CN114385873A - Furniture scene reconstruction and recommendation method based on long text semantic consistency - Google Patents
Furniture scene reconstruction and recommendation method based on long text semantic consistency Download PDFInfo
- Publication number
- CN114385873A CN114385873A CN202111667409.5A CN202111667409A CN114385873A CN 114385873 A CN114385873 A CN 114385873A CN 202111667409 A CN202111667409 A CN 202111667409A CN 114385873 A CN114385873 A CN 114385873A
- Authority
- CN
- China
- Prior art keywords
- furniture
- entity
- text
- user
- scene
- 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
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000004458 analytical method Methods 0.000 claims abstract description 29
- 238000012545 processing Methods 0.000 claims abstract description 7
- 230000011218 segmentation Effects 0.000 claims description 23
- 238000000605 extraction Methods 0.000 claims description 16
- 230000008676 import Effects 0.000 claims description 3
- 239000003086 colorant Substances 0.000 claims description 2
- 238000011160 research Methods 0.000 abstract description 5
- 230000014509 gene expression Effects 0.000 abstract description 4
- 238000013461 design Methods 0.000 abstract description 3
- 238000012800 visualization Methods 0.000 description 18
- 238000010276 construction Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 3
- 230000018109 developmental process Effects 0.000 description 3
- 238000003058 natural language processing Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000004883 computer application Methods 0.000 description 1
- 238000013079 data visualisation Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000003012 network analysis Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000007794 visualization technique Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9032—Query formulation
- G06F16/90332—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9035—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/9038—Presentation of query results
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/205—Parsing
- G06F40/211—Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Databases & Information Systems (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a furniture scene reconstruction and recommendation method based on long text semantic consistency, which comprises the following steps: step 1, extracting a furniture entity according to a text lexical method and a syntax processing result; step 2, extracting an algorithm according to the spatial relationship of the text syntactic relationship; step 3, forming a semantic network according to the home entity and the text data of the spatial relationship thereof and visualizing the semantic network; step 4, recommending existing furniture in the scene according to the furniture incidence relation and the co-occurrence; and 5, after the entity furniture and the spatial relationship between the entity furniture are extracted and obtained, generating a three-dimensional scene for the formed semantic network. The recommendation method enables a user to obtain the wanted home layout design only through spoken language expression through text semantic analysis research in the home field.
Description
Technical Field
The invention relates to the technical field of computer application of scene generation, in particular to a furniture scene reconstruction and recommendation method based on long text semantic consistency.
Background
In recent years, with the development of computer technology, people become more convenient and faster in various aspects of daily life. For the design of furniture layouts, it is desirable that the user obtain the desired layout by spoken language expression only. With the continuous development of natural language processing technology in China, a plurality of mature Chinese processing tools are available, but due to the complexity of Chinese language and the diversity of application scenes, the Chinese text extraction method under a certain specific situation is not mature enough. Natural language processing is rapidly developed at home and abroad, and a mature system is provided in the aspect of generating scenes of English texts, but the research on the application of Chinese texts, particularly in the aspect of furniture scenes, is relatively deficient.
For the research from text to scenes, there are currently more sophisticated systems: the PUT system is a language-based object-interactive operating system, the input being restricted to a subset consisting of expressions of the form PUT (X py), where X and Y are objects and P is a spatial preposition; the WordsEye system can automatically convert the text into a representative 3D scene, so that a common user can quickly create the 3D scene without learning special software, mastering artistic skills or even clicking a desktop window interface; the core of the 3DSV system is to develop a tool for converting natural language input into a graphical representation; the AVDT system can automatically convert any descriptive text into a representative three-dimensional scene, and realize the creation from natural language to virtual environment.
The key technology of the furniture scene reconstruction and recommendation method based on the consistency of the long text semantics is the construction of a text-based furniture scene semantic network and the recommendation of a furniture scene based on the semantic network. The key point of the text-based furniture scene semantic network construction is to realize furniture entity extraction, furniture spatial relationship extraction and semantic network generation; the key point of recommendation and reconstruction of furniture scenes based on the semantic network is recommendation based on furniture incidence relation and co-occurrence relation and home scene construction based on recommendation results.
The furniture entity extraction and the furniture spatial relationship extraction depend on lexical analysis and syntactic analysis of natural language processing technology. Lexical analysis generally includes four analysis methods, rule-based, statistic-based, semantic-based and understanding-based: the rule-based method is that matching is carried out in a dictionary according to a certain rule; the statistical-based method is to consider the combination of words with high frequency of occurrence in the text as possibly-formed words; the semantic-based approach is to help segment words by semantic analysis; the understanding-based approach is to segment words through syntactic and semantic analysis, looking like to understand a sentence so that words can be recognized. The dependency parsing of the parsing type can visually represent a sentence in the form of a dependency parse book, and related participles are connected by a dependency arc with a relationship tag, thereby reflecting the syntactic relationship between words.
Disclosure of Invention
The invention aims to provide a furniture scene reconstruction and recommendation method based on long text semantic consistency, which is used for solving the problem of furniture scene reconstruction and recommendation based on long text semantic consistency of visual furniture character layout and furniture three-dimensional scene generated from a text.
The invention adopts the following technical scheme:
a furniture scene reconstruction and recommendation method based on long text semantic consistency comprises the following steps:
step 1, extracting a furniture entity according to a text lexical method and a syntax processing result;
step 2, extracting an algorithm according to the spatial relationship of the text syntactic relationship;
step 4, recommending existing furniture in the scene according to the furniture incidence relation and the co-occurrence;
and 5, after the entity furniture and the spatial relationship between the entity furniture are extracted and obtained, generating a three-dimensional scene for the formed semantic network.
Further, in step 1, regarding a participle which is a noun in the text and has a syntactic relationship of one of 'VOB', 'POB', 'SBV', 'COO', 'ATT' as a furniture entity, including:
step 1.1, performing lexical analysis and syntactic analysis on the text to obtain word segmentation, part of speech and syntactic relations;
step 1.2, traversing the word segmentation result list, judging whether the word segmentation is added, if so, executing the step 1.2, otherwise, executing the step 1.3;
step 1.3, judging whether the participle is a noun or not, if so, executing step 1.4, otherwise, executing step 1.2;
step 1.4, judging whether the participle belongs to one of the defined syntactic relations, if so, executing step 1.5, otherwise, executing step 1.2;
and 1.5, adding the participles into a furniture entity list, finishing the algorithm if the traversal is finished, and otherwise, repeating the step 1.2.
Further, in step 2, the part-of-speech tagged with 'nd' participle is regarded as "azimuth word" according to the part-of-speech tagged and dependency syntactic relationship, and the landmark and the shooter in the spatial relationship are found in the predecessor and successor of the azimuth word dependency tree, including:
step 2.1, performing lexical analysis and syntactic analysis on the text to obtain word segmentation, part of speech and syntactic relations;
step 2.2, traversing the word segmentation result list, judging whether the word segmentation part of speech is 'nd', if so, adding the word segmentation part of speech to the direction word list, and otherwise, skipping;
step 2.3, if the traversal is finished, the algorithm is finished, otherwise, the step 2.2 is repeated;
step 2.4, traversing the azimuth word list, judging whether the successor of the azimuth word is in the furniture entity list, if so, dividing the word into the landmark, otherwise, skipping; if the landmark and the shooter are not empty, adding the three groups of the landmark, the shooter and the azimuth word into the relation series table, and ending the algorithm;
step 2.5, traversing the participle result list, and if finding the homothetic language of the identified shooter, adding the participle into the shooter list;
and 2.6, if the landmark and the shooter are not empty, adding the three groups of the landmark, the shooter and the azimuth word into the relation series table, and ending the algorithm.
Further, in step 3, an entity extraction and spatial relationship extraction algorithm is used for processing the database and importing the processed database into a database, all furniture nouns and spatial relationships thereof are constructed into a semantic network and are visualized, wherein the database consists of 12 tables, which are respectively 1 vertex table, 1 edge table, 1 total data table, 1 binary co-occurrence data table and 8 data tables of different rooms, and the method comprises the following steps:
step 3.1, adding the first two elements in all triples in the edge table to the node, and setting the third element as the weight of the edge;
step 3.2, add all elements of the vertex table to the node.
Further, step 4 includes:
step 4.1, acquiring a furniture entity added by a user through text analysis, and importing the furniture entity into a database;
and 4.2, acquiring all furniture entities added by the user from the database, calculating the recommendation probability of each furniture in the database through a recommendation algorithm, and recommending the furniture with the highest probability rank to the user.
Further, step 4 includes:
step 4.1, recommending according to the furniture incidence relation;
step 4.11, acquiring a furniture entity added by the user through text analysis, importing the furniture entity into a database, and marking the furniture entity as data from the user;
step 4.12, acquiring all furniture entities under the scene S added by the user in the database, and traversing the furniture entities;
step 4.13, recommending a furniture entity list { I) which is associated with the furniture entity O and has a front frequency in the scene S according to the existing furniture association relation table in the database1,I2,I3};
Step 4.14, inquiring a furniture incidence relation table in a database, and acquiring a landmark O and a shooter I in a scene Sj∈{I1,I 2,I3Importing the relation triple into a data table of a user and marking the relation triple as data from recommendation;
4.2, carrying out Bayesian recommendation of furniture entity co-occurrence;
step 4.21, acquiring the furniture entity added by the user through text analysis, importing the furniture entity into a database, and marking the furniture entity as data from the user;
step 4.22, acquiring all furniture entity lists { A) under the scene S added by the user in the database1,A2,...,An};
Step 4.23, traversing furniture entities in the furniture database, and calculating the furniture entity B to obtain the probability P (B) of occurrence;
step 4.24, traverse { A1,A2,...,AnAnd combining with a furniture entity B, calculating through a database table
Step 4.25, calculateThe recommendation degree of the furniture entity B under the user can be obtained;
step 4.26, acquiring furniture entities { B with recommendation degrees ranked in the first three1,B2,B3Acquiring all B in the scene S according to an existing furniture incidence relation table in a databasei∈{B1,B2,B3With all Aj∈{A1,A2,...,AnThe spatial relationship with the most frequent frequency in all the spatial relationships of the three (A) forms a triple (A)j,BiR), import into the user's data table and mark it as data from recommendations.
Further, in step 5, generating a three-dimensional scene for the formed semantic network includes:
step 5.1, visualizing a recommendation result text;
step 5.11, acquiring a scene, a furniture entity and a spatial relation according to the text input by the user;
step 5.12, recommending through a recommendation method of entity furniture incidence relation and a Bayes recommendation method based on furniture entity co-occurrence according to the obtained information;
step 5.13, visualizing the recommended text result;
step 5.2, visualizing the three-dimensional scene;
step 5.21, acquiring a scene, a furniture entity and a spatial relationship according to the text input by the user;
step 5.22, preliminarily determining the coordinates of the furniture entity on the premise of ensuring that the furniture entities do not collide with each other;
step 5.23, according to the spatial relationship among all the entities input by the user, storing the furniture entities input by the user in an ordered array according to depth-first traversal;
step 5.24, sequentially adjusting the coordinates of each furniture entity according to the array and the spatial relationship;
and 5.25, drawing the furniture entities in the three-dimensional grid by using different colors according to the coordinates of the furniture entities.
The invention has the following advantages: on the basis of a scene description text corpus obtained through voice recognition, a system from text to scene is constructed, and the research of a Chinese text extraction method under a specific situation is promoted, particularly in the aspect of furniture scenes; through the development and research of text semantic analysis in the household field, a user can obtain a desired household layout design only through spoken language expression without learning special software, mastering artistic skills or even any interface and interactive interface.
Drawings
FIG. 1 is a flowchart of a method for reconstructing and recommending furniture scenes based on semantic consistency of long texts according to an embodiment of the present invention;
FIG. 2 is a dependency tree of the sentence "[ object 1] with [ object 2 ]" and the sentence "[ object 1] with [ object 2 ]", in an embodiment of the present invention;
FIG. 3 is a dependency tree of the sentence "in [ object 1] the [ azimuth word ] is [ object 2 ]";
FIG. 4 is a dependency tree in an embodiment of the present invention "[ Azimuth word of [ object 1] is [ object 2] and [ object 3 ]";
FIG. 5 is a text visualization of a triplet (obj1, obj2, relation) in the semantic network construction algorithm according to the embodiment of the present invention;
FIG. 6 is a partial semantic network visualization of vertex tables and edge tables in an embodiment of the present invention;
FIG. 7 is a visualization in directed graph form of furniture and associations that a user adds in a bedroom in an embodiment of the invention;
FIG. 8 is a text visualization before recommendation based on furniture correlations in an embodiment of the present invention;
FIG. 9 is a text visualization after recommendation based on furniture correlations in an embodiment of the present invention;
FIG. 10 is a text visualization prior to furniture co-occurrence based recommendation in an embodiment of the present invention;
FIG. 11 is a text visualization after a recommendation based on furniture co-occurrence in an embodiment of the present invention;
FIG. 12 is a three-dimensional scene visualization of a scene in an embodiment of the invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments, it being understood that the embodiments and features of the embodiments of the present application can be combined with each other without conflict.
Examples
The basic technical thought of the furniture scene reconstruction and recommendation method based on the consistency of the long text semantics is as follows: on the basis of having a furniture scene description text corpus obtained through voice recognition, a set of appropriate semantic information extraction algorithm is summarized by using a Language Technology Platform (LTP) of Harbin university of industry, the extraction result is recommended by using the recommendation algorithm, and finally, visual furniture character layout and furniture three-dimensional scene are generated by using the user input and the recommended result.
As shown in fig. 1, the furniture scene reconstruction and recommendation method based on long text semantic consistency includes:
step 1, furniture entity extraction based on statistics;
in this embodiment, a rule-based furniture entity extraction method is used, in which a participle that is a noun in a text and has a syntactic relationship of one of 'VOB', 'POB', 'SBV', 'COO', 'ATT' is regarded as a furniture entity, and specifically includes:
step 1.1, performing lexical analysis and syntactic analysis on the Chinese text to obtain word segmentation, part of speech and syntactic relations;
step 1.2, traversing the word segmentation result list, judging whether the word segmentation is added, if so, executing the step 1.2, otherwise, executing the step 1.3;
step 1.3, judging whether the participle is a noun or not, if so, executing step 1.4, otherwise, executing step 1.2;
step 1.4, judging whether the participle belongs to one of the defined syntactic relations, if so, executing step 1.5, otherwise, executing step 1.2;
and 1.5, adding the participles into a furniture entity list, finishing the algorithm if the traversal is finished, and otherwise, repeating the step 1.2.
Step 2, extracting the furniture entity spatial relationship based on the rule;
in this embodiment, a set of rules is used to extract spatial relationships among all entities, where the rules regard a word segmentation with a part of speech tag being 'nd' as an "azimuth word" based on a part of speech tag and a dependency syntactic relationship, and a landmark and a shooter in the spatial relationship are found in a predecessor and a successor in a dependency tree of the azimuth word, and the rules specifically include:
step 2.1, performing lexical analysis and syntactic analysis on the Chinese text to obtain word segmentation, part of speech and syntactic relations;
step 2.2, traversing the word segmentation result list, judging whether the word segmentation part of speech is 'nd', if so, adding the word segmentation part of speech to the direction word list, and otherwise, skipping;
step 2.3, if the traversal is finished, the algorithm is finished, otherwise, the step 2.2 is repeated;
step 2.4, traversing the azimuth word list, and judging whether the successor of the azimuth word is in the furniture entity list, if so, the word segmentation is the landmark, otherwise, skipping; judging whether all successors (except self) of the orientation word predecessor are in the furniture entity list, if so, the word segmentation is one of the shots, otherwise, judging whether all successors of the orientation word predecessor are in the furniture entity list, and if so, the word segmentation is one of the shots;
step 2.5, traversing the participle result list, and if finding the homothetic language of the identified shooter, adding the participle into the shooter list;
and 2.6, if the landmark and the shooter are not empty, adding the three-tuple (landmark, shooter and azimuth word) into the relation series table, and ending the algorithm.
In this embodiment, the specific determination of the landmarks and shots is as follows: as shown in fig. 2, in the dependency tree of the sentence "[ object 1] whose [ azimuth word ] is [ object 2 ]" and the sentence "[ object 1] in [ azimuth word ] of [ object 2], the successor node of the azimuth word is the landmark, and the successor node of the predecessor node of the azimuth word is the shooter; as shown in fig. 3, in the dependency tree whose sentence "in [ object 1] the [ azimuth word ] is [ object 2 ]", the successor node of the azimuth word is the landmark, and the successor node of the predecessor node of the azimuth word is the shooter; as shown in fig. 4, the [ azimuth words ] for the sentence pattern "[ object 1] are [ object 2] and [ object 3 ]. "one landmark corresponds to a plurality of shots, in the dependency tree, the successor node of the azimuth word is the landmark, the successor node of the predecessor node of the azimuth word is the shot, and the participle whose dependency relationship with the shot is 'COO' is also the shot. The dependency parsing notation and part-of-speech tagging notation referred to in fig. 2, 3, and 4 are illustrated in the following table:
TABLE 1 dependency parsing notation and part-of-speech tag notation specification
in this embodiment, a corpus is processed and imported into a database through an entity extraction and spatial relationship extraction algorithm, all furniture nouns and their spatial relationships (if any) are constructed into a semantic network, and the semantic network is visualized, wherein the database is composed of 12 tables, which are respectively 1 vertex table, 1 edge table, 1 total data table, 1 binary co-occurrence data table, and data tables of 8 different rooms, and specifically includes:
step 3.1, adding the first two elements in all triples in the edge table to nodes (not repeated), and setting the third element as the weight of the edge;
step 3.2, add all elements of the vertex table to the node (not repeated);
the triple form is (obj1, obj2, relation), and the text visualization form is shown in fig. 5.
Performing text visualization on the basis of a semantic network, including performing whole and local semantic network visualization on the basis of a vertex table and an edge table in a database, performing scene text visualization on each scene, and performing local semantic network visualization on the vertex table and the edge table as shown in fig. 6; as shown in fig. 7, is part of a visualization in the form of a directed graph of furniture and their associations that a user adds in a bedroom.
Step 4, recommending furniture scenes based on a semantic network;
in the embodiment, the recommendation based on the furniture incidence relation and the furniture co-occurrence is completed;
acquiring a furniture entity added by a user through text analysis, and importing the furniture entity into a database;
acquiring all furniture entities added by a user in a database, calculating the recommendation probability of each furniture in the database through a certain recommendation algorithm, and recommending some furniture with the highest probability rank to the user;
step 4.1, recommending furniture with a certain spatial relationship in a certain scene to a user based on the entity furniture incidence relation, namely recommending furniture with a certain spatial relationship possibly appearing in the scene to each furniture appearing in the scene when the user inputs certain furniture in the scene, specifically comprising:
step 4.11, acquiring a furniture entity added by the user through text analysis, importing the furniture entity into a database, and marking the furniture entity as data from the user;
step 4.12, acquiring all furniture entities in a certain scene S added by a user from the database, and traversing the furniture entities;
step 4.13, recommending a furniture entity list { I) which is associated with the furniture entity O and has a front frequency in the scene S according to the existing furniture association relation table in the database1,I2,I3};
Step 4.14, inquiring a furniture incidence relation table in a database, and acquiring a landmark O and a shooter I in a scene Sj∈{I1,I2,I3The relation triple of the user is imported into a data table of the user and marked as data from recommendation;
step 4.2, based on the Bayesian recommendation of furniture entity co-occurrence, that is, when the user inputs some furniture in a certain scene, recommending the furniture possibly appearing in the scene simultaneously with the furniture to the user, specifically including:
step 4.21, acquiring the furniture entity added by the user through text analysis, importing the furniture entity into a database, and marking the furniture entity as data from the user;
step 4.22, acquiring all furniture entity lists { A) under a certain scene S added by a user from the database1,A2,...,An};
Step 4.23, traversing furniture entities in the furniture database, and calculating the furniture entity B to obtain the probability P (B) of occurrence;
step 4.24, traverse { A1,A2,...,AnAnd combining with a furniture entity B, calculating through a database table
Step 4.25, calculateThe recommendation degree of the furniture entity B under the user can be obtained;
step 4.26, acquiring furniture entities { B with recommendation degrees ranked in the first three1,B2,B3Acquiring all B in the scene S according to an existing furniture incidence relation table in a databasei∈{B1,B2,B3With all Aj∈{A1,A2,...,AnThe spatial relationship with the most frequent frequency in all the spatial relationships of the three (A) forms a triple (A)j,BiR), import into the user's data table and mark it as data from recommendations.
Step 5, building a home scene based on a recommendation result;
step 5.1, visualizing the recommended result text, wherein the operation steps are as follows:
step 5.11, acquiring a text input by a user, and acquiring a scene, a furniture entity and a spatial relationship;
step 5.12, based on the information obtained in the step 5.11, recommending by a recommending method based on entity furniture incidence relation and a Bayes recommending method based on furniture entity co-occurrence;
as shown in fig. 8 and 9, before recommendation based on the furniture incidence relation and after recommendation based on the furniture incidence relation are respectively visualized; as shown in fig. 10 and 11, text visualization before and after the furniture co-occurrence recommendation is performed, respectively;
step 5.13, visualizing the recommended text result;
step 5.2, visualizing the three-dimensional scene, wherein the operation steps are as follows:
step 5.21, acquiring a scene, a furniture entity and a spatial relationship according to the text input by the user;
step 5.22, based on certain priori knowledge, preliminarily determining the coordinates of the furniture entity on the premise of ensuring that the furniture entities do not collide with each other;
step 5.23, according to the spatial relationship among all the entities input by the user, storing the furniture entities input by the user in an ordered array according to depth-first traversal;
step 5.24, sequentially adjusting the coordinates of each furniture entity based on the array and the spatial relationship;
step 5.25, all furniture entities are rendered in a three-dimensional grid based on their coordinates, and fig. 12 shows a three-dimensional scene visualization of a scene.
In this embodiment, the furniture scene construction depends on a scene visualization method, which includes a large-scale scene fast drawing method and a basic scene data visualization method, and may be operated by using Gephi software, where Gephi is an open source software that can be used for graph and network analysis, and completes real-time display of a large network by using a 3D rendering engine, so as to accelerate the exploration speed. Graphciz is a heterogeneous graphic drawing working set, which comprises a batch layout program (dot, neighbor, fdp, twopi), an incremental layout platform (Dynagraph), a custom graphic editor (dotty, Grappa), a server (Webdot) containing graphics in a webpage, supporting graphics as COM objects (Montage), a graphic visualization utility program and a library for property maps, and provides three batch processing algorithms, namely a hierarchical layout algorithm, a symmetric layout algorithm and a circular layout algorithm, which all allow wide parameterization.
The above description is only for the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the present disclosure should be covered within the scope of the present invention claimed in the appended claims.
Claims (7)
1. A furniture scene reconstruction and recommendation method based on long text semantic consistency is characterized by comprising the following steps:
step 1, extracting a furniture entity according to a text lexical method and a syntax processing result;
step 2, extracting an algorithm according to the spatial relationship of the text syntactic relationship;
step 3, forming a semantic network according to the home entity and the text data of the spatial relationship thereof and visualizing the semantic network;
step 4, recommending existing furniture in the scene according to the furniture incidence relation and the co-occurrence;
and 5, after the entity furniture and the spatial relationship between the entity furniture are extracted and obtained, generating a three-dimensional scene for the formed semantic network.
2. The method as claimed in claim 1, wherein the step 1 of regarding the participles with syntactic relation being one of VOB, POB, SBV, COO, ATT in text as furniture entities comprises:
step 1.1, performing lexical analysis and syntactic analysis on the text to obtain word segmentation, part of speech and syntactic relations;
step 1.2, traversing the word segmentation result list, judging whether the word segmentation is added, if so, executing the step 1.2, otherwise, executing the step 1.3;
step 1.3, judging whether the participle is a noun or not, if so, executing step 1.4, otherwise, executing step 1.2;
step 1.4, judging whether the participle belongs to one of the defined syntactic relations, if so, executing step 1.5, otherwise, executing step 1.2;
and 1.5, adding the participles into a furniture entity list, finishing the algorithm if the traversal is finished, and otherwise, repeating the step 1.2.
3. The method for reconstructing and recommending furniture scenes based on long-text semantic consistency according to claim 1, wherein in step 2, the part of speech tagged with nd participles is regarded as an orientation word according to the syntactic relation of part of speech tagging and dependency, and landmarks and arches in the spatial relation are found in predecessors and successors in an orientation word dependency tree, and the method comprises the following steps:
step 2.1, performing lexical analysis and syntactic analysis on the text to obtain word segmentation, part of speech and syntactic relations;
step 2.2, traversing the word segmentation result list, judging whether the part of speech of the segmented word is nd, if so, adding the segmented word to the direction word list, and otherwise, skipping;
step 2.3, if the traversal is finished, the algorithm is finished, otherwise, the step 2.2 is repeated;
step 2.4, traversing the azimuth word list, judging whether the successor of the azimuth word is in the furniture entity list, if so, dividing the word into the landmark, otherwise, skipping; if the landmark and the shooter are not empty, adding the three groups of the landmark, the shooter and the azimuth word into the relation series table, and ending the algorithm;
step 2.5, traversing the participle result list, and if finding the homothetic language of the identified shooter, adding the participle into the shooter list;
and 2.6, if the landmark and the shooter are not empty, adding the three groups of the landmark, the shooter and the azimuth word into the relation series table, and ending the algorithm.
4. The method for reconstructing and recommending furniture scenes based on consistent long text semantics according to claim 1, wherein in step 3, an entity extraction and spatial relationship extraction algorithm is used for processing a corpus and importing the corpus into a database, and all furniture nouns and spatial relationships thereof are constructed into a semantic network and visualized, wherein the database is composed of 12 tables, which are 1 vertex table, 1 edge table, 1 total data table, 1 binary co-occurrence data table and 8 data tables of different rooms, and comprises: step 3.1, adding the first two elements in all triples in the edge table to the node, and setting the third element as the weight of the edge; step 3.2, add all elements of the vertex table to the node.
5. The method for reconstructing and recommending furniture scenes based on semantic consistency of long texts according to claim 4, wherein the step 4 comprises:
step 4.1, acquiring a furniture entity added by a user through text analysis, and importing the furniture entity into a database;
and 4.2, acquiring all furniture entities added by the user from the database, calculating the recommendation probability of each furniture in the database through a recommendation algorithm, and recommending the furniture with the highest probability rank to the user.
6. The method for reconstructing and recommending furniture scenes based on semantic consistency of long texts according to claim 5, wherein the step 4 comprises:
step 4.1, recommending according to the furniture incidence relation;
step 4.11, acquiring a furniture entity added by the user through text analysis, importing the furniture entity into a database, and marking the furniture entity as data from the user;
step 4.12, acquiring all furniture entities under the scene S added by the user in the database, and traversing the furniture entities;
step 4.13, recommending a furniture entity list { I) which is associated with the furniture entity O and has a front frequency in the scene S according to the existing furniture association relation table in the database1,I2,I3};
Step 4.14, inquiring a furniture incidence relation table in a database, and acquiring a landmark O and a shooter I in a scene Sj∈{I1,I2,I3Importing the relation triple into a data table of a user and marking the relation triple as data from recommendation;
4.2, carrying out Bayesian recommendation of furniture entity co-occurrence;
step 4.21, acquiring the furniture entity added by the user through text analysis, importing the furniture entity into a database, and marking the furniture entity as data from the user;
step 4.22, acquiring all furniture entity lists { A) under the scene S added by the user in the database1,A2,...,An};
Step 4.23, traversing furniture entities in the furniture database, and calculating the furniture entity B to obtain the probability P (B) of occurrence;
step 4.24, traverseA1,A2,...,AnAnd combining with a furniture entity B, calculating through a database table
Step 4.25, calculateThe recommendation degree of the furniture entity B under the user can be obtained;
step 4.26, acquiring furniture entities { B with recommendation degrees ranked in the first three1,B2,B3Acquiring all B in the acquired scene S according to an existing furniture incidence relation table in a databasei∈{B1,B2,B3With all Aj∈{A1,A2,...,AnThe spatial relationship with the most frequent frequency in all the spatial relationships of the three (A) forms a triple (A)j,BiR), import into the user's data table and mark it as data from recommendations.
7. The method for reconstructing and recommending furniture scenes based on semantic consistency of long texts according to claim 5, wherein in step 5, generating three-dimensional scenes for the formed semantic network comprises:
step 5.1, visualizing a recommendation result text;
step 5.11, acquiring a scene, a furniture entity and a spatial relation according to the text input by the user;
step 5.12, recommending through a recommendation method of entity furniture incidence relation and a Bayes recommendation method based on furniture entity co-occurrence according to the obtained information;
step 5.13, visualizing the recommended text result;
step 5.2, visualizing the three-dimensional scene;
step 5.21, acquiring a scene, a furniture entity and a spatial relationship according to the text input by the user;
step 5.22, preliminarily determining the coordinates of the furniture entity on the premise of ensuring that the furniture entities do not collide with each other;
step 5.23, according to the spatial relationship among all the entities input by the user, storing the furniture entities input by the user in an ordered array according to depth-first traversal;
step 5.24, sequentially adjusting the coordinates of each furniture entity according to the array and the spatial relationship;
and 5.25, drawing the furniture entities in the three-dimensional grid by using different colors according to the coordinates of the furniture entities.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111667409.5A CN114385873A (en) | 2021-12-31 | 2021-12-31 | Furniture scene reconstruction and recommendation method based on long text semantic consistency |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111667409.5A CN114385873A (en) | 2021-12-31 | 2021-12-31 | Furniture scene reconstruction and recommendation method based on long text semantic consistency |
Publications (1)
Publication Number | Publication Date |
---|---|
CN114385873A true CN114385873A (en) | 2022-04-22 |
Family
ID=81200134
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111667409.5A Pending CN114385873A (en) | 2021-12-31 | 2021-12-31 | Furniture scene reconstruction and recommendation method based on long text semantic consistency |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114385873A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9910914B1 (en) * | 2016-05-05 | 2018-03-06 | Thomas H. Cowley | Information retrieval based on semantics |
CN112144918A (en) * | 2020-08-19 | 2020-12-29 | 广州光建通信技术有限公司 | 5G Internet of things visual intelligent learning and entertainment space |
US20210110457A1 (en) * | 2019-10-09 | 2021-04-15 | Target Brands, Inc. | Compatibility based furniture recommendations |
US20210304072A1 (en) * | 2020-03-26 | 2021-09-30 | Tata Consultancy Services Limited | Method and system for unsupervised multi-modal set completion and recommendation |
-
2021
- 2021-12-31 CN CN202111667409.5A patent/CN114385873A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9910914B1 (en) * | 2016-05-05 | 2018-03-06 | Thomas H. Cowley | Information retrieval based on semantics |
US20210110457A1 (en) * | 2019-10-09 | 2021-04-15 | Target Brands, Inc. | Compatibility based furniture recommendations |
US20210304072A1 (en) * | 2020-03-26 | 2021-09-30 | Tata Consultancy Services Limited | Method and system for unsupervised multi-modal set completion and recommendation |
CN112144918A (en) * | 2020-08-19 | 2020-12-29 | 广州光建通信技术有限公司 | 5G Internet of things visual intelligent learning and entertainment space |
Non-Patent Citations (1)
Title |
---|
周全;卞景帅;饶永明;罗月童;: "基于布局模板的家具自动布局方法研究与应用", 合肥工业大学学报(自然科学版), no. 06, 28 June 2020 (2020-06-28) * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111259653B (en) | Knowledge graph question-answering method, system and terminal based on entity relationship disambiguation | |
EP4040309A1 (en) | Systems and methods of applying pragmatics principles for interaction with visual analytics | |
Li et al. | Composing simple image descriptions using web-scale n-grams | |
Velardi et al. | How to encode semantic knowledge: a method for meaning representation and computer-aided acquisition | |
CN102955848B (en) | A kind of three-dimensional model searching system based on semanteme and method | |
AU2019226135A1 (en) | Natural language processing for language-driven synthesis of 3d scenes | |
Curto et al. | Question generation based on lexico-syntactic patterns learned from the web | |
Smith et al. | Evaluating visual representations for topic understanding and their effects on manually generated topic labels | |
JP2003308320A (en) | System for realizing sentence | |
CN111177591A (en) | Knowledge graph-based Web data optimization method facing visualization demand | |
CN112784598A (en) | Method, device and equipment for generating thinking guide graph and storage medium | |
CN112148886A (en) | Method and system for constructing content knowledge graph | |
CN113449066B (en) | Method, processor and storage medium for storing cultural relic data by using knowledge graph | |
Reif et al. | Visualizing linguistic diversity of text datasets synthesized by large language models | |
Jayashree et al. | Multimodal web page segmentation using self-organized multi-objective clustering | |
Navigli et al. | Enriching a formal ontology with a thesaurus: an application in the cultural heritage domain | |
Lu et al. | A novel knowledge-based system for interpreting complex engineering drawings: Theory, representation, and implementation | |
Kordomatis et al. | Web object identification for web automation and meta-search | |
Pinheiro et al. | ChartText: Linking Text with Charts in Documents | |
CN114385873A (en) | Furniture scene reconstruction and recommendation method based on long text semantic consistency | |
Kolthoff et al. | GUI2WiRe: rapid wireframing with a mined and large-scale GUI repository using natural language requirements | |
Riehmann et al. | The NETSPEAK WORDGRAPH: Visualizing keywords in context | |
US20050154750A1 (en) | Methods and apparatus for generating automated graphics using stored graphics examples | |
Liang et al. | A semantic‐driven generation of 3D Chinese opera performance scenes | |
Yamazaki et al. | Learning hierarchies from ambiguous natural language data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |