CN102222075A - Tree-structure-based language bank compression method and system - Google Patents
Tree-structure-based language bank compression method and system Download PDFInfo
- Publication number
- CN102222075A CN102222075A CN2010101646361A CN201010164636A CN102222075A CN 102222075 A CN102222075 A CN 102222075A CN 2010101646361 A CN2010101646361 A CN 2010101646361A CN 201010164636 A CN201010164636 A CN 201010164636A CN 102222075 A CN102222075 A CN 102222075A
- Authority
- CN
- China
- Prior art keywords
- subtree
- tree
- node
- compression
- storage unit
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a data compression method and a data compression system, in particular to a tree-structure-based language bank compression method and a tree-structure-based language bank compression system. The method comprises the following steps of: A, traversing a main tree stored in a main tree storage unit, and selecting a sub-tree with the highest compression ratio by using a searching unit; B, connecting a sub-tree example searched by the step A to a sub-tree node of a sub-tree set storage unit, endowing the sub-tree node with a current number, replacing the node the sub-tree and similar thereof on the main tree, and adding 1 to a sub-tree node number counter; and C, cyclically executing the step A until the sub-tree cannot be found. By the method and the system, the technical problem that the conventional directed acyclic graph (DAWG) compression method is low in efficiency and unsuitable for the compression and rapid searching of a large database is mainly solved.
Description
Technical field
The present invention relates to data compression method and system, particularly a kind of language library compression method and system based on tree construction.
Background technology
Tree construction depends on the equivalence conversion of general tree and binary tree and deposit tree node by deep search in storer in the realization on the carrier.The former be used to set the data structure of node and the latter to this structure optimization and make it to move towards practical application.For example from vocabulary add, adding, added, the general tree that adds forms is converted to the data structure that binary tree can get egress.From general tree (Fig. 1) be:＜KP to the data structure that corresponding binary tree (Fig. 2) can get node〉＜BP〉＜EOW〉＜L 〉.This binary tree is deposited (Fig. 3) by deep search, can optimize the node data structure and be:＜K＜B〉＜EOW〉＜L 〉, saved the pointer expense.But to cross all child node of this node from certain node to its brotgher of node, such as passing through (i) all child node (n), (g) from node (i) to the brotgher of node (e).Can certainly directly generate binary tree from dictionary.Because the longest vocabulary is relevant and total amount speech in width and the dictionary is relevant in the degree of depth of language library tree construction and this storehouse, so the characteristics of language library tree construction are that degree of depth finite width is very big.Along with the increasing of vocabulary is all the more so.At last, the method that engine is realized on carrier is to be used as storehouse as workbench with some arrays or vector, enter from the appointed place of tree, by depth-first search scanning tree node, it is stacked and find out word or the vocabulary that satisfies specified requirements to deposit the node of coupling or pointer information in.
Definition:
＜K〉be the child node zone bit, mean Kid;
＜B〉be brotgher of node zone bit, mean Brother;
＜EOW〉be the speech zone bit, mean End of Word;
＜L〉be letter, mean Letter;
＜ST〉be the subtree zone bit, mean Sub Tree;
＜pK〉be offspring's pointer zone bit;
＜pB〉be fraternal pointer zone bit;
[] represents that the content in this bracket does not occur or occurs at most 1 time;
{ } represents that the content in this bracket does not occur or occurs repeatedly;
＜Pointer 〉=＜1, * * * * * * *,＜0, * * * * * * *, the pointer definition of forming by one or more bytes.＜0/1, * * * * * * * expression byte (Byte), * expression binary digit (bit).The byte of wherein representing the back when the first place is 1 still is this pointer part, and to the last a byte first place is till 0 o'clock; All * as to have formed this pointer value.
＜KP〉be the child node pointer;＜Pointer〉structure.
＜BP〉be brotgher of node pointer;＜Pointer〉structure.
Existing DAWG compression method based on tree construction is applied on its text input method product eZiText by Zi Corp.Its node data structure is＜K〉＜B〉＜EOW〉＜L〉＜pK〉[＜KP 〉]＜pB〉[＜BP 〉].Its basic generating principle is for beginning from left to right from the bottom leaf node of tree or from the father node of leaf node, reduction progressively from top to bottom: seek subtree example (first sub tree), that will find thereafter then similarly removes its all nodes and with this example of pointed, thereby reaches the purpose of database compression.
Illustrate the compression process of directed acyclic graph compression method (DAWG) by following (table 1) one group of vocabulary:
| convention | congressesional | extremist | sensation |
| conventional | congressesionally | extremists | sensational |
| conventionalism | essential | extremities | sensationalise |
| conventionalist | essentialism | extremity | sensationally |
| conventionalists | essentialist | secularism | vocation |
| conventionalities | essentialists | secularist | vocational |
| conventionality | essentialities | secularists | vocationalisms |
| conventionally | essentiality | secularities | vocationally |
| congressesion | extremism | secularity |
Its binary tree structure (Fig. 4) has 118 nodes, and the data structure of exceeding with 32 letters is an example, holds 3 zone bits and 1 letter:＜K with a byte〉＜B〉＜EOW〉＜L 〉, need 118 bytes altogether.
The compression process of DAWG is from the second from the bottom layer of bottom reduction (Fig. 5), reduction (Fig. 6) up gradually again, so repeatedly up end product (Fig. 7), be compressible to 83 nodes and 6 pointers.Press DAWG node definition:＜K〉＜B〉＜EOW〉＜L〉＜pK〉[＜KP 〉]＜pB〉[＜BP 〉], each node needs 5/4 byte, and each pointer needs 110 bytes thus altogether by a byte after the DAWG compression here, has compressed 7% of elite tree.
Limitation by discovering that there is theoretic major defect in this compression method and uses is in particular in:
(1) ending of DAWG compression method can only the compressed tree structure on design concept also is the suffix portion of vocabulary, the vast zone in the middle of can not the compressed tree structure, and also this design concept can't be improved in essence.
(2) each subtree example (first sub tree) is fixed in the dictionary tree construction, all pointers from its its first node of similar sensing are cost with the address of this node in the storehouse, if address big pointer more takes up space also big more. after example leans in tree relatively, then certainly will increase all and point to the cost of its pointer and cause ratio of compression to descend, in addition invalid.For example when dictionaries such as compression Arabic, Dard, just there is this situation, particularly evident during compression Dard dictionary.
(3) the different language property difference is very big and the expectation ratio of compression is also different.Subtree example node number, similar number and pointer size all are the key factors that influences ratio of compression.Under the situation of different language even the different vocabularies of same languages, not all subtree ex hoc genus anne all should or be necessary by reduction.Be not that each subtree example all should or be necessary by similar reduction, both do not reached the expection ratio of compression, also increased the search burden.Obviously, the DAWG compression method is too single, and is difficult to optimize.
(4) node among the DAWG need be taken a lot of aspects into account, except original＜child node zone bit〉＜brotgher of node zone bit〉＜the speech zone bit〉＜letter 〉, also will increase＜offspring's pointer zone bit＜fraternal pointer zone bit 〉, and the offspring's pointer that when it is not 0, follows closely thereafter and or fraternal pointer be byte or multibyte, in search, increased the traversal cost of engine to node.Therefore and since the node data structure too complexity cause the DAWG structure too fat to move, thereby directly influenced the search speed of engine.
In sum, DAWG compression method efficient is high and be not suitable for the compression of large database and search fast.
Summary of the invention
The object of the present invention is to provide a kind of language library compression method and system based on tree construction, it is high and be not suitable for compression and Fast search technique problem to large database mainly to solve existing DAWG compression method efficient.
For achieving the above object, the present invention is achieved in that
A kind of language library compression method based on tree construction is characterized in that it comprises the steps:
The master that A deposits in the main tree storage unit by the search unit traversal sets, and selects the best subtree of ratio of compression;
The subtree example that B searches steps A is connected on the subtree node of subtree set storage unit, gives this subtree node and substitutes the appearance that main tree goes up this subtree ex hoc genus anne with this node with current numbering, and then subtree node serial number counter is added 1;
C circulation execution in step A is till can't searching out subtree.
Described language library compression method based on tree construction, the system of selection that it is characterized in that described optimum compression ratio subtree are the subtrees of choosing the similar subtree number of subtree example node number * end value maximum.
Described language library compression method based on tree construction is characterized in that described subtree comprises the No way out subtree and the outlet subtree is arranged.
A kind of enforcement is the language library compressibility based on tree construction of method as mentioned above, it is characterized in that it comprises:
Be used for selecting the search unit of the best subtree of ratio of compression in main tree storage unit;
Be used to deposit the main tree storage unit of main tree;
Be used to deposit the subtree set storage unit of subtree, this subtree set storage unit comprises a subtree node serial number counter;
Be used to control search, subtree storage and the alternative scheduler module of subtree node of optimum compression ratio subtree.
Compression method of the present invention is with the beneficial effect that existing DAWG compression method is compared:
1, the address overhead of pointing to subtree in main tree has been alleviated in the disengaging of subtree from main tree in the inventive method, also makes the succinct scale down of main tree construction.Thereby for scanning and dynamic link lay the first stone fast.The dynamic link mode is complementary with the subtree compression method for the present invention's initiative, makes search procedure flexible and efficient.
2, subtree has both simultaneously and has or No way out has improved similar expansion face in the inventive method, and then has also improved the compressibility of tree construction.
3, separate then meaning of the subtree example in the set can make main tree obtain maximum subtree pipe nipple points according to its frequency of utilization series arrangement from high to low the subtree example by optimization means in the inventive method, and then minification.
4, but the selection method of best subtree is various to adapt to the database of different language and different size in the inventive method.
In sum, the beneficial effect of compression method of the present invention all is that the DAWG compression method is not available from design concept is still put into practice.
Description of drawings
Fig. 1 is the structural representation of general tree;
Fig. 2 is the structural representation of the binary tree of Fig. 1 correspondence;
Fig. 3 is the structural representation of binary tree by the deep search storage mode;
Fig. 4 is the binary tree structure synoptic diagram that comprises table 1 vocabulary group information;
Fig. 5 is DAWG compression step 1 synoptic diagram from bottom layer beginning second from the bottom;
Fig. 6 is DAWG compression step 2 synoptic diagram from bottom layer beginning second from the bottom;
Fig. 7 is DAWG compression step 3 synoptic diagram from bottom layer beginning second from the bottom;
Fig. 8 is the data structure synoptic diagram of compression method of the present invention;
Fig. 9 A is a subtree data structure No way out subtree synoptic diagram;
Fig. 9 B is that the subtree data structure has outlet subtree synoptic diagram;
Figure 10 comprises the initial data structure synoptic diagram of table 1 vocabulary group information in the compression method of the present invention;
Figure 11 is the data structure synoptic diagram during execution in step A first in the compression method of the present invention;
Figure 12 is the data structure synoptic diagram during execution in step B first in the compression method of the present invention;
Figure 13 is the data structure synoptic diagram during execution in step A for the second time in the compression method of the present invention;
Figure 14 is the data structure synoptic diagram during execution in step B for the second time in the compression method of the present invention;
Figure 15 is the data structure synoptic diagram during execution in step A for the third time in the compression method of the present invention;
Figure 16 is the data structure synoptic diagram during execution in step B for the third time in the compression method of the present invention;
Figure 17 is the structural representation of system of the present invention.
Embodiment:
Because there is not the space of improvement in essence in existing DAWG compression method, the invention provides a kind of language library compression method and system based on tree construction, it not only from practice but also overcome disadvantages of background technology in theory, makes it to be applicable to the compression of large database and search fast.Particular content is as follows:
One, definition data structure:
All subtree examples in the tree construction database are all proposed to set outer formation subtree set Sub TreeSet, and in main tree Main Tree, remove all subtrees ex hoc genus anne and replace (Fig. 8) with the subtree node in the place that they occur.The node of main tree and subtree is all deposited by deep search in storer.Ordinary node and subtree node that whole main tree only has the corporate data structure, if with byte is unit, when certain language database is in limited time by 32 letters, a byte unit can be held 3 zone bits and 1 letter, its common structure is:＜ST 〉,＜K〉＜B〉＜EOW〉＜L 〉, [＜Byte 〉], here＜Byte〉byte unit of expression.And press 64 letters in limited time, a byte unit can be held 2 zone bits and 1 letter, and its common structure is:＜K 〉,＜B 〉,＜ST〉＜EOW〉＜L 〉, [＜Byte 〉].When＜ST〉this node was an ordinary node when value was 0.Its expense is 1 or 2 zone bit+1 byte, when＜ST〉when being 1 the subtree node.The subtree node is divided into pipe nipple point and meropodium point again.As＜EOW〉when being 0 and ordinary node isometric, be the pipe nipple point, its＜L content be the subtree numbering of this subtree in system.And as＜EOW when being 1 the meropodium point, show that thereafter byte is with＜L〉formed the subtree numbering of this subtree in system.Small and exquisite, simple for structure with this main tree shape that defines.Engine needs only maximum 2 judgements just can identify node size, crosses fast.All subtree examples are proposed the outer composition set of tree significantly reduced the expense of subtree address code in main tree.Separate then meaning of subtree example in the set can make main tree obtain maximum subtree pipe nipple points according to its frequency of utilization series arrangement from high to low the subtree example by optimization means, and then minification.Have＜K〉＜B the subtree node also mean its suffix portion in can not only the compressed tree structure, can also compress center section, filled up the compression blank of DAWG.
Subtree Sub Tree i (i=1,2,3 ..., N) by ordinary node or add one as the outlet subtree pipe nipple point form, its first node can have the brother to mean that it is to be made of a subtree or a fraternal subtree group, the abbreviation subtree.The generating structure of subtree has following regulation:
1, the subtree of forming by ordinary node be the No way out subtree (the suffix subtree: ending sub tree), as Fig. 9 A, can only with main tree in＜K null subtree node is complementary.The subtree of being made up of ordinary node and subtree pipe nipple point is that the outlet subtree is arranged, as Fig. 9 B.
2, group tree node＜B〉be not equal to zero mean that this node has the brotgher of node to link to each other in main tree.Engine enters after the corresponding subtree general＜B in search procedure〉value dynamically is linked on the first node of subtree or its last brotgher of node that does not have the brother.Just at main tree neutron tree node＜B〉value only with in the corresponding subtree do not have subtree head node or its last brotgher of node of brother relevant.
3, do not allow to exist the subtree node of any representative subtree set neutron tree in the subtree, do not have recursive definition, the subtree pipe nipple point in it only plays transition, location and is connected effect.It neither allows to appear on the position of the first node of subtree or its brotgher of node, does not also allow the child node that belongs to it, its＜K〉value inside is 0 in order to engine dynamic link as the case may be when scanning.
4, as the subtree pipe nipple point that exports corresponding subtree node＜K in main tree〉inoperative when equalling zero, thus make subtree become the suffix subtree.If corresponding subtree node＜K in main tree〉be not equal to zero, then this excessive node step down and at this general＜K the child node in main tree of indication is continuous.
All nodes are isometric in the subtree means that simple for structure and engine can effective search, and the subtree that has the subtree node has both simultaneously to be had or the No way out subtree has improved similar expansion face, and then has also improved the compressibility of tree construction.In fact, engine realizes that on carrier the logical process of scanning is by judging, cross no articulation point, deposit in the nodal information of coupling stacked and when in the matched node＜EOW find out the word or the vocabulary that satisfy specified requirements when being not equal to 0.By investigating subtree, especially investigation＜K〉when being not equal to the first node of 0 the corresponding subtree example of subtree node extremely the brotgher of node having or not match information, broken away from tying down of its child node in the main tree, judged rapidly and efficiently, irrelevant common or subtree node one more and mistake.And have only the child node link that after engine has entered the subtree of coupling, in storehouse, just dynamically will lead in the tree to put in place along with the propelling of scanning process, go back the true colours of elite tree.This dynamic link mode is complementary with the subtree compression method for the present invention's initiative, makes flexible and efficient, the clear-cut individuality of search procedure.From design concept is still put into practice all is that the DAWG compression method is not available.
Two, the concrete steps of compression method of the present invention are as follows:
The master that A deposits in the main tree storage unit by the search unit traversal sets, and selects the best subtree of ratio of compression;
The subtree example that B searches steps A is connected on the subtree node of subtree set storage unit, gives this subtree node and substitutes the appearance that main tree goes up this subtree ex hoc genus anne with this node with current numbering, and then subtree node serial number counter is added 1;
C circulation execution in step A is till can't searching out subtree.
The optimal selection method of described optimum compression ratio subtree is to choose the subtree of the similar subtree number of subtree example node number * end value maximum.Described subtree comprises the No way out subtree and the outlet subtree is arranged.
Certainly, also can be according to the subtree of choosing tree example node number maximum earlier, when the result of this condition selection was unique, then this subtree was the optimal compression subtree; And when this condition selection result was not unique, the quantity of then judging similar subtree number again was to determine the subtree of optimum compression ratio.
We still illustrate the inventive method with one group of listed vocabulary of table 1, and its main tree construction that is positioned at main tree storage unit as shown in figure 10.
When the inventive method of circulation execution for the first time steps A, as shown in figure 11; When circulating execution in step B for the first time, as shown in figure 12.
When the inventive method of circulation execution for the second time steps A, as shown in figure 13; When circulating execution in step B for the second time, as shown in figure 14.
When the inventive method of circulation execution for the third time steps A, as shown in figure 15; When circulating execution in step B for the first time, as shown in figure 16.
Three, the system architecture that is used for the inventive method, as shown in figure 17:
It comprises:
Be used for selecting the search unit of the best subtree of ratio of compression in main tree storage unit;
Be used to deposit the main tree storage unit of main tree;
Be used to deposit the subtree set storage unit of subtree, this subtree set storage unit comprises a subtree node serial number counter;
Be used to control search, subtree storage and the alternative scheduler module of subtree node of optimum compression ratio subtree.
Described scheduler module can be selected TMS320C6416, and it is the serial High Performance DSP s chip of kernel newly of a C64XX of having that TI company released in 2000.TMS320C6416 adopts a kind of high performance advanced person's VLIW (very CLIW) structure, and its inside has 8 parallel processing elements.Because of the single instrction word length is 32,8 instructions can be formed the instruction bag that reaches 256, are assigned to 8 processing units simultaneously by the special instruction distribution module in inside and move simultaneously.Therefore when the 600MHz dominant frequency, the maximum processing capability of TMS320C6416 is up to 4800MIPS (MIPS).The TMS320C6416 core voltage is 1.2V, and peripheral voltage is 3.3V, and dominant frequency is 400MHz ~ 1GHz, and under the 600MHz dominant frequency, can provide 833B level device.
Described master's tree storage unit and subtree set storage unit can be used the above mass-memory unit of 100M, also a mass-memory unit can be divided into two high capacity storage areas to set storage unit and subtree set storage unit as the master respectively.
Described search unit also can use TMS320C6416.
Four, the implementation result of the inventive method:
Following table 2 is to compare with the result of DAWG and the inventive method respectively with the part language database:
| Language | Vocabulary | The subtree compression method | The DAWG compression method |
| Arabic | 60000 | 132 | 160 |
| Croatian | 97995 | 92 | 106 |
| Dutch | 36496 | 92 | 104 |
| English | 60000 | 112 | 128 |
| Persian | 47845 | 124 | 139 |
| Finnish | 45000 | 109 | 122 |
| French | 50000 | 87 | 96 |
| German | 33893 | 98 | 109 |
| Hebrew | 47295 | 99 | 125 |
| Dard | 49001 | 139 | 173 |
| Indonesian | 39976 | 99 | 117 |
| Malay | 33926 | 100 | 119 |
| Spanish (Europe) | 50134 | 95 | 105 |
| Spanish (South America) | 46161 | 97 | 106 |
| Thai | 27000 | 132 | 143 |
| Urdu | 28493 | 81 | 95 |
| Vietnamese | 33030 | 112 | 133 |
| Serbian | 159999 | 106 | 118 |
| Slovene | 233994 | 109 | 117 |
In sum, be preferred embodiment of the present invention only, be not to be used for limiting practical range of the present invention, promptly all equivalences of doing according to the content of the present patent application claim change and modify, and all should be technology category of the present invention.
Claims (4)
1. the language library compression method based on tree construction is characterized in that it comprises the steps:
The master that A deposits in the main tree storage unit by the search unit traversal sets, and selects the best subtree of ratio of compression;
The subtree example that B searches steps A is connected on the subtree node of subtree set storage unit, gives this subtree node and substitutes the appearance that main tree goes up this subtree ex hoc genus anne with this node with current numbering, and then subtree node serial number counter is added 1;
C circulation execution in step A is till can't searching out subtree.
2. the language library compression method based on tree construction according to claim 1, the system of selection that it is characterized in that described optimum compression ratio subtree are the subtrees of choosing the similar subtree number of subtree example node number * end value maximum.
3. the language library compression method based on tree construction according to claim 1 is characterized in that described subtree comprises the No way out subtree and the outlet subtree is arranged.
4. an enforcement is characterized in that as the language library compressibility based on tree construction of method as described in claim 1 or 2 or 3 it comprises:
Be used for selecting the search unit of the best subtree of ratio of compression in main tree storage unit;
Be used to deposit the main tree storage unit of main tree;
Be used to deposit the subtree set storage unit of subtree, this subtree set storage unit comprises a subtree node serial number counter;
Be used to control search, subtree storage and the alternative scheduler module of subtree node of optimum compression ratio subtree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010101646361A CN102222075A (en) | 2010-04-15 | 2010-04-15 | Tree-structure-based language bank compression method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010101646361A CN102222075A (en) | 2010-04-15 | 2010-04-15 | Tree-structure-based language bank compression method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102222075A true CN102222075A (en) | 2011-10-19 |
Family
ID=44778630
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010101646361A Pending CN102222075A (en) | 2010-04-15 | 2010-04-15 | Tree-structure-based language bank compression method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102222075A (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1564991A (en) * | 2001-10-02 | 2005-01-12 | 索尼国际(欧洲)股份有限公司 | Word database compression |
| CN1737791A (en) * | 2005-09-08 | 2006-02-22 | 无敌科技(西安)有限公司 | Data compression method by finite exhaustive optimization |
-
2010
- 2010-04-15 CN CN2010101646361A patent/CN102222075A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1564991A (en) * | 2001-10-02 | 2005-01-12 | 索尼国际(欧洲)股份有限公司 | Word database compression |
| CN1737791A (en) * | 2005-09-08 | 2006-02-22 | 无敌科技(西安)有限公司 | Data compression method by finite exhaustive optimization |
Non-Patent Citations (1)
| Title |
|---|
| 孔鹏程 等: "基于离散区间的频繁嵌入式子树挖掘算法", 《计算机应用》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102184169B (en) | Method, device and equipment used for determining similarity information among character string information | |
| CN101937430B (en) | Method for extracting event sentence pattern from Chinese sentence | |
| JPH0675992A (en) | Limited-state transducer in related work pattern for indexing and retrieving text | |
| CN106202010A (en) | The method and apparatus building Law Text syntax tree based on deep neural network | |
| JP5241738B2 (en) | Method and apparatus for building tree structure data from tables | |
| CN102693310A (en) | Resource description framework querying method and system based on relational database | |
| SG168416A1 (en) | Improvements relating to graphical user interfaces | |
| CN101576929B (en) | Fast vocabulary entry prompting realization method | |
| WO2004109492A1 (en) | Object representing and processing method and apparatus | |
| CN101876995A (en) | Method for calculating similarity of XML documents | |
| CN110046348A (en) | Main body recognition methods in a kind of rule-based and dictionary metro design code | |
| CN107844608A (en) | A kind of sentence similarity comparative approach based on term vector | |
| CN105677638A (en) | Web information extraction method | |
| CN103530316A (en) | Science subject extraction method based on multi-view learning | |
| CN109783819A (en) | Regular expression generation method and system | |
| CN109857458A (en) | The method for transformation of the flattening of AltaRica 3.0 based on ANTLR | |
| Zhou et al. | Master: Multi-task pre-trained bottlenecked masked autoencoders are better dense retrievers | |
| CN101398830A (en) | Thesaurus fuzzy enquiry method and thesaurus fuzzy enquiry system | |
| Binding et al. | Migrating a complex classification scheme to the semantic web: expressing the Integrative Levels Classification using SKOS RDF | |
| CN108491381A (en) | A kind of syntactic analysis method of Chinese bipartite structure | |
| CN101561800B (en) | Character pattern generation method of Mongolian | |
| CN101488127B (en) | Bit mark character string fuzzy retrieval method for grouping character and labellng with bit | |
| CN102736552A (en) | Method for converting ladder diagram developed by programmable logic controller (PLC) into statement list | |
| CN101499056A (en) | Backward reference sentence pattern language analysis method | |
| CN1604075A (en) | Method for conducting words reading sequence recovery for newspaper pages |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20111019 |
|
| C20 | Patent right or utility model deemed to be abandoned or is abandoned |