CN102955848B - A kind of three-dimensional model searching system based on semanteme and method - Google Patents
A kind of three-dimensional model searching system based on semanteme and method Download PDFInfo
- Publication number
- CN102955848B CN102955848B CN201210421340.2A CN201210421340A CN102955848B CN 102955848 B CN102955848 B CN 102955848B CN 201210421340 A CN201210421340 A CN 201210421340A CN 102955848 B CN102955848 B CN 102955848B
- Authority
- CN
- China
- Prior art keywords
- semantic
- concept
- semantic concept
- dimensional model
- information
- 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.)
- Active
Links
Abstract
The invention provides a kind of three-dimensional model searching system based on semanteme and method, described system comprises semantic retrieval module and semantic database.Wherein semantic retrieval module is used for carrying out semantic participle, filtration to retrieve statement, thus obtains semantic concept; In WordNet, the semantic concept with its synonym is inquired about according to described semantic concept, and the similarity calculated between described semantic concept and the semantic concept inquired, carry out semantic concept screening according to this similarity, obtain other semantic concepts relevant to described semantic concept; In semantic database, retrieve corresponding three-dimensional model information in conjunction with described semantic concept and relative semantic concept, and obtain result for retrieval.Three-dimensional model searching system provided by the present invention and method compensate for high-layer semantic information and low-level feature information semantic gap, expand three-dimensional model search scope, improve retrieval precision.
Description
Technical field
The present invention relates generally to technical field of information retrieval, is specifically related to a kind of three-dimensional model searching system based on semanteme and method.
Background technology
Along with the rapid growth of internet information and multimedia technology in daily life universal, information retrieval is not only confined to text retrieval, but multimedia retrieval can be carried out according to multimedia file, namely progressively to Directional Extensions such as image retrieval, audio retrieval, video frequency searchings.Meanwhile, for the ease of improving people to the use of searching system, text retrieval also progressively should merge semantic meaning, thus reduces the semantic gap of high-level semantic and low level data.
Three-dimensional model is progressively applied to industry, design multiple field such as industry, microorganism industry.Three-dimensional model can be vivid the structure of expression object and composition, but due to data volume large, therefore it can not retrieve fast as text message.Which results in a large amount of three-dimensional models to repeat to generate, store, be not easy to reusing of three-dimensional model.Present stage, the retrieval of three-dimensional model mainly adopts content-based retrieval method.Its essence is and set up aspect indexing according to the visual signature of three-dimensional model, the aspect indexing set up meets human vision requirements, and namely result for retrieval is visually similar.Content-based three-dimensional model search mainly adopts following steps: first, by carrying out pre-service to the visual signature of three-dimensional model to be retrieved, extracts the data of vacuate three-dimensional model; Subsequently three-dimensional model feature extraction is carried out to the data of vacuate, extract the proper vector of this three-dimensional model; Finally calculate the similarity degree of the proper vector in three-dimensional model to be retrieved and characteristic vector data storehouse, in 3 d model library, obtain result for retrieval according to similarity degree.
Current, three-dimensional model searching system mainly comprises text retrieval, two-dimentional Sketch Searching and content-based three-dimensional model search three kinds of modes.Wherein, two-dimentional Sketch Searching and content-based three-dimensional model search are all retrieve based on the essential characteristic of image graphics.And the semantic concept mark that text retrieval still rests on three-dimensional model has carries out keyword retrieval, it is not semantic retrieval truly.Wherein, semantic concept mark is the one of text key word mark, is that user or a kind of text with semantic concept of expert to three-dimensional model describe, is often made up of brief word, can be used for three-dimensional model semantic retrieval.Its objective is the mark by carrying out three-dimensional model on semantic concept, namely can be three-dimensional model and increasing semantic concept, finally can provide method for searching three-dimension model based on semanteme for user.For text retrieval, if two complete dissimilar three-dimensional models, but but have identical semantic concept mark, when so retrieving this semantic concept mark, these two three-dimensional models will be exported as result for retrieval jointly.On the contrary, if the semantic concept mark of three-dimensional model is not identical, but relevant on semantic level, such as, retrieval " furniture " one word, should have cupboard, chair, desk etc., and is not the class three-dimensional model only occurring having " furniture " semantic concept mark.This situation, can be described to " semantic gap " problem, i.e. the generation gap of high-level semantic and low layer three-dimensional model.This situation often comes across in the middle of text retrieval, thus causes the result precision of text retrieval not high.
The works and expressions for everyday use that three-dimensional model search based on semanteme can be avoided " semantic gap " problem, can be close to the users, such as: user inputs " quadrupeds ", system will retrieve the three-dimensional model such as " dog ", " cat " and " horse ", and the content that the semantic concept mark of these three-dimensional models comprises may be the item name such as " dog ", " cat " and " horse ", and do not mark " quadrupeds ", but belong to " quadrupeds " this semantic concept in semantically these contents.At present, the mark that existing 3D modelling system carries out for the semantic concept of three-dimensional model is single, or the mark just based on expert's angle, three-dimensional model carried out, due to when user really uses, because everyone is to the difference of the view of things, angle, stock of knowledge, and the mode that user is convenient to take to press close to oneself works and expressions for everyday use is to provide Search Requirement, thus cause, when carrying out the retrieval of semantic concept mark, larger deviation being produced.
Existing use has three-dimensional model searching system disclosed in Chinese Patent Application No. 200810115698.6 and method based on the three-dimensional model search of semanteme, and Fig. 1 shows the structural drawing of this searching system.The patent proposes a kind of multi-level relevance feedback algorithm MRF, MRF mainly utilizes feedback mechanism to obtain the high-layer semantic information of user feedback, by constantly decomposing, high-layer semantic information is converted into the correlation information of weight relationship between different characteristic vector and proper vector inside, namely according to the feedback information of user, the weighted value of each proper vector of retrieval model and correlation information is upgraded.Utilize weighted value and the correlation information of each proper vector after upgrading, calculate the similarity distance between retrieval model and Matching Model, obtain the three-dimensional model similar to retrieval model according to similarity distance.Thus, this method is still based on the retrieval in three-dimensional model content, judges to assist a ruler in governing a country and revises three-dimensional model proper vector, can not complete based on the retrieval of semanteme to three-dimensional model by the correlativity of user.
Summary of the invention
For the problems referred to above, the present invention proposes a kind of three-dimensional model searching system based on semanteme and method, with the semantic gap problem adopting the mode of semantic retrieval truly to solve high-layer semantic information and low layer three-dimensional model characteristic information.
According to one embodiment of the invention, provide a kind of three-dimensional model searching system, comprise semantic retrieval module and semantic database, wherein:
Described semantic retrieval module also comprises: participle submodule, semantic association calculating sub module and information transmit submodule, wherein
Described participle submodule is used for carrying out semantic participle and filtration to the retrieve statement that user inputs, and obtains semantic concept;
Described semantic association calculating sub module is used in semantic dictionary WordNet, inquiring about the multiple semantic concepts with its synonym according to described semantic concept, and described semantic concept and the semantic concept inquired are carried out Similarity Measure, carry out semantic concept screening according to this similarity, obtain the semantic concept relevant to described semantic concept;
Described information is transmitted submodule and is used in semantic database, retrieving corresponding three-dimensional model information in conjunction with described semantic concept and relative semantic concept, and obtains result for retrieval;
Described semantic database is for depositing the corresponding relation of semantic concept information and semantic concept and three-dimensional model.
In one embodiment, described system also comprises three-dimensional modeling data storehouse and display module, wherein
Described three-dimensional modeling data storehouse, for storing the data about three-dimensional model;
This retrieve statement for supporting that user submits retrieve statement to, and is passed to described semantic retrieval module by described display module; And for transmitting the result for retrieval that submodule obtains according to described information, read the three-dimensional modeling data in described three-dimensional modeling data storehouse, and show three-dimensional modeling data;
Described information is transmitted submodule and result for retrieval is passed to display module.
In one embodiment, described semantic concept and the semantic concept that inquires are carried out Similarity Measure and adopt following formula by described semantic association calculating sub module:
001"/>
Wherein, A, B represent two semantic concepts that will calculate, lcs (A, B) most recent co mmon ancestor node in the WordNet semantic dictionary of A, B is represented, IC (lcs (A, B)) represents the information content value that this common ancestor's node has, structure factor
Wherein, depth (A) and depth (B) represents the degree of depth that semantic concept A, B are residing in WordNet semantic dictionary respectively.
In one embodiment, described semantic association calculating sub module carries out semantic concept screening according to similarity, obtains the semantic concept relevant to described semantic concept and comprises:
According to the similarity threshold preset, by semantic association calculating sub module, the similarity result obtained by semantic association computational algorithm is screened, selects the semantic concept of the similarity had higher than the described similarity threshold preset,
In one embodiment, described information is transmitted submodule and in semantic database, is retrieved corresponding three-dimensional model information in conjunction with described semantic concept and relative semantic concept, and acquisition result for retrieval comprises:
If transmitting the semantic concept come has been present in described semantic database, then retrieves and returned the three-dimensional model information corresponding to this semantic concept;
If transmit next semantic concept not in semantic database, then do not return three-dimensional model information.
In one embodiment, described system also comprises semantic tagger module, and this semantic tagger module comprises user interface;
The three-dimensional model semantic concept mark that user submits to is audited by data base administrator by user interface, check whether mark meets standard and do not repeat with existing mark, if by examination & verification, then this semantic concept mark is added in the semantic concept mark of corresponding three-dimensional model in described semantic database; Otherwise, then miscue information will be returned to display module.
In a further embodiment, described display module is also for supporting that user submits to semantic concept to mark, and semantic concept mark user submitted to is passed to semantic tagger module.
According to one embodiment of the invention, propose a kind of method for searching three-dimension model based on described three-dimensional model searching system, comprising:
Step 1), semantic participle and filtration are carried out to retrieve statement, obtain semantic concept;
Step 2), obtain relative semantic concept according to described semantic concept, in semantic database, retrieve corresponding three-dimensional model information according to described semantic concept and relative semantic concept, obtain result for retrieval.
In one embodiment, also comprise before step 1):
Step 0), display module submit to retrieve statement.
According to one embodiment of the invention, step 2) after also comprise:
Step 3), result for retrieval to be inquired about in three-dimensional modeling data storehouse, obtain the data of three-dimensional model, and show this three-dimensional modeling data.
In one embodiment, step 2) in obtain relative semantic concept according to described semantic concept and comprise:
Step 21), in WordNet semantic dictionary, inquire the multiple semantic concept vocabulary with its synonym according to described semantic concept;
Step 22), adopt following formula to carry out Similarity Measure described semantic concept and the semantic concept that inquires:
003"/>
Wherein, A, B represent two semantic concepts that will calculate, lcs (A, B) most recent co mmon ancestor node in the WordNet semantic dictionary of A, B is represented, IC (lcs (A, B)) represents the information content value that this common ancestor's node has, structure factor
Wherein, depth (A) and depth (B) represents the degree of depth that semantic concept A, B are residing in WordNet semantic dictionary respectively.
Step 23) carry out semantic concept screening according to the similarity calculating gained, obtain the semantic concept relevant to described semantic concept.
In a further embodiment, step 23) in carry out semantic concept screening comprise according to calculating the similarity of gained:
According to the similarity threshold preset, the similarity of gained is screened, select to have semantic concept higher than the similarity of this similarity threshold as relevant semantic concept.
In one embodiment, step 2) in semantic database, retrieve corresponding three-dimensional model information according to described semantic concept and relative semantic concept, obtain result for retrieval and comprise:
If these semantic concepts have been present in semantic database, then from semantic data library searching and the three-dimensional model information returned corresponding to this semantic concept; If do not existed, then do not return three-dimensional model information
In sum, the three-dimensional model searching system based on semanteme provided by the invention and method have following beneficial effect:
(1) avoid the limitation only based on the content characteristic of three-dimensional model and the retrieval mode of text marking, consider the Global Information of three-dimensional model from different level, expand range of search
（2）。Compensate for the semantic gap of high-layer semantic information and low layer three-dimensional model characteristic information, improve retrieval precision.
Accompanying drawing explanation
Fig. 1 is the existing three-dimensional model searching system structural drawing based on semanteme;
Fig. 2 is according to an embodiment of the invention based on the three-dimensional model searching system structural drawing of semanteme;
Fig. 3 is the semantic concept mark schematic diagram adopting RDF markup language;
Fig. 4 is according to an embodiment of the invention based on the method for searching three-dimension model process flow diagram of semanteme.
Embodiment
Below in conjunction with the drawings and specific embodiments, present invention is described.
According to one embodiment of the invention, provide a kind of three-dimensional model searching system based on semanteme.Fig. 2 shows the general structure of this system, comprises server end and client.Described server end comprises semantic retrieval module, semantic database, three-dimensional modeling data storehouse and semantic tagger module; Described client comprises display module.Wherein
Described semantic retrieval module also comprises: participle submodule, carries out semantic participle and filtration, and obtain semantic concept for the retrieve statement inputted user; Semantic association calculating sub module, for inquiring multiple semantic concepts (i.e. semantic concept vocabulary) of synonym with it in semantic dictionary WordNet according to described semantic concept, and described semantic concept and the semantic concept inquired are carried out Similarity Measure, similarity according to calculating carries out semantic concept screening, obtains other semantic concepts relevant to described semantic concept; Information transmits submodule, for retrieving corresponding three-dimensional model information in conjunction with described semantic concept and the semantic concept relevant to described semantic concept in semantic database, obtaining result for retrieval, and result is passed to display module.
Described semantic database, for the corresponding relation of the information (i.e. semantic concept mark) and semantic concept and three-dimensional model that store semantic concept, i.e. a semantic concept, corresponding multiple three-dimensional model.
Described three-dimensional modeling data storehouse, for storing about the data of three-dimensional model, three-dimensional model classified information, the two dimensional image of three-dimensional model and the essential information etc. of three-dimensional model.
Described semantic tagger module, three-dimensional model semantic concept mark for submitting to user is audited, check whether this mark meets format standard (such as, mark can not comprise numbers and symbols etc.) and with existing mark (namely, the semantic concept mark that in semantic database, this three-dimensional model has) do not repeat, if by examination & verification, then by this semantic concept mark be added in described semantic database to should three-dimensional model semantic concept mark in; Otherwise, then miscue information will be returned.
Described display module is used for every interactive operation of completing user and system, and it comprises: provide the system page, supports that user submits retrieve statement to, and the retrieve statement that user submits to is passed to semantic retrieval module; Display page (can comprise the three-dimensional model information page) is provided, the next result of the submodule transmission (three-dimensional model information of retrieval gained is transmitted according to the information in semantic retrieval module, namely the results list be made up of three-dimensional model title) three-dimensional modeling data in digital independent three-dimensional modeling data storehouse, and show this three-dimensional modeling data; Described display module also provides and adds the semantic concept tagging page, supports that user submits to semantic concept to mark, and semantic concept mark user submitted to is passed to semantic tagger module.Described display module arranges prompt facility, and user can be instructed to operate.
Below the modules of the three-dimensional model searching system based on semanteme is described respectively.
One. semantic retrieval module
Semantic retrieval module comprises participle submodule, semantic association calculating sub module and information and transmits submodule.
1. participle submodule described in comprises participle instrument and filter word list.Described participle instrument supports Chinese word segmentation and English string segmentation, and existing participle instrument comprises: ictclas4j, IKAnalyzer, dismember an ox as skillfully as a butcher.The groundwork of participle instrument is split according to part of speech and the meaning of a word by the word of input, and the result after fractionation is a series of words, and automatically omit the word that ", " etc. does not have actual semantic meaning.Such as, participle submodule reads the retrieve statement that user submits at the system page: " three-dimensional model of quadrupeds ", utilizes participle instrument that this statement is split into multiple word, as " four limbs ", " animal ", " three-dimensional ", " model ".Then, the stop word in the word after fractionation filters according to filter word list by participle submodule, and after filtering, the word (as " four limbs ", " animal ") of gained can be used as semantic concept, and this semantic concept is passed to semantic association calculating sub module.
2. first semantic association calculating sub module is inquiring multiple semantic concept vocabulary (such as 2-5) according to the semantic concept obtained from word-dividing mode among semantic dictionary WordNet.WordNet is the English semantic dictionary of a kind of hierarchical structure developed by Princeton university.In WordNet, be by synonym (synonym), antisense (antonym), overall (holonym), partly (meronym), upper (hypernym), the next (hypony) between semantic concept, contain multiple semantic relations such as (entailment) to build relational network, whole WordNet is exactly the semantic network be made into by numerous net.Among WordNet, the synonym semantic concept about described semantic concept (semantic concept from word-dividing mode obtains) is inquired by interpretative tool (such as obtainable interpretative tool JAWL etc. from internet).Then, the semantic concept obtained from participle submodule and the semantic concept inquired from WordNet are carried out Similarity Measure, the similarity according to calculating gained carries out semantic concept screening, obtains other semantic concepts relevant to described semantic concept.
And then the information that the semantic concept obtained from participle submodule and relative semantic concept is all passed to transmits submodule.Wherein
Semantic association computational algorithm can adopt the algorithm, such as based on the information content: the characteristic model etc. that Lin algorithm and Rodriguez and Egenhofer propose.
According to one embodiment of present invention, there is provided W & IC algorithm as the semantic association computational algorithm of semantic association calculating sub module, on this W & IC algorithm is structured in Rodriguez and Egenhofer characteristic model based on WordNet semantic dictionary (similarity degree that characteristic model described here judges between semantic concept according to the quantity of the common feature set of semantic concept).The information content (IC) refers to TongYiCi CiLin (synset, minimum unit in WordNet, it is the integrated definition for a vocabulary, comprise concept, synonym, directional information and the side-play amount etc. of this word itself) frequency that the information that comprises occurs in specific corpus, the shared rate of the information content between two synset can be utilized to calculate similarity.The method of current existing computing semantic similarity has based on semantic dictionary and the semantic similarity calculation method based on the information content.Wherein, semantic computation method based on semantic dictionary utilizes the physical path length between two concepts to judge Similarity value between the two, the nearlyer similarity between the two of semantic distance is larger, but but there is certain restrictive condition in this theory, such as same path length concept between, abstract concept between semantic distance should be greater than concrete concept between semantic distance.And based on the information content Measures compare two concepts between the information content thus calculate semantic similarity between the two, but the value of traditional information content is all the frequency that a certain concept of calculating appears in large corpora (such as Blang's corpus, LOB, KolhapurCorpus etc.), contain much information in large corpora, frequency values accurately can not reflect semantic information, depends on the defect that corpus is the method unduly.For the problem of prior art, W & IC algorithm provided by the invention, when the value of computing information content, utilizes characteristic model, only carries out the calculating of semantic distance for the hierarchical structure of WordNet.WordNet data volume is less, and semantic structure is clear and semantic concept accurate, the accuracy that can reduce computation complexity, accelerate computing velocity, promote result of calculation.
W & IC algorithm provided by the invention is based on traditional characteristic model proposed by Rodriguez and Egenhofer, and this model utilizes feature quantity to calculate similarity between two concepts, and computing formula is as follows:
Wherein A, B are two semantic concepts that will calculate similarity, the word namely comprised in two semantic concepts as characteristic item, the characteristic set of composition; A ∩ B represents and not only appears in A but also appear at the feature set in B, and A-B indicates in present A, but does not appear at the characteristic set in B, and in like manner B-A to indicate in present B but the characteristic set do not appeared in A;
in this scope is selected
value be to make the importance of public characteristic to similarity be greater than not common feature, namely reducing the impact of not common feature on similarity, improve the accuracy of similarity.
Above-mentioned formula is the characteristic model that a concept similarity calculates, wherein, feature quantity and IC value have common part, all represent the quantity of information that a concept contains, the feature quantity of imagination semantic concept A is its IC value, the common trait quantity of concept A, B is the IC value of both most recent co mmon ancestor nodes (LeastCommonSubsumer is called for short LCS), then A ∩ B, | A-B|, | the computing formula of B-A| is as follows:
A∩B≈sim
res(A,B)=IC(lcs(A,B))
|A-B|=Feature((A)-(A∩B))≈IC(A)-IC(lcs(A,B))
|B-A|=Feature((B)-(A∩B))≈IC(B)-IC(lcs(A,B))
Bring above three formula into sim (A, B) and W & IC algorithmic formula can be obtained:
006"/>
Wherein lcs (A, B) represents semantic concept A, B most recent co mmon ancestor node in WordNet semantic dictionary, and IC (lcs (A, B)) represents the information content (IC) value that this common ancestor's node has.Because IC value described herein calculating prerequisite does not use corpus, wherein, the computing formula of structure factor a is as follows:
Wherein, the span of a is still 0 to 0.5, namely suppose it is still that the importance of public characteristic to similarity is greater than not common feature, wherein A, B are the characteristic sets of two semantic concepts, and depth (A) and depth (B) represents the degree of depth that A, B are residing in WordNet semantic dictionary respectively.
W & IC algorithm is compared with traditional algorithm based on the information content, the hierarchical structure in WordNet is only relied on to carry out computing information content (IC) value, decrease cost and the complexity of calculating, considering that public characteristic factor and not common characteristic factor are on the impact of Similarity value, the subjective judgement of final calculating income value and people is also more close simultaneously.
According to one embodiment of present invention, according to the similarity threshold preset, by semantic association calculating sub module, the similarity result obtained by semantic association computational algorithm is screened, select the semantic concept with the similarity presetting similarity threshold higher than this, selected semantic concept is relevant to splitting according to retrieve statement the semantic concept obtained.The semantic concept that these relevant semantic concepts and retrieve statement split all is passed to information and transmits submodule.Such as, this similarity threshold can be redefined for 85%, and split according to retrieve statement after the semantic concept obtained carries out Similarity Measure, select higher than the semantic concept corresponding to the similarity of 85% as relevant semantic concept.
3. information is transmitted submodule and is received the next content of semantic association calculating sub module transmission (semantic concept that retrieve statement splits and relative semantic concept), in semantic database, the semantic content that three-dimensional model has is retrieved, search semantic concept and three-dimensional model (namely according to existing semantic concept, finding the model name of corresponding three-dimensional model) corresponding to relative semantic concept that retrieve statement splits.
Because the semantic concept in semantic database is limited, be present in semantic database if transmit the semantic concept come, then retrieved and return the three-dimensional model information (namely as the model name of the three-dimensional model of result for retrieval) corresponding to this semantic concept; If transmit next semantic concept not in semantic database, then do not return three-dimensional model information.If result need be shown at display page, then result (model name of three-dimensional model) is passed to display module, then carries out the display of three-dimensional model semantic retrieval result by display module.
Two. semantic database
Semantic database for depositing the corresponding relation of semantic concept information and semantic concept and three-dimensional model, i.e. a semantic concept, corresponding multiple three-dimensional model, its storage mode is < semantic concept, model name 1 ..., model name n>.A semantic concept can be had by multiple model, and multiple model also can corresponding multiple semantic concept, but semantic concept unique existence in a database, can not repeat.According to semantic concept, three-dimensional model corresponding to this semantic concept can be retrieved in semantic database.
Three. three-dimensional modeling data storehouse
Three-dimensional modeling data storehouse is for storing three-dimensional modeling data, three-dimensional model classified information, the two dimensional image of three-dimensional model and the essential information of three-dimensional model.Wherein said essential information comprises: the boundary value (x-axis maximin, y-axis maximin, z-axis maximin) of three-dimensional model name, three-dimensional model, three-dimensional model barycentric coordinates value, three-dimensional model main shaft etc.In this database, model name, as major key, unique to exist, not reproducible.According to unique model name, the information of three-dimensional model can be retrieved, obtain three-dimensional model file, two dimensional image in the Url of server end and essential information, utilize these information, can be used as result for retrieval and be passed to display module and show.
Four. semantic tagger module
Semantic tagger module support user is semantic concept tagging for three-dimensional model adds, first the semantic concept mark come from display module transmission is audited, checks and audit this mark whether meet standard and whether repeat with the already present mark about this model in semantic database.If comprise numeral or symbol in this mark, or comprise bad, invalid information in this mark, then think that this mark does not meet standard, not by examination & verification; In addition, if this model has had this mark in semantic database, then this mark has been thought not by examination & verification; Otherwise, then by examination & verification.If by examination & verification, then this semantic concept mark is added in the semantic concept mark of corresponding three-dimensional model in semantic database: first in semantic database, add this semantic concept, then according to < semantic concept, model name 1, ..., the storage mode of model name n>, after being added on this semantic concept by the model name of model, finally successful for interpolation information is back to display module, points out user to add the success of semantic concept tagging by display module; If not by examination & verification, then do not do interpolation operation, interpolation failure information is back to display module, point out user to add this semantic concept by display module and mark unsuccessfully, and require user again to add or abandon adding.
According to one embodiment of present invention, semantic tagger module also comprises user interface, for semantic concept mark is supplied to semantic database managerial personnel examination & verification, and examination result is returned semantic tagger module.
Five. display module
Display module is used for every interactive operation of completing user and system, and display module arranges prompt facility, user can be instructed to carry out operation and comprise:
1. the system page is provided, supports that user submits retrieve statement to, and the retrieve statement that user submits to is passed to semantic retrieval module.The system page provides input retrieve statement square frame, places and submit retrieve statement button to around square frame.After user inputs retrieve statement, click on submission button, is committed to the semantic retrieval module of server end by statement.
2. provide display page and the three-dimensional model information page to show three-dimensional model.The result (model name of three-dimensional model) that display module can come according to the transmission of semantic retrieval module retrieves three-dimensional modeling data in three-dimensional modeling data storehouse: the model name utilizing three-dimensional model, searches the information such as two dimensional image Url of three-dimensional model file Url corresponding to this model name, three-dimensional model in a database; According to Url, when showing this three-dimensional model, the three-dimensional picture of three-dimensional model two dimensional image and this model can be demonstrated at the page; In addition, in the process that the semantic concept of user to three-dimensional model marks, also Presentation Function is provided.
According to one embodiment of the invention, the mode of the display page display three-dimensional model of described display module can be divided into three-dimensional model independently to show and three-dimensional model list display.Wherein:
Three-dimensional model list display is mainly used in result for retrieval output and category checks three-dimensional model.When showing result for retrieval, three-dimensional model can carry out sort (such as sorting from high to low) by large young pathbreaker's result for retrieval according to calculated similarity, and the result for retrieval of highlighted display the most similar (such as 1-5).The two dimensional image that original list display three-dimensional model is corresponding, and show three-dimensional model relevant information (such as three-dimensional model name etc.) and part of semantic concept tagging.The two dimensional image that user clicks in original list can enter the three-dimensional model information page, have the details of three-dimensional model (comprising three-dimensional model spatial information, coordinate axis value, center of gravity etc.) and two dimensional image in this page, button has interpolation semantic concept tagging, 3-D display and returns.
Check three-dimensional model for the ease of user's category, in one embodiment, three-dimensional model can be read three-dimensional modeling data, utilize WebGL to browse three-dimensional model with tree structure classification display, support user, and support mouse rotation process by display module.
Three-dimensional model independently shows and is mainly used in checking independent three-dimensional model, calls WebGL control auxiliary display three-dimensional model.Mainly trigger by 3-D display button.In three-dimensional model stereo display frame, three-dimensional model has been removed the much informations such as color, texture, pattern, reduces the data volume of server process, is convenient to user and checks model structure fast.Meanwhile, display mode supports that user checks the angle of three-dimensional model by mouse action adjustment, namely supports rotation process, translation, zoom operations etc.
3. provide and add the semantic concept tagging page, support that user submits to semantic concept to mark, and semantic concept mark user submitted to is passed to semantic tagger module.
Wherein, the system page provides the button selecting semantic concept mark, enters the system page and select semantic concept to mark as user.System enters the display page of three-dimensional model automatically, such as three-dimensional model original list, and be the tree-shaped classification chart of three-dimensional model on the left of the page, right side is for showing three-dimensional model.The tree-shaped classification chart of three-dimensional model is according to the sort file display in three-dimensional modeling data storehouse, and each classification is as a file, and all categories is under root root directory.Click arbitrary classification, the two dimensional image of the right side page all three-dimensional models under showing this classification and essential information, comprise three-dimensional model name, three-dimensional model space characteristics etc.User now can click arbitrary two dimensional image, enter the three-dimensional model information page corresponding to this two dimensional image, there are the details of three-dimensional model in this page, and have the semantic concept tagging of interpolation and 3-D display two buttons, can three-dimensional model be checked by three-dimensional display key.
In addition, user can enter the semantic concept tagging page of interpolation by clicking the semantic concept tagging button of interpolation.In an embodiment of the invention, the semantic concept mark page comprises three-dimensional model stereo display mode, supports that mouse action and unspecified angle check model; Also comprise semantic concept mark and add frame, need input text in this frame, text should not comprise symbol, but can comprise Chinese and English and numeral, and wherein multiple semantic concept mark application space separates; Semantic concept mark adds above frame Chinese and English information, helps user correctly to add semantic concept tagging.User is after click confirms to add button, and this semantic concept mark marks by semantic concept the semantic tagger module that the page is passed to server end,
Semantic concept mark is the means by one mark (Tag), the metadata of resource in HTML or XML, i.e. and the semantic information process of getting up with corresponding resource relationship.The markup language used can be the markup language being applicable to this well known in the art such as such as XML, RDF etc.
Illustrate in Fig. 3 and adopted RDF technology three-dimensional model to be added to the example of semantic concept tagging.In the three-dimensional model, by RDF tlv triple < resource, attribute, value > describes three-dimensional model, and for Fig. 3, following form provides the example adopting RDF dimensioning of three-dimensional model semantics concept:
According to one embodiment of present invention, additionally provide a kind of method for searching three-dimension model based on above-mentioned three-dimensional model searching system, Fig. 4 shows the process flow diagram of the method, comprising:
The first step, at system page input retrieve statement, click on submission button, submits retrieve statement to or selects to click semantic concept labelled buttons and carry out semantic concept mark;
If input retrieve statement in the first step and submit to, then perform following steps:
Second step, carries out semantic participle to described retrieve statement, namely utilizes participle instrument to carry out participle, filtration to retrieve statement, thus obtains semantic concept.
3rd step, obtains relative semantic concept according to described semantic concept, retrieves corresponding three-dimensional model information in conjunction with described semantic concept and relative semantic concept in semantic database.
If these semantic concepts have been present in semantic database, then from semantic data library searching and the three-dimensional model information (this result for retrieval is the model name of three-dimensional model) returned corresponding to this semantic concept; If do not existed, then do not return three-dimensional model information.
Wherein, obtain relative semantic concept according to described semantic concept to comprise:
1. inquire in WordNet semantic dictionary and multiple semantic concept vocabulary of its synonym (i.e. semantic concept, such as 2-5) according to described semantic concept, can such as be inquired about by the available interpretative tool of network.
2. described semantic concept and the semantic concept inquired are carried out Similarity Measure.According to one embodiment of present invention, computing semantic similarity can adopt the method based on the information content, such as W & IC algorithm.
3. carry out semantic concept screening according to the similarity calculating gained, obtain other semantic concepts relevant to described semantic concept.According to one embodiment of present invention, according to the similarity threshold preset, similarity result is screened, select to have semantic concept higher than the similarity of this similarity threshold as relevant semantic concept.Such as, 85% is set as similarity threshold.
4th step, according to the model name of the three-dimensional model that the 3rd step obtains, inquires about, obtains the data of this three-dimensional model, then it can be used as result for retrieval to show at the display module of client in three-dimensional modeling data storehouse.
Wherein, three-dimensional model list can be adopted to show.Can sort by large young pathbreaker's result for retrieval according to calculated similarity, and the result for retrieval that highlighted display is the most similar.The form that display page (first level pages) can be arranged by 5 row 4 forms, form comprises the two dimensional image of three-dimensional model and the model name of three-dimensional model, click this image and can enter this three-dimensional model information page (the secondary page), the secondary page comprises the 3-D display of this three-dimensional model, support to carry out the operations such as convergent-divergent to three-dimensional model, be included in the essential information stored in three-dimensional modeling data storehouse, as boundary value, main shaft, center of gravity etc. simultaneously.
If select semantic concept mark in the first step, then perform following steps:
Second step, user selects the three-dimensional model that will mark, and the semantic concept tagging page of the interpolation entering display module.
3rd step, user submits semantic concept mark at the semantic concept mark page, marks the page semantic concept of submission mark is passed to semantic tagger module by semantic concept.
The markup language of such as XML, RDF wherein can be adopted to carry out the semantic tagger of three-dimensional model.
4th step, in semantic tagger module, and judge whether the semantic concept mark obtained meets standard and whether repeat (such as with the already present mark about this model in semantic database, by user interface, this semantic concept mark is supplied to semantic data library manager to audit), if comprise numeral or symbol in this mark, or change mark and comprise bad, invalid information, then think that this mark does not meet standard, not by examination & verification; Or this model has had this mark in semantic database, then think this mark not by examination & verification; Otherwise, then by examination & verification.
If by examination & verification, by semantic tagger module this semantic concept marked and to be added in semantic database in corresponding three-dimensional model semantic concept mark, comprising:
First in semantic database, this semantic concept is added, then according to < semantic concept, model name 1, ..., the storage mode of model name n>, after the model name of model is added on this semantic concept, finally successful for interpolation information is back to display module, points out user to add the success of semantic concept tagging by display module; If not by examination & verification, then do not do interpolation operation, interpolation failure information is back to display module, point out user to add the failure of semantic concept tagging by display module, and user can also be required again to add or abandon adding.
In above-mentioned semantic concept annotation process, the user of second step adds the ability to express that semantic concept tagging process need describes three-dimensional model by people, carries out treatment and analysis, usually need through following process for corresponding three-dimensional model mark object:
1. user profile, mark three-dimensional model to be marked.User can according to the understanding of the information such as the vision had three-dimensional model, content, structure, provides paragraph and describe and the dimensioning of three-dimensional model attribute that should possess, provide property value.By to the analysis and the description that mark object, the key words content marking object can be obtained.
2. extract the concept described in content.For described content, by with the semantic concept in Wordnet one by one comparison after, it is extracted to process further.Such as in three-dimensional model semantic tagger, " quadrupeds ", " ox ", " furniture ", etc. noun be all concept in Wordnet, when analyzing, they are extracted.The series of concepts now obtained really can not embody the semantic description content of three-dimensional model, but simple concept is enumerated and piles up, and therefore also needs to further process.
3. mark the extraction of property value in object.For the content described in three-dimensional model, by a series of process, obtain the property value described by them, thus obtain the semantic object of three-dimensional model.In the extracting method of semantic instance, the semantic concept marked content of three-dimensional model is exactly the property value composition of multiple attribute, such as, use RDF language to mark, property value concrete in the semantic concept corresponding three-dimensional model of each mark.
It should be noted that and understand, when not departing from the spirit and scope of the present invention required by accompanying claim, various amendment and improvement can be made to the present invention of foregoing detailed description.Therefore, the scope of claimed technical scheme is not by the restriction of given any specific exemplary teachings.
Claims (11)
1. a three-dimensional model searching system, comprises semantic retrieval module and semantic database, wherein
Described semantic retrieval module also comprises: participle submodule, semantic association calculating sub module and information transmit submodule, wherein
Described participle submodule is used for carrying out semantic participle and filtration to the retrieve statement that user inputs, and obtains semantic concept;
Described semantic association calculating sub module is used in semantic dictionary WordNet, inquiring about the multiple semantic concepts with its synonym according to described semantic concept, and described semantic concept and the semantic concept inquired are carried out Similarity Measure, carry out semantic concept screening according to this similarity, obtain the semantic concept relevant to described semantic concept; Wherein, carry out Similarity Measure and adopt following formula:
Wherein, A, B represent two semantic concepts that will calculate, lcs (A, B) most recent co mmon ancestor node in the WordNet semantic dictionary of A, B is represented, IC (lcs (A, B)) represents the information content value that this common ancestor's node has, structure factor
Wherein, depth (A) and depth (B) represents the degree of depth that semantic concept A, B are residing in WordNet semantic dictionary respectively;
Described information is transmitted submodule and is used in semantic database, retrieving corresponding three-dimensional model information in conjunction with described semantic concept and relative semantic concept, and obtains result for retrieval;
Described semantic database is for depositing the corresponding relation of semantic concept information and semantic concept and three-dimensional model.
2. system according to claim 1, also comprises three-dimensional modeling data storehouse and display module, wherein
Described three-dimensional modeling data storehouse, for storing the data about three-dimensional model;
This retrieve statement for supporting that user submits retrieve statement to, and is passed to described semantic retrieval module by described display module; And for transmitting the result for retrieval that submodule obtains according to described information, read the three-dimensional modeling data in described three-dimensional modeling data storehouse, and show three-dimensional modeling data;
Described information is transmitted submodule and result for retrieval is passed to display module.
3. system according to claim 1 and 2, wherein, described semantic association calculating sub module carries out semantic concept screening according to similarity, obtains the semantic concept relevant to described semantic concept and comprises:
According to the similarity threshold preset, by semantic association calculating sub module, the similarity result obtained by semantic association computational algorithm is screened, select the semantic concept of the similarity had higher than the described similarity threshold preset.
4. system according to claim 1 and 2, wherein, described information is transmitted submodule and in semantic database, is retrieved corresponding three-dimensional model information in conjunction with described semantic concept and relative semantic concept, and acquisition result for retrieval comprises:
If transmitting the semantic concept come has been present in described semantic database, then retrieves and returned the three-dimensional model information corresponding to this semantic concept;
If transmit next semantic concept not in semantic database, then do not return three-dimensional model information.
5. system according to claim 2, also comprises semantic tagger module, and this semantic tagger module comprises user interface;
The three-dimensional model semantic concept mark that user submits to is audited by data base administrator by user interface, check whether mark meets standard and do not repeat with existing mark, if by examination & verification, then this semantic concept mark is added in the semantic concept mark of corresponding three-dimensional model in described semantic database; Otherwise, then miscue information will be returned to display module.
6. system according to claim 5, wherein, described display module is also for supporting that user submits to semantic concept to mark, and semantic concept mark user submitted to is passed to semantic tagger module.
7., based on a method for searching three-dimension model for the three-dimensional model searching system of claim 1, comprising:
Step 1), semantic participle and filtration are carried out to retrieve statement, obtain semantic concept;
Step 2), obtain relative semantic concept according to described semantic concept, in semantic database, retrieve corresponding three-dimensional model information according to described semantic concept and relative semantic concept, obtain result for retrieval; Wherein, obtain relative semantic concept according to described semantic concept to comprise:
Step 21), in WordNet semantic dictionary, inquire the multiple semantic concept vocabulary with its synonym according to described semantic concept;
Step 22), adopt following formula to carry out Similarity Measure described semantic concept and the semantic concept that inquires:
Wherein, A, B represent two semantic concepts that will calculate, lcs (A, B) most recent co mmon ancestor node in the WordNet semantic dictionary of A, B is represented, IC (lcs (A, B)) represents the information content value that this common ancestor's node has, structure factor
Wherein, depth (A) and depth (B) represents the degree of depth that semantic concept A, B are residing in WordNet semantic dictionary respectively;
Step 23), carry out semantic concept screening according to calculating the similarity of gained, obtain the semantic concept relevant to described semantic concept.
8. method according to claim 7, wherein step 1) also comprise before:
Step 0), display module submit to retrieve statement.
9. the method according to claim 7 or 8, wherein step 2) after also comprise:
Step 3), result for retrieval is inquired about in three-dimensional modeling data storehouse, obtain the data of three-dimensional model, and show this three-dimensional modeling data.
10. method according to claim 7, step 23) in carry out semantic concept screening according to the similarity calculating gained and comprise:
According to the similarity threshold preset, the similarity of gained is screened, select to have semantic concept higher than the similarity of this similarity threshold as relevant semantic concept.
11. methods according to claim 7 or 8, step 2) in semantic database, retrieve corresponding three-dimensional model information according to described semantic concept and relative semantic concept, obtain result for retrieval and comprise:
If these semantic concepts have been present in semantic database, then from semantic data library searching and the three-dimensional model information returned corresponding to this semantic concept; If do not existed, then do not return three-dimensional model information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210421340.2A CN102955848B (en) | 2012-10-29 | 2012-10-29 | A kind of three-dimensional model searching system based on semanteme and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210421340.2A CN102955848B (en) | 2012-10-29 | 2012-10-29 | A kind of three-dimensional model searching system based on semanteme and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102955848A CN102955848A (en) | 2013-03-06 |
| CN102955848B true CN102955848B (en) | 2015-11-18 |
Family
ID=47764654
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210421340.2A Active CN102955848B (en) | 2012-10-29 | 2012-10-29 | A kind of three-dimensional model searching system based on semanteme and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102955848B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105453086B (en) * | 2013-08-15 | 2020-01-10 | 国际商业机器公司 | Concept utilization to obtain user-specific data of interest from one or more data storage locations |
| US10223401B2 (en) | 2013-08-15 | 2019-03-05 | International Business Machines Corporation | Incrementally retrieving data for objects to provide a desired level of detail |
| US9767222B2 (en) | 2013-09-27 | 2017-09-19 | International Business Machines Corporation | Information sets for data management |
| CN104461525B (en) * | 2014-11-27 | 2018-01-23 | 韩慧健 | A kind of intelligent consulting platform generation system that can customize |
| CN105335487A (en) * | 2015-10-16 | 2016-02-17 | 北京农业信息技术研究中心 | Agricultural specialist information retrieval system and method on basis of agricultural technology information ontology library |
| KR101880275B1 (en) * | 2017-01-09 | 2018-08-16 | 김선중 | Search system and method for biological system information |
| CN107169051B (en) * | 2017-04-26 | 2019-09-24 | 山东师范大学 | Based on relevant method for searching three-dimension model semantic between ontology and system |
| CN107908760A (en) * | 2017-11-23 | 2018-04-13 | 北京航空航天大学 | A kind of method for searching three-dimension model for considering design knowledge and the aspect of model |
| CN108021683B (en) * | 2017-12-11 | 2021-06-22 | 首都航天机械公司 | Three-dimensional labeling-based similar model retrieval implementation method |
| CN108170823B (en) * | 2018-01-04 | 2021-10-15 | 江西师范大学 | Hand-drawn interactive three-dimensional model retrieval method based on high-level semantic attribute understanding |
| CN109753655B (en) * | 2018-12-29 | 2023-02-28 | 上海半坡网络技术有限公司 | Document retrieval method, system, storage medium and terminal based on semantic concept association |
| CN110348470B (en) * | 2019-05-21 | 2022-11-22 | 湖南大学 | Semantic retrieval method for industrial fault information rapid matching |
| CN110704653A (en) * | 2019-09-09 | 2020-01-17 | 上海慧之建建设顾问有限公司 | Method for searching component by graph in BIM (building information modeling) model and graph-text searching system |
| CN112487239B (en) * | 2020-11-27 | 2024-04-05 | 北京百度网讯科技有限公司 | Video retrieval method, model training method, device, equipment and storage medium |
| CN112765363B (en) * | 2021-01-19 | 2022-11-22 | 昆明理工大学 | Demand map construction method for scientific and technological service demand |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102207973A (en) * | 2011-06-22 | 2011-10-05 | 上海互联网软件有限公司 | Fuzzy search system and search method |
| CN102722501A (en) * | 2011-03-31 | 2012-10-10 | 北京百度网讯科技有限公司 | Search engine and realization method thereof |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8301633B2 (en) * | 2007-10-01 | 2012-10-30 | Palo Alto Research Center Incorporated | System and method for semantic search |
-
2012
- 2012-10-29 CN CN201210421340.2A patent/CN102955848B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102722501A (en) * | 2011-03-31 | 2012-10-10 | 北京百度网讯科技有限公司 | Search engine and realization method thereof |
| CN102207973A (en) * | 2011-06-22 | 2011-10-05 | 上海互联网软件有限公司 | Fuzzy search system and search method |
Non-Patent Citations (1)
| Title |
|---|
| 基于同义词扩展的文本检索改进研究;赵医娟等;《中国教育技术装备》;20100630(第18期);第95页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102955848A (en) | 2013-03-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102955848B (en) | A kind of three-dimensional model searching system based on semanteme and method | |
| CN109800284B (en) | Task-oriented unstructured information intelligent question-answering system construction method | |
| CN107180045B (en) | Method for extracting geographic entity relation contained in internet text | |
| CN100416570C (en) | FAQ based Chinese natural language ask and answer method | |
| CN109271505A (en) | A kind of question answering system implementation method based on problem answers pair | |
| CN103955529B (en) | A kind of internet information search polymerize rendering method | |
| KR102158352B1 (en) | Providing method of key information in policy information document, Providing system of policy information, and computer program therefor | |
| CN111143479A (en) | Knowledge graph relation extraction and REST service visualization fusion method based on DBSCAN clustering algorithm | |
| CN111177591B (en) | Knowledge graph-based Web data optimization method for visual requirements | |
| Zheng et al. | Template-independent news extraction based on visual consistency | |
| CN102663129A (en) | Medical field deep question and answer method and medical retrieval system | |
| CN111190900B (en) | JSON data visualization optimization method in cloud computing mode | |
| US20080016087A1 (en) | Interactively crawling data records on web pages | |
| US8577887B2 (en) | Content grouping systems and methods | |
| CN103229223A (en) | Providing answers to questions using multiple models to score candidate answers | |
| CN110750995B (en) | File management method based on custom map | |
| CN112925901B (en) | Evaluation resource recommendation method for assisting online questionnaire evaluation and application thereof | |
| CN114090861A (en) | Education field search engine construction method based on knowledge graph | |
| Kisilevich et al. | “Beautiful picture of an ugly place”. Exploring photo collections using opinion and sentiment analysis of user comments | |
| Dewi et al. | XAI for Image Captioning using SHAP. | |
| CN112015907A (en) | Method and device for quickly constructing discipline knowledge graph and storage medium | |
| CN102982072B (en) | A kind of method of visual concept detection device and constructing semantic field | |
| CN114490930A (en) | Cultural relic question-answering system and question-answering method based on knowledge graph | |
| Fudholi et al. | Code (common ontology development): A knowledge integration approach from multiple ontologies | |
| Zhou et al. | Research on mechanism of the information retrieval based on ontology label |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Li Haisheng Inventor after: Cao Qian Inventor after: Cao Jian Inventor after: Cai Qiang Inventor before: Li Haisheng Inventor before: Liu Xuan Inventor before: Cao Jian Inventor before: Cai Qiang |
|
| COR | Change of bibliographic data | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |