CN113705198A - Scene graph generation method and device, electronic equipment and storage medium - Google Patents
Scene graph generation method and device, electronic equipment and storage medium Download PDFInfo
- Publication number
- CN113705198A CN113705198A CN202111224754.1A CN202111224754A CN113705198A CN 113705198 A CN113705198 A CN 113705198A CN 202111224754 A CN202111224754 A CN 202111224754A CN 113705198 A CN113705198 A CN 113705198A
- Authority
- CN
- China
- Prior art keywords
- word
- entity
- pair
- relationship
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000011218 segmentation Effects 0.000 claims abstract description 47
- 230000001419 dependent effect Effects 0.000 claims description 55
- 238000012545 processing Methods 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 10
- 230000000875 corresponding effect Effects 0.000 description 60
- 238000010586 diagram Methods 0.000 description 17
- 238000013136 deep learning model Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000001404 mediated effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000007935 neutral effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—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
- G06F40/20—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
- G06F40/295—Named entity recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Machine Translation (AREA)
Abstract
The utility model relates to a scene graph generation method, a device, an electronic device and a storage medium, wherein the method comprises the steps of obtaining a word segmentation sequence of a target Chinese sentence and part-of-speech information of words in the word segmentation sequence; inputting the word segmentation sequence and the part-of-speech information into a syntactic dependency model for syntactic dependency analysis to obtain word pairs with dependency relationships and word relationship data corresponding to the dependency relationships; identifying words belonging to the entity in the word segmentation sequence according to the part of speech information; determining a first target word pair with a medium dependency relationship and a second target word pair with a target dependency relationship according to the word relationship data; generating entity attribute information based on the first target word pair and words belonging to the entity; generating entity relation information representing the incidence relation between the entities based on the second target word pair; and generating a scene graph corresponding to the target Chinese sentence based on the entity attribute information and the entity relation information. By utilizing the method and the device, the scene graph of the Chinese sentence can be generated quickly and accurately.
Description
Technical Field
The present disclosure relates to the field of computer vision technologies, and in particular, to a method and an apparatus for generating a scene graph, an electronic device, and a storage medium.
Background
With the development of computer vision technology, the understanding of scenes is developed from classification, detection and segmentation to the description of natural language, which has higher-level understanding and reasoning on visual scenes. Unlike unstructured natural language, a scene graph is a structured representation of a scene, which facilitates the processing of machine learning models.
In the related art, a syntactic dependency tree is often generated from an english sentence, and then the relationship in the syntactic dependency tree is converted into the relationship required in a scene graph according to the english grammar, but the rules used in extracting the scene graph in the related art are based on the english grammar and are not suitable for chinese. Therefore, a method for generating a scene graph for a chinese sentence is needed.
Disclosure of Invention
The present disclosure provides a scene graph generation method, apparatus, electronic device, and storage medium, to at least solve the problem in the related art that a scene graph of a chinese sentence cannot be generated. The technical scheme of the disclosure is as follows:
according to a first aspect of the embodiments of the present disclosure, there is provided a scene graph generation method, including:
acquiring a word segmentation sequence of a target Chinese sentence and part-of-speech information of words in the word segmentation sequence;
inputting the word segmentation sequence and the part of speech information into a syntactic dependency model for syntactic dependency analysis to obtain word relation data, wherein the word relation data comprises word pairs with dependency relations and dependency relations corresponding to the word pairs;
identifying words belonging to the entity in the word segmentation sequence according to the part of speech information;
determining a first target word pair with a medium dependency relationship and a second target word pair with a target dependency relationship according to the word relationship data; the target dependency relationship is the dependency relationship that the corresponding word pair contains entity associated words, and the entity associated words are words representing the association relationship among the entities;
generating entity attribute information based on the first target word pair and the words belonging to the entity;
generating entity relation information based on the second target word pair, wherein the entity relation information represents the incidence relation between every two entities;
and generating a scene graph corresponding to the target Chinese statement based on the entity attribute information and the entity relation information.
Optionally, the first target word pair includes: a word pair comprising a core word and a dependency word; the generating entity attribute information based on the first target word pair and the words belonging to the entity comprises:
determining a third target word pair of the dependency word belonging to the entity from the first target word pair according to the words belonging to the entity;
and generating the entity attribute information by taking the dependent words in the third target word pair as entities and the core words in the third target word pair as attributes of corresponding entities.
Optionally, the association relationship includes an action association relationship, and the target dependency relationship includes: a main and subordinate relationship and an action and guest relationship; the second target word pair includes: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary-to-predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-to-guest relationship;
the generating entity relationship information based on the second target term pair comprises:
matching the dependency word in each first subword pair with the core word in each second subword pair to obtain a matching word pair group, wherein the dependency word in the first subword pair contained in the matching word pair group is consistent with the core word in the second subword pair contained in the matching word pair group;
and taking a core word in a first sub-word pair contained in the matching word pair group and a dependent word in a second sub-word contained in the matching word pair group as a first entity pair, and taking a dependent word in the first sub-word pair contained in the matching word pair group or a core word in the second sub-word contained in the matching word pair group as an action incidence relation corresponding to the first entity pair to generate the entity relation information.
Optionally, the association relationship includes a location association relationship, and the target dependency relationship includes: a guest-mediating relationship; the second target word pair includes: the dependency relationship is a word pair containing a core word and a dependency word of the intermediary relationship;
the generating entity relationship information based on the second target term pair comprises:
determining that the dependent word is a fourth target word pair of the dependent words in the second target word pair and the dependent word is a fifth target word pair of the core word in the second target word pair according to the word relation data;
and generating the entity relationship information by taking the core word in the fourth target word pair and the core word in the fifth target word pair as a second entity pair, and taking the core word in the second target word pair as a position association relationship corresponding to the second entity pair.
Optionally, the entity attribute information includes an entity and an attribute of the entity, and the entity relationship information includes an entity pair and an association relationship corresponding to the entity pair; the generating a scene graph corresponding to the target chinese statement based on the entity attribute information and the entity relationship information includes:
taking an entity in the entity attribute information as a first type node, and taking an attribute of the entity in the entity attribute information as a second type node;
and constructing the scene graph by taking the corresponding association relationship of the entity pair in the entity relationship information as the corresponding edge between the first type nodes, and taking the dependency relationship between the attributes of the entity and the entity as the corresponding edge between the first type nodes and the second type nodes.
Optionally, the method further includes:
acquiring a Chinese text to be processed;
dividing the Chinese text to be processed into a plurality of sentences based on a regular expression with characters to be matched as preset punctuations;
and taking any one of the sentences as the target Chinese sentence.
Optionally, the part-of-speech information of the words in the word segmentation sequence is obtained by adopting the following method:
and inputting the word segmentation sequence into a part-of-speech recognition model for part-of-speech recognition processing to obtain part-of-speech information of words in the word segmentation sequence.
According to a second aspect of the embodiments of the present disclosure, there is provided a scene graph generating apparatus, including:
the information acquisition module is configured to execute acquisition of a word segmentation sequence of a target Chinese sentence and part-of-speech information of words in the word segmentation sequence;
a syntactic dependency analysis module configured to perform syntactic dependency analysis on the segmented word sequence and the part-of-speech information input into a syntactic dependency model to obtain word relationship data, wherein the word relationship data includes word pairs with dependency relationships and dependency relationships corresponding to the word pairs;
the entity identification module is configured to identify words belonging to the entity in the word segmentation sequence according to the part of speech information;
a target word pair determination module configured to perform determining, according to the word relation data, a first target word pair whose dependency relationship is a fixed relationship and a second target word pair whose dependency relationship is a target dependency relationship; the target dependency relationship is the dependency relationship that the corresponding word pair contains entity associated words, and the entity associated words are words representing the association relationship among the entities;
an entity attribute information generation module configured to perform generation of entity attribute information based on the first target word pair and the words belonging to the entity;
the entity relation information generation module is configured to execute generation of entity relation information based on the second target word pair, and the entity relation information represents an incidence relation between every two entities;
and the scene graph generating module is configured to execute generating a scene graph corresponding to the target Chinese statement based on the entity attribute information and the entity relation information.
Optionally, the first target word pair includes: a word pair comprising a core word and a dependency word; the entity attribute information generation module includes:
a third target word pair determination unit configured to perform determination of a third target word pair, in which a dependent word belongs to an entity, from the first target word pair according to the word belonging to the entity;
and the entity attribute information generating unit is configured to execute the step of generating the entity attribute information by taking the dependent words in the third target word pair as entities and the core words in the third target word pair as attributes of corresponding entities.
Optionally, the association relationship includes an action association relationship, and the target dependency relationship includes: a main and subordinate relationship and an action and guest relationship; the second target word pair includes: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary-to-predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-to-guest relationship;
the entity relationship information generation module comprises:
the word pair matching unit is configured to perform matching processing on a dependent word in each first subword pair and a core word in each second subword pair to obtain a matching word pair group, wherein the dependent word in the first subword pair contained in the matching word pair group is consistent with the core word in the second subword pair contained in the matching word pair group;
a first entity relationship information generating unit, configured to execute an action association relationship that a core word in a first subword pair included in the matching word pair group and a dependent word in a second subword included in the matching word pair group are a first entity pair, and a dependent word in the first subword pair included in the matching word pair group or a core word in the second subword included in the matching word pair group is a corresponding first entity pair, so as to generate the entity relationship information.
Optionally, the association relationship includes a location association relationship, and the target dependency relationship includes: a guest-mediating relationship; the second target word pair includes: the dependency relationship is a word pair containing a core word and a dependency word of the intermediary relationship;
the entity relationship information generation module comprises:
a target word pair determination unit configured to perform determining, according to the word relation data, that a dependent word is a fourth target word pair of the dependent words in the second target word pair, and that a dependent word is a fifth target word pair of the core word in the second target word pair;
a second entity relationship information generating unit, configured to execute a position association relationship that the core word in the fourth target word pair and the core word in the fifth target word pair are the second entity pair and the core word in the second target word pair is the second entity pair, so as to generate the entity relationship information.
Optionally, the entity attribute information includes an entity and an attribute of the entity, and the entity relationship information includes an entity pair and an association relationship corresponding to the entity pair; the scene graph generation module comprises:
a node determining unit configured to perform the operation of regarding an entity in the entity attribute information as a first type node and regarding an attribute of the entity in the entity attribute information as a second type node;
and the scene graph building unit is configured to execute building of the scene graph by taking the corresponding association relationship of the entity pair in the entity relationship information as the corresponding edge between the first type nodes, and taking the dependency relationship between the attributes of the entity and the entity as the corresponding edge between the first type nodes and the second type nodes.
Optionally, the apparatus further comprises:
the to-be-processed Chinese text acquisition module is configured to execute acquisition of the to-be-processed Chinese text;
the sentence dividing module is configured to execute a regular expression which takes the characters to be matched as preset punctuation marks, and divide the Chinese text to be processed into a plurality of sentences;
a target Chinese sentence determination module configured to execute taking any one of the plurality of sentences as the target Chinese sentence.
Optionally, the information obtaining module includes:
and the part-of-speech recognition processing unit is configured to execute part-of-speech recognition processing of the part-of-speech recognition model input by the word segmentation sequence to obtain part-of-speech information of the words in the word segmentation sequence.
According to a third aspect of the embodiments of the present disclosure, there is provided an electronic apparatus including: a processor; a memory for storing the processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method of any of the first aspects above.
According to a fourth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium, wherein instructions, when executed by a processor of an electronic device, enable the electronic device to perform the method of any one of the first aspects of the embodiments of the present disclosure.
According to a fifth aspect of the embodiments of the present disclosure, there is provided a computer program product containing instructions which, when run on a computer, cause the computer to perform the method of any one of the first aspects of the embodiments of the present disclosure.
The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:
the method comprises the steps of performing syntactic dependency analysis by combining word segmentation sequences and part-of-speech information of a target Chinese sentence to obtain word pairs with dependency relationships in the target Chinese sentence and dependency relationships corresponding to the word pairs, positioning words belonging to entities by combining the part-of-speech information, and then generating entity attribute information and entity relationship information representing the association relationship between every two entities based on the result of the syntactic dependency analysis.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the disclosure and are not to be construed as limiting the disclosure.
FIG. 1 is a flow diagram illustrating a method for generating a scene graph in accordance with an exemplary embodiment;
FIG. 2 is a flow diagram illustrating generation of entity attribute information based on a first target term pair and terms belonging to an entity, according to an example embodiment;
FIG. 3 is a flow diagram illustrating the generation of entity relationship information based on a second target word pair, according to an example embodiment;
FIG. 4 is a flow diagram illustrating another generation of entity relationship information based on a second target word pair, according to an example embodiment;
FIG. 5 is a flowchart illustrating a method for generating a scene graph corresponding to a target Chinese statement based on entity attribute information and entity relationship information, according to an example embodiment;
FIG. 6 is a schematic diagram of a scene graph provided in accordance with an exemplary embodiment;
FIG. 7 is a schematic diagram of another scene graph provided in accordance with an exemplary embodiment;
FIG. 8 is a block diagram illustrating a scene graph generation apparatus according to an exemplary embodiment;
FIG. 9 is a block diagram illustrating an electronic device for scenegraph generation in accordance with an exemplary embodiment;
FIG. 10 is a block diagram illustrating another electronic device for scenegraph generation in accordance with an exemplary embodiment.
Detailed Description
In order to make the technical solutions of the present disclosure better understood by those of ordinary skill in the art, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.
It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the above-described drawings are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the disclosure described herein are capable of operation in sequences other than those illustrated or otherwise described herein. The implementations described in the exemplary embodiments below are not intended to represent all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present disclosure, as detailed in the appended claims.
It should be noted that, the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for presentation, analyzed data, etc.) referred to in the present disclosure are information and data authorized by the user or sufficiently authorized by each party.
Referring to fig. 1, fig. 1 is a flowchart illustrating a scene graph generating method according to an exemplary embodiment, where as shown in fig. 1, the scene graph generating method is used in a terminal and a server electronic device, and includes the following steps.
In step S101, a word segmentation sequence of a target chinese sentence and part-of-speech information of words in the word segmentation sequence are acquired.
In practical applications, the target chinese sentence may be any chinese sentence for which a corresponding scene graph needs to be generated. Alternatively, the target chinese sentence may be a sentence extracted from a chinese text. Accordingly, in an optional embodiment, the method may further include:
acquiring a Chinese text to be processed;
dividing the Chinese text to be processed into a plurality of sentences based on the regular expression with the characters to be matched as the preset punctuations;
any one of the plurality of sentences is used as a target Chinese sentence.
In an alternative embodiment, the chinese text to be processed may be any chinese text, such as a novel, news, etc. The predetermined punctuation can be a chinese punctuation. Specifically, in the process of dividing the to-be-processed chinese text into a plurality of sentences based on the regular expression in which the to-be-matched character is the preset punctuation mark, the preset punctuation mark may be matched from the to-be-processed chinese text, and optionally, if the preset punctuation mark is matched in the to-be-processed chinese text, the chinese character from the last preset punctuation mark of the currently matched preset punctuation mark to the currently matched preset punctuation mark may be used as a sentence.
In the embodiment, the characters to be matched are combined with the regular expression with the preset punctuations, so that the Chinese text to be processed can be quickly and accurately divided into a plurality of sentences, and further the scene graph of different sentences can be conveniently generated subsequently.
In an alternative embodiment, the word segmentation sequence may include a plurality of words in the target chinese sentence, and specifically, any word may include at least one word. Optionally, a word segmentation tool, a HanLP (Han Language Processing package) or other word segmentation tools may be combined to perform word segmentation Processing on the target chinese sentence, so as to obtain a word segmentation sequence.
In an optional embodiment, the part-of-speech information of the words in the word segmentation sequence may be obtained by:
and inputting the word segmentation sequence into a part-of-speech recognition model for part-of-speech recognition processing to obtain part-of-speech information of the words in the word segmentation sequence.
In an optional embodiment, the part-of-speech recognition model may be obtained by training a preset deep learning model in advance based on the sample word segmentation sequence and part-of-speech information corresponding to the sample word segmentation sequence.
In a specific embodiment, it is assumed that the word segmentation sequence is:
wherein
is n words in the word segmentation sequence. Correspondingly, the part-of-speech information output by the part-of-speech recognition model may be:
wherein,
are respectively as
Corresponding part of speech information.
In the embodiment, the part-of-speech recognition processing is performed on the part-of-speech sequence by combining with the part-of-speech recognition model, so that part-of-speech information of words in the part-of-speech sequence can be quickly and accurately recognized, and data support is further provided for the subsequent scene graph generation.
In step S103, the segmentation sequence and the part-of-speech information are input to a syntactic dependency model for syntactic dependency analysis, and word relationship data is obtained.
In this specification embodiment, the word relationship data may include word pairs having dependency relationships and dependency relationships corresponding to the word pairs. In one particular embodiment, the word relationship data may be data in the form of triples, e.g.,
wherein,
meaning term
Words and phrases
The dependency relationship between the two is
And, optionally,
is a core word and is a word with a certain meaning,
(and
words with dependencies) are dependent words.
In a specific embodiment, the dependency relationship corresponding to a word pair may be a dependency relationship between two words in the word pair, and specifically, the dependency relationship may include, but is not limited to, a predicate relationship, a dynamic guest relationship, a mediated guest relationship, a middle-form relationship, a centering relationship, a parallel relationship, and the like.
In a specific embodiment, the syntactic dependency model may be obtained by training a preset deep learning model in advance based on the sample word segmentation sequence, the corresponding part-of-speech information, and word relationship data corresponding to the sample word segmentation sequence.
In step S105, identifying words belonging to the entity in the word segmentation sequence according to the part of speech information;
in a specific embodiment, a word whose part-of-speech information is a noun may be taken as an entity. Specifically, the nouns may include, but are not limited to, common nouns, other proper names, and the like.
In a specific embodiment, the entity may be an object in a certain scene, and the object may be information of people, objects, and the like, which constitute the scene.
In step S107, a first target word pair whose dependency relationship is a neutral relationship and a second target word pair whose dependency relationship is a target dependency relationship are determined from the word relationship data.
In this specification, the target dependency relationship is a dependency relationship in which a corresponding word pair includes an entity related word, and specifically, the entity related word may be a word representing a relationship between entities.
In a particular embodiment, the association relationship between the entities may include, but is not limited to, at least one of an action association relationship and a location association relationship. Specifically, a connection established based on a certain action exists between entities with action association relation; the relation established based on a certain position relation exists between the entities with the position association relation.
In a specific embodiment, in the case that the association relationship includes an action association relationship, the target dependency relationship may be a predicate relationship and an action-guest relationship. Accordingly, the second target word pair may include: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-guest relationship.
In a particular embodiment, where the associations include location associations, the target dependencies may be intermediaries. Accordingly, the second target word pair may include: the dependency relationship is a word pair containing a core word and a dependency word of the intervent relationship.
In step S109, entity attribute information is generated based on the first target word pair and the words belonging to the entity;
in an alternative embodiment, the first target word pair may include: a word pair comprising a core word and a dependency word; optionally, as shown in fig. 2, the generating entity attribute information based on the first target term pair and the terms belonging to the entity may include the following steps:
in step S1091, determining, from the first target word pair, a third target word pair in which the dependent word belongs to the entity, according to the word belonging to the entity;
in step S1093, entity attribute information is generated by using the dependent word in the third target word pair as the entity and the core word in the third target word pair as the attribute of the corresponding entity.
In a particular embodiment, the entity attribute information may include entities and attributes of the entities. In particular, an attribute of an entity may be a word used to modify the entity. Specifically, the word belonging to the entity may be combined to determine that the first target word pair and the dependent word belong to a third target word pair of the entity, and the dependent word in the third target word pair is used as the entity, and the corresponding core word is used as the attribute for modifying the entity.
In the foregoing embodiment, the third target word pair of the entity to which the dependent word belongs is determined from the first target word pair having the centering relationship, and then the core word used for modifying the dependent word in the centering relationship may be used as the attribute of the entity (the dependent word), so as to implement accurate description on the entity in the scene.
In step S111, entity relationship information is generated based on the second target word pair.
In the embodiment of the present specification, the entity relationship information may represent an association relationship between every two entities; in a specific embodiment, the entity relationship information includes an entity pair and an association relationship corresponding to the entity pair;
in an alternative embodiment, where the associations include action associations, the target dependency includes: a main and subordinate relationship and an action and guest relationship; the second target word pair includes: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary-to-predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-to-guest relationship; accordingly, as shown in fig. 3, the generating of the entity relationship information based on the second target word pair may include the following steps:
in step S301, matching the dependent word in each first subword pair with the core word in each second subword pair to obtain a matching word pair group, where the dependent word in the first subword pair included in the matching word pair group is consistent with the core word in the second subword pair included in the matching word pair group;
in step S303, the core word in the first subword pair included in the matching word pair group and the dependent word in the second subword included in the matching word pair group are taken as the first entity pair, and the dependent word in the first subword pair included in the matching word pair group or the core word in the second subword included in the matching word pair group are taken as the action association relationship corresponding to the first entity pair, so as to generate entity relationship information.
In one specific embodiment, the target Chinese statement is: i eat an apple, the first sub-word pair is: [ I, eat ], the second sub-word pair is: [ eat, apple ]. Wherein, the dependency word "eat" in the first sub-word pair matches (is consistent with) the core word "eat" in the second sub-word pair, and correspondingly, the first sub-word pair: [ I, eat ] and the second sub-word pair: [ eat, apple ] is a first sub-word pair and a second sub-word pair which are matched pairwise (namely a matching word pair group), further, "me" and "apple" can be used as a first entity pair, and "eat" is a corresponding action incidence relation of the first entity pair "me" and "apple"; accordingly, the entity relationship information may be "i-eat-apple".
In the embodiment, through matching between the dependency word in the first sub-word pair with the dependency relationship as the main predicate relationship and the core word in the second sub-word pair with the dependency relationship as the action guest relationship, the word capable of representing the action association relationship between the entities can be accurately extracted, and further entity relationship information capable of accurately describing the action association relationship between the entities in the scene is constructed.
In an optional embodiment, in a case that the association relationship includes a location association relationship, the target dependency relationship includes: a guest-mediating relationship; the second target word pair includes: the dependency relationship is a word pair containing a core word and a dependency word of the intermediary relationship; accordingly, as shown in fig. 4, the generating of the entity relationship information based on the second target word pair may include:
in step S401, according to the word relation data, it is determined that the dependent word is a fourth target word pair of the dependent word in the second target word pair, and the dependent word is a fifth target word pair of the core word in the second target word pair;
in step S403, the core word in the fourth target word pair and the core word in the fifth target word pair are used as the second entity pair, and the core word in the second target word pair is used as the position association relationship corresponding to the second entity pair, so as to generate entity relationship information.
In one specific embodiment, assuming that the target chinese sentence is "a book on a table", the corresponding second target word pair may be: then, the fourth target word pair with the dependency relationship corresponding to the word pair with the dependency relationship of 'one book on the desk' is traversed, and the dependency word is determined to be 'on': [ book, on ], and the dependency word is "upper" fifth target word pair: [ Table, on ]; accordingly, the fourth target word pair may be: the core word "book" and the fifth target word pair in [ book, in ]: the core word "table" in [ table, on ] is the second entity object, and the position association relationship between "table" and "book" is taken as "on", so as to obtain entity relationship information: "desk-on-book".
In the above embodiment, the fourth target word pair and the fifth target word pair are respectively matched by matching the dependent word and the core word in the second target word pair whose dependent relationship is the betin relationship, and two entities whose position association relationships are the core word in the second target word pair are accurately determined, so as to construct entity relationship information capable of accurately describing the position association relationships among the entities in the scene.
In step S113, a scene graph corresponding to the target chinese sentence is generated based on the entity attribute information and the entity relationship information.
In a particular embodiment, a scene graph (scene graph) is a structured representation of a scene that may be used to describe entities, attributes, and relationships between entities in the scene. Correspondingly, the scene graph corresponding to the target chinese sentence may use the structured data to represent the entities, attributes, and relationships between the entities in the scene corresponding to the target chinese sentence.
In an optional embodiment, as shown in fig. 5, the generating a scene graph corresponding to the target chinese statement based on the entity attribute information and the entity relationship information may include the following steps:
in step S501, an entity in the entity attribute information is used as a first type node, and an attribute of the entity in the entity attribute information is used as a second type node;
in step S503, a scene graph is constructed by using the association relationship corresponding to the entity pair in the entity relationship information as the edge between the corresponding first-type nodes, and the dependency relationship between the attributes of the entity and the entity as the edge between the corresponding first-type nodes and the second-type nodes.
In a specific embodiment, the first type node may be a node corresponding to an entity, and the second type node may be a node corresponding to an attribute of the entity. Optionally, the edge between the entities may be a directed edge, and specifically, the direction of the edge may be determined by combining the association relationship between the two entities. The edge between the entity and the attribute can be pointed to by the entity.
In a specific embodiment, as shown in fig. 6, fig. 6 is a schematic diagram of a scenario diagram provided according to an exemplary embodiment, specifically, taking the above "i eat an apple" as an example, the first type node includes an entity "i" and an entity "apple", and the second type node includes an attribute "one" of the entity "apple"; the orientation of the edge (eating) between the entity "i" and the entity "apple" may be from "i" to "apple"; the edge between the entity "apple" and the attribute "one" may be "pointed to" one book "by the book.
In a specific embodiment, as shown in fig. 7, fig. 7 is a schematic diagram of another scenario provided according to an exemplary embodiment, specifically, taking the above "book on table" as an example, the first type node includes an entity "table" and an entity "book", and the second type node includes an attribute "book" of the entity "book"; the orientation of the edge (on) between the entity "table" and the entity "book" may be from "table" to "book"; the edge between the entity "book" and the attribute "book" may be "book" pointed to by the book.
In the above embodiment, by combining the entity attribute information including the entity and the attribute of the entity and the entity relationship information including the entity pair and the association relationship corresponding to the entity pair, the entity, the attribute and the relationship between the entities in the natural language can be accurately converted from unstructured data into structured data, so as to accurately depict the scene in the target chinese sentence and provide data support for the processing of the downstream task.
In an optional embodiment, the scene graph may be converted into a corresponding scene image in combination with an image generation model, so as to generate information that more vividly embodies a scene, such as an animation or a small video, in a related application.
In an optional embodiment, a corresponding knowledge graph can also be generated by combining scene graphs corresponding to a large number of statements in a certain field, so as to provide further data support for subsequent applications.
As can be seen from the technical solutions provided by the embodiments of the present specification, the present specification performs syntactic dependency analysis in combination with the word segmentation sequence and the part-of-speech information of the target chinese sentence, so as to obtain a word pair having a dependency relationship in the target chinese sentence and a dependency relationship corresponding to the word pair, and locates words belonging to the entity in combination with the part-of-speech information, and then, based on a result of the syntactic dependency analysis, generates entity attribute information and entity relationship information representing an association relationship between two entities, and can accurately convert the entities, attributes, and relationships between the entities in the natural language from unstructured data to structured data, thereby realizing accurate depiction of a scene in the target chinese sentence, improving accuracy and vividness of scene description, and also providing data support for processing of downstream tasks.
Fig. 8 is a block diagram illustrating a scene graph generation apparatus according to an example embodiment. Referring to fig. 8, the apparatus includes:
an information obtaining module 810 configured to perform obtaining a word segmentation sequence of a target chinese sentence and part-of-speech information of words in the word segmentation sequence;
a syntactic dependency analysis module 820 configured to perform syntactic dependency analysis by inputting the segmented word sequences and the part-of-speech information into a syntactic dependency model, so as to obtain word relationship data, where the word relationship data includes word pairs with dependency relationships and dependency relationships corresponding to the word pairs;
an entity identification module 830 configured to perform identifying words belonging to an entity in the word segmentation sequence according to the part of speech information;
a target word pair determination module 840 configured to perform determining, according to the word relationship data, a first target word pair whose dependency relationship is a fixed relationship and a second target word pair whose dependency relationship is a target dependency relationship; the target dependency relationship is the dependency relationship of corresponding word pairs containing entity associated words, and the entity associated words are words representing the association relationship among the entities;
an entity attribute information generating module 850 configured to perform generating entity attribute information based on the first target word pair and words belonging to the entity;
an entity relationship information generation module 860 configured to perform generating entity relationship information based on the second target word pair, the entity relationship information representing an association relationship between two entities;
and the scene graph generation module 870 is configured to execute generating a scene graph corresponding to the target chinese statement based on the entity attribute information and the entity relationship information.
Optionally, the first target word pair includes: a word pair comprising a core word and a dependency word; the entity attribute information generating module 850 includes:
a third target word pair determination unit configured to perform determination of a third target word pair, to which the dependent word belongs, from the first target word pair according to the words belonging to the entity;
and the entity attribute information generating unit is configured to execute attribute generation by taking the dependency word in the third target word pair as the entity and the core word in the third target word pair as the corresponding entity.
Optionally, the association relationship includes an action association relationship, and the target dependency relationship includes: a main and subordinate relationship and an action and guest relationship; the second target word pair includes: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary-to-predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-to-guest relationship;
the entity relationship information generation module 860 includes:
the word pair matching unit is configured to perform matching processing on a dependent word in each first subword pair and a core word in each second subword pair to obtain a matching word pair group, wherein the dependent word in the first subword pair contained in the matching word pair group is consistent with the core word in the second subword pair contained in the matching word pair group;
a first entity relationship information generating unit, configured to execute an action association relationship that a core word in a first subword pair included in the matching word pair group and a dependent word in a second subword included in the matching word pair group are a first entity pair, and a dependent word in the first subword pair included in the matching word pair group or a core word in the second subword included in the matching word pair group is a corresponding first entity pair, so as to generate the entity relationship information.
Optionally, the association relationship includes a location association relationship, and the target dependency relationship includes: a guest-mediating relationship; the second target word pair includes: the dependency relationship is a word pair containing a core word and a dependency word of the intermediary relationship;
the entity relationship information generation module 860 includes:
a target word pair determination unit configured to perform determining, according to the word relation data, that the dependent word is a fourth target word pair of the dependent word in the second target word pair, and that the dependent word is a fifth target word pair of the core word in the second target word pair;
and the second entity relationship information generating unit is configured to execute position association relationship corresponding to the second entity pair by taking the core word in the fourth target word pair and the core word in the fifth target word pair as the second entity pair and taking the core word in the second target word pair as the second entity pair, and generate entity relationship information.
Optionally, the entity attribute information includes an entity and an attribute of the entity, and the entity relationship information includes an association relationship between an entity pair and the entity pair; the scene graph generation module 870 includes:
a node determination unit configured to perform the operation of regarding the entity in the entity attribute information as a first type node and regarding the attribute of the entity in the entity attribute information as a second type node;
and the scene graph building unit is configured to execute building of the scene graph by taking the corresponding association relationship of the entity pair in the entity relationship information as the corresponding edge between the first type nodes, and taking the dependency relationship between the attributes of the entity and the entity as the corresponding edge between the first type nodes and the second type nodes.
Optionally, the apparatus further comprises:
the to-be-processed Chinese text acquisition module is configured to execute acquisition of the to-be-processed Chinese text;
the sentence dividing module is configured to execute a regular expression taking the character to be matched as a preset punctuation mark, and divide the Chinese text to be processed into a plurality of sentences;
a target Chinese sentence determination module configured to execute taking any one of the plurality of sentences as a target Chinese sentence.
Optionally, the information obtaining module 810 includes:
and the part-of-speech recognition processing unit is configured to execute part-of-speech recognition processing of the part-of-speech sequence input part-of-speech recognition model to obtain part-of-speech information of the words in the part-of-speech sequence.
With regard to the apparatus in the above-described embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment related to the method, and will not be elaborated here.
Fig. 9 is a block diagram illustrating an electronic device for scene graph generation, which may be a terminal, according to an exemplary embodiment, and an internal structure diagram of the electronic device may be as shown in fig. 9. The electronic device comprises a processor, a memory, a network interface, a display screen and an input device which are connected through a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic equipment comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the electronic device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement a scenegraph generation method. The display screen of the electronic equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the electronic equipment, an external keyboard, a touch pad or a mouse and the like.
Fig. 10 is a block diagram illustrating another electronic device for scenegraph generation according to an exemplary embodiment, where the electronic device may be a server, and the internal structure thereof may be as shown in fig. 10. The electronic device includes a processor, a memory, and a network interface connected by a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic equipment comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the electronic device is used for connecting and communicating with an external terminal through a network. The computer program is executed by a processor to implement a scenegraph generation method.
It will be understood by those skilled in the art that the configurations shown in fig. 9 or fig. 10 are only block diagrams of some configurations relevant to the present disclosure, and do not constitute a limitation on the electronic device to which the present disclosure is applied, and a particular electronic device may include more or less components than those shown in the drawings, or combine some components, or have a different arrangement of components.
In an exemplary embodiment, there is also provided an electronic device including: a processor; the processor is configured to execute the instructions to realize the scene graph generation method in the embodiment of the disclosure.
In an exemplary embodiment, a computer-readable storage medium is also provided, and when instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute a scene graph generation method in the embodiments of the present disclosure.
In an exemplary embodiment, a computer program product containing instructions is also provided, which when run on a computer, causes the computer to perform a scene graph generation method in the embodiments of the present disclosure.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the disclosure disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosure following, in general, the principles of the disclosure and including such departures from the present disclosure as come within known or customary practice within the art to which the disclosure pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.
It will be understood that the present disclosure is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.
Claims (15)
1. A scene graph generation method is characterized by comprising the following steps:
acquiring a word segmentation sequence of a target Chinese sentence and part-of-speech information of words in the word segmentation sequence;
inputting the word segmentation sequence and the part of speech information into a syntactic dependency model for syntactic dependency analysis to obtain word relation data, wherein the word relation data comprises word pairs with dependency relations and dependency relations corresponding to the word pairs;
identifying words belonging to the entity in the word segmentation sequence according to the part of speech information;
determining a first target word pair with a medium dependency relationship and a second target word pair with a target dependency relationship according to the word relationship data; the target dependency relationship is the dependency relationship that the corresponding word pair contains entity associated words, and the entity associated words are words representing the association relationship among the entities;
generating entity attribute information based on the first target word pair and the words belonging to the entity;
generating entity relation information based on the second target word pair, wherein the entity relation information represents the incidence relation between every two entities;
and generating a scene graph corresponding to the target Chinese statement based on the entity attribute information and the entity relation information.
2. The scenegraph generation method of claim 1, wherein the first target term pair comprises: a word pair comprising a core word and a dependency word; the generating entity attribute information based on the first target word pair and the words belonging to the entity comprises:
determining a third target word pair of the dependency word belonging to the entity from the first target word pair according to the words belonging to the entity;
and generating the entity attribute information by taking the dependent words in the third target word pair as entities and the core words in the third target word pair as attributes of corresponding entities.
3. The method of generating a scene graph according to claim 1, wherein the association relationship comprises an action association relationship, and the target dependency relationship comprises: a main and subordinate relationship and an action and guest relationship; the second target word pair includes: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary-to-predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-to-guest relationship;
the generating entity relationship information based on the second target term pair comprises:
matching the dependency word in each first subword pair with the core word in each second subword pair to obtain a matching word pair group, wherein the dependency word in the first subword pair contained in the matching word pair group is consistent with the core word in the second subword pair contained in the matching word pair group;
and taking a core word in a first sub-word pair contained in the matching word pair group and a dependent word in a second sub-word contained in the matching word pair group as a first entity pair, and taking a dependent word in the first sub-word pair contained in the matching word pair group or a core word in the second sub-word contained in the matching word pair group as an action incidence relation corresponding to the first entity pair to generate the entity relation information.
4. The method of generating a scene graph according to claim 1, wherein the association relationship comprises a location association relationship, and the target dependency relationship comprises: a guest-mediating relationship; the second target word pair includes: the dependency relationship is a word pair containing a core word and a dependency word of the intermediary relationship;
the generating entity relationship information based on the second target term pair comprises:
determining that the dependent word is a fourth target word pair of the dependent words in the second target word pair and the dependent word is a fifth target word pair of the core word in the second target word pair according to the word relation data;
and generating the entity relationship information by taking the core word in the fourth target word pair and the core word in the fifth target word pair as a second entity pair, and taking the core word in the second target word pair as a position association relationship corresponding to the second entity pair.
5. The scene graph generating method according to claim 1, wherein the entity attribute information includes an entity and an attribute of the entity, and the entity relationship information includes an entity pair and an association relationship corresponding to the entity pair; the generating a scene graph corresponding to the target chinese statement based on the entity attribute information and the entity relationship information includes:
taking an entity in the entity attribute information as a first type node, and taking an attribute of the entity in the entity attribute information as a second type node;
and constructing the scene graph by taking the corresponding association relationship of the entity pair in the entity relationship information as the corresponding edge between the first type nodes, and taking the dependency relationship between the attributes of the entity and the entity as the corresponding edge between the first type nodes and the second type nodes.
6. The scene graph generation method according to any one of claims 1 to 5, characterized by further comprising:
acquiring a Chinese text to be processed;
dividing the Chinese text to be processed into a plurality of sentences based on a regular expression with characters to be matched as preset punctuations;
and taking any one of the sentences as the target Chinese sentence.
7. A scene graph generation apparatus, comprising:
the information acquisition module is configured to execute acquisition of a word segmentation sequence of a target Chinese sentence and part-of-speech information of words in the word segmentation sequence;
a syntactic dependency analysis module configured to perform syntactic dependency analysis on the segmented word sequence and the part-of-speech information input into a syntactic dependency model to obtain word relationship data, wherein the word relationship data includes word pairs with dependency relationships and dependency relationships corresponding to the word pairs;
the entity identification module is configured to identify words belonging to the entity in the word segmentation sequence according to the part of speech information;
a target word pair determination module configured to perform determining, according to the word relation data, a first target word pair whose dependency relationship is a fixed relationship and a second target word pair whose dependency relationship is a target dependency relationship; the target dependency relationship is the dependency relationship that the corresponding word pair contains entity associated words, and the entity associated words are words representing the association relationship among the entities;
an entity attribute information generation module configured to perform generation of entity attribute information based on the first target word pair and the words belonging to the entity;
the entity relation information generation module is configured to execute generation of entity relation information based on the second target word pair, and the entity relation information represents an incidence relation between every two entities;
and the scene graph generating module is configured to execute generating a scene graph corresponding to the target Chinese statement based on the entity attribute information and the entity relation information.
8. The apparatus of claim 7, wherein the first target term pair comprises: a word pair comprising a core word and a dependency word; the entity attribute information generation module includes:
a third target word pair determination unit configured to perform determination of a third target word pair, in which a dependent word belongs to an entity, from the first target word pair according to the word belonging to the entity;
and the entity attribute information generating unit is configured to execute the step of generating the entity attribute information by taking the dependent words in the third target word pair as entities and the core words in the third target word pair as attributes of corresponding entities.
9. The apparatus according to claim 7, wherein the association comprises an action association, and the target dependency comprises: a main and subordinate relationship and an action and guest relationship; the second target word pair includes: the dependency relationship is a first sub-word pair containing the core word and the dependency word of the primary-to-predicate relationship and a second sub-word pair containing the core word and the dependency word of the dependency relationship of the mobile-to-guest relationship;
the entity relationship information generation module comprises:
the word pair matching unit is configured to perform matching processing on a dependent word in each first subword pair and a core word in each second subword pair to obtain a matching word pair group, wherein the dependent word in the first subword pair contained in the matching word pair group is consistent with the core word in the second subword pair contained in the matching word pair group;
a first entity relationship information generating unit, configured to execute an action association relationship that a core word in a first subword pair included in the matching word pair group and a dependent word in a second subword included in the matching word pair group are a first entity pair, and a dependent word in the first subword pair included in the matching word pair group or a core word in the second subword included in the matching word pair group is a corresponding first entity pair, so as to generate the entity relationship information.
10. The apparatus according to claim 7, wherein the association relationship comprises a location association relationship, and the target dependency relationship comprises: a guest-mediating relationship; the second target word pair includes: the dependency relationship is a word pair containing a core word and a dependency word of the intermediary relationship;
the entity relationship information generation module comprises:
a target word pair determination unit configured to perform determining, according to the word relation data, that a dependent word is a fourth target word pair of the dependent words in the second target word pair, and that a dependent word is a fifth target word pair of the core word in the second target word pair;
a second entity relationship information generating unit, configured to execute a position association relationship that the core word in the fourth target word pair and the core word in the fifth target word pair are the second entity pair and the core word in the second target word pair is the second entity pair, so as to generate the entity relationship information.
11. The apparatus according to claim 7, wherein the entity attribute information includes entities and attributes of the entities, and the entity relationship information includes entity pairs and associated relationships corresponding to the entity pairs; the scene graph generation module comprises:
a node determining unit configured to perform the operation of regarding an entity in the entity attribute information as a first type node and regarding an attribute of the entity in the entity attribute information as a second type node;
and the scene graph building unit is configured to execute building of the scene graph by taking the corresponding association relationship of the entity pair in the entity relationship information as the corresponding edge between the first type nodes, and taking the dependency relationship between the attributes of the entity and the entity as the corresponding edge between the first type nodes and the second type nodes.
12. The apparatus according to any one of claims 7 to 11, characterized in that the apparatus further comprises:
the to-be-processed Chinese text acquisition module is configured to execute acquisition of the to-be-processed Chinese text;
the sentence dividing module is configured to execute a regular expression which takes the characters to be matched as preset punctuation marks, and divide the Chinese text to be processed into a plurality of sentences;
a target Chinese sentence determination module configured to execute taking any one of the plurality of sentences as the target Chinese sentence.
13. An electronic device, comprising:
a processor;
a memory for storing the processor-executable instructions;
wherein the processor is configured to execute the instructions to implement the scenegraph generation method of any of claims 1-6.
14. A computer-readable storage medium, wherein instructions in the storage medium, when executed by a processor of an electronic device, enable the electronic device to perform the scene graph generation method of any one of claims 1 to 6.
15. A computer program product comprising computer instructions, wherein the computer instructions, when executed by a processor, implement the scenegraph generation method of any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111224754.1A CN113705198B (en) | 2021-10-21 | 2021-10-21 | Scene graph generation method and device, electronic equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111224754.1A CN113705198B (en) | 2021-10-21 | 2021-10-21 | Scene graph generation method and device, electronic equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113705198A true CN113705198A (en) | 2021-11-26 |
| CN113705198B CN113705198B (en) | 2022-03-25 |
Family
ID=78646892
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111224754.1A Active CN113705198B (en) | 2021-10-21 | 2021-10-21 | Scene graph generation method and device, electronic equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113705198B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662493A (en) * | 2022-02-28 | 2022-06-24 | 腾讯科技（深圳）有限公司 | Nested entity identification method and device, electronic equipment and storage medium |
| CN115994536A (en) * | 2023-03-24 | 2023-04-21 | 浪潮电子信息产业股份有限公司 | Text information processing method, system, equipment and computer storage medium |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02266469A (en) * | 1989-02-01 | 1990-10-31 | Bso Buro Voor Syst Ontwikkeling Bv | Method and apparatus for obtaining expression pattern of multiple syntax analysis in dependence grammer and parser for generating expression pattern |
| CN106777275A (en) * | 2016-12-29 | 2017-05-31 | 北京理工大学 | Entity attribute and property value extracting method based on many granularity semantic chunks |
| US20170300470A1 (en) * | 2016-04-15 | 2017-10-19 | Marca Research & Development International, Llc | Systems and methods for identifying evidentiary information |
| CN107291687A (en) * | 2017-04-27 | 2017-10-24 | 同济大学 | It is a kind of based on interdependent semantic Chinese unsupervised open entity relation extraction method |
| CN109241538A (en) * | 2018-09-26 | 2019-01-18 | 上海德拓信息技术股份有限公司 | Based on the interdependent Chinese entity relation extraction method of keyword and verb |
| CN111177315A (en) * | 2019-12-19 | 2020-05-19 | 北京明略软件系统有限公司 | Knowledge graph updating method and device and computer readable storage medium |
| CN111339751A (en) * | 2020-05-15 | 2020-06-26 | 支付宝(杭州)信息技术有限公司 | Text keyword processing method, device and equipment |
| CN111597351A (en) * | 2020-05-14 | 2020-08-28 | 上海德拓信息技术股份有限公司 | Visual document map construction method |
| CN111597794A (en) * | 2020-05-11 | 2020-08-28 | 浪潮软件集团有限公司 | Dependency relationship-based 'yes' word and sentence relationship extraction method and device |
| CN111832316A (en) * | 2020-06-30 | 2020-10-27 | 北京小米松果电子有限公司 | Semantic recognition method and device, electronic equipment and storage medium |
| CN112232074A (en) * | 2020-11-13 | 2021-01-15 | 完美世界控股集团有限公司 | Entity relationship extraction method and device, electronic equipment and storage medium |
-
2021
- 2021-10-21 CN CN202111224754.1A patent/CN113705198B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02266469A (en) * | 1989-02-01 | 1990-10-31 | Bso Buro Voor Syst Ontwikkeling Bv | Method and apparatus for obtaining expression pattern of multiple syntax analysis in dependence grammer and parser for generating expression pattern |
| US20170300470A1 (en) * | 2016-04-15 | 2017-10-19 | Marca Research & Development International, Llc | Systems and methods for identifying evidentiary information |
| CN106777275A (en) * | 2016-12-29 | 2017-05-31 | 北京理工大学 | Entity attribute and property value extracting method based on many granularity semantic chunks |
| CN107291687A (en) * | 2017-04-27 | 2017-10-24 | 同济大学 | It is a kind of based on interdependent semantic Chinese unsupervised open entity relation extraction method |
| CN109241538A (en) * | 2018-09-26 | 2019-01-18 | 上海德拓信息技术股份有限公司 | Based on the interdependent Chinese entity relation extraction method of keyword and verb |
| CN111177315A (en) * | 2019-12-19 | 2020-05-19 | 北京明略软件系统有限公司 | Knowledge graph updating method and device and computer readable storage medium |
| CN111597794A (en) * | 2020-05-11 | 2020-08-28 | 浪潮软件集团有限公司 | Dependency relationship-based 'yes' word and sentence relationship extraction method and device |
| CN111597351A (en) * | 2020-05-14 | 2020-08-28 | 上海德拓信息技术股份有限公司 | Visual document map construction method |
| CN111339751A (en) * | 2020-05-15 | 2020-06-26 | 支付宝(杭州)信息技术有限公司 | Text keyword processing method, device and equipment |
| CN111832316A (en) * | 2020-06-30 | 2020-10-27 | 北京小米松果电子有限公司 | Semantic recognition method and device, electronic equipment and storage medium |
| CN112232074A (en) * | 2020-11-13 | 2021-01-15 | 完美世界控股集团有限公司 | Entity relationship extraction method and device, electronic equipment and storage medium |
Non-Patent Citations (5)
| Title |
|---|
| B. KARAOĞLAN等: "Contribution of syntactic and semantic attributes in paraphrase identification", 《2018 26TH SIGNAL PROCESSING AND COMMUNICATIONS APPLICATIONS CONFERENCE (SIU)》 * |
| 张建华等: "基于依存句法和二叉树模型的评价对象抽取", 《计算机技术与发展》 * |
| 曹文斌等: "基于文本语料的涉恐事件实体属性抽取", 《工程科学学报》 * |
| 虞欢欢等: "基于合一句法和实体语义树的中文语义关系抽取", 《中文信息学报》 * |
| 郭振等: "基于字符的中文分词、词性标注和依存句法分析联合模型", 《中文信息学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114662493A (en) * | 2022-02-28 | 2022-06-24 | 腾讯科技（深圳）有限公司 | Nested entity identification method and device, electronic equipment and storage medium |
| CN115994536A (en) * | 2023-03-24 | 2023-04-21 | 浪潮电子信息产业股份有限公司 | Text information processing method, system, equipment and computer storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113705198B (en) | 2022-03-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112417102B (en) | Voice query method, device, server and readable storage medium | |
| RU2607416C2 (en) | Crowd-sourcing vocabulary teaching systems | |
| JP5356197B2 (en) | Word semantic relation extraction device | |
| CN109800307B (en) | Product evaluation analysis method and device, computer equipment and storage medium | |
| US20110238413A1 (en) | Domain dictionary creation | |
| US20130177893A1 (en) | Method and Apparatus for Responding to an Inquiry | |
| CN113705198B (en) | Scene graph generation method and device, electronic equipment and storage medium | |
| CN111680634A (en) | Document file processing method and device, computer equipment and storage medium | |
| CN105378706B (en) | Entity extraction is fed back | |
| US10496751B2 (en) | Avoiding sentiment model overfitting in a machine language model | |
| CN110162771A (en) | The recognition methods of event trigger word, device, electronic equipment | |
| Alhassan et al. | A novel framework for Arabic dialect chatbot using machine learning | |
| CN110309513B (en) | Text dependency analysis method and device | |
| JP2005181928A (en) | System and method for machine learning, and computer program | |
| JP2017151678A (en) | Topic estimation device, topic estimation method, and program | |
| Duran et al. | Some issues on the normalization of a corpus of products reviews in Portuguese | |
| CN112559711A (en) | Synonymous text prompting method and device and electronic equipment | |
| CN114676699A (en) | Entity emotion analysis method and device, computer equipment and storage medium | |
| Srinivasan | Chatbot application for tourism using natural language tool kit | |
| CN113704452A (en) | Data recommendation method, device, equipment and medium based on Bert model | |
| WO2021221535A1 (en) | System and method for augmenting a training set for machine learning algorithms | |
| Langlais et al. | Issues in analogical inference over sequences of symbols: A case study on proper name transliteration | |
| Rychtyckyj et al. | Applying automated language translation at a global enterprise level | |
| Jiacuo et al. | Tibetan Case Grammar Error Correction Method Based on Neural Networks | |
| CN116701640B (en) | Watermark identification model generation method, watermark identification device and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |