of each word in " X1X2X3X4 "
With each Automobile driving probability in probability and " Y1Y2Y3 ", the word for determining that Automobile driving probability is greater than predetermined probabilities threshold value is same
Adopted word pair.
In one embodiment, the Automobile driving probability can indicate that Fig. 3 is by thermodynamic chart as shown in Figure 3
A kind of schematic diagram of the thermodynamic chart of Automobile driving probability provided in an embodiment of the present invention.As shown in figure 3, abscissa is and input
The corresponding each field of text sequence " how folding arms nice ", respectively " how ", " doing ", " folding arms ", " nice ", ordinate
For the corresponding each field of text sequence " it is nice how chaos is done " with output, respectively " chaos ", " how ", " doing ", " good
It eats ".Wherein, color is deeper in thermodynamic chart shown in Fig. 3, represents that Automobile driving probability is bigger, and as can be seen from Figure 3, color is most deep
Region 31 be abscissa " folding arms " region corresponding with ordinate " chaos ".It therefore, can be true according to the color in thermodynamic chart
Making " folding arms " and " chaos " is synonym pair.
In other embodiments, the synonym discrimination model proposed by the present invention for introducing attention mechanism can also use IBM
The other machines such as Model1, IBM Model2 translation model excavates to complete the building of synonymicon, does not do specific limit herein
It is fixed.
The determination method of synonym provided in an embodiment of the present invention is schematically illustrated with reference to the accompanying drawing.
Fig. 4 is specifically referred to, Fig. 4 is a kind of process signal of the determination method of synonym provided in an embodiment of the present invention
Figure, the method can be executed by the determining device of the synonym in server, wherein for example preceding institute of the specific explanations of server
It states.Specifically, described method includes following steps for the embodiment of the present invention.
S401: it obtains concurrent and hits search text sequence set, it includes multiple that the concurrent, which is hit in search text sequence set,
Concurrent hits search text sequence, and each concurrent hits each search text sequence correspondence in search text sequence, and there are associated search
As a result.
In the embodiment of the present invention, the available concurrent of server hits search text sequence set, and the concurrent hits search text
It include that multiple concurrents hit search text sequence in this arrangement set, each concurrent hits each search text sequence in search text sequence
Column are corresponding, and there are associated search results.
For example, it is assumed that it includes that 2 concurrents hit search text sequence that concurrent, which is hit in search text sequence set, respectively " kind
The recipe of eggplant " and " recipe of tomato ", wherein " recipe of tomato " corresponding search result is " tomato scrambled eggs ", " western red
The corresponding search result of the recipe of persimmon " is " Tomato omelette ", and " the tomato scrambled eggs " and " Tomato omelette " is to close
The search result of connection.
In certain embodiments, the concurrent hit search text sequence can be obtained in wechat search it is a large amount of associated
The search text sequence of search result is that concurrent hits search text sequence, for example, largely clicking the first search in wechat search
There are associated search results for text sequence and the second search of click text sequence, can such as obtain the associated search knot
First search text sequence of the number of clicks of fruit greater than 50 times or more and the second search text sequence are that concurrent hits search text
Sequence, it can it is being semantically similar for determining that the first search text sequence searches for text sequence with second.
In one embodiment, the associated search result can by calculating the similarity between each search result,
The search result for determining that the similarity is greater than similarity threshold is associated search result.
In one embodiment, server can determine different search when acquisition concurrent hits search text sequence set
Rope text sequence is corresponding, and there are the numbers of associated search result, and there is pass according to the different search text sequence is corresponding
The number of the search result of connection determines that the concurrent hits search text sequence set.
In one embodiment, the different search text sequence includes the first search text sequence and the second search text
This sequence；Server is corresponding there are the number of associated search result according to the different search text sequence, determine described in
When concurrent hits search text sequence set, the search result of the available first search text sequence, and described in acquisition
The search result of second search text sequence.Server can determine it is described first search text sequence search result with it is described
There are the number of associated search result in the search result of second search text sequence, if the associated search result
Number is greater than preset times threshold value, then can determine that the first search text sequence and the second search text sequence are described total
Click search text sequence set.
For example, it is assumed that the first search text sequence is " recipe of tomato ", the second search text sequence is " tomato
Recipe ", get it is described first search text sequence " recipe of tomato " search result be " tomato scrambled eggs " number
It is 20 times, and getting the search result of second search text sequence " recipe of tomato " is " Tomato omelette "
Number be 25, if preset times threshold value is 18 times, the search result of " recipe of tomato " is " tomato scrambled eggs "
Number be to be greater than preset times threshold value 18 20 times, the search result of " recipe of tomato " is " Tomato omelette "
Number is 25 greater than preset times threshold value 18, thus may determine that " recipe of tomato " and " recipe of tomato " is described
Concurrent hits search text sequence set.
S402: determine that the Automobile driving for each field for including in each search text sequence is general based on attention model
Rate.
In the embodiment of the present invention, server, which can be determined based on attention model in each search text sequence, includes
The Automobile driving probability of each field.
In one embodiment, server determines each word for including in each search text sequence based on attention model
When the Automobile driving probability of section, deconsolidation process can be carried out to each search text sequence, obtain each search text
At least one corresponding field of sequence, and determine the initial Automobile driving probability of each field at least one described field, with
And the initial Automobile driving probability is carried out according to the number that each field occurs in each search text sequence
Update processing, with the Automobile driving probability of determination each field.
In one embodiment, the initial Automobile driving of server each field in determining at least one described field is general
When rate, the part of speech of each field at least one described field can be determined, and according to preset part of speech and Automobile driving probability
Corresponding relationship, determine corresponding with the part of speech of each field initial Automobile driving probability.
For example, by search for text sequence " recipe of tomato " be split as " tomato ", " ", " recipe " three fields,
Wherein, the part of speech of " tomato " be noun, " " part of speech be adverbial word, " recipe " be noun, according to preset part of speech and pay attention to
The corresponding relationship of power allocation probability determines that initial Automobile driving probability corresponding with noun " tomato " is 0.45, adverbial word
" " corresponding initial Automobile driving probability is 0.1, the corresponding initial Automobile driving probability of noun " recipe " is 0.45.
In one embodiment, the initial Automobile driving of server each field in determining at least one described field is general
When rate, the quantity of field in each search text sequence can be determined, and according to field in each search text sequence
Equal initial Automobile driving probability is arranged for each field, wherein the initial Automobile driving probability of each field in quantity
The sum of be 1.
For example, by search for text sequence " recipe of tomato " be split as " tomato ", " ", " recipe " three fields,
It is respectively 1/3,1/3,1/3 that equal initial Automobile driving probability, which is arranged, for these three fields.
In one embodiment, the initial Automobile driving of server each field in determining at least one described field is general
When rate, 0 can also be set by the initial Automobile driving probability of each field in each search text sequence.
S403: synonym is differentiated according to the Automobile driving probability for each field for including in each search text sequence
Model is trained, and obtains the synonym discrimination model for introducing attention mechanism.
In the embodiment of the present invention, server can be according to the attention for each field for including in each search text sequence
Allocation probability is trained synonym discrimination model, obtains the synonym discrimination model for introducing attention mechanism.
In one embodiment, server is in the attention point according to each field for including in each search text sequence
Synonym discrimination model is trained with probability, it, can basis when obtaining introducing the synonym discrimination model of attention mechanism
The Automobile driving probability that concurrent hits each field in search text sequence is trained synonym discrimination model, obtains each field
Automobile driving probability.If the concurrent hits the Automobile driving probability of the synonym pair in search each field of text sequence
No more than predetermined probabilities threshold value, then corresponding parameter in synonym discrimination model is adjusted, and the basis after adjusting parameter
The Automobile driving probability that concurrent hits each field in search text sequence is trained synonym discrimination model, when in each field
When the Automobile driving probability of synonym is greater than predetermined probabilities threshold value, the synonym discrimination model for introducing attention mechanism is obtained.
S404: concurrent to be discriminated is hit into search text sequence and inputs the synonym differentiation mould for introducing attention mechanism
Type, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.
In the embodiment of the present invention, concurrent to be discriminated can be hit search text sequence and input the introducing attention by server
The synonym discrimination model of power mechanism, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.
In one embodiment, concurrent to be discriminated is hit search text sequence and inputs the introducing attention machine by server
The synonym discrimination model of system can be with to determine synonym clock synchronization that the concurrent to be discriminated is hit in search text sequence
The concurrent to be discriminated is hit into search text sequence and inputs the synonym discrimination model for introducing attention mechanism, obtains institute
It states concurrent to be discriminated and hits the corresponding Automobile driving probability of each field in search text sequence, and determine that each field is corresponding
Automobile driving probability be greater than predetermined probabilities threshold value field be synonym pair.
For example, it is assumed that it is " how folding arms nice " that concurrent to be discriminated, which hits search text sequence, paid attention to by the introducing
The synonym discrimination model of power mechanism obtains the Automobile driving probability of each field in text sequence " how folding arms nice "
For (how, 0.1), (doing, 0.1), (folding arms, 0.7), (nice, 0.1), and obtain in output sequence " it is nice how chaos is done "
The Automobile driving probability of each field is (chaos, 0.7), (how, 0.1), (doing, 0.1), (nice, 0.1), it is assumed that default general
Rate threshold value is 0.5, then can determine that " folding arms " and " chaos " is synonym pair.
As it can be seen that can be determined in each search text sequence to be discriminated of different field sequence by this embodiment
Synonym pair, expand the range of synonym, improve the accuracy rate of determining synonym.
In the embodiment of the present invention, the available concurrent of server hits search text sequence set, true based on attention model
The fixed concurrent hits in search text sequence set the Automobile driving probability for respectively searching for each field for including in text sequence, and
Synonym discrimination model is trained according to the Automobile driving probability for each field for including in each search text sequence,
It obtains introducing the synonym discrimination model of attention mechanism, and concurrent to be discriminated is hit and is drawn described in search text sequence input
Enter the synonym discrimination model of attention mechanism, to determine that the concurrent to be discriminated hits the synonym searched in text sequence
It is right.By this embodiment, the accuracy of determining synonym pair is improved, expands the semantic coverage of synonym pair.
Fig. 5 is referred to, Fig. 5 is a kind of structural schematic diagram of the determining device of synonym provided in an embodiment of the present invention.Tool
Body, described device includes: to obtain module 501, distribution module 502, training module 503, determining module 504；
Module 501 is obtained, hits search text sequence set for obtaining concurrent, the concurrent hits search text sequence set
In include that multiple concurrents hit search text sequence, each search text sequence that each concurrent is hit in search text sequence is corresponding to be existed
Associated search result；
Distribution module 502, for determining each field for including in each search text sequence based on attention model
Automobile driving probability；
Training module 503, for the Automobile driving probability according to each field for including in each search text sequence
Synonym discrimination model is trained, the synonym discrimination model for introducing attention mechanism is obtained；
Determining module 504 inputs the introducing attention mechanism for concurrent to be discriminated to be hit search text sequence
Synonym discrimination model, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.
Further, when the acquisition of acquisition module 501 concurrent hits search text sequence set, it is specifically used for:
Determine that there are the numbers of associated search result for different search text sequence correspondences；
It is corresponding there are the number of associated search result according to the different search text sequence, determine that the concurrent is hit
Search for text sequence set.
Further, the different search text sequence includes the first search text sequence and the second search text sequence
Column；The module 501 that obtains is corresponded to according to the different search text sequence there are the number of associated search result, is determined
When the concurrent hits search text sequence set, it is specifically used for:
Obtain the search result of the first search text sequence；
Obtain the search result of the second search text sequence；
It determines in the search result of the first search text sequence and the search result of the second search text sequence
There are the numbers of associated search result；
If the number of the associated search result is greater than preset times threshold value, it is determined that the first search text sequence
Column and the second search text sequence are that the concurrent hits search text sequence set.
Further, the distribution module 502 is determined in each search text sequence based on attention model includes
When the Automobile driving probability of each field, it is specifically used for:
Deconsolidation process is carried out to each search text sequence, obtain each search text sequence it is corresponding at least one
Field；
Determine the initial Automobile driving probability of each field at least one described field；
The number occurred in each search text sequence according to each field is general to the initial Automobile driving
Rate is updated processing, with the Automobile driving probability of determination each field.
Further, the distribution module 502 determines the initial Automobile driving of each field at least one described field
When probability, it is specifically used for:
Determine the part of speech of each field at least one described field；
According to the corresponding relationship of preset part of speech and Automobile driving probability, determination is corresponding with the part of speech of each field
Initial Automobile driving probability.
Further, the distribution module 502 determines the initial Automobile driving of each field at least one described field
When probability, it is specifically used for:
Determine the quantity of field in each search text sequence；
According to the quantity of field in each search text sequence, it is general that equal initial Automobile driving is set for each field
Rate, wherein the sum of initial Automobile driving probability of each field is 1.
Further, concurrent to be discriminated is hit search text sequence and inputs the introducing attention by the determining module 504
The synonym discrimination model of power mechanism, to determine that the concurrent to be discriminated hits the synonym clock synchronization searched in text sequence,
It is specifically used for:
The concurrent to be discriminated is hit into search text sequence and inputs the synonym differentiation mould for introducing attention mechanism
Type obtains the concurrent to be discriminated and hits the corresponding Automobile driving probability of each field in search text sequence；
The field for determining that the corresponding Automobile driving probability of each field is greater than predetermined probabilities threshold value is synonym pair.
The available concurrent of the embodiment of the present invention hits search text sequence set, determines the concurrent based on attention model
It hits in search text sequence set and respectively searches for the Automobile driving probability for each field for including in text sequence, and according to described each
The Automobile driving probability for each field for including in search text sequence is trained synonym discrimination model, obtains introducing note
The synonym discrimination model for power mechanism of anticipating, and concurrent to be discriminated is hit into search text sequence and inputs the introducing attention machine
The synonym discrimination model of system, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.Pass through this
Kind embodiment, improves the accuracy of determining synonym pair, expands the semantic coverage of synonym pair.
Fig. 6 is referred to, Fig. 6 is a kind of structural schematic diagram of server provided in an embodiment of the present invention.Specifically, the clothes
Business device includes: memory 601, processor 602.
In one embodiment, the server further includes data-interface 603, the data-interface 603, same for transmitting
Data information between the determining device of adopted word and other devices.
The memory 601 may include volatile memory (volatile memory)；Memory 601 also can wrap
Include nonvolatile memory (non-volatile memory)；Memory 601 can also include the group of the memory of mentioned kind
It closes.The processor 602 can be central processing unit (central processing unit, CPU).The processor 602 is also
It may further include hardware chip.Above-mentioned hardware chip can be specific integrated circuit (application-specific
Integrated circuit, ASIC), programmable logic device (programmable logic device, PLD) or its group
It closes.Above-mentioned PLD can be Complex Programmable Logic Devices (complex programmable logic device, CPLD), existing
Field programmable logic gate array (field-programmable gate array, FPGA) or any combination thereof.
The memory 601 can call the journey stored in storage 601 for storing program instruction, the processor 602
Sequence instruction, for executing following steps:
It obtains concurrent and hits search text sequence set, it includes that multiple concurrents are hit that the concurrent, which is hit in search text sequence set,
Text sequence is searched for, each concurrent hits each search text sequence correspondence in search text sequence, and there are associated search results；
The Automobile driving probability for each field for including in each search text sequence is determined based on attention model；
According to the Automobile driving probability for each field for including in each search text sequence to synonym discrimination model
It is trained, obtains the synonym discrimination model for introducing attention mechanism；
Concurrent to be discriminated is hit into search text sequence and inputs the synonym discrimination model for introducing attention mechanism, with
Determine the synonym pair that the concurrent to be discriminated is hit in search text sequence.
Further, when the acquisition of processor 602 concurrent hits search text sequence set, it is specifically used for:
Determine that there are the numbers of associated search result for different search text sequence correspondences；
It is corresponding there are the number of associated search result according to the different search text sequence, determine that the concurrent is hit
Search for text sequence set.
Further, the different search text sequence includes the first search text sequence and the second search text sequence
Column；There are the numbers of associated search result according to the different search text sequence correspondence for the processor 602, determine institute
State concurrent hit search text sequence set when, be specifically used for:
Obtain the search result of the first search text sequence；
Obtain the search result of the second search text sequence；
It determines in the search result of the first search text sequence and the search result of the second search text sequence
There are the numbers of associated search result；
If the number of the associated search result is greater than preset times threshold value, it is determined that the first search text sequence
Column and the second search text sequence are that the concurrent hits search text sequence set.
Further, the processor 602 based on attention model determine it is described it is each search text sequence in include it is each
When the Automobile driving probability of field, it is specifically used for:
Deconsolidation process is carried out to each search text sequence, obtain each search text sequence it is corresponding at least one
Field；
Determine the initial Automobile driving probability of each field at least one described field；
The number occurred in each search text sequence according to each field is general to the initial Automobile driving
Rate is updated processing, with the Automobile driving probability of determination each field.
Further, the processor 602 determines that the initial Automobile driving of each field at least one described field is general
When rate, it is specifically used for:
Determine the part of speech of each field at least one described field；
According to the corresponding relationship of preset part of speech and Automobile driving probability, determination is corresponding with the part of speech of each field
Initial Automobile driving probability.
Further, the processor 602 determines that the initial Automobile driving of each field at least one described field is general
When rate, it is specifically used for:
Determine the quantity of field in each search text sequence；
According to the quantity of field in each search text sequence, it is general that equal initial Automobile driving is set for each field
Rate, wherein the sum of initial Automobile driving probability of each field is 1.
Further, concurrent to be discriminated is hit search text sequence and inputs the introducing attention by the processor 602
The synonym discrimination model of mechanism, to determine that the concurrent to be discriminated hits the synonym clock synchronization searched in text sequence, tool
Body is used for:
The concurrent to be discriminated is hit into search text sequence and inputs the synonym differentiation mould for introducing attention mechanism
Type obtains the concurrent to be discriminated and hits the corresponding Automobile driving probability of each field in search text sequence；
The field for determining that the corresponding Automobile driving probability of each field is greater than predetermined probabilities threshold value is synonym pair.
In the embodiment of the present invention, the available concurrent of server hits search text sequence set, true based on attention model
The fixed concurrent hits in search text sequence set the Automobile driving probability for respectively searching for each field for including in text sequence, and
Synonym discrimination model is trained according to the Automobile driving probability for each field for including in each search text sequence,
It obtains introducing the synonym discrimination model of attention mechanism, and concurrent to be discriminated is hit and is drawn described in search text sequence input
Enter the synonym discrimination model of attention mechanism, to determine that the concurrent to be discriminated hits the synonym searched in text sequence
It is right.By this embodiment, the accuracy of determining synonym pair is improved, expands the semantic coverage of synonym pair.
The embodiments of the present invention also provide a kind of computer readable storage medium, the computer readable storage medium is deposited
Computer program is contained, is realized described in embodiment corresponding to Fig. 4 of the present invention when the computer program is executed by processor
Method can also realize the device of embodiment corresponding to the present invention described in Fig. 5, and details are not described herein.
The computer readable storage medium can be the internal storage unit of equipment described in aforementioned any embodiment, example
Such as the hard disk or memory of equipment.The computer readable storage medium is also possible to the External memory equipment of the equipment, such as
The plug-in type hard disk being equipped in the equipment, intelligent memory card (Smart Media Card, SMC), secure digital (Secure
Digital, SD) card, flash card (Flash Card) etc..Further, the computer readable storage medium can also be wrapped both
The internal storage unit for including the equipment also includes External memory equipment.The computer readable storage medium is described for storing
Other programs and data needed for computer program and the terminal.The computer readable storage medium can be also used for temporarily
When store the data that has exported or will export.
Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with
Relevant hardware is instructed to complete by computer program, the program can be stored in a computer-readable storage medium
In, the program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, the storage medium can be magnetic
Dish, CD, read-only memory (Read-Only Memory, ROM) or random access memory (Random Access
Memory, RAM) etc..
Above disclosed is only section Example of the invention, cannot limit the right of the present invention with this certainly
Range, those skilled in the art can understand all or part of the processes for realizing the above embodiment, and according to right of the present invention
Equivalent variations made by it is required that, still belongs to the scope covered by the invention.
Claims (10)
1. a kind of determination method of synonym characterized by comprising
It obtains concurrent and hits search text sequence set, it includes that multiple concurrents hit search that the concurrent, which is hit in search text sequence set,
Text sequence, each concurrent hits each search text sequence correspondence in search text sequence, and there are associated search results；
The Automobile driving probability for each field for including in each search text sequence is determined based on attention model；
Synonym discrimination model is carried out according to the Automobile driving probability for each field for including in each search text sequence
Training obtains the synonym discrimination model for introducing attention mechanism；
Concurrent to be discriminated is hit into search text sequence and inputs the synonym discrimination model for introducing attention mechanism, with determination
The concurrent to be discriminated hits the synonym pair in search text sequence out.
2. the method according to claim 1, wherein the acquisition concurrent hits search text sequence set, comprising:
Determine that there are the numbers of associated search result for different search text sequence correspondences；
It is corresponding there are the number of associated search result according to the different search text sequence, determine that the concurrent hits search
Text sequence set.
3. according to the method described in claim 2, it is characterized in that, the different search text sequence includes the first search text
This sequence and the second search text sequence；It is described that according to the different search text sequence correspondence, there are associated search results
Number, determine the concurrent hit search text sequence set, comprising:
Obtain the search result of the first search text sequence；
Obtain the search result of the second search text sequence；
It determines and exists in the search result of the first search text sequence and the search result of the second search text sequence
The number of associated search result；
If the number of the associated search result be greater than preset times threshold value, it is determined that it is described first search text sequence and
Second search text sequence is that the concurrent hits search text sequence set.
4. the method according to claim 1, wherein described determine each search text based on attention model
The Automobile driving probability for each field for including in sequence, comprising:
Deconsolidation process is carried out to each search text sequence, obtains at least one corresponding word of each search text sequence
Section；
Determine the initial Automobile driving probability of each field at least one described field；
According to each field in each search text sequence the number that occurs to the initial Automobile driving probability into
Row update processing, with the Automobile driving probability of determination each field.
5. according to the method described in claim 4, it is characterized in that, at least one field described in the determination each field just
Beginning Automobile driving probability, comprising:
Determine the part of speech of each field at least one described field；
According to the corresponding relationship of preset part of speech and Automobile driving probability, determine corresponding with the part of speech of each field initial
Automobile driving probability.
6. according to the method described in claim 4, it is characterized in that, at least one field described in the determination each field just
Beginning Automobile driving probability, comprising:
Determine the quantity of field in each search text sequence；
According to the quantity of field in each search text sequence, equal initial Automobile driving probability is set for each field,
Wherein, the sum of initial Automobile driving probability of each field is 1.
7. the method according to claim 1, wherein described hit search text sequence input for concurrent to be discriminated
The synonym discrimination model for introducing attention mechanism, to determine that the concurrent to be discriminated is hit in search text sequence
Synonym pair, comprising:
The concurrent to be discriminated is hit into search text sequence and inputs the synonym discrimination model for introducing attention mechanism, is obtained
The corresponding Automobile driving probability of each field in search text sequence is hit to the concurrent to be discriminated；
The field for determining that the corresponding Automobile driving probability of each field is greater than predetermined probabilities threshold value is synonym pair.
8. a kind of determining device of synonym, which is characterized in that described device includes:
Module is obtained, hits search text sequence set for obtaining concurrent, the concurrent, which is hit in search text sequence set, includes
Multiple concurrents hit search text sequence, and each concurrent hits each search text sequence correspondence in search text sequence, and there are associated
Search result；
Distribution module, for determining the attention point for each field for including in each search text sequence based on attention model
With probability；
Training module, for according to it is described it is each search text sequence in include each field Automobile driving probability to synonym
Discrimination model is trained, and obtains the synonym discrimination model for introducing attention mechanism；
Determining module, for by concurrent to be discriminated hit search text sequence input it is described introduce attention mechanism synonym sentence
Other model, to determine that the concurrent to be discriminated hits the synonym pair searched in text sequence.
9. a kind of server, which is characterized in that including processor and storage device, the processor is mutually interconnected with storage device
It connects, wherein the storage device is for storing computer program, and the computer program includes program instruction, the processor
It is configured for calling described program instruction, executes the method according to claim 1 to 7.
10. a kind of computer readable storage medium, which is characterized in that be stored with program in the computer readable storage medium and refer to
It enables, which is performed, for realizing the method according to claim 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910699704.5A CN110413737B (en) | 2019-07-29 | 2019-07-29 | Synonym determination method, synonym determination device, server and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910699704.5A CN110413737B (en) | 2019-07-29 | 2019-07-29 | Synonym determination method, synonym determination device, server and readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110413737A true CN110413737A (en) | 2019-11-05 |
| CN110413737B CN110413737B (en) | 2022-10-14 |
Family
ID=68364503
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910699704.5A Active CN110413737B (en) | 2019-07-29 | 2019-07-29 | Synonym determination method, synonym determination device, server and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110413737B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111046966A (en) * | 2019-12-18 | 2020-04-21 | 江南大学 | Image subtitle generating method based on measurement attention mechanism |
| CN111881255A (en) * | 2020-06-24 | 2020-11-03 | 百度在线网络技术（北京）有限公司 | Synonymy text acquisition method and device, electronic equipment and storage medium |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101226532A (en) * | 2007-12-28 | 2008-07-23 | 腾讯科技（北京）有限公司 | Method and system for extracting homoionym in network |
| CN102750282A (en) * | 2011-04-19 | 2012-10-24 | 北京百度网讯科技有限公司 | Synonym template mining method and device as well as synonym mining method and device |
| CN102760127A (en) * | 2011-04-26 | 2012-10-31 | 北京百度网讯科技有限公司 | Method, device and equipment for determining resource type based on extended text information |
| CN105095203A (en) * | 2014-04-17 | 2015-11-25 | 阿里巴巴集团控股有限公司 | Methods for determining and searching synonym, and server |
| US20160004766A1 (en) * | 2006-10-10 | 2016-01-07 | Abbyy Infopoisk Llc | Search technology using synonims and paraphrasing |
| CN105279252A (en) * | 2015-10-12 | 2016-01-27 | 广州神马移动信息科技有限公司 | Related word mining method, search method and search system |
| CN105426379A (en) * | 2014-10-22 | 2016-03-23 | 武汉理工大学 | Keyword weight calculation method based on position of word |
| CN106202382A (en) * | 2016-07-08 | 2016-12-07 | 南京缘长信息科技有限公司 | Link instance method and system |
| US20170078308A1 (en) * | 2015-09-16 | 2017-03-16 | Mastercard International Incorporated | Systems and Methods for Use in Scoring Entities in Connection With Preparedness of the Entities for Cyber-Attacks |
| CN107491518A (en) * | 2017-08-15 | 2017-12-19 | 北京百度网讯科技有限公司 | Method and apparatus, server, storage medium are recalled in one kind search |
| CN107679030A (en) * | 2017-09-04 | 2018-02-09 | 北京京东尚科信息技术有限公司 | Method and apparatus based on user's operation behavior data extraction synonym |
| CN108052520A (en) * | 2017-11-01 | 2018-05-18 | 平安科技（深圳）有限公司 | Conjunctive word analysis method, electronic device and storage medium based on topic model |
| CN108509474A (en) * | 2017-09-15 | 2018-09-07 | 腾讯科技（深圳）有限公司 | Search for the synonym extended method and device of information |
| CN109241294A (en) * | 2018-08-29 | 2019-01-18 | 国信优易数据有限公司 | A kind of entity link method and device |
| CN109902144A (en) * | 2019-01-11 | 2019-06-18 | 杭州电子科技大学 | A kind of entity alignment schemes based on improvement WMD algorithm |
| CN109902154A (en) * | 2018-11-30 | 2019-06-18 | 华为技术有限公司 | Information processing method, device, service equipment and computer readable storage medium |
| CN109918661A (en) * | 2019-03-04 | 2019-06-21 | 腾讯科技（深圳）有限公司 | Synonym acquisition methods and device |
-
2019
- 2019-07-29 CN CN201910699704.5A patent/CN110413737B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160004766A1 (en) * | 2006-10-10 | 2016-01-07 | Abbyy Infopoisk Llc | Search technology using synonims and paraphrasing |
| CN101226532A (en) * | 2007-12-28 | 2008-07-23 | 腾讯科技（北京）有限公司 | Method and system for extracting homoionym in network |
| CN102750282A (en) * | 2011-04-19 | 2012-10-24 | 北京百度网讯科技有限公司 | Synonym template mining method and device as well as synonym mining method and device |
| CN102760127A (en) * | 2011-04-26 | 2012-10-31 | 北京百度网讯科技有限公司 | Method, device and equipment for determining resource type based on extended text information |
| CN105095203A (en) * | 2014-04-17 | 2015-11-25 | 阿里巴巴集团控股有限公司 | Methods for determining and searching synonym, and server |
| CN105426379A (en) * | 2014-10-22 | 2016-03-23 | 武汉理工大学 | Keyword weight calculation method based on position of word |
| US20170078308A1 (en) * | 2015-09-16 | 2017-03-16 | Mastercard International Incorporated | Systems and Methods for Use in Scoring Entities in Connection With Preparedness of the Entities for Cyber-Attacks |
| CN105279252A (en) * | 2015-10-12 | 2016-01-27 | 广州神马移动信息科技有限公司 | Related word mining method, search method and search system |
| CN106202382A (en) * | 2016-07-08 | 2016-12-07 | 南京缘长信息科技有限公司 | Link instance method and system |
| CN107491518A (en) * | 2017-08-15 | 2017-12-19 | 北京百度网讯科技有限公司 | Method and apparatus, server, storage medium are recalled in one kind search |
| CN107679030A (en) * | 2017-09-04 | 2018-02-09 | 北京京东尚科信息技术有限公司 | Method and apparatus based on user's operation behavior data extraction synonym |
| CN108509474A (en) * | 2017-09-15 | 2018-09-07 | 腾讯科技（深圳）有限公司 | Search for the synonym extended method and device of information |
| CN108052520A (en) * | 2017-11-01 | 2018-05-18 | 平安科技（深圳）有限公司 | Conjunctive word analysis method, electronic device and storage medium based on topic model |
| CN109241294A (en) * | 2018-08-29 | 2019-01-18 | 国信优易数据有限公司 | A kind of entity link method and device |
| CN109902154A (en) * | 2018-11-30 | 2019-06-18 | 华为技术有限公司 | Information processing method, device, service equipment and computer readable storage medium |
| CN109902144A (en) * | 2019-01-11 | 2019-06-18 | 杭州电子科技大学 | A kind of entity alignment schemes based on improvement WMD algorithm |
| CN109918661A (en) * | 2019-03-04 | 2019-06-21 | 腾讯科技（深圳）有限公司 | Synonym acquisition methods and device |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111046966A (en) * | 2019-12-18 | 2020-04-21 | 江南大学 | Image subtitle generating method based on measurement attention mechanism |
| CN111881255A (en) * | 2020-06-24 | 2020-11-03 | 百度在线网络技术（北京）有限公司 | Synonymy text acquisition method and device, electronic equipment and storage medium |
| CN111881255B (en) * | 2020-06-24 | 2023-10-27 | 百度在线网络技术（北京）有限公司 | Synonymous text acquisition method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110413737B (en) | 2022-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yao et al. | An improved LSTM structure for natural language processing | |
| US20200301954A1 (en) | Reply information obtaining method and apparatus | |
| CN109844743B (en) | Generating responses in automated chat | |
| Zhou et al. | End-to-end learning of semantic role labeling using recurrent neural networks | |
| Chen et al. | Learning a recurrent visual representation for image caption generation | |
| CN109840287A (en) | A kind of cross-module state information retrieval method neural network based and device | |
| CN110427463A (en) | Search statement response method, device and server and storage medium | |
| US9213936B2 (en) | Electronic brain model with neuron tables | |
| CN109992773B (en) | Word vector training method, system, device and medium based on multi-task learning | |
| US11729120B2 (en) | Generating responses in automated chatting | |
| CN111143576A (en) | Event-oriented dynamic knowledge graph construction method and device | |
| CN110032632A (en) | Intelligent customer service answering method, device and storage medium based on text similarity | |
| CN110532571A (en) | Text handling method and relevant apparatus | |
| CN110309511B (en) | Shared representation-based multitask language analysis system and method | |
| CN111046661A (en) | Reading understanding method based on graph convolution network | |
| CN110457718A (en) | A kind of document creation method, device, computer equipment and storage medium | |
| AU2021225262A1 (en) | Scene graph modification based on natural language commands | |
| CN107679225A (en) | A kind of reply generation method based on keyword | |
| CN112287656B (en) | Text comparison method, device, equipment and storage medium | |
| CN108846000A (en) | A kind of common sense semanteme map construction method and device based on supernode and the common sense complementing method based on connection prediction | |
| CN110413737A (en) | A kind of determination method, apparatus, server and the readable storage medium storing program for executing of synonym | |
| Xiong et al. | DGI: recognition of textual entailment via dynamic gate matching | |
| CN113095072B (en) | Text processing method and device | |
| CN110929532B (en) | Data processing method, device, equipment and storage medium | |
| CN110852066B (en) | Multi-language entity relation extraction method and system based on confrontation training mechanism |
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 |