CN108763206A - A method of quicksort is carried out to single text keyword - Google Patents
A method of quicksort is carried out to single text keyword Download PDFInfo
- Publication number
- CN108763206A CN108763206A CN201810491735.7A CN201810491735A CN108763206A CN 108763206 A CN108763206 A CN 108763206A CN 201810491735 A CN201810491735 A CN 201810491735A CN 108763206 A CN108763206 A CN 108763206A
- Authority
- CN
- China
- Prior art keywords
- vector
- numerical value
- value
- sequence
- single text
- 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
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/284—Lexical analysis, e.g. tokenisation or collocates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a kind of methods carrying out quicksort to single text keyword, which is characterized in that the method includes the following steps：S1：List text is selected and by single text conversion at corresponding graph model structure, then according to graph model structural generation candidate word adjacency matrix；S2：Use the value of power method grey iterative generation candidate word adjacency matrix for the approximation of the corresponding feature vector of 1 characteristic value；S3：Qualitative analysis is used in step s 2, and qualitative analysis is carried out to the feature vector of each power method grey iterative generation, generates partial ordering's vector；S4：A judgment threshold is set, the backward numerical value between the ordering vector of adjacent grey iterative generation twice is calculated, compares the size of backward numerical value backward numerical value corresponding with last round of iteration, while the backward numerical value of more last round of iteration and the size of judgment threshold；The method of the present invention can be such that iterative process restrains rapidly, can effectively reduce the time complexity of calculating, while have extraction accuracy height, the high feature of sequence correctness.
Description
Technical field
The invention belongs to natural language processing fields, include mainly the process for extracting and sorting for single text, especially
It is related to a kind of method carrying out quicksort to single text keyword.
Background technology
One of the target of natural language processing task is to simplify magnanimity document to indicate and effectively organize, so that people search
It is used with research.Specifically, it is how to indicate a document with one section of word, in short even several words, how in people
Propose quickly respond demand and provide more accurate response contents after lookup demand.Appoint on the basis of this target
Business is the keyword extraction task of document, and by the processing to document, algorithm generates keyword sequences, the keyword in sequence according to
It is secondary to form expression and dependence with respective document, with there are two types of the relevant data of this task：Keyword rank result and pass
The vector of key word indicates.
The several method of mainstream is in single text keyword Sorting task at present：Based on the method for lexical item frequency, it is based on
The combination of the method for candidate word position, the method for text based node graph model, and the above several method.
The method of text based node graph model, theoretical and realization approach have followed to ultra-large internet section
The method of point importance ranking is denoted as node graph model, and carry out using power method by carrying out model conversion to text
Iterative calculation；Judge whether iteration terminates by the value situation of change for weighing iterative vectorized, and exports corresponding keyword rank.
The power method alternative manner of node graph model has a completeness in theory, the research carried out at present to it also not mistake
The more model for focusing on method, parameter and its iterative process, and method is more focused in each natural language processing
The application of tasks in areas and combination with other algorithms, for this angle, the power method alternative manner of node graph model is more
Mostly it is counted as a kind of " atom method " to be applied in natural language processing task.
In actual keyword extraction task, the output of the node diagram Model Calculating Method of text is vector space mould
Type, wherein element represent corresponding word and its weight with value, and after sequence, the high preceding m word of weight is document
Keyword, the difference of this m keyword weight shows there is different representative degree to document each other；Meanwhile as one kind
" atom method ", quantitative output also can be easily combined with other algorithms by vector calculating, to generate more
The effective output result of adduction reason；In addition, for other tasks, quantifying for single document set of keywords and its weight is needed
It is denoted as inputting, preferably to carry out further work, such as the cluster of document, division, sentiment analysis etc..
And keyword vector value need not provided, it is only necessary to keyword sequences are provided, or even only need to provide keyword
In the task of set, such methods are required for higher time complexity, so that iterative process therein becomes less
It is applicable in.
Invention content
The main purpose of the present invention is to provide a kind of method carrying out quicksort to single text keyword, the present invention exists
Under the premise of ensureing higher keyword extraction precision, additionally it is possible to be carried to single text keyword with the completion of lower time complexity
It takes and sorting operation, specific technical solution is as follows：
A method of quicksort being carried out to single text keyword, the method includes the following steps：
S1：List text is selected and by single text conversion at corresponding graph model structure, then according to graph model structural generation
Candidate word adjacency matrix；
S2：Use the value of power method grey iterative generation candidate word adjacency matrix for the approximation of the corresponding feature vector of 1 characteristic value
Value；
S3：Qualitative analysis is used in step s 2, and qualitative analysis is carried out to the feature vector of each power method grey iterative generation, it is raw
At partial ordering's vector；
S4：A judgment threshold is set, the backward numerical value between the ordering vector of adjacent grey iterative generation twice is calculated, compares
The size of backward numerical value backward numerical value corresponding with last round of iteration, at the same the backward numerical value of more last round of iteration with sentence
The size of disconnected threshold value compares the size of backward numerical value and interpretation threshold value.
Further, in step sl, single text is considered as word packet first, then according to described in institute's predicate packet composition
The graph model thaumatropy is finally generated the candidate word adjacency matrix corresponding with the list text by graph model structure.
Further, the generating process of the candidate word adjacency matrix includes step：First, it is assumed that there is single text
T enables T={ C1,C2,...Cm, wherein CiFor specific sentence in T,sijFor certain candidate word, node
Figure figure G=(P, V), if there is word s ∈ P, s ∈ Ci；After carrying out lexical item cutting to single text T again, with candidate word it
Between co-occurrence probabilities be node diagram G in side tax weights, in this approach generate text graph model the adjacency matrix, by matrix
M=(wij) ∈ Rn × n expressions.
Further, in step s 4, if the backward numerical value is less than the corresponding backward numerical value of last round of iteration, while two
Person is both less than the judgment threshold, then stops iteration, exports the ordering vector, otherwise, repeats step S2~S4.
Further, the power method iteration includes step：First, if sequence vectorMake
With unit vector P (0)=eT=(1,1...1)T∈RnFor initial value, formula P (t+1)=M is usedT× P (t), t=0,1,2...
∞ is iterated calculating and to sequence vector P (t) assignment；Then, in every wheel iterative process and to current P (t) in member
Element carries out m simple selection sequence, and serial number and sequence of the element big m in P (t), sequence vector is generated with this before choosingCalculate backward numerical value K (Q (t), Q (t- of the vector to (Q (t), Q (t-1))
1) it) and with given threshold value ε is compared；Finally according to formula
The comparison result of K (Q (t), Q (t-1))≤K (Q (t-1), Q (t-2)) < ε judges whether to terminate iteration, if K (Q
(t), Q (t-1))≤K (Q (t-1), Q (t-2)) < ε set up, then output vector Q (t)；Otherwise, step S2~S4 is repeated.
Further, in the calculation formula K of the backward numerical value (Q (t), Q (t-1)), sequence σ and sequence τ is set, and
Enable the sequence serial number σ (i), the sequence serial number τ in sequence τ of number is i in the node diagram G node in sequence σ
(i), it can thus be concluded that functional expressionIn functional expression, when the mould of σ and τ are
When odd number, the value of MP is | σ |2- 1, when the mould of σ and τ is even number, the value of MP is | σ |2, for all sections appeared in σ and τ
Point all judges (i, j) value of this function is between [0,1].
The quick method that single text keyword is ranked up of the present invention, the graph model structure formed first according to text
Generate candidate word adjacency matrix, use the value of power method grey iterative generation candidate word adjacency matrix for the corresponding feature of 1 characteristic value
The approximation of vector, and the feature vector to generating every time carries out qualitative analysis, generates partial ordering's vector, finally calculates adjacent
Backward numerical value between the ordering vector of grey iterative generation twice carries out the backward numerical value of backward numerical value and last round of grey iterative generation
Compare, while the backward numerical value of last round of grey iterative generation being compared with the judgment threshold of setting, if epicycle backward numerical value is small
In last round of backward numerical value, while last round of backward numerical value is less than the judgment threshold of setting, then can stop iteration, output row
Sequence vector, otherwise re-starts iterative step to the Overall Steps process for meeting above-mentioned comparison condition；Compared with prior art, originally
Advantageous effect of the invention is：It is big compared to the existing power method alternative manner based on node graph model with lower time complexity
Width reduces iterations, generally reduces 10%~25% to the operation times for testing this paper；It can guarantee keyword sequences
Stability, it is ensured that keyword rank, extraction accuracy will not reduce.
Description of the drawings
Fig. 1 is the flow diagram signal of single text keyword sort method of the present invention；
Fig. 2 is the required data structure schematic diagram of single text generation adjacency matrix of the present invention.
Specific implementation mode
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described.
For existing power method, with | | P (t)-P (t-1) | |1≤ ε as decision condition be to it is iterative vectorized itself
The quantized values of numerical value compare, and can be referred to as the quantitative analysis to vectorial P (t) sequence, only each member in this sequence
When the sum of plain variable quantity absolute value is less than some threshold epsilon, it just can determine that iteration terminates.
And for vectorial P (t), wherein " magnitude relationship " between each element is likely to close by quantitative analysis
It is just had determined that when before lattice very early." magnitude relationship " between element, i.e., vectorial sequence stability will not
Generate variation because of continuing for iteration, or generate part very small variation (such as：For one have 500 it is non-
The sequence vector of weight word, the 299th word and the 300th the word magnitude relationship after certain iteration are exchanged after sequence), this
Shi Xiangliang enters " sequence is stablized " stage；In this stage, algorithm, which still also wants iteration many times, can just make to divide the quantitative of P (t)
Analysis is qualified, and for the task of actual value for not needing result vector, subsequent iteration is all useless；In this regard, this hair
It is bright to propose a kind of method that quicksort is carried out to single text keyword, qualitative analysis is introduced in power method iterative process, and
The convergent standard of iterative process is redefined with this, specifically sees Fig. 1 and Fig. 2.
Refering to fig. 1, it is illustrated that for a kind of flow for the method carrying out quicksort to single text keyword proposed by the present invention
Block diagram specifically includes the following steps：S1：Selected list text and by single text conversion at corresponding graph model structure, then basis
Graph model structural generation candidate word adjacency matrix；S2：Use the value of power method grey iterative generation candidate word adjacency matrix for 1 characteristic value
The approximation of corresponding feature vector；S3：Qualitative analysis is used in step s 2, to the feature vector of each power method grey iterative generation
Qualitative analysis is carried out, partial ordering's vector is generated；And S4：A judgment threshold is set, adjacent grey iterative generation twice is calculated
Backward numerical value between ordering vector, the size of backward numerical value backward numerical value corresponding with last round of iteration, simultaneously
The backward numerical value of more last round of iteration and the size of judgment threshold.
In addition, in step s 4, if the backward numerical value is less than the corresponding backward numerical value of last round of iteration, while the two is all
Less than judgment threshold, then stop iteration, export ordering vector, otherwise, repeats step S2~S4.
Specifically, in step sl, single text is considered as word packet first, graph model structure is then constituted according to word packet, most
Graph model thaumatropy is generated into candidate word adjacency matrix afterwards, in conjunction with Fig. 2, detailed process can be described as follows：
First, the single text T, wherein T={ C of setting one1,C2,...,Cm, CiFor specific sentence in T,sijFor certain candidate word,G=(P, V) is schemed, if there are word s ∈ P, s ∈ Ci；Then text is passed through
The processes such as cutting word, lexical item segmentation, convert single text to word packet model, i.e. matrix S, by the duplicate removal, statistics, meter to matrix S
The operations such as conditional probability are calculated, adjacency matrix M=(w are ultimately generatedij)∈Rn×n；Then create sequence vectorMake P
(0)=eT=(1,1...1)T∈Rn, it is iterated and calculates P (t+1)=MT× P (t), t=0,1,2... ∞；And to P (t) into
M simple selection sequences generate Q (t) before row, calculate
Whether K (Q (t), Q (t-1))≤K (Q (t-1), Q (t-2)) < ε are true, if if true, stopping iteration, exports Q
(t), otherwise repeat to iterate to calculate P (t+1)=MTThe step of × P (t), t=0,1,2... ∞ and to P (t) carry out before m
Simple selection sequence generates Q (t), calculates
Whether K (Q (t), Q (t-1))≤K (Q (t-1), Q (t-2)) < ε are genuine step.
Vectorial P (t)=(p for iterating to calculate generation each time of single text keyword sort method of the present inventioni)∈RnIn
The maximum preceding m element of value be according to value ranked up from big to small, and then obtained vectorWherein qi
=f (pi) be P (t) in i-th of element by from big to small sort after sequence Q (t) in ranking, i.e. injection f:P → R,
f(pi) > f (pj), then with the increase of t, we obtain sequence vector
P (t), t=0,1,2... ∞ have also obtained sequence vector Q (t), t=0,1,2... ∞.
Wherein, the specific calculating process of backward numerical value and its calculating function K (σ, τ) are described as follows：
First, if sequence serial number σ (i) of the node that number is i in node diagram G in sequence σ, the row in sequence τ
Sequence serial number τ (i), then there is function：
When the mould of σ and τ is odd number, the value of MP is | σ |2- 1,
When the mould of σ and τ is even number, the value of MP is | σ |2, all nodes appeared in σ and τ all sentence (i, j)
Fixed, this function is the value between 0 to 1.
The present invention is to the analysis of complexity that keyword is ranked up in single text：If single text needs to close first m
Key word is ranked up to take which part keyword as the keyword of document, then in each iterative process, needs to do m wheels
Selected and sorted, this needs additional m*n times to calculate, then, the backward numerical computations function K (Q (t), Q (t-1)) of qualitative analysis
Need the secondary calculating of additional mlog (m), it is assumed that need t1 end of iteration, count the data processing of constant term rank in, then execute calculation
Method to iteration terminates to need to carry out in total
t1[n2+mn+mlog(m)+const1]+const2Secondary calculating, wherein const1And const2Indicate two meanings not
Same constant.
The method that quicksort is carried out to single text keyword of the present invention, the graph model structure formed first according to text
Generate candidate word adjacency matrix, use the value of power method grey iterative generation candidate word adjacency matrix for the corresponding feature of 1 characteristic value
The approximation of vector, and the feature vector to generating every time carries out qualitative analysis, generates partial ordering's vector, finally calculates adjacent
The backward numerical value that adjacent two-wheeled iterates to calculate is compared to each other by the backward numerical value between the ordering vector of grey iterative generation twice, and
The two is compared with the judgment threshold, if the backward numerical value is less than the corresponding backward numerical value of last round of iteration, while two
Person is both less than the judgment threshold, then can stop iteration, exports the Overall Steps process of ordering vector；Compared with prior art,
Beneficial effects of the present invention are：With lower time complexity, the existing power method alternative manner based on node graph model is compared
Iterations are substantially reduced, 10%~25% generally is reduced to the operation times for testing this paper；It can guarantee keyword sequence
The stability of row, it is ensured that keyword rank, extraction accuracy will not reduce.
The foregoing is merely a prefered embodiment of the invention, the scope of the claims of the present invention is not intended to limit, although with reference to aforementioned reality
Applying example, invention is explained in detail, still can be to aforementioned each tool for those skilled in the art comes
Technical solution recorded in body embodiment is modified, or carries out equivalence replacement to which part technical characteristic.Every profit
The equivalent structure made of description of the invention and accompanying drawing content is directly or indirectly used in other related technical areas,
Similarly within scope of patent protection of the present invention.
Claims (6)
1. a kind of method carrying out quicksort to single text keyword, which is characterized in that the method includes the following steps：
S1：Select list text and by single text conversion at corresponding graph model structure, it is then candidate according to graph model structural generation
Word adjacency matrix；
S2：Use the value of power method grey iterative generation candidate word adjacency matrix for the approximation of the corresponding feature vector of 1 characteristic value；
S3：Qualitative analysis is used in step s 2, and qualitative analysis, generation office are carried out to the feature vector of each power method grey iterative generation
Portion's ordering vector；
S4：Set a judgment threshold, calculate the backward numerical value between the ordering vector of adjacent grey iterative generation twice, relatively described in
The size of backward numerical value backward numerical value corresponding with last round of iteration, while the backward numerical value of more last round of iteration and judging threshold
The size of value.
2. a kind of method carrying out quicksort to single text keyword according to claim 1, which is characterized in that in step
In rapid S1, single text is considered as word packet first, the graph model structure is then constituted according to institute's predicate packet, it finally will be described
Graph model thaumatropy generates the candidate word adjacency matrix corresponding with the list text.
3. a kind of method carrying out quicksort to single text keyword according to claim 2, which is characterized in that described
The generating process of candidate word adjacency matrix includes step：First, it is assumed that there is single text T, T={ C are enabled1,C2,...Cm,
Wherein, CiFor specific sentence in T,sijFor certain candidate word, node diagram G=(P, V), if there is word s
∈ P, then s ∈ Ci；After carrying out lexical item cutting to single text T again, using the co-occurrence probabilities between candidate word as in node diagram G
The tax weights on side generate the adjacency matrix of text graph model, by matrix M=(w in this approachij)∈Rn×nIt indicates.
4. a kind of method carrying out quicksort to single text keyword according to claim 3, which is characterized in that in step
In rapid S4, if the backward numerical value is less than the corresponding backward numerical value of last round of iteration, while the two is both less than the judgment threshold,
Then stop iteration, export the ordering vector, otherwise, repeats step S2~S4.
5. a kind of method carrying out quicksort to single text keyword according to claim 4, which is characterized in that
The power method iteration includes step：First, if sequence vectorIt uses unit vector P (0)
=eT=(1,1...1)T∈RnFor initial value, formula P (t+1)=M is usedT× P (t), t=0,1,2... ∞ are iterated meter
It calculates and to sequence vector P (t) assignment；Then, in every wheel iterative process and to current P (t) in element carry out m time it is simple
Single selected and sorted, serial number and sequence of the element big m in P (t), sequence vector is generated with this before choosingCalculate backward numerical value K (Q (t), Q (t- of the vector to (Q (t), Q (t-1))
1) it) and with given threshold value ε is compared；Finally according to the ratio of formula K (Q (t), Q (t-1))≤K (Q (t-1), Q (t-2)) < ε
Relatively result judges whether to terminate iteration, if K (Q (t), Q (t-1))≤K (Q (t-1), Q (t-2)) < ε are set up, then output vector Q
(t)；Otherwise, step S2~S4 is repeated.
6. a kind of method carrying out quicksort to single text keyword according to claim 5, which is characterized in that in institute
In the calculation formula K (Q (t), Q (t-1)) for stating backward numerical value, sequence σ and sequence τ is set, and enables numbering in the node diagram G and is
The node of i the sequence serial number σ (i) in sequence σ, the sequence serial number τ (i) in sequence τ, it can thus be concluded that functional expressionIn functional expression, when the mould of σ and τ is odd number, the value of MP is | σ
|2- 1, when the mould of σ and τ is even number, the value of MP is | σ |2, all nodes appeared in σ and τ all carry out (i, j)
Between the value whether early [0,1] for judging this function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810491735.7A CN108763206B (en) | 2018-05-22 | 2018-05-22 | Method for quickly sequencing keywords of single text |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810491735.7A CN108763206B (en) | 2018-05-22 | 2018-05-22 | Method for quickly sequencing keywords of single text |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108763206A true CN108763206A (en) | 2018-11-06 |
| CN108763206B CN108763206B (en) | 2022-04-05 |
Family
ID=64007703
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810491735.7A Active CN108763206B (en) | 2018-05-22 | 2018-05-22 | Method for quickly sequencing keywords of single text |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108763206B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101398814A (en) * | 2007-09-26 | 2009-04-01 | 北京大学 | Method and system for simultaneously abstracting document summarization and key words |
| CN101751425A (en) * | 2008-12-10 | 2010-06-23 | 北京大学 | Method for acquiring document set abstracts and device |
| CN103530402A (en) * | 2013-10-23 | 2014-01-22 | 北京航空航天大学 | Method for identifying microblog key users based on improved Page Rank |
| CN104462253A (en) * | 2014-11-20 | 2015-03-25 | 武汉数为科技有限公司 | Topic detection or tracking method for network text big data |
| CN104537835A (en) * | 2015-01-30 | 2015-04-22 | 北京航空航天大学 | Macroscopic view and microscopic view combined circular transportation reliability simulation method and system |
| CN104835181A (en) * | 2015-05-23 | 2015-08-12 | 温州大学 | Object tracking method based on ordering fusion learning |
| CN107464032A (en) * | 2017-06-23 | 2017-12-12 | 昆明理工大学 | A kind of online service measures of reputation method based on KendallTau distances |
-
2018
- 2018-05-22 CN CN201810491735.7A patent/CN108763206B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101398814A (en) * | 2007-09-26 | 2009-04-01 | 北京大学 | Method and system for simultaneously abstracting document summarization and key words |
| CN101751425A (en) * | 2008-12-10 | 2010-06-23 | 北京大学 | Method for acquiring document set abstracts and device |
| CN103530402A (en) * | 2013-10-23 | 2014-01-22 | 北京航空航天大学 | Method for identifying microblog key users based on improved Page Rank |
| CN104462253A (en) * | 2014-11-20 | 2015-03-25 | 武汉数为科技有限公司 | Topic detection or tracking method for network text big data |
| CN104537835A (en) * | 2015-01-30 | 2015-04-22 | 北京航空航天大学 | Macroscopic view and microscopic view combined circular transportation reliability simulation method and system |
| CN104835181A (en) * | 2015-05-23 | 2015-08-12 | 温州大学 | Object tracking method based on ordering fusion learning |
| CN107464032A (en) * | 2017-06-23 | 2017-12-12 | 昆明理工大学 | A kind of online service measures of reputation method based on KendallTau distances |
Non-Patent Citations (2)
| Title |
|---|
| CYNTHIA DWORK等: "Rank Aggregation Methods for the Web", 《WWW10》 * |
| 柳林青等: "一种基于TextＲank 的单文本关键字提取算法", 《计算机应用研究》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108763206B (en) | 2022-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11341330B1 (en) | Applied artificial intelligence technology for adaptive natural language understanding with term discovery | |
| CN106649260B (en) | Product characteristic structure tree construction method based on comment text mining | |
| CN104598611B (en) | The method and system being ranked up to search entry | |
| CN108647226B (en) | Hybrid recommendation method based on variational automatic encoder | |
| CN106547864A (en) | A kind of Personalized search based on query expansion | |
| Boyabatli et al. | Parameter selection in genetic algorithms | |
| CN107885749A (en) | Ontology extends the process knowledge search method with collaborative filtering Weighted Fusion | |
| CN105786799A (en) | Web article originality judgment method | |
| CN111582325B (en) | Multi-order feature combination method based on automatic feature coding | |
| CN111027776A (en) | Sewage treatment water quality prediction method based on improved long-short term memory LSTM neural network | |
| Gupte et al. | Role discovery in graphs using global features: Algorithms, applications and a novel evaluation strategy | |
| Bortnikova et al. | Search Query Classification Using Machine Learning for Information Retrieval Systems in Intelligent Manufacturing. | |
| CN114756656A (en) | Hydraulic engineering potential safety hazard description association rule mining method based on improved Apriori algorithm | |
| Shirazi et al. | EDA++: Estimation of distribution algorithms with feasibility conserving mechanisms for constrained continuous optimization | |
| CN112905906B (en) | Recommendation method and system fusing local collaboration and feature intersection | |
| CN116342167B (en) | Intelligent cost measurement method and device based on sequence labeling named entity recognition | |
| CN117252186A (en) | XAI-based information processing method, device, equipment and storage medium | |
| CN108763206A (en) | A method of quicksort is carried out to single text keyword | |
| Ramakrishnan et al. | Hypergraph based clustering for document similarity using FP growth algorithm | |
| CN116304726A (en) | Material similarity analysis method based on semantic library and knowledge graph | |
| CN111651512B (en) | Multisource heterogeneous commodity feature weight solving method and device based on semi-supervised learning | |
| CN114547294A (en) | Rumor detection method and system based on comprehensive information of propagation process | |
| Tang et al. | Automatic Classification of Matching Rules in Pattern Matching | |
| TWI730536B (en) | A system for question recommendation and a method thereof | |
| CN117556118B (en) | Visual recommendation system and method based on scientific research big data prediction |
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 |