CN108446408A - Short text summarization method based on PageRank - Google Patents

Short text summarization method based on PageRank Download PDF

Info

Publication number
CN108446408A
CN108446408A CN201810329318.2A CN201810329318A CN108446408A CN 108446408 A CN108446408 A CN 108446408A CN 201810329318 A CN201810329318 A CN 201810329318A CN 108446408 A CN108446408 A CN 108446408A
Authority
CN
China
Prior art keywords
word
item
item collection
state
collection
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
Application number
CN201810329318.2A
Other languages
Chinese (zh)
Other versions
CN108446408B (en
Inventor
曹斌
吴佳伟
王思超
范菁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN201810329318.2A priority Critical patent/CN108446408B/en
Publication of CN108446408A publication Critical patent/CN108446408A/en
Application granted granted Critical
Publication of CN108446408B publication Critical patent/CN108446408B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/34Browsing; Visualisation therefor
    • G06F16/345Summarisation for human users
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/279Recognition of textual entities
    • G06F40/289Phrasal analysis, e.g. finite state techniques or chunking

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

The invention relates to a short text summarization method based on PageRank. The method comprises the following steps: the method comprises the following steps: generating a frequent item set; modeling item set relations; and (4) calculating and abstracting an item set model. The method is based on the PageRank model, processes short texts of the event, forms a keyword set, simulates the importance degree of a plurality of sets through model calculation, and selects the most general set as the keyword abstract of the event. In practical application, main contents of events are clearly described, and the purposes of saving labor cost and improving working efficiency are achieved.

Description

A kind of short text method of abstracting based on PageRank
Technical field
The present invention relates to a kind of short text method of abstracting based on PageRank mainly solves the problems, such as have for same class Under conditions of description, the problem of how selecting representative problem to describe.More particularly to a kind of text item row Sequence method.In this way, relatively representative description can be selected in a variety of descriptions of same class problem.
Background introduction
It is well known that text is one of most important information carrier in life production.Therefore, in many fields, text Classification is paid much attention to and is widely used.Under normal conditions, it is believed that a certain class text is that special thing is corresponded to it The description of part, and this class text is typically some short texts, it is opposite that there is generality, contain abundant information.Therefore, to these Text carries out analyzing processing and makes a summary, and forms generality description, plays the role of very positive and meaning to production and living, in turn Also at a urgent problem to be solved.
Through investigation, existing short text method of abstracting has theme modeling, autoabstract, but above method still has Defect.Common theme modeler model, LDA models are relative complex, relatively poor to short text treatment effect, and accuracy is not high; There are mainly two types of patterns for autoabstract:One is extraction-type, i.e., select certain sentences as abstract from text;Another kind is understanding Formula, by understanding that context is made a summary.The mode of relative maturity is extraction-type at present, but often effect is barely satisfactory, and usually It applies in single long text scene rather than multiple short text language material scenes.
In the practical application scene of the present invention, need that all kinds of demands of enterprise are analyzed and made a summary, so as to Enterprise targetedly solves user's demand, improves service quality.And in actual operation, it is existing since user's demand amount is huge There is processing method to take excessive time and be easy error, leads to inefficiency, it is difficult to follow-up work is promoted, it finally cannot be timely Give user feedback handling result.Simultaneously human resources it is limited be difficult to allocate human hand participate in the work, it is effective there is an urgent need for one These complicated cumbersome operating process are automated with computer technology, reduce mistake, improve efficiency, save people by solution Power resource.
Invention content
The present invention will overcome the disadvantages mentioned above of the prior art, provide a kind of short text sports representative's property based on PageRank Degree sort method is formed by keyword set to treated demand and is ranked up, and chooses most recapitulative set Keyword as the demand describes, and so that analyst is clearly understood the main contents of demand, saves labour turnover and improve work Make efficiency.
The method of abstracting according to an aspect of the invention, there is provided a kind of short text based on PageRank is classified, including: Frequent item set generates;Item collection relationship modeling;Item collection model is calculated and is made a summary.
Step 1:Frequent item set generates
Including the following contents:Pending text is segmented and filtered, stop words is removed, replaces synonym, generating should The set of the initial word of text;After having handled all texts, the word frequency of each word is counted in text word segmentation result and according to word Frequency is ranked up all words;Word order inside text word segmentation result is adjusted, is arranged by word frequency descending;Given threshold MinSupport leaves out the word that word frequency in word segmentation result is less than the threshold value;Data knot based on frequent pattern tree (fp tree) (FP-tree) Structure generates frequent item set using frequent mode growing method (FP-growth).
Step 2:Item collection relationship modeling
Need the analysis by data to count and carry out simple computation, build PageRank relational models, it is specific comprising with Lower step:
Step 2.1:Initialize item collects weights
The frequent item set sum n of a kind of problem is generated in statistic procedure 1, statistical items concentrate the word frequency tf of each wordi, i ∈ [1, N], collecting terms concentrate the case where containing word, and the initial weight for counting each item collection in set of computations is as follows:
What i.e. item collection contained word and its word frequency product is accumulated in accounting in total word frequency.
And then the initial weight vector P gathered0={ p1,p2,…,pn}T
Step 2.2:Build state transition probability matrix
Because there is the word of overlapping in set between each frequent item set, and the purpose of this method be just by structure figures come Association between frequent item set is described.Therefore the word number of the intersection by frequent item set in set of computations between any two is come generation Numeralization relationship between corresponding two frequent item sets of table.In the digraph that all item collections i.e. in set are constituted, calculate Its side right value.Item collection can be considered as to a specific state, the physical significance of side right value is i.e. another to be converted to by a certain state Shape probability of state, i.e. transition probability.
For each item collection SiAnd Sj, there is intersection term vector Xij={ xi1,xi2,…,xin}T, wherein xijIndicate item collection Si With item collection SjIntersection word word frequency, as i=j value be 0, and then constitute matrix W (because weigh object be all frequent episodes Collection, therefore tie up matrix for n):
WhereinThat is item collection SiWith item collection SjIntersection word frequency to item collection SiIt is handed over remaining all item collection The ratio for collecting word frequency sum represents the side right value between each item collection with this, forms state transition probability matrix.
Step 2.3:Correct state transition probability matrix
It is an object of the present invention to which a representative item collects weights are calculated by model.In the above process In it can be seen that, since there are the associations of intersection word between item collection, it is not difficult to visualize in calculating process, the weights meeting of item collection Variation is generated according to the weights of other item collections.Therefore it needs to calculate correction model so that a stable value can be calculated.
According to Markov convergence theorem, when meeting the following conditions:
1. finite state number;2. fixed state transition probability;
3. can change in any way between state;4. state branch mode is not unique;
The markoff process will converge to an equilibrium state, and this is balanced unique.
The present invention in the case where meeting the following conditions,
①:State number is item collection number n;②:State transition probability matrix is determined and be will not change by item collection;④: Collection intersection constituted when being all two-way, in the case that up to there are many branch mode between each state;There is still a need for progress It corrects, to meet condition 3..
Consider special circumstances, when the intersection of a certain item collection and remaining item collection is sky, i.e., can not build side, herein should Item collection is known as the item collection of isolated state.Then when the item collection is accessed, state can not be shifted.To adapt to this case, into One step correction matrix W is W1
From the perspective of scheming, the modified physical significance be so that figure be connection, meet condition 3.;
Wherein α is empirical value, the probability that isolated state in an iterative process carries out state transfer is represented, in combination with practical feelings Condition is voluntarily corrected.E is unit matrix, therefore the latter half of formula represents the probability for directly accessing the isolated state.
Step 3:Model calculates
Given number of iterations max_iter and threshold value min_diff.According to Pn+1=W1Pn, initial value Pn=P0It carries out Operation.When twice, iteration result difference is less than threshold value, i.e. Pn+1-Pn<When min_diff;Or work as iterations and be more than predetermined number of times, i.e., n>Max_iter, you can be considered as operation result convergence, ranking can be exported on demand.
The present invention is based on PageRank models to form keyword set and merge through model calculating to event short essay present treatment The significance level of these multiple set is simulated, most recapitulative set is chosen and makes a summary as the keyword of the event. In practical applications, the main contents for clearly describing event achieve the purpose that save labour turnover and improve working efficiency.
It is an advantage of the invention that:The present invention can be automatically to the relevant assigned short text set symphysis of a kind of event at multiple candidates' Event keyword is made a summary, and by PageRank models, calculates the significance level of each event summary, is finally obtained and is most summarized Property event keyword abstract.Obtained event keyword abstract can clearly describe the main contents of event, and the present invention Computational efficiency it is higher, fed back through practical application, the event description that the present invention obtains can help people to be best understood from event, to reach To the purpose to save labour turnover.
Description of the drawings
Fig. 1 is that the present invention is based on the overall flow figures of the short text of PageRank classification one example of method of abstracting.
Fig. 2 is that the present invention is based on the flows of the short text of PageRank classification one generation frequent item set example of method of abstracting Figure.
Fig. 3 is that the present invention is based on the flow charts that the short text of PageRank classification method of abstracting initializes weights.
Fig. 4 is that the present invention is based on the physics that the short text of PageRank classification method of abstracting builds state transition probability matrix Meaning schematic diagram.
Fig. 5 is that the present invention is based on the physics that the short text of PageRank classification method of abstracting corrects state transition probability matrix Meaning schematic diagram.
Fig. 6 is that the present invention is based on the short text of PageRank classification method of abstracting model calculation flow charts.
Specific implementation mode
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 technology description to be briefly described, it should be apparent that, the one of the only present invention is described below A little embodiments for those of ordinary skill in the art without having to pay creative labor, can also be according to this A little embodiments obtain other embodiments.
An example of the present invention is to carry out short text work order using the short text classification method of abstracting based on PageRank Abstract.
The schematic diagram of one example of the short text classification method of abstracting based on PageRank, referring to Fig. 1, including following Step:
S101, frequent item set generate;
S102, set relation model foundation;
Result is calculated in S103, model;
101 step is specially:Pending text is segmented and filtered, stop words is removed, replaces synonym, it is raw At the set of the initial word of the text;After having handled all texts, count in text word segmentation result the word frequency of each word and according to All words are ranked up according to word frequency;Word order inside text word segmentation result is adjusted, is arranged by word frequency descending;Given threshold MinSupport leaves out the word that word frequency in word segmentation result is less than the threshold value;Data knot based on frequent pattern tree (fp tree) (FP-tree) Structure generates frequent item set using frequent mode growing method (FP-growth).
102 step is specially:
S201, initialization item collects weights;
S202, structure state transition probability matrix;
S203, state transition probability matrix is corrected;
201 step is specially:
The frequent item set sum n of a kind of problem is generated in statistic procedure 1, statistical items concentrate the word frequency tf of each wordi, i ∈ [1, N], collecting terms concentrate the case where containing word, and the initial weight for counting each item collection in set of computations is as follows: What i.e. item collection contained word and its word frequency product is accumulated in accounting in total word frequency.And then the initial weight vector P gathered0={ p1, p2,…,pn}T
202 step is specially:
According to the intersection between each frequent item set in set, represented between corresponding two frequent item sets with intersection situation Numeralization relationship, build relational matrix.
For each item collection SiAnd Sj, there is intersection term vector Xij={ xi1,xi2,…,xin}T, wherein xijIndicate item collection Si With item collection SjIntersection word word frequency, as i=j value be 0, and then constitute matrix W (because weigh object be all frequent episodes Collection, therefore tie up matrix for n):
WhereinThat is item collection SiWith item collection SjIntersection word frequency to item collection SiIt is handed over remaining all item collection The ratio for collecting word frequency sum represents the side right value between each item collection with this, forms state transition probability matrix.
203 step is specially:
Due to the index alternatively made a summary it is an object of the invention to acquire a stable weights, while this model accords with The convergent partial condition of Markov is closed, needing to be modified makes its Markov convergence theorem, to adapt to the purpose of the present invention.
Consider special circumstances, it is sky to gather intersection with remaining when a certain set, i.e., side can not be built, herein by the collection It is collectively referred to as isolated state.Then when the set is accessed, state can not be shifted.To adapt to this case, further correct Matrix W is W1
From the perspective of scheming, the modified physical significance be so that figure be connection, meet condition 3.;
Wherein α is empirical value, the probability that isolated state in an iterative process carries out state transfer is represented, in combination with practical feelings Condition is voluntarily corrected.
103 step is specially:Given number of iterations max_iter and threshold value min_diff.According to Pn+1=W1Pn, Initial value Pn=P0Carry out operation.When twice, iteration result difference is less than threshold value, i.e. Pn+1-Pn<When min_diff;Or when iteration time Number is more than predetermined number of times, i.e. n>Max_iter, you can be considered as operation result convergence, ranking can be exported on demand.

Claims (1)

  1. The method of abstracting 1. a kind of short text based on PageRank is classified, includes the following steps:
    Step 1:Frequent item set generates;
    Including the following contents:Pending text is segmented and filtered, stop words is removed, replaces synonym, generates the text Initial word set;After having handled all texts, the word frequency of each word is counted in text word segmentation result and according to word frequency pair All words are ranked up;Word order inside text word segmentation result is adjusted, is arranged by word frequency descending;Given threshold minSupport, is deleted Word frequency in word segmentation result is gone to be less than the word of the threshold value;Based on the data structure of frequent pattern tree (fp tree) FP-tree, increased using frequent mode Long method FP-growth generates frequent item set;
    Step 2:Item collection relationship modeling;
    It needs the analysis by data to count and carries out simple computation, build PageRank relational models, include specifically following step Suddenly:
    Step 2.1:Initialize item collects weights;
    The frequent item set sum n of a kind of problem is generated in statistic procedure 1, statistical items concentrate the word frequency tf of each wordi, i ∈ [1, n], knot The case where containing word in item collection is closed, the initial weight for counting each item collection in set of computations is as follows:
    What i.e. item collection contained word and its word frequency product is accumulated in accounting in total word frequency;
    And then the initial weight vector P gathered0={ p1,p2,…,pn}T
    Step 2.2:Build state transition probability matrix;
    Because there is the word of overlapping in set between each frequent item set, and the purpose of this method is just to describe by structure figures Association between frequent item set;Therefore the word number of intersection by frequent item set in set of computations between any two represents phase Answer the numeralization relationship between two frequent item sets;In the digraph that all item collections i.e. in set are constituted, its side is calculated Weights;Item collection can be considered as to a specific state, the physical significance of side right value is to be converted to another state by a certain state Probability, i.e. transition probability;
    For each item collection SiAnd Sj, there is intersection term vector Xij={ xi1,xi2,…,xin}T, wherein xijIndicate item collection SiWith item Collect SjIntersection word word frequency, as i=j value be 0, and then constitute matrix W (because weigh object be all frequent item sets, Therefore tie up matrix for n):
    WhereinThat is item collection SiWith item collection SjIntersection word frequency to item collection SiWith remaining all item collection intersection word The ratio of frequency sum represents the side right value between each item collection with this, forms state transition probability matrix;
    Step 2.3:Correct state transition probability matrix;
    Since there are the associations of intersection word between item collection, it is not difficult to visualize in calculating process, the weights of item collection can be according to it The weights of his item collection and generate variation;Therefore it needs to calculate correction model so that a stable value can be calculated;
    According to Markov convergence theorem, when meeting the following conditions:
    1. finite state number;2. fixed state transition probability;
    3. can change in any way between state;4. state branch mode is not unique;
    The markoff process will converge to an equilibrium state, and this is balanced unique;
    In the case where meeting the following conditions,
    ①:State number is item collection number n;②:State transition probability matrix is determined and be will not change by item collection;④:Item collection is handed over Collection is constituted when being all two-way, in the case that up to there are many branch modes between each state;There is still a need for being modified, To meet condition 3.;
    Consider special circumstances, when the intersection of a certain item collection and remaining item collection is sky, i.e., side can not be built, herein by the item collection The referred to as item collection of isolated state;Then when the item collection is accessed, state can not be shifted;To adapt to this case, further Correction matrix W is W1
    From the perspective of scheming, the modified physical significance be so that figure be connection, meet condition 3.;
    Wherein α is empirical value, represents the probability that isolated state in an iterative process carries out state transfer, certainly in combination with actual conditions Row is corrected;E is unit matrix, therefore the latter half of formula represents the probability for directly accessing the isolated state.
    Step 3:Item collection model is calculated and is made a summary;
    Given number of iterations max_iter and threshold value min_diff;According to Pn+1=W1Pn, initial value Pn=P0Carry out operation; When twice, iteration result difference is less than threshold value, i.e. Pn+1-Pn<When min_diff;Or when iterations are more than predetermined number of times, i.e. n> Max_iter, you can be considered as operation result convergence, ranking can be exported on demand.
CN201810329318.2A 2018-04-13 2018-04-13 Short text summarization method based on PageRank Active CN108446408B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810329318.2A CN108446408B (en) 2018-04-13 2018-04-13 Short text summarization method based on PageRank

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810329318.2A CN108446408B (en) 2018-04-13 2018-04-13 Short text summarization method based on PageRank

Publications (2)

Publication Number Publication Date
CN108446408A true CN108446408A (en) 2018-08-24
CN108446408B CN108446408B (en) 2021-04-06

Family

ID=63199842

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810329318.2A Active CN108446408B (en) 2018-04-13 2018-04-13 Short text summarization method based on PageRank

Country Status (1)

Country Link
CN (1) CN108446408B (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109739953A (en) * 2018-12-30 2019-05-10 广西财经学院 The text searching method extended based on chi-square analysis-Confidence Framework and consequent
CN110533194A (en) * 2019-03-25 2019-12-03 东北大学 The optimization method of maintenance system construction
US10579894B1 (en) * 2019-07-17 2020-03-03 Capital One Service, LLC Method and system for detecting drift in text streams
US10657416B1 (en) 2019-07-17 2020-05-19 Capital One Services, Llc Method and system for detecting drift in image streams
CN111797945A (en) * 2020-08-21 2020-10-20 成都数联铭品科技有限公司 Text classification method
CN111984688A (en) * 2020-08-19 2020-11-24 中国银行股份有限公司 Method and device for determining business knowledge association relation
CN112256801A (en) * 2020-10-10 2021-01-22 深圳力维智联技术有限公司 Method, system and storage medium for extracting key entities in entity relationship graph
CN112883080A (en) * 2021-02-22 2021-06-01 重庆邮电大学 UFIM-Matrix algorithm-based improved uncertain frequent item set marketing data mining algorithm
CN116777525A (en) * 2023-06-21 2023-09-19 深圳市创致联创科技有限公司 Popularization and delivery system based on group optimization algorithm

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090169097A1 (en) * 2007-12-30 2009-07-02 Jianguo Li Markov stationary color descriptor
CN101727437A (en) * 2009-11-26 2010-06-09 上海大学 Method for computing importance degree of events in text set
CN102043851A (en) * 2010-12-22 2011-05-04 四川大学 Multiple-document automatic abstracting method based on frequent itemset
CN103699611A (en) * 2013-12-16 2014-04-02 浙江大学 Microblog flow information extracting method based on dynamic digest technology

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090169097A1 (en) * 2007-12-30 2009-07-02 Jianguo Li Markov stationary color descriptor
CN101727437A (en) * 2009-11-26 2010-06-09 上海大学 Method for computing importance degree of events in text set
CN102043851A (en) * 2010-12-22 2011-05-04 四川大学 Multiple-document automatic abstracting method based on frequent itemset
CN103699611A (en) * 2013-12-16 2014-04-02 浙江大学 Microblog flow information extracting method based on dynamic digest technology

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ELENA BARALIS: "GraphSum: Discovering correlations among multiple terms for graph-based summarization", 《ELSEVIER》 *
林莉媛: "基于PageRank的中文多文档文本情感摘要", 《中文信息学报》 *

Cited By (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109739953A (en) * 2018-12-30 2019-05-10 广西财经学院 The text searching method extended based on chi-square analysis-Confidence Framework and consequent
CN109739953B (en) * 2018-12-30 2021-07-20 广西财经学院 Text retrieval method based on chi-square analysis-confidence framework and back-part expansion
CN110533194A (en) * 2019-03-25 2019-12-03 东北大学 The optimization method of maintenance system construction
US10657416B1 (en) 2019-07-17 2020-05-19 Capital One Services, Llc Method and system for detecting drift in image streams
US10579894B1 (en) * 2019-07-17 2020-03-03 Capital One Service, LLC Method and system for detecting drift in text streams
US11138458B2 (en) 2019-07-17 2021-10-05 Capital One Services, Llc Method and system for detecting drift in text streams
US11386286B2 (en) 2019-07-17 2022-07-12 Capital One Services, Llc Method and system for detecting drift in image streams
US11694457B2 (en) 2019-07-17 2023-07-04 Capital One Services, Llc Method and system for detecting drift in image streams
CN111984688A (en) * 2020-08-19 2020-11-24 中国银行股份有限公司 Method and device for determining business knowledge association relation
CN111984688B (en) * 2020-08-19 2023-09-19 中国银行股份有限公司 Method and device for determining business knowledge association relationship
CN111797945A (en) * 2020-08-21 2020-10-20 成都数联铭品科技有限公司 Text classification method
CN111797945B (en) * 2020-08-21 2020-12-15 成都数联铭品科技有限公司 Text classification method
CN112256801A (en) * 2020-10-10 2021-01-22 深圳力维智联技术有限公司 Method, system and storage medium for extracting key entities in entity relationship graph
CN112256801B (en) * 2020-10-10 2024-04-09 深圳力维智联技术有限公司 Method, system and storage medium for extracting key entity in entity relation diagram
CN112883080A (en) * 2021-02-22 2021-06-01 重庆邮电大学 UFIM-Matrix algorithm-based improved uncertain frequent item set marketing data mining algorithm
CN112883080B (en) * 2021-02-22 2022-10-18 重庆邮电大学 UFIM-Matrix algorithm-based improved uncertain frequent item set marketing data mining algorithm
CN116777525A (en) * 2023-06-21 2023-09-19 深圳市创致联创科技有限公司 Popularization and delivery system based on group optimization algorithm
CN116777525B (en) * 2023-06-21 2024-06-28 深圳市创致联创科技有限公司 Popularization and delivery system based on group optimization algorithm

Also Published As

Publication number Publication date
CN108446408B (en) 2021-04-06

Similar Documents

Publication Publication Date Title
CN108446408A (en) Short text summarization method based on PageRank
EP3605358A1 (en) Olap precomputed model, automatic modeling method, and automatic modeling system
US10902022B2 (en) OLAP pre-calculation model, automatic modeling method, and automatic modeling system
CN110209807A (en) A kind of method of event recognition, the method for model training, equipment and storage medium
CN109960763B (en) Photography community personalized friend recommendation method based on user fine-grained photography preference
CN104700190B (en) One kind is for project and the matched method and apparatus of professional
EP3131021A1 (en) Hybrid data storage system and method and program for storing hybrid data
CN103488662A (en) Clustering method and system of parallelized self-organizing mapping neural network based on graphic processing unit
CN105005589A (en) Text classification method and text classification device
CN106021364A (en) Method and device for establishing picture search correlation prediction model, and picture search method and device
CN103714084A (en) Method and device for recommending information
CN110148023A (en) The electric power integral Method of Commodity Recommendation and system that logic-based returns
CN106874292A (en) Topic processing method and processing device
CN103257921A (en) Improved random forest algorithm based system and method for software fault prediction
CN114647465B (en) Single program splitting method and system for multi-channel attention map neural network clustering
CN108665148B (en) Electronic resource quality evaluation method and device and storage medium
CN106600067A (en) Method and device for optimizing multidimensional cube model
CN109857457B (en) Function level embedding representation method in source code learning in hyperbolic space
CN103324765A (en) Multi-core synchronization data query optimization method based on column storage
CN107046557A (en) The intelligent medical calling inquiry system that dynamic Skyline is inquired about under mobile cloud computing environment
CN109255012A (en) A kind of machine reads the implementation method and device of understanding
CN106570173A (en) High-dimensional sparse text data clustering method based on Spark
CN115795131A (en) Electronic file classification method and device based on artificial intelligence and electronic equipment
CN115757735A (en) Intelligent retrieval method and system for power grid digital construction result resources
CN115270921A (en) Power load prediction method, system and storage medium based on combined prediction model

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