CN114048748A - Named entity recognition system, method, electronic device, and medium - Google Patents
Named entity recognition system, method, electronic device, and medium Download PDFInfo
- Publication number
- CN114048748A CN114048748A CN202111363516.9A CN202111363516A CN114048748A CN 114048748 A CN114048748 A CN 114048748A CN 202111363516 A CN202111363516 A CN 202111363516A CN 114048748 A CN114048748 A CN 114048748A
- Authority
- CN
- China
- Prior art keywords
- text
- layer
- entity recognition
- processed
- named entity
- 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 47
- 239000013598 vector Substances 0.000 claims abstract description 57
- 238000012216 screening Methods 0.000 claims abstract description 54
- 238000000605 extraction Methods 0.000 claims abstract description 33
- 230000004927 fusion Effects 0.000 claims description 26
- 238000012549 training Methods 0.000 claims description 23
- 230000011218 segmentation Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 6
- 238000013135 deep learning Methods 0.000 claims description 5
- 238000010801 machine learning Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 abstract description 31
- 230000008569 process Effects 0.000 description 10
- 238000010586 diagram Methods 0.000 description 7
- 230000008901 benefit Effects 0.000 description 6
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 230000000306 recurrent effect Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000003058 natural language processing Methods 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 239000012634 fragment Substances 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000006467 substitution reaction Methods 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
- G06F40/295—Named entity recognition
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention provides a named entity recognition system, a named entity recognition method, electronic equipment and a medium, wherein the named entity recognition system comprises the following steps: the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer; the word embedding layer is used for converting the text to be processed into a characteristic word vector and inputting the characteristic word vector into the coarse-granularity screening layer; the coarse-granularity screening layer is used for screening the feature word vectors based on the target scene and inputting the screened feature word vectors into the fine-granularity screening layer; and the fine-grained extraction layer is used for extracting a target entity corresponding to the target scene from the screened feature word vectors. The invention can shorten the processing period of massive texts and improve the processing efficiency under the condition of meeting the requirement of certain accuracy.
Description
Technical Field
The present invention relates to the field of artificial intelligence technologies, and in particular, to a named entity recognition system, method, electronic device, and medium.
Background
The Named Entity Recognition (NER) task is one of basic tasks related to Natural Language Processing (NLP) technology in the field of artificial intelligence, and aims to extract predefined target entities from texts, such as names of people, names of institutions, predefined professional names and the like, so as to provide rich information for subsequent various deeper text tasks. In the current big data era, massive trivial information and text fragments can be rapidly stacked in a short time, and although various algorithms related to NER are numerous at present, the requirement of certain accuracy rate on rapid processing of massive texts in a short time is difficult to meet.
Disclosure of Invention
In view of the above, the present invention provides a named entity recognition system, method, electronic device and medium, which can shorten the processing period of massive texts and improve the processing efficiency under the requirement of a certain accuracy.
In order to achieve the above purpose, the embodiment of the present invention adopts the following technical solutions:
in a first aspect, an embodiment of the present invention provides a named entity identification system, including: the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer; the word embedding layer is used for converting the text to be processed into a characteristic word vector and inputting the characteristic word vector into the coarse-granularity screening layer; the coarse-granularity screening layer is used for screening the feature word vectors based on the target scene and inputting the screened feature word vectors into the fine-granularity screening layer; and the fine-grained extraction layer is used for extracting a target entity corresponding to the target scene from the screened feature word vectors.
In one embodiment, the word embedding layer includes: a word segmentation model and a pre-training language model; the word segmentation model is used for segmenting the text to be processed; the pre-training language model is used for converting the text to be processed after word segmentation into a characteristic word vector.
In one embodiment, the coarse-grained screening layer includes one or more of the following scene classifiers: rule classifier, machine learning classifier, deep learning classifier.
In one embodiment, the fine-grained decimation layer includes a one-dimensional convolutional network, the one-dimensional convolutional network comprising: the system comprises a first sub-network, a second sub-network, a feature fusion module and an entity identification module; the first sub-network is used for performing first convolution operation on the screened feature word vectors to obtain a first convolution result; the second sub-network is used for performing second convolution operation on the first convolution result to obtain a second convolution result; the characteristic fusion module is used for carrying out characteristic fusion operation on the first convolution result and the second convolution result to obtain a characteristic fusion result; and the entity identification module is used for carrying out entity identification based on the feature fusion result to obtain a target entity.
In a second aspect, an embodiment of the present invention provides a named entity identification method, including: acquiring a text to be processed; inputting a text to be processed into a preset named entity recognition system; wherein the named entity recognition system is the system of any one of the above-mentioned first aspect; and carrying out entity recognition on the text to be processed through a named entity recognition system to obtain a target entity.
In one embodiment, the step of performing entity recognition on the text to be processed by the named entity recognition system to obtain the target entity includes: converting the text to be processed into a feature word vector through a named entity recognition system; screening the feature word vectors based on the target scene to obtain screened feature word vectors; and extracting a target entity corresponding to the target scene from the screened feature word vectors.
In one embodiment, the step of converting the text to be processed into the feature word vector by the named entity recognition system includes: segmenting a text to be processed by a named entity recognition system; and converting the text to be processed after word segmentation into a feature word vector.
In one embodiment, the step of extracting a target entity corresponding to a target scene from the filtered feature word vectors includes: performing a first convolution operation on the feature word vector to obtain a first convolution result; performing a second convolution operation on the first convolution result to obtain a second convolution result; performing feature fusion operation on the first convolution result and the second convolution result to obtain a feature fusion result; and carrying out entity identification based on the feature fusion result to obtain a target entity corresponding to the target scene.
In a third aspect, an embodiment of the present invention provides an electronic device, which includes a processor and a memory, where the memory stores computer-executable instructions capable of being executed by the processor, and the processor executes the computer-executable instructions to implement the steps of any one of the methods provided in the second aspect.
In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to perform the steps of any one of the methods provided in the second aspect.
The embodiment of the invention has the following beneficial effects:
the embodiment of the invention provides the named entity identification system, the named entity identification method, the electronic equipment and the medium, wherein the system comprises: the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer; the word embedding layer is used for converting the text to be processed into a characteristic word vector and inputting the characteristic word vector into the coarse-granularity screening layer; the coarse-granularity screening layer is used for screening the feature word vectors based on the target scene and inputting the screened feature word vectors into the fine-granularity screening layer; and the fine-grained extraction layer is used for extracting a target entity corresponding to the target scene from the screened feature word vectors. According to the system, the recognition process is divided into coarse-grained screening and fine-grained extraction of related entity texts, the text to be processed far away from a target scene is quickly separated out as far as possible through a coarse-grained screening layer, and then the target entity is obtained through fine-grained extraction layer extraction so as to accelerate the processing speed of the entity text.
Additional features and advantages of the invention will be set forth in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
In order to make the aforementioned and other objects, features and advantages of the present invention comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a schematic structural diagram of a named entity recognition system according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of another named entity recognition system according to an embodiment of the present invention;
FIG. 3 is a schematic structural diagram of a coarse-grained screening layer according to an embodiment of the present invention;
fig. 4 is a training flowchart of a named entity recognition system according to an embodiment of the present invention;
fig. 5 is a flowchart of a named entity recognition method according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
To make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is apparent that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the modern big data era, massive trivial information and text fragments can be rapidly stacked in a short time, and although various algorithms related to NER are developed, in practical engineering application, particularly in the field of financial wind control, under the condition of meeting the requirement of certain accuracy, how to rapidly process massive texts in a short time is a pain point of engineering application.
Although the current mainstream NER algorithm such as a Bert + Bi-LSTM, a Bert + Transformer and other recurrent neural network or graph neural network structures can meet higher accurate effect, when a billionth-level short text is rapidly processed, the application of the network structures needs a processing machine to have very high GPU configuration resources, and a Bert memory or recurrent neural network model cannot meet the performance requirement of rapid processing under the condition of limited implementation cost. Based on this, the hole convolution model IDCNN/IDCNN + CRF algorithm enters application workbenches of many NLP engineers, under the same-proportion condition, the calculation performance is greatly improved (the text processing speed is about 2 times of the former), but the accuracy rate is not much more superior than that of various recurrent neural networks under Bert, taking CoNL-2003 as an example, the accuracy rate of Bi-LSTM + CRF/Bi-LSTM is 90.43/89.34, and the accuracy rate of IDCNN + CRF/IDCNN is 90.54/90.32. The purpose of the hole convolution layer is to increase the perception field of view and avoid the loss of resolution of the conventional CNN due to the Pooling method integrating the multi-scale context information, but for short texts, the text length itself is limited, and in the case of considering the calculation performance, the two-dimensional hole convolution layer is slower than the one-dimensional one. Therefore, the existing NER algorithm is difficult to meet the requirement of certain accuracy rate and quickly process massive texts in a short time.
Based on this, the named entity recognition system, the named entity recognition method, the electronic device and the medium provided by the embodiments of the present invention can shorten the processing period of a large amount of texts and improve the processing efficiency under the condition that the requirement of a certain accuracy rate is met.
To facilitate understanding of the embodiment, first, a named entity recognition system disclosed in the embodiment of the present invention is described in detail, referring to a schematic structural diagram of a named entity recognition system shown in fig. 1, which illustrates that the system includes: the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer are sequentially connected.
The word embedding layer is used for converting the text to be processed into a feature word vector and inputting the feature word vector into the coarse-granularity screening layer; the coarse-granularity screening layer is used for screening the feature word vectors based on a target scene, quickly filtering out scenes irrelevant to a target entity or texts with weak relevance, and inputting the screened feature word vectors into the fine-granularity screening layer; and the fine-grained extraction layer is used for extracting a target entity corresponding to the target scene from the screened feature word vectors.
The named entity recognition system provided by the embodiment of the invention divides the recognition process into coarse-grained screening and fine-grained extraction of related entity texts, firstly quickly separates out the texts to be processed far away from a target scene as far as possible through a coarse-grained screening layer, and then obtains the target entities through the fine-grained extraction layer, so as to accelerate the processing speed of the entity texts.
Further, referring to the schematic structural diagram of another named entity recognition system shown in fig. 2, the overall mechanism of the named entity recognition system is schematically divided into three layers: the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer.
Specifically, the word embedding layer includes: a word segmentation model and a pre-training language model; the word segmentation model is used for segmenting the text to be processed; the pre-training language model is used for converting the text to be processed after word segmentation into a characteristic word vector. The Word embedding layer carries out Word segmentation (Tokenize) on the batch of texts, and then the texts after Word segmentation are converted into batch characteristic Word vectors through a Word vector layer generated under a pre-training language model (Word2 vec). In terms of selecting the pre-training language model of the word embedding layer, the word embedding layer relates to large-scale data batch processing, and the performance of a second generation pre-training language model and the performance of a first generation pre-training language model are compared, the second generation pre-training language model is represented by Bert, although the second generation pre-training language model has better context semantic understanding performance, huge performance processing pressure and configuration pressure are brought at the same time, and the word embedding layer is always a difficult problem in the industrial batch processing speed. For short texts, the text content of the short texts is often single-segment or few-segment sentence content, and the advantage of the language model represented by the second generation pre-training language model Bert is mainly reflected in the sense-level, that is, the language model plays a more significant effect on long texts of multiple segments of sentences. Based on this, the application can adopt the pre-training language model Word2vec, and exert the advantages thereof in short text application.
In one embodiment, a coarse-grained screening layer, namely a Scene Classifier, is used to quickly filter out scenes that are not relevant to the target extraction entity or text that is weakly relevant, such as: to extract entities of the financial wind control short texts, the coarse-grained screening layer can filter out relevant texts in scenes such as schools, hospitals and the like.
In consideration, the named entity recognition task is a classification task, namely, fine-grained classification of each word or phrase in a text is performed, and the position, the part of speech and the like of the target entity in the text are determined by final classification. In massive texts, the application of some professional entities often appears in one or several types of scenes, and the conventional named entity identification method can directly classify words/words in each text in fine granularity, and achieves the purpose of identifying the positions and the parts of speech of the entities by identifying information such as target word-B (initial position label), target word-I (middle position label), target word-E (end position label), target word-S (single word/word position label), 0 (irrelevant word/word label) and the like. In hundreds of millions of random texts, most texts are irrelevant scene texts for the predefined professional entity targets needing extraction, namely the fine-grained classification result of most texts is always a full 0 series. Therefore, in order to increase the data processing speed, referring to fig. 3, the coarse-grained screening layer in the embodiment of the present invention may select one or more of the following scene classifiers according to different task sources: rule Classifier (Syntax Rules), machine learning Classifier (ML-Classifier), and deep learning Classifier (DL-Classifier). The rule classifier may be based on keywords or regular rules for screening, and the deep learning classifier may be a textcnn deep learning model.
The coarse-grained screening layer in the embodiment of the invention is similar to a coarse screening flow device, and directly separates out texts irrelevant to professional target entities as much as possible. For the financial wind control type short text scenes, only a single CPU processor (adopting AMD Ryzen7 PRO AVX 28 threads) has the processing speed of 3.3 pieces/ms, namely 8.42 h/hundred million, can completely separate scene groups of related classes and scene groups of unrelated entity recognition tasks, and the micro-f1 of the scene classes can reach 95.6%, so that the data processing speed is greatly increased.
In one embodiment, a fine-grained extraction layer, that is, a NER Classifier Block entity extraction layer, that is, a classification and extraction of words or words within a single text, is used to extract target entity information. In the embodiment of the present invention, in combination with the actual situation of the short text, the fine-grained extraction layer is reconstructed by using a one-dimensional convolution network structure, and as shown in fig. 2, the one-dimensional convolution network of the fine-grained extraction layer includes: the system comprises a first sub-network, a second sub-network, a feature fusion module and an entity identification module; the first sub-network is used for performing first convolution operation on the screened feature word vectors to obtain a first convolution result; the second sub-network is used for performing second convolution operation on the first convolution result to obtain a second convolution result; the characteristic fusion module is used for carrying out characteristic fusion operation on the first convolution result and the second convolution result to obtain a characteristic fusion result; and the entity identification module is used for carrying out entity identification based on the feature fusion result to obtain a target entity.
The fine-grained extraction layer in the embodiment of the invention adopts a convolution network structure more suitable for short text reconstruction, is different from a common recurrent neural network (RNN, LSTM, GRU and the like) or a Transfomer structure, and has higher calculation speed; meanwhile, the Pooling mode used by the conventional CNN network is skillfully avoided during final feature processing, so that the information of the feature fusion module is enriched as much as possible to improve the accuracy of the model and avoid overfitting.
The named entity recognition system provided by the embodiment of the invention is skillfully hierarchically reconstructed based on the classification angle by combining the condition of short texts, the recognition process is divided into coarse-grained screening and entity fine-grained extraction of related entity texts, and scenes far away from a target entity are quickly separated out as far as possible in coarse-grained classification (Scene Classifier) so as to accelerate the processing speed of actual unrelated texts; meanwhile, in a fine-grained Classifier (NER Classifier Block), the method avoids the conventional method of using a cyclic neural network to identify the named entity, uses a one-dimensional convolutional network structure layer to perform the named entity task, constructs a convolutional network structure suitable for short texts, accelerates the calculation processing speed, and skillfully avoids using Pooling mode in the traditional CNN network in a final feature processing Block by combining the characteristics of the short texts, so that the information of a feature fusion module is enriched as much as possible to improve the model accuracy and is not over-fitted. Under the matching use of the coarse-grained screening layer and the fine-grained extraction layer, the processing performance of massive texts can be accelerated, and particularly, the texts with the size of more than one hundred million are related.
In addition, on a single common CPU processor (adopting AMD Ryzen7 PRO AVX 28 core 16 threads), the processing speed of the rough classification screening in the Scene Classifier link can reach 3.3 pieces/ms, namely 8.42 h/hundred million level, and the micro-f1 can reach 0.956; fine-grained classifier processing speeds are around 0.5 bars/ms, i.e. 55.56 h/billion, and f1 can reach 0.9223. In practical application, due to the change of scene correlation related to batch text, the processing performance fluctuates in the range of 8.42 h/hundred million-55.56 h/hundred million, and if multiple CPUs/GPUs are clustered for processing, the processing performance is improved by times.
For the named entity recognition system, before formal work, training, optimizing and evaluating the whole system model are required, referring to a training flow chart of the named entity recognition system shown in fig. 4, it is illustrated that the training process mainly includes the following steps S402 to S408:
step S402: and acquiring massive text data.
Step S404: marking the acquired text data.
Specifically, the marking of the text data is mainly divided into two parts: the first part is the marking of scenes of a coarse-granularity screening layer, the text data is marked according to the scenes, and when the coarse-granularity classification layer only adopts Syntax Rules to judge Rules, the data marking is not needed; when the coarse grain classification layer adopts a machine learning classifier or a deep learning classifier, scene marking of the text data is required; and the second part is entity marking of a fine-grained extraction layer, namely sequence marking of text entity positions and parts of speech. Therefore, if the coarse granularity classification layer comprises the machine learning classifier or the depth learning classifier, each marking text simultaneously comprises marking information of the scene marking and the entity marking, otherwise, only marking information of the entity marking is needed.
Step S406: and carrying out data preprocessing and data set division according to the marked text data.
The marked text data is divided into a training set, a verification set and a test set according to a certain proportion (such as 8:1:1), and abstract text data is converted into structural data which can be learned by a model after passing through a word embedding layer.
Step S408: and training the coarse-grained screening layer and the fine-grained extraction layer based on the divided data sets to obtain an optimal model.
Specifically, the training set is used for training and learning the model, the verification set is used for capturing the optimized model in the training process, and the test set is used for evaluating the condition of the trained model and the accuracy.
Further, the trained or optimized Scene Classifier and NER Classifier Block are applied to the hierarchical structure shown in fig. 1, and the process of quickly screening and extracting the text of the related entity is completed according to the three-layer structure (the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer) shown in fig. 1.
As for the named entity recognition system, an embodiment of the present invention further provides a named entity recognition method, which is shown in a flowchart of the named entity recognition method shown in fig. 5, and illustrates that the method mainly includes the following steps S502 to S506:
step S502: and acquiring a text to be processed.
The text to be processed may include a large amount of short texts.
Step S504: and inputting the text to be processed into a preset named entity recognition system.
The named entity recognition system comprises a word embedding layer, a coarse-grained screening layer and a fine-grained extraction layer.
Step S506: and carrying out entity recognition on the text to be processed through a named entity recognition system to obtain a target entity.
Specifically, the named entity recognition system firstly converts a text to be processed into a feature word vector through a word embedding layer; then, screening the feature word vectors based on the target scene through a coarse-grained screening layer, quickly filtering out scenes irrelevant to the target entity or texts with weak relevance, and inputting the screened feature word vectors into a fine-grained screening layer; and finally, extracting a target entity corresponding to the target scene from the screened feature word vectors through a fine-grained extraction layer.
The named entity recognition method provided by the embodiment of the invention can be used for carrying out named entity recognition through the named entity recognition system, the named entity recognition system divides the recognition process into coarse-grained screening and fine-grained extraction of related entity texts, firstly, the text to be processed far away from a target scene is quickly separated out as far as possible through the coarse-grained screening layer, and then, the target entity is obtained through the fine-grained extraction layer, so that the processing speed of the entity text is accelerated.
In one embodiment, when the text to be processed is converted into the feature word vector by the named entity recognition system, the following ways may be adopted, but not limited to: firstly, performing word segmentation on a text to be processed through a named entity recognition system; and then converting the text to be processed after word segmentation into a feature word vector.
In an embodiment, when extracting a target entity corresponding to a target scene from the filtered feature word vectors, the following methods may be adopted, but are not limited to: firstly, performing first convolution operation on a feature word vector to obtain a first convolution result; then, carrying out second convolution operation on the first convolution result to obtain a second convolution result; then, performing feature fusion operation on the first convolution result and the second convolution result to obtain a feature fusion result; and finally, carrying out entity identification based on the feature fusion result to obtain a target entity corresponding to the target scene.
The method provided by the embodiment of the present invention has the same implementation principle and technical effect as the system embodiment, and for the sake of brief description, reference may be made to the corresponding contents in the system embodiment for the parts that are not mentioned in the method embodiment.
It should be noted that all the embodiments mentioned in the embodiments of the present invention are merely exemplary, and may be different from the embodiments in practical applications, and are not limited herein.
The embodiment of the invention also provides electronic equipment, which specifically comprises a processor and a storage device; the storage means has stored thereon a computer program which, when executed by the processor, performs the method of any of the above embodiments.
Fig. 6 is a schematic structural diagram of an electronic device 100 according to an embodiment of the present invention, where the electronic device 100 includes: a processor 60, a memory 61, a bus 62 and a communication interface 63, wherein the processor 60, the communication interface 63 and the memory 61 are connected through the bus 62; the processor 60 is arranged to execute executable modules, such as computer programs, stored in the memory 61.
The Memory 61 may include a high-speed Random Access Memory (RAM) and may also include a non-volatile Memory (non-volatile Memory), such as at least one disk Memory. The communication connection between the network element of the system and at least one other network element is realized through at least one communication interface 63 (which may be wired or wireless), and the internet, a wide area network, a local network, a metropolitan area network, and the like can be used.
The bus 62 may be an ISA bus, PCI bus, EISA bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one double-headed arrow is shown in FIG. 6, but that does not indicate only one bus or one type of bus.
The memory 61 is used for storing a program, the processor 60 executes the program after receiving an execution instruction, and the method executed by the apparatus defined by the flow process disclosed in any of the foregoing embodiments of the present invention may be applied to the processor 60, or implemented by the processor 60.
The processor 60 may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method may be performed by integrated logic circuits of hardware or instructions in the form of software in the processor 60. The Processor 60 may be a general-purpose Processor, and includes a Central Processing Unit (CPU), a Network Processor (NP), and the like; the device can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other Programmable logic devices, discrete Gate or transistor logic devices, discrete hardware components. The various methods, steps and logic blocks disclosed in the embodiments of the present invention may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of the method disclosed in connection with the embodiments of the present invention may be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software module may be located in ram, flash memory, rom, prom, or eprom, registers, etc. storage media as is well known in the art. The storage medium is located in a memory 61, and the processor 60 reads the information in the memory 61 and, in combination with its hardware, performs the steps of the above method.
The computer program product of the readable storage medium provided in the embodiment of the present invention includes a computer readable storage medium storing a program code, where instructions included in the program code may be used to execute the method described in the foregoing method embodiment, and specific implementation may refer to the foregoing method embodiment, which is not described herein again.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and other various media capable of storing program codes.
In addition, in the description of the embodiments of the present invention, unless otherwise explicitly specified or limited, the terms "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "inner", "outer", and the like indicate orientations or positional relationships based on those shown in the drawings, and are only for convenience of description and simplification of description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and thus, should not be construed as limiting the present invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and are not to be construed as indicating or implying relative importance.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A named entity recognition system, comprising: the word embedding layer, the coarse-grained screening layer and the fine-grained extraction layer;
the word embedding layer is used for converting the text to be processed into a feature word vector and inputting the feature word vector to the coarse-granularity screening layer;
the coarse-granularity screening layer is used for screening the feature word vectors based on a target scene and inputting the screened feature word vectors into the fine-granularity screening layer;
and the fine-grained extraction layer is used for extracting a target entity corresponding to the target scene from the screened feature word vectors.
2. The named entity recognition system of claim 1, wherein the word embedding layer comprises: a word segmentation model and a pre-training language model;
the word segmentation model is used for segmenting the text to be processed;
the pre-training language model is used for converting the text to be processed after word segmentation into the feature word vector.
3. The system of claim 1, wherein the coarse-grained screening layer comprises one or more of the following scene classifiers: rule classifier, machine learning classifier, deep learning classifier.
4. The system of claim 1, wherein the fine-grained decimation layer comprises a one-dimensional convolutional network comprising: the system comprises a first sub-network, a second sub-network, a feature fusion module and an entity identification module;
the first sub-network is used for performing first convolution operation on the screened feature word vectors to obtain a first convolution result;
the second sub-network is used for performing second convolution operation on the first convolution result to obtain a second convolution result;
the feature fusion module is used for performing feature fusion operation on the first convolution result and the second convolution result to obtain a feature fusion result;
and the entity identification module is used for carrying out entity identification based on the feature fusion result to obtain the target entity.
5. A named entity recognition method, comprising:
acquiring a text to be processed;
inputting the text to be processed into a preset named entity recognition system; wherein the named entity recognition system is the system of any one of claims 1 to 4;
and carrying out entity recognition on the text to be processed through the named entity recognition system to obtain a target entity.
6. The method according to claim 5, wherein the step of performing entity recognition on the text to be processed by the named entity recognition system to obtain a target entity comprises:
converting the text to be processed into a feature word vector through the named entity recognition system;
screening the feature word vectors based on a target scene to obtain the screened feature word vectors;
and extracting a target entity corresponding to the target scene from the screened feature word vectors.
7. The method of claim 6, wherein the step of converting the text to be processed into feature word vectors by the named entity recognition system comprises:
segmenting the text to be processed by the named entity recognition system;
and converting the text to be processed after word segmentation into a feature word vector.
8. The method according to claim 6, wherein the step of extracting the target entity corresponding to the target scene from the filtered feature word vector comprises:
performing a first convolution operation on the feature word vector to obtain a first convolution result;
performing a second convolution operation on the first convolution result to obtain a second convolution result;
performing feature fusion operation on the first convolution result and the second convolution result to obtain a feature fusion result;
and performing entity identification based on the feature fusion result to obtain a target entity corresponding to the target scene.
9. An electronic device comprising a processor and a memory, the memory storing computer-executable instructions executable by the processor to perform the steps of the method of any one of claims 6 to 8.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of the preceding claims 6 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111363516.9A CN114048748B (en) | 2021-11-17 | 2021-11-17 | Named entity recognition system, named entity recognition method, named entity recognition electronic equipment and named entity recognition medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111363516.9A CN114048748B (en) | 2021-11-17 | 2021-11-17 | Named entity recognition system, named entity recognition method, named entity recognition electronic equipment and named entity recognition medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114048748A true CN114048748A (en) | 2022-02-15 |
| CN114048748B CN114048748B (en) | 2024-04-05 |
Family
ID=80209919
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111363516.9A Active CN114048748B (en) | 2021-11-17 | 2021-11-17 | Named entity recognition system, named entity recognition method, named entity recognition electronic equipment and named entity recognition medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114048748B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109918500A (en) * | 2019-01-17 | 2019-06-21 | 平安科技（深圳）有限公司 | File classification method and relevant device based on convolutional neural networks |
| CN111368541A (en) * | 2018-12-06 | 2020-07-03 | 北京搜狗科技发展有限公司 | Named entity identification method and device |
| WO2020232861A1 (en) * | 2019-05-20 | 2020-11-26 | 平安科技（深圳）有限公司 | Named entity recognition method, electronic device and storage medium |
| CN112711948A (en) * | 2020-12-22 | 2021-04-27 | 北京邮电大学 | Named entity recognition method and device for Chinese sentences |
| CN112800764A (en) * | 2020-12-31 | 2021-05-14 | 江苏网进科技股份有限公司 | Entity extraction method in legal field based on Word2Vec-BilSTM-CRF model |
| CN112818694A (en) * | 2021-03-02 | 2021-05-18 | 浙江工业大学 | Named entity recognition method based on rules and improved pre-training model |
| CN112989829A (en) * | 2021-02-10 | 2021-06-18 | 海尔数字科技（上海）有限公司 | Named entity identification method, device, equipment and storage medium |
-
2021
- 2021-11-17 CN CN202111363516.9A patent/CN114048748B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111368541A (en) * | 2018-12-06 | 2020-07-03 | 北京搜狗科技发展有限公司 | Named entity identification method and device |
| CN109918500A (en) * | 2019-01-17 | 2019-06-21 | 平安科技（深圳）有限公司 | File classification method and relevant device based on convolutional neural networks |
| WO2020232861A1 (en) * | 2019-05-20 | 2020-11-26 | 平安科技（深圳）有限公司 | Named entity recognition method, electronic device and storage medium |
| CN112711948A (en) * | 2020-12-22 | 2021-04-27 | 北京邮电大学 | Named entity recognition method and device for Chinese sentences |
| CN112800764A (en) * | 2020-12-31 | 2021-05-14 | 江苏网进科技股份有限公司 | Entity extraction method in legal field based on Word2Vec-BilSTM-CRF model |
| CN112989829A (en) * | 2021-02-10 | 2021-06-18 | 海尔数字科技（上海）有限公司 | Named entity identification method, device, equipment and storage medium |
| CN112818694A (en) * | 2021-03-02 | 2021-05-18 | 浙江工业大学 | Named entity recognition method based on rules and improved pre-training model |
Non-Patent Citations (1)
| Title |
|---|
| 盛剑 等: "多场景文本的细粒度命名实体识别", 《中文信息学报》, pages 1 - 8 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114048748B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112115267B (en) | Training method, device, equipment and storage medium of text classification model | |
| CN108319668A (en) | Generate the method and apparatus of text snippet | |
| CN110309301B (en) | Enterprise category classification method and device and intelligent terminal | |
| CN109543826A (en) | A kind of activation amount quantization method and device based on deep neural network | |
| CN111475622A (en) | Text classification method, device, terminal and storage medium | |
| WO2023134088A1 (en) | Video summary generation method and apparatus, electronic device, and storage medium | |
| CN112528637A (en) | Text processing model training method and device, computer equipment and storage medium | |
| CN110968689A (en) | Training method of criminal name and law bar prediction model and criminal name and law bar prediction method | |
| CN112232070A (en) | Natural language processing model construction method, system, electronic device and storage medium | |
| CN109784207B (en) | Face recognition method, device and medium | |
| CN115374325A (en) | Website classification method and device, classification equipment and storage medium | |
| CN116796288A (en) | Industrial document-oriented multi-mode information extraction method and system | |
| CN116955644A (en) | Knowledge fusion method, system and storage medium based on knowledge graph | |
| CN114048748B (en) | Named entity recognition system, named entity recognition method, named entity recognition electronic equipment and named entity recognition medium | |
| CN115984633A (en) | Gate-level circuit component identification method, system, storage medium and equipment | |
| CN115438658A (en) | Entity recognition method, recognition model training method and related device | |
| CN112100986B (en) | Voice text clustering method and device | |
| CN112131384A (en) | News classification method and computer-readable storage medium | |
| CN113868417A (en) | Sensitive comment identification method and device, terminal equipment and storage medium | |
| CN114065749A (en) | Text-oriented Guangdong language recognition model and training and recognition method of system | |
| CN108021918B (en) | Character recognition method and device | |
| CN113836297A (en) | Training method and device for text emotion analysis model | |
| Chang et al. | Code Transform Model Producing High-Performance Program. | |
| KR20200135044A (en) | Apparatus and method of defect classification using image transformation based on machine-learning | |
| CN115550684B (en) | Improved video content filtering method and system |
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 |