CN110705301B - Entity relationship extraction method and device, storage medium and electronic equipment - Google Patents
Entity relationship extraction method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN110705301B CN110705301B CN201910941040.9A CN201910941040A CN110705301B CN 110705301 B CN110705301 B CN 110705301B CN 201910941040 A CN201910941040 A CN 201910941040A CN 110705301 B CN110705301 B CN 110705301B
- Authority
- CN
- China
- Prior art keywords
- entity
- relationship
- text
- triple
- entity relationship
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/355—Class or cluster creation or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Machine Translation (AREA)
Abstract
The present disclosure provides an entity relationship extraction method and apparatus, an electronic device, and a storage medium; relates to the technical field of natural language processing. The method comprises the following steps: acquiring a text of an entity relationship to be extracted, and processing the text through an entity and relationship combined extraction model to obtain a first entity relationship triple; determining words and labels corresponding to the words in the text, matching the words and the labels according to the relationship rule set, and determining a second entity relationship triple; constructing an entity pair according to the words and the labels, processing the text through a relation classification model, and determining a third entity relation triple aiming at the entity pair; and determining an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple and the third entity relationship triple. The method and the device can improve the accuracy of entity relationship extraction.
Description
Technical Field
The present disclosure relates to the field of natural language processing technologies, and in particular, to an entity relationship extraction method, an entity relationship extraction apparatus, an electronic device, and a computer-readable storage medium.
Background
In recent years, with the continuous development of information technology, a large amount of unstructured text data, such as press releases and the like, is generated, and the unstructured data is not conducive to rapid analysis and processing. Information extraction researches how to extract structured data from the texts, and the information extraction comprises entity extraction and entity relation extraction, wherein the entity refers to words with strong objective meanings or referring properties, and comprises a name of a person, a place name, an organization name, time and date, a proper noun and the like. In the related art, the accuracy is low when the entity relationship is extracted.
It is to be noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the present disclosure, and thus may include information that does not constitute prior art known to those of ordinary skill in the art.
Disclosure of Invention
The present disclosure is directed to an entity relationship extraction method, an entity relationship extraction apparatus, an electronic device, and a computer-readable storage medium, which overcome the problem of low accuracy in extracting entity relationships due to limitations and defects of the related art to some extent.
According to a first aspect of the present disclosure, there is provided an entity relationship extraction method, including:
acquiring a text of an entity relationship to be extracted, and processing the text through an entity and relationship combined extraction model to obtain a first entity relationship triple;
determining words in the text and labels corresponding to the words, matching the words and the labels in the text according to a relationship rule set, and determining a second entity relationship triple;
constructing an entity pair according to the words and the labels, processing the text through a relation classification model, and determining a third entity relation triple aiming at the entity pair;
and determining an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple and the third entity relationship triple.
In an exemplary embodiment of the present disclosure, the determining an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple, and the third entity relationship triple includes:
selecting target entity relationship triples which are the same among the first entities and the second entities from the first entity relationship triples, the second entity relationship triples and the third entity relationship triples;
and determining an entity relationship extraction result according to the target entity relationship triple.
In an exemplary embodiment of the present disclosure, the determining an entity relationship extraction result according to the target entity relationship triple includes:
carrying out weighted average on the entity relationship vectors in the target entity relationship triple to obtain a target entity relationship vector;
and taking the first entity and the second entity in the target entity relationship triple and the entity relationship corresponding to the maximum value in the target entity relationship vector as an entity relationship extraction result.
In an exemplary embodiment of the present disclosure, the matching the words and the tags in the text according to a relationship rule set to determine a second entity relationship triple includes:
judging whether a single word in the text is the same as a relation triggering word in the relation rule or not aiming at each relation rule in the relation rule set;
searching whether a first target label identical to a first entity label in the relation rule and a second target label identical to a second entity label in the relation rule exist in the labels when the single word is identical to the relation triggering word;
and when the first target label and the second target label exist, constructing a second entity relationship triple according to the word corresponding to the first target label, the word corresponding to the second target label and the entity relationship type in the relationship rule.
In an exemplary embodiment of the present disclosure, the processing the text through an entity and relationship joint extraction model to obtain a first entity-relationship triple includes:
performing word segmentation processing on the text to obtain a plurality of words;
converting the plurality of words into semantic vectors, inputting the semantic vectors into an entity and relation combined extraction model, and obtaining word labels;
and constructing a first entity relationship triple according to the word label.
In one exemplary embodiment of the present disclosure,
after the building of entity pairs from the words and tags, the method further comprises:
marking entity pairs in the text to obtain a marked text;
the processing the text through the relationship classification model to determine a third entity relationship triple specifically includes:
and processing the marked text through a relation classification model to determine a third entity relation triple.
In an exemplary embodiment of the present disclosure, the processing the labeled text through the relationship classification model to determine a third entity relationship triple includes:
converting the marked text into a corresponding text semantic vector;
inputting the text semantic vector into a relation classification model to obtain a probability distribution vector of an entity relation;
and constructing a third entity relationship triple according to the entity pair and the probability distribution vector.
In an exemplary embodiment of the present disclosure, the method further comprises:
and performing reference resolution on the obtained initial text to obtain the text of the entity relationship to be extracted.
According to a second aspect of the present disclosure, there is provided an entity relationship extraction apparatus, including:
the system comprises a first entity relationship triple determining module, a second entity relationship triple determining module and a third entity relationship triple determining module, wherein the first entity relationship triple determining module is used for acquiring a text of an entity relationship to be extracted, and processing the text through an entity and relationship combined extraction model to obtain a first entity relationship triple;
the second entity relationship triple determining module is used for determining words in the text and labels corresponding to the words, matching the words and the labels in the text according to a relationship rule set and determining a second entity relationship triple;
a third entity relationship triple determining module, configured to construct an entity pair according to the word and the tag, process the text through a relationship classification model, and determine a third entity relationship triple for the entity pair;
and the entity relationship extraction result determining module is used for determining an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple and the third entity relationship triple.
In an exemplary embodiment of the present disclosure, the entity relationship extraction result determining module includes:
the target triple selecting unit is used for selecting target entity relationship triples which are the same among the first entities and the second entities from the first entity relationship triples, the second entity relationship triples and the third entity relationship triples;
and the extraction result determining unit is used for determining an entity relationship extraction result according to the target entity relationship triple.
In an exemplary embodiment of the present disclosure, the extraction result determining unit is specifically configured to perform weighted average on the entity relationship vectors in the target entity relationship triple to obtain a target entity relationship vector; and taking the first entity and the second entity in the target entity relationship triple and the entity relationship corresponding to the maximum value in the target entity relationship vector as an entity relationship extraction result.
In an exemplary embodiment of the disclosure, the first entity relationship triple determining module includes:
the word segmentation unit is used for carrying out word segmentation processing on the text to obtain a plurality of words;
the word label determining unit is used for converting the words into semantic vectors and inputting the semantic vectors into an entity and relation combined extraction model to obtain word labels;
and the first triple construction unit is used for constructing a first entity relationship triple according to the word label.
In an exemplary embodiment of the disclosure, the second entity relationship triple determining module includes:
a triggering word judgment unit, configured to judge, for each relationship rule in the relationship rule set, whether a single word in the text is the same as a relationship triggering word in the relationship rule;
the target label judging unit is used for searching whether a first target label identical to a first entity label in the relation rule and a second target label identical to a second entity label in the relation rule exist in the labels when the single word is identical to the relation triggering word;
and the second triple construction unit is used for constructing a second entity relationship triple according to the word corresponding to the first target label, the word corresponding to the second target label and the entity relationship type in the relationship rule when the first target label and the second target label both exist.
In an exemplary embodiment of the present disclosure, the apparatus further includes:
the marking module is used for marking the entity pairs in the text to obtain a marked text;
the third entity relationship triple determining module is specifically configured to process the marked text through a relationship classification model to determine a third entity relationship triple.
In an exemplary embodiment of the disclosure, the third entity relationship triple determining module implements the processing of the tagged text by the relationship classification model to determine a third entity relationship triple by:
converting the marked text into a corresponding text semantic vector;
inputting the text semantic vector into a relation classification model to obtain a probability distribution vector of an entity relation;
and constructing a third entity relationship triple according to the entity pair and the probability distribution vector.
In an exemplary embodiment of the present disclosure, the apparatus further includes:
and the reference resolution module is used for performing reference resolution on the obtained initial text to obtain the text of the entity relationship to be extracted.
According to a third aspect of the present disclosure, there is provided an electronic device comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method of any one of the above via execution of the executable instructions.
According to a fourth aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of any one of the above.
Exemplary embodiments of the present disclosure may have at least some or all of the following benefits:
in the entity relationship extraction method provided by an example embodiment of the present disclosure, the entity relationship is extracted through the entity and relationship joint extraction model, and the accuracy of entity relationship extraction can be improved according to the interaction constraint information of named entity identification and relationship extraction. When the entity types and the relationship types are more, the entity relationships extracted by the relationship classification model and the relationship rule set are combined with the entity relationships extracted by the entity and relationship combined extraction model to determine the final entity relationship, so that the accuracy of the extraction result can be improved, and the stability of the system performance can be improved.
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. It is to be understood that the drawings in the following description are merely exemplary of the disclosure, and that other drawings may be derived from those drawings by one of ordinary skill in the art without the exercise of inventive faculty.
Fig. 1 is a schematic diagram illustrating an exemplary system architecture to which an entity relationship extraction method and apparatus according to an embodiment of the present disclosure may be applied;
FIG. 2 illustrates a flow diagram of an entity relationship extraction method according to one embodiment of the present disclosure;
FIG. 3 is a flow chart illustrating entity relationship extraction through a combined entity and relationship extraction model in an embodiment of the present disclosure;
FIG. 4 illustrates a flow diagram of a rule-based method for extracting entity relationships in accordance with an embodiment of the present disclosure;
FIG. 5 shows a block diagram of an entity relationship extraction apparatus according to one embodiment of the present disclosure;
FIG. 6 illustrates a schematic structural diagram of a computer system suitable for use in implementing an electronic device of an embodiment of the present disclosure.
Detailed Description
Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a thorough understanding of embodiments of the disclosure. One skilled in the relevant art will recognize, however, that the subject matter of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, devices, steps, and the like. In other instances, well-known technical solutions have not been shown or described in detail to avoid obscuring aspects of the present disclosure.
Furthermore, the drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and thus their repetitive description will be omitted. Some of the block diagrams shown in the figures are functional entities and do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices.
Fig. 1 is a schematic diagram illustrating a system architecture of an exemplary application environment to which an entity relationship extraction method and apparatus according to an embodiment of the present disclosure may be applied.
As shown in fig. 1, the system architecture 100 may include one or more of terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, to name a few. The terminal devices 101, 102, 103 may be various electronic devices having a display screen, including but not limited to desktop computers, portable computers, smart phones, tablet computers, and the like. It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative, and that there may be any number of terminal devices, networks, and servers, as desired for an implementation. For example, server 105 may be a server cluster comprised of multiple servers, or the like.
The entity relationship extracting method provided by the embodiment of the present disclosure is generally executed by the server 105, and accordingly, the entity relationship extracting apparatus is generally disposed in the server 105. However, it is easily understood by those skilled in the art that the entity relationship extracting method provided in the embodiment of the present disclosure may also be executed by the terminal devices 101, 102, and 103, and accordingly, the entity relationship extracting apparatus may also be disposed in the terminal devices 101, 102, and 103, which is not particularly limited in this exemplary embodiment. For example, in an exemplary embodiment, a user may upload a text of an entity relationship to be extracted to the server 105 through the terminal devices 101, 102, and 103, and the server 105 processes the text by using the entity relationship extraction method provided in the embodiment of the present disclosure, and sends an obtained entity relationship extraction result to the terminal devices 101, 102, and 103.
The technical solution of the embodiment of the present disclosure is explained in detail below:
in the information extraction technology, named entity recognition is mature for researching how to extract key information such as a person name, a place name, an organization name and the like from a text, and in contrast, a relationship extraction technology for researching and extracting a relationship between two entities is still in need of development. Currently, entity relationships can be extracted through relationship classification models, entity and relationship joint extraction models, rule-based methods, and the like.
The method comprises the following steps that a relation classification model determines the relation category of an entity pair by selecting different features and classification models, the relation classification model assumes an entity giving a relation to be extracted, the entity to be classified is not given in practical application, named entity identification needs to be carried out firstly, and the named entity identification and relation extraction are divided by the method, so that joint modeling of the two tasks is lacked; compared with the semantic information with complex texts, the common text features and classification model capabilities are still limited, and the accuracy of the extracted entity relationships is low.
The entity and relation combined extraction model is formed by arranging and combining entity categories and relation categories by defining a special label set. The method comprises the steps of combining an entity and a relation extraction task to be generalized into a sequence labeling task, labeling each character or word in a text with a label containing an entity type and a relation type through a model, and extracting the entity and the relation according to the label of each character or word. The joint extraction model performs joint modeling on entity identification and relation extraction, can consider interaction constraint information between the two tasks, and has lower accuracy of extracted entity relation when the entity and relation types are more. If the accuracy of entity relationship extraction is improved by increasing the tag set, the complexity of the model is increased, and the performance is affected.
Rule-based methods may determine the relationship of entities based on automatically learned or manually customized rules, once a rule is found that meets a certain rule. However, this method cannot identify the case that is not in the rule set, and is less accurate.
In order to solve the above problem, an embodiment of the present disclosure provides an entity relationship extraction method, which can improve accuracy of entity relationship extraction.
Referring to fig. 2, fig. 2 shows a flowchart of an entity relationship extraction method according to an embodiment of the present disclosure, including the following steps:
step S210, obtaining a text of the entity relationship to be extracted, and processing the text through the entity and relationship combined extraction model to obtain a first entity relationship triple.
Step S220, determining words and labels corresponding to the words in the text, matching the words and the labels corresponding to the words in the text according to the relationship rule set, and determining a second entity relationship triple.
Step S230, an entity pair is constructed according to the words and the labels corresponding to the words in the text, the text is processed through the relationship classification model, and a third entity relationship triple for the entity pair is determined.
Step S240, determining an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple, and the third entity relationship triple.
According to the entity relationship extraction method, the entity relationship is extracted through the entity and relationship combined extraction model, the accuracy of entity relationship extraction can be improved according to the interaction constraint information of named entity identification and relationship extraction. When the entity types and the relationship types are more, the entity relationships extracted by the relationship classification model and the relationship rule set are combined with the entity relationships extracted by the entity and relationship combined extraction model to determine the final entity relationship, so that the accuracy of the extraction result can be improved, and the stability of the system performance can be improved.
The above steps of the present exemplary embodiment will be described in more detail below.
In step S210, a text of the entity relationship to be extracted is obtained, and the text is processed through the entity and relationship joint extraction model to obtain a first entity relationship triple.
In the embodiment of the disclosure, the text of the entity relationship to be extracted may be a text, a word text, and the like, and the information in the text may be news information, social network information, and the like. It should be noted that the text may be a text after the reference resolution, that is, before this step, an initial text may be obtained first, and then the text of the entity relationship to be extracted is obtained after the reference resolution is performed on the obtained initial text. Reference resolution may eliminate reference relationships in text. Such as using the human pronouns "he", "she", etc. to refer to a person entity, or "AA", etc. to refer to the same person, etc. Through the reference resolution, the reference in the initial text can be replaced by the real name of the entity, the text after the reference resolution is used as the text of the entity relationship to be extracted, and the complexity of the subsequent entity and relationship extraction task can be effectively reduced through the reference resolution.
After obtaining the text of the entity relationship to be extracted, the entity relationship may be extracted through an entity and relationship joint extraction model, referring to fig. 3, where fig. 3 shows a flowchart of extracting the entity relationship through the entity and relationship joint extraction model, and the method may include the following steps:
step S310, performing word segmentation processing on the text to obtain a plurality of words.
In the embodiment of the disclosure, word segmentation, word stop processing and the like can be performed on the text to obtain a plurality of words. The word segmentation is a process of recombining texts into word sequences according to certain specifications, for example, the text information is "today's weather is clear", and after the word segmentation, a plurality of words "today", "weather" and "clear" can be obtained. Stop words can be removed, for example, English characters, numbers, mathematical characters, punctuation marks, single Chinese characters with high use frequency and the like can be filtered.
Step S320, converting the plurality of words into semantic vectors, inputting the semantic vectors into the entity and relationship combined extraction model, and obtaining word labels.
For each word, a vectorized representation may be performed, converting the word into a corresponding word vector. For example, each word may be converted to a corresponding word vector by word2vec, where word2vec is a correlation model that may convert words to vectors. And then, inputting the word vectors of the words into the entity and relation combined extraction model to obtain a word label corresponding to each word, wherein the word label can represent the entity category and the relation category of the words.
Optionally, the number of the entity and relationship joint extraction models may be one or more, for example, the number may be one or more, and when the number of the entity and relationship joint extraction models is multiple, different weights may be set for different entity and relationship joint extraction models. The long-short term memory network-conditional random field model combines Softmax with a CRF model at an output end, the problem of extracting sequence features can be solved through the long-short term memory network, and sentence-level marking information is effectively utilized through the random field model. Under the long-short term memory network-conditional random field model, the output will not be mutually independent tags, but an optimal tag sequence.
The bidirectional coder representation model from the Transformer (Bert model) is a language model trained on 33 hundred million text corpora, and the Bert model jointly adjusts based on left and right contexts in all layers to pre-train deep bidirectional representations. Therefore, only one output layer needs to be added, the pre-trained Bert representation can be finely adjusted, the current optimal model can be created for more tasks, such as question answering and language inference tasks, substantial framework modification is not needed in the whole process for specific tasks, and the Bert model obtains better results in different tasks.
And step S330, constructing a first entity relationship triple according to the word label.
Specifically, because the word label may represent the entity category and the relationship category of the word, entities corresponding to word labels having the same relationship category may be merged into one entity relationship triple as the final result. If a sentence contains more than one word label in the same relation category, the pairing can be performed by adopting the principle of proximity.
It should be noted that, the number of the first entity relationship triples may be multiple, and each entity relationship triplet includes: a first entity, a second entity, and a relationship between the first entity and the second entity. In the embodiment of the disclosure, the first entity can be paired
And a second entity
Is encoded to obtain an encoding result L1，L1Is an entity relationship vector. For example, it may be for a first entity
And a second entity
After the relation of (A) is coded by one-hot coding or other coding modes, L is obtained1. Then, the first entity-relationship triplet may pass
Is represented by L1∈RnAnd n is the number of relationship categories. Wherein, the one-hot coding, namely the one-bit effective coding, mainly adopts an R-bit state register to code R states, each state has an independent register bit, and only one bit is effective at any time.
In step S220, words and labels corresponding to the words in the text are determined, and the words and the labels corresponding to the words in the text are matched according to the relationship rule set, so as to determine a second entity relationship triple.
As previously described, the information extraction includes an entity extraction, and the named entity recognition technique can identify entities in the text that have a particular meaning. The named entity recognition technology is mature, and in the current named entity recognition technology, the named entity recognition technology can be realized by a sequence labeling method, namely, different labels are set for each word in a text to judge whether the words belong to one part of a certain class of entities. Through the named entity recognition technology, words and labels corresponding to the words in the text can be determined, that is, word sequences (marked as tokens) and corresponding label sequences (marked as tags) of the text are obtained, and the word sequences and the label sequences are in one-to-one correspondence:
tokens＝{token1,token2,…,tokenm}；
tags＝{tag1,tag2,…,tagm}；
wherein m is the number of words tokens in the text, tagiIs tokeniA corresponding label. It should be noted that the word sequence may include words belonging to an entity, or may include words not belonging to an entity.
After determining the word sequence and the tag sequence, the entity relationships may be extracted according to a rule-based method, as can be seen in fig. 4, which may include the following steps for each relationship rule in the set of relationship rules:
step S410, determine whether the single word in the text is the same as the relationship triggering word in the relationship rule.
It should be noted that the relationship rule set is a set of relationship rules, and may be stored in a rule file, and the rule file is used to store rules, which facilitates quick addition and modification. In each relationship rule, the relationship rule may include: the relationship triggers the terms, the relationship type, the first entity label, and the second entity label. The relation triggering words represent words triggering the extraction of the entity relation, that is, for any word in the text, when the word is the same as the relation triggering words, the extraction of the entity relation can be carried out; otherwise, the entity relation is not extracted. The first entity tag and the second entity tag are tags representing entities, and correspond to the entities one to one. The entity relationship type may represent a relationship between an entity corresponding to the first entity tag and an entity corresponding to the second entity tag.
Before extracting the entity relationship, the relationship rule in the rule file can be read out and stored in the memory for use. When extracting the entity relationship, the processing may be performed in sequence according to the word sequence, and since the processing method of each word in the text is the same, a single word is taken as an example for description here.
For a single term, the entity relationship may be determined by matching the term to each relationship rule in turn. And for each relation rule, judging whether the word is the same as the relation triggering word in the relation rule, if the word is different from the relation triggering word in the relation rule, matching the next relation rule, and if the word is different from the relation triggering words in all the relation rules, indicating that the word does not accord with the entity relation in the relation rule set, processing the next word in the text. And when the word is the same as the relationship triggering word, performing step S420. It should be noted that the number of relationship rules in the relationship rule set, in which a relationship triggering term is the same as the term, may be one or multiple.
Step S420, search whether there exists a first target tag identical to the first entity tag in the relationship rule and a second target tag identical to the second entity tag in the relationship rule in the tags.
In this disclosure, when the word is the same as the relationship triggering word in the relationship rule, it indicates that the word may conform to the entity relationship in the relationship rule, and it may be further determined whether a first target tag that is the same as a first entity tag in the relationship rule and a second target tag that is the same as a second entity tag in the relationship rule exist in tags corresponding to the word in the text. When both the first target label and the second target label exist, the entity relationship can be extracted, and step S430 is executed; and when the first target label and the second target label do not exist or only exist one label, the entity relationship triple corresponding to the single word cannot be extracted.
Wherein, the relationship rule may further include: a first entity search policy and a second entity search policy. In the searching process, whether a first target label identical to a first entity label in the relation rule exists in the label sequence can be searched through a first entity searching strategy, and whether a second target label identical to a second entity label in the relation rule exists in the label sequence can be searched through a second entity searching strategy. The first entity search policy and the second entity search policy may include: and simultaneously triggering the forward and backward search of the label corresponding to the word from the relationship in the label sequence, or performing forward and backward search, forward and backward search and the like.
Step S430, a second entity relationship triple is constructed according to the word corresponding to the first target tag, the word corresponding to the second target tag, and the entity relationship type in the relationship rule.
When both the first target label and the second target label exist, the entity relationship type in the relationship rule is the entity relationship between the entity corresponding to the first target label and the entity corresponding to the second target label. Therefore, the constructed second entity relationship triple is: the words corresponding to the first target label, the words corresponding to the second target label and the entity relationship type in the relationship rule. Similarly, the present disclosure may also encode the entity relationship type to obtain L2The second entity relationship triplet may pass
Is represented by L2Representing a first entity
And a second entity
Relation of (1), L2∈Rn。
Therefore, the rule-based method disclosed by the embodiment of the disclosure has the advantages that the complexity of the relationship rule is low, the process of comparing the words with each relationship rule is simple and quick when the entity relationship is extracted, the accuracy is high, and the performance is stable.
Step S230, an entity pair is constructed according to the words and the labels corresponding to the words in the text, the text is processed through the relationship classification model, and a third entity relationship triple for the entity pair is determined.
In step S220, the word and the tag corresponding to the word in the text are obtained, and whether the word belongs to an entity can be determined according to the type of the tag. Therefore, words corresponding to the tags belonging to the entity can be selected to construct an entity pair. In the embodiment of the present disclosure, after determining any entity pair, the entity pair in the text may be marked to obtain a marked text. Alternatively, the pair of entities in the text may be replaced by special symbols, or special mark symbols may be added in front of and behind two entities in the pair of entities in the text, or the like. For example, if two entities in a pair of entities are: the names A and B, after being labeled, may be [ name A ] and # name B #, or < h > name A < h > and < t > name B < t >, etc.
Accordingly, the tagged text may be processed through the relationship classification model to determine a third entity relationship triple. Specifically, the marked-up text is converted into a corresponding text semantic vector, and the conversion method is similar to step S320 and will not be described in detail here. And then, inputting the text semantic vector into a relation classification model to obtain a probability distribution vector of the entity relation. The number of the relationship classification models can be one or more, and when the number of the relationship classification models is multiple, different weights can be set for different relationship classification models. The elements in the probability distribution vector respectively represent the probability of belonging to the corresponding relationship class, and the correspondence between the elements in the probability distribution vector and the relationship class may be preset when training the relationship classification model, for example, the first element represents the probability of belonging to a friendship, the second element represents the probability of belonging to a parent-child relationship, and so on.
A third entity-relationship triple may be constructed from the entity pairs and the probability distribution vector. The third entity relationship triplet may pass
Is represented by L3The probability distribution vector represents a probability that the larger the value of an element in the probability distribution vector is, the more the probability distribution vector belongs to the corresponding relationship classThe higher the rate, the smaller the value of the element in the probability distribution vector, and the lower the probability of belonging to the corresponding relationship class, and therefore, L3Can be used to represent a first entity
And a second entity
The relationship (2) of (c).
Step S240, determining an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple, and the third entity relationship triple.
After the first entity relationship triplet, the second entity relationship triplet and the third entity relationship triplet are obtained, in order to improve the accuracy of entity relationship extraction, target entity relationship triplets which are the same between the first entities and between the second entities may be selected from the first entity relationship triplet, the second entity relationship triplet and the third entity relationship triplet. That is, selecting
And is
The selected entity relationship triple is the target entity relationship triple. It should be noted that, since the target entity relationship triples are respectively selected from the first entity relationship triplet, the second entity relationship triplet, and the third entity relationship triplet, the number of the selected target entity relationship triples may be 3 for each set of target entity relationship triples.
And then, obtaining an extraction result of the determined entity relationship of each target entity according to the target entity relationship triple. Optionally, the entity relationship vectors in the target entity relationship triple may be weighted and averaged to obtain the target entity relationship vector. For example, the following may be expressed according to the formula: l ═ w1L′1+w2L′2+w3L′3To calculate the eyesAnd (5) marking entity relation vectors.
Wherein, L'1Representing a relationship triplet L from a first entity1In the selected target entity relationship triple, L'2Representing a relationship triplet L from a first entity2In the selected target entity relationship triple, L'3Representing a relationship triplet L from a first entity3In the selected target entity relationship triple, w1,w2,w3Are respectively L'1、L2'and L'3The corresponding weight can be learned by various integration methods or set by experience.
Finally, the entity relationship corresponding to the maximum value in the target entity relationship vector may be used as the relationship between the first entity and the second entity in the target entity relationship triple, and the entity relationship corresponding to the maximum value in the target entity relationship vector may be used as the entity relationship extraction result.
The method and the device can also display the entity relationship in the target entity relationship triple selected from the first entity relationship triple, the second entity relationship triple and the third entity relationship triple to the user respectively, so that the user can select the final entity relationship from the target entity relationship triple. And then taking the first entity and the second entity in the target entity relationship triple and the final entity relationship selected by the user as an entity relationship extraction result.
According to the entity relationship extraction method, the entity relationship is extracted through the entity and relationship combined extraction model, the accuracy of entity relationship extraction can be improved according to the interaction constraint information of named entity identification and relationship extraction. When the entity types and the relationship types are more, the entity relationships extracted by the relationship classification model and the relationship rule set are combined with the entity relationships extracted by the entity and relationship combined extraction model to determine the final entity relationship, so that the accuracy of the extraction result can be improved, and the stability of the system performance can be improved. Also, the present disclosure is highly applicable compared to rule-based methods.
It should be noted that although the various steps of the methods of the present disclosure are depicted in the drawings in a particular order, this does not require or imply that these steps must be performed in this particular order, or that all of the depicted steps must be performed, to achieve desirable results. Additionally or alternatively, certain steps may be omitted, multiple steps combined into one step execution, and/or one step broken down into multiple step executions, etc.
Further, in the present exemplary embodiment, there is also provided an entity relationship extracting apparatus 500, as shown in fig. 5, including:
a first entity relationship triple determining module 510, configured to obtain a text of an entity relationship to be extracted, and process the text through an entity and relationship joint extraction model to obtain a first entity relationship triple;
a second entity relationship triple determining module 520, configured to determine words and labels corresponding to the words in the text, match the words and labels corresponding to the words in the text according to the relationship rule set, and determine a second entity relationship triple;
a third entity relationship triple determining module 530, configured to construct an entity pair according to the words and the tags corresponding to the words in the text, process the text through the relationship classification model, and determine a third entity relationship triple for the entity pair;
and the entity relationship extraction result determining module 540 is configured to determine an entity relationship extraction result according to the first entity relationship triple, the second entity relationship triple, and the third entity relationship triple.
In an exemplary embodiment of the present disclosure, the entity relationship extraction result determining module includes:
the target triple selecting unit is used for selecting target entity relationship triples which are the same among the first entities and the second entities from the first entity relationship triples, the second entity relationship triples and the third entity relationship triples;
and the extraction result determining unit is used for determining the entity relationship extraction result according to the target entity relationship triple.
In an exemplary embodiment of the present disclosure, the extraction result determining unit is specifically configured to perform weighted average on entity relationship vectors in a target entity relationship triple to obtain a target entity relationship vector; and taking the first entity and the second entity in the target entity relationship triple and the entity relationship corresponding to the maximum value in the target entity relationship vector as an entity relationship extraction result.
In an exemplary embodiment of the present disclosure, the first entity relationship triplet determining module includes:
the word segmentation unit is used for carrying out word segmentation processing on the text to obtain a plurality of words;
the word label determining unit is used for converting a plurality of words into semantic vectors, inputting the semantic vectors into the entity and relationship combined extraction model, and obtaining word labels;
and the first triple construction unit is used for constructing a first entity relationship triple according to the word label.
In an exemplary embodiment of the disclosure, the second entity relationship triplet determining module includes:
the triggering word judgment unit is used for judging whether a single word in the text is the same as a relation triggering word in the relation rule or not aiming at each relation rule in the relation rule set;
the target label judging unit is used for searching whether a first target label identical to a first entity label in the relation rule and a second target label identical to a second entity label in the relation rule exist in the labels when the single word is identical to the relation triggering word;
and the second triple construction unit is used for constructing a second entity relationship triple according to the word corresponding to the first target label, the word corresponding to the second target label and the entity relationship type in the relationship rule when the first target label and the second target label both exist.
In an exemplary embodiment of the present disclosure, the entity relationship extracting apparatus further includes:
the marking module is used for marking the entity pairs in the text to obtain a marked text;
the third entity relationship triple determining module is specifically configured to process the marked text through the relationship classification model to determine a third entity relationship triple;
in an exemplary embodiment of the disclosure, the third entity relationship triple determining module implements the processing of the tagged text by the relationship classification model to determine a third entity relationship triple by:
converting the marked text into a corresponding text semantic vector;
inputting the text semantic vector into a relation classification model to obtain a probability distribution vector of an entity relation;
and constructing a third entity relation triple according to the entity pair and the probability distribution vector.
In an exemplary embodiment of the present disclosure, the entity relationship extracting apparatus further includes:
and the reference resolution module is used for performing reference resolution on the obtained initial text to obtain a text of the entity relationship to be extracted.
The specific details of each module or unit in the above device have been described in detail in the corresponding entity relationship extraction method, and therefore are not described herein again.
It should be noted that although in the above detailed description several modules or units of the device for action execution are mentioned, such a division is not mandatory. Indeed, the features and functionality of two or more modules or units described above may be embodied in one module or unit, according to embodiments of the present disclosure. Conversely, the features and functions of one module or unit described above may be further divided into embodiments by a plurality of modules or units.
In an exemplary embodiment of the present disclosure, there is also provided an electronic device including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute all or part of the steps of the entity relationship extraction method in the present exemplary embodiment.
Fig. 6 shows a schematic structural diagram of a computer system of an electronic device for implementing an embodiment of the present disclosure. It should be noted that the computer system 600 of the electronic device shown in fig. 6 is only an example, and should not bring any limitation to the functions and the scope of the application of the embodiments of the present disclosure.
As shown in fig. 6, the computer system 600 includes a Central Processing Unit (CPU)601 that can perform various appropriate actions and processes according to a program stored in a Read Only Memory (ROM)602 or a program loaded from a storage section 608 into a Random Access Memory (RAM) 603. In the RAM 603, various programs and data necessary for system operation are also stored. The CPU 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input/output (I/O) interface 605 is also connected to bus 604.
The following components are connected to the I/O interface 605: an input portion 606 including a keyboard, a mouse, and the like; an output portion 607 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 608 including a hard disk and the like; and a communication section 609 including a network interface card such as a Local Area Network (LAN) card, a modem, or the like. The communication section 609 performs communication processing via a network such as the internet. The driver 610 is also connected to the I/O interface 605 as needed. A removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 610 as necessary, so that a computer program read out therefrom is mounted in the storage section 608 as necessary.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program may be downloaded and installed from a network through the communication section 609, and/or installed from the removable medium 611. When the computer program is executed by a Central Processing Unit (CPU)601, various functions defined in the apparatus of the present application are executed.
In an exemplary embodiment of the disclosure, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of any one of the above.
It should be noted that the computer readable storage medium shown in the present disclosure can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present disclosure, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In contrast, in the present disclosure, a computer-readable signal medium may include a propagated data signal with computer-readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, radio frequency, etc., or any suitable combination of the foregoing.
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 (9)
1. An entity relationship extraction method, the method comprising:
acquiring a text of an entity relationship to be extracted, and processing the text through an entity and relationship combined extraction model to obtain a first entity relationship triple;
determining words in the text and labels corresponding to the words, matching the words and the labels in the text according to a relationship rule set, and determining a second entity relationship triple;
constructing an entity pair according to the words and the labels, processing the text through a relation classification model, and determining a third entity relation triple aiming at the entity pair;
selecting target entity relationship triples which are the same between first entities and between second entities from the first entity relationship triples, the second entity relationship triples and the third entity relationship triples, and determining an entity relationship extraction result according to the target entity relationship triples;
wherein, the relationship rule set is a set of relationship rules, and each relationship rule comprises: the relationship triggering words, the relationship types, the first entity labels and the second entity labels;
matching the words and the labels in the text according to the relationship rule set to determine a second entity relationship triple, including:
judging whether a single word in the text is the same as a relation triggering word in the relation rule or not aiming at each relation rule in the relation rule set;
searching whether a first target label identical to a first entity label in the relation rule and a second target label identical to a second entity label in the relation rule exist in the labels when the single word is identical to the relation triggering word;
and when the first target label and the second target label exist, constructing a second entity relationship triple according to the word corresponding to the first target label, the word corresponding to the second target label and the entity relationship type in the relationship rule.
2. The method of claim 1, wherein determining an entity relationship extraction result according to the target entity relationship triple comprises:
carrying out weighted average on the entity relationship vectors in the target entity relationship triple to obtain a target entity relationship vector;
and taking the first entity and the second entity in the target entity relationship triple and the entity relationship corresponding to the maximum value in the target entity relationship vector as an entity relationship extraction result.
3. The method of claim 1, wherein the processing the text through the entity and relationship joint extraction model to obtain a first entity-relationship triple comprises:
performing word segmentation processing on the text to obtain a plurality of words;
converting the plurality of words into semantic vectors, inputting the semantic vectors into an entity and relation combined extraction model, and obtaining word labels;
and constructing a first entity relationship triple according to the word label.
4. The method of claim 1, wherein after said building an entity pair from said terms and said tags, said method further comprises:
marking entity pairs in the text to obtain a marked text;
the processing the text through the relationship classification model to determine a third entity relationship triple specifically includes:
and processing the marked text through a relation classification model to determine a third entity relation triple.
5. The method of claim 4, wherein the processing the tagged text through a relationship classification model to determine a third entity-relationship triple comprises:
converting the marked text into a corresponding text semantic vector;
inputting the text semantic vector into a relation classification model to obtain a probability distribution vector of an entity relation;
and constructing a third entity relationship triple according to the entity pair and the probability distribution vector.
6. The method of claim 1, further comprising:
and performing reference resolution on the obtained initial text to obtain the text of the entity relationship to be extracted.
7. An entity relationship extraction apparatus, the apparatus comprising:
the system comprises a first entity relationship triple determining module, a second entity relationship triple determining module and a third entity relationship triple determining module, wherein the first entity relationship triple determining module is used for acquiring a text of an entity relationship to be extracted, and processing the text through an entity and relationship combined extraction model to obtain a first entity relationship triple;
the second entity relationship triple determining module is used for determining words in the text and labels corresponding to the words, matching the words and the labels in the text according to a relationship rule set and determining a second entity relationship triple;
a third entity relationship triple determining module, configured to construct an entity pair according to the word and the tag, process the text through a relationship classification model, and determine a third entity relationship triple for the entity pair;
an entity relationship extraction result determining module, configured to select, from the first entity relationship triple, the second entity relationship triple, and the third entity relationship triple, a target entity relationship triple in which both the first entity and the second entity are the same, and determine an entity relationship extraction result according to the target entity relationship triple;
wherein, the relationship rule set is a set of relationship rules, and each relationship rule comprises: the relationship triggering words, the relationship types, the first entity labels and the second entity labels;
the second entity relationship triple determining module is used for matching the words and the labels in the text according to a relationship rule set and determining a second entity relationship triple:
judging whether a single word in the text is the same as a relation triggering word in the relation rule or not aiming at each relation rule in the relation rule set;
searching whether a first target label identical to a first entity label in the relation rule and a second target label identical to a second entity label in the relation rule exist in the labels when the single word is identical to the relation triggering word;
and when the first target label and the second target label exist, constructing a second entity relationship triple according to the word corresponding to the first target label, the word corresponding to the second target label and the entity relationship type in the relationship rule.
8. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the method of any one of claims 1 to 6.
9. An electronic device, comprising:
a processor; and
a memory for storing executable instructions of the processor;
wherein the processor is configured to perform the method of any of claims 1-6 via execution of the executable instructions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910941040.9A CN110705301B (en) | 2019-09-30 | 2019-09-30 | Entity relationship extraction method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910941040.9A CN110705301B (en) | 2019-09-30 | 2019-09-30 | Entity relationship extraction method and device, storage medium and electronic equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110705301A CN110705301A (en) | 2020-01-17 |
| CN110705301B true CN110705301B (en) | 2021-01-26 |
Family
ID=69197350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910941040.9A Active CN110705301B (en) | 2019-09-30 | 2019-09-30 | Entity relationship extraction method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110705301B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111339314B (en) * | 2020-02-19 | 2024-02-13 | 北京百度网讯科技有限公司 | Ternary group data generation method and device and electronic equipment |
| CN113378568B (en) * | 2020-03-09 | 2024-05-14 | 株式会社理光 | Relation extraction method, device, system and computer readable storage medium |
| CN111400448A (en) * | 2020-03-12 | 2020-07-10 | 中国建设银行股份有限公司 | Method and device for analyzing incidence relation of objects |
| CN113536766B (en) * | 2020-04-16 | 2024-04-12 | 浙江大搜车软件技术有限公司 | Analysis method and device for automobile maintenance records |
| CN111581376B (en) * | 2020-04-17 | 2024-04-19 | 中国船舶重工集团公司第七一四研究所 | Automatic knowledge graph construction system and method |
| CN111651575B (en) * | 2020-05-29 | 2023-09-12 | 泰康保险集团股份有限公司 | Session text processing method, device, medium and electronic equipment |
| CN112001180A (en) * | 2020-07-14 | 2020-11-27 | 北京百度网讯科技有限公司 | Multi-mode pre-training model acquisition method and device, electronic equipment and storage medium |
| CN112463960B (en) * | 2020-10-30 | 2021-07-27 | 完美世界控股集团有限公司 | Entity relationship determination method and device, computing equipment and storage medium |
| CN112507688A (en) * | 2020-12-16 | 2021-03-16 | 咪咕数字传媒有限公司 | Text similarity analysis method and device, electronic equipment and readable storage medium |
| CN112765330A (en) * | 2020-12-31 | 2021-05-07 | 科沃斯商用机器人有限公司 | Text data processing method and device, electronic equipment and storage medium |
| CN113160917B (en) * | 2021-05-18 | 2022-11-01 | 山东浪潮智慧医疗科技有限公司 | Electronic medical record entity relation extraction method |
| CN114595686B (en) * | 2022-03-11 | 2023-02-03 | 北京百度网讯科技有限公司 | Knowledge extraction method, and training method and device of knowledge extraction model |
| CN114613515B (en) * | 2022-03-28 | 2023-02-07 | 医渡云（北京）技术有限公司 | Medical entity relationship extraction method and device, storage medium and electronic equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107273349A (en) * | 2017-05-09 | 2017-10-20 | 清华大学 | A kind of entity relation extraction method and server based on multilingual |
| CN107480191A (en) * | 2017-07-12 | 2017-12-15 | 清华大学 | A kind of entity alignment model of iteration |
| CN107506486A (en) * | 2017-09-21 | 2017-12-22 | 北京航空航天大学 | A kind of relation extending method based on entity link |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8886522B2 (en) * | 2012-03-30 | 2014-11-11 | The Florida State University Research Foundation, Inc. | Automated extraction of bio-entity relationships from literature |
| US10339143B2 (en) * | 2014-05-08 | 2019-07-02 | Koninklijke Philips N.V. | Systems and methods for relation extraction for Chinese clinical documents |
| CN104657750B (en) * | 2015-03-23 | 2018-04-27 | 苏州大学张家港工业技术研究院 | A kind of method and apparatus extracted for character relation |
| CN106844413B (en) * | 2016-11-11 | 2020-12-08 | 南京柯基数据科技有限公司 | Method and device for extracting entity relationship |
| CN108733792B (en) * | 2018-05-14 | 2020-12-01 | 北京大学深圳研究生院 | Entity relation extraction method |
| CN108875000B (en) * | 2018-06-14 | 2021-12-28 | 广东工业大学 | Semantic relation classification method fusing multi-syntax structure |
| CN109408812A (en) * | 2018-09-30 | 2019-03-01 | 北京工业大学 | A method of the sequence labelling joint based on attention mechanism extracts entity relationship |
| CN110046351B (en) * | 2019-04-19 | 2022-06-14 | 福州大学 | Text relation extraction method based on features under rule driving |
-
2019
- 2019-09-30 CN CN201910941040.9A patent/CN110705301B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107273349A (en) * | 2017-05-09 | 2017-10-20 | 清华大学 | A kind of entity relation extraction method and server based on multilingual |
| CN107480191A (en) * | 2017-07-12 | 2017-12-15 | 清华大学 | A kind of entity alignment model of iteration |
| CN107506486A (en) * | 2017-09-21 | 2017-12-22 | 北京航空航天大学 | A kind of relation extending method based on entity link |
Non-Patent Citations (2)
| Title |
|---|
| Improvement of n-ary Relation Extraction by Adding Lexical Semantics to Distant-Supervision Rule Learning;Hong Li等;《ICAART 2015: Proceedings of the International Conference on Agents and Artificial Intelligence》;20151231;1-8 * |
| 知识图谱研究进展;漆桂林等;《情报工程》;20170215;第3卷(第1期);4-25 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110705301A (en) | 2020-01-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110705301B (en) | Entity relationship extraction method and device, storage medium and electronic equipment | |
| US11501182B2 (en) | Method and apparatus for generating model | |
| CN111444340B (en) | Text classification method, device, equipment and storage medium | |
| EP4060565A1 (en) | Method and apparatus for acquiring pre-trained model | |
| CN112270196B (en) | Entity relationship identification method and device and electronic equipment | |
| CN111737476A (en) | Text processing method and device, computer readable storage medium and electronic equipment | |
| CN113392209B (en) | Text clustering method based on artificial intelligence, related equipment and storage medium | |
| CN111339765B (en) | Text quality assessment method, text recommendation method and device, medium and equipment | |
| CN112988963B (en) | User intention prediction method, device, equipment and medium based on multi-flow nodes | |
| CN111666500A (en) | Training method of text classification model and related equipment | |
| CN114861889A (en) | Deep learning model training method, target object detection method and device | |
| CN112052424B (en) | Content auditing method and device | |
| CN111414561A (en) | Method and apparatus for presenting information | |
| CN111967267A (en) | XLNET-based news text region extraction method and system | |
| CN113761190A (en) | Text recognition method and device, computer readable medium and electronic equipment | |
| CN112926341A (en) | Text data processing method and device | |
| CN117114063A (en) | Method for training a generative large language model and for processing image tasks | |
| CN115357710B (en) | Training method and device for table description text generation model and electronic equipment | |
| CN116450829A (en) | Medical text classification method, device, equipment and medium | |
| CN116092101A (en) | Training method, image recognition method apparatus, device, and readable storage medium | |
| CN114911940A (en) | Text emotion recognition method and device, electronic equipment and storage medium | |
| CN115168609A (en) | Text matching method and device, computer equipment and storage medium | |
| CN114357195A (en) | Knowledge graph-based question-answer pair generation method, device, equipment and medium | |
| CN114490969A (en) | Question and answer method and device based on table and electronic equipment | |
| CN113806541A (en) | Emotion classification method and emotion classification model training method and device |
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 | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20200721 Address after: Room 806, 8 / F, Zhongguancun International Innovation Building, Haidian District, Beijing 100080 Applicant after: Beijing Jingdong intelligent city big data research institute Address before: 100086 No.76 Zhichun Road, Haidian District, Beijing, Building No.1, Building No.9, Floor 1-7-5 Applicant before: Jingdong City (Beijing) Digital Technology Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |