CN109753653B - Entity name recognition method, entity name recognition device, computer equipment and storage medium - Google Patents
Entity name recognition method, entity name recognition device, computer equipment and storage medium Download PDFInfo
- Publication number
- CN109753653B CN109753653B CN201811592664.6A CN201811592664A CN109753653B CN 109753653 B CN109753653 B CN 109753653B CN 201811592664 A CN201811592664 A CN 201811592664A CN 109753653 B CN109753653 B CN 109753653B
- Authority
- CN
- China
- Prior art keywords
- vector
- word
- name
- neural network
- text
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 239000013598 vector Substances 0.000 claims abstract description 305
- 238000013528 artificial neural network Methods 0.000 claims abstract description 123
- 238000012216 screening Methods 0.000 claims abstract description 75
- 230000002457 bidirectional effect Effects 0.000 claims abstract description 50
- 238000006243 chemical reaction Methods 0.000 claims description 35
- 238000002372 labelling Methods 0.000 claims description 23
- 238000004590 computer program Methods 0.000 claims description 13
- 238000000605 extraction Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 18
- 238000012549 training Methods 0.000 description 12
- 238000001514 detection method Methods 0.000 description 9
- 238000010586 diagram Methods 0.000 description 9
- 230000008569 process Effects 0.000 description 6
- 239000000284 extract Substances 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 238000009795 derivation Methods 0.000 description 2
- 238000012015 optical character recognition Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 230000001960 triggered effect Effects 0.000 description 2
- 230000001934 delay Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002441 reversible effect Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- 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
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Character Discrimination (AREA)
Abstract
The application relates to an entity name identification method, an entity name identification device, computer equipment and a storage medium. The method comprises the following steps: acquiring word vectors corresponding to words in the text to be recognized; inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network; screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set; extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively; and identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network. The method firstly screens the candidate entity names, then identifies the entity names from the candidate entity names, and improves the accuracy of entity name identification through the processing of the two-way long and short memory neural network.
Description
Technical Field
The present invention relates to the field of pattern recognition, and in particular, to a method, an apparatus, a computer device, and a storage medium for recognizing an entity name.
Background
With the development of pattern recognition technology, named entity recognition technology (Named Entity Recognition, NER) has emerged, which aims to recognize proper nouns in natural language text, such as named entities of personal names, place names, company names, organization names, and the like.
However, in the conventional named entity recognition technology, a statistical-based model, such as a hidden markov model and a support vector machine, is usually adopted, training is required by using manually labeled corpus, the modeling capability of text context is weak, the influence of the front and back sequences of words in the text is not considered, and the accuracy of entity name recognition is low.
Disclosure of Invention
In view of the foregoing, it is desirable to provide an entity name recognition method, apparatus, computer device, and storage medium capable of improving recognition accuracy.
A method of entity name identification, the method comprising:
acquiring word vectors corresponding to words in the text to be recognized;
inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network;
screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set;
Extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively;
and identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
An entity name recognition apparatus, the apparatus comprising:
the vector acquisition module is used for acquiring word vectors corresponding to each word in the text to be identified;
the feature obtaining module is used for inputting the obtained word vector into a first two-way long and short memory neural network to obtain the vector feature output by the first two-way long and short memory neural network;
the collection obtaining module is used for screening candidate entity names from the text to be identified according to the vector characteristics through a noun screening model to obtain a name candidate collection;
the vector extraction module is used for extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively;
and the name recognition module is used for recognizing entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of:
Acquiring word vectors corresponding to words in the text to be recognized;
inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network;
screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set;
extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively;
and identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of:
acquiring word vectors corresponding to words in the text to be recognized;
inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network;
screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set;
extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively;
And identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
The entity name recognition method, the entity name recognition device, the computer equipment and the storage medium are used for obtaining word vectors corresponding to each word in the text to be recognized respectively, inputting the obtained word vectors into the first two-way long and short memory neural network, obtaining vector characteristics of the word vectors through the first two-way long and short memory neural network, marking the words according to the vector characteristics through the noun screening model, screening candidate entity names from the text to be recognized, and obtaining a name candidate set, namely, a set of all words which are possibly entity names in the text to be recognized; and extracting word vectors corresponding to the candidate entity names in the name candidate set respectively, inputting the extracted word vectors into a second bidirectional long and short memory neural network for recognition again, and recognizing the entity names in the name candidate set according to the extracted word vectors, so that the accuracy of entity name recognition is improved through the processing of the two-time bidirectional long and short memory neural network.
Drawings
FIG. 1 is a diagram of an application environment for a method of entity name identification in one embodiment;
FIG. 2 is a flow chart of a method for identifying entity names in one embodiment;
FIG. 3 is a flowchart illustrating steps for obtaining word vectors in one embodiment;
FIG. 4 is a schematic diagram of a vector conversion model in one embodiment;
FIG. 5 is a flowchart illustrating steps for obtaining vector features in one embodiment;
FIG. 6 is a flowchart illustrating steps for obtaining a candidate set of names in one embodiment;
FIG. 7 is a flowchart illustrating steps for identifying entity names in one embodiment;
FIG. 8 is a diagram of entity name identification in one embodiment;
FIG. 9 is a schematic diagram of an entity name detection model in one embodiment;
FIG. 10 is a schematic diagram of a structure of an entity name determination model in one embodiment;
FIG. 11 is a block diagram of an entity name recognition device in one embodiment;
fig. 12 is an internal structural diagram of a computer device in one embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
The entity name identification method provided by the application can be applied to an application environment shown in fig. 1, wherein the application environment can comprise a terminal 102 and a server 104, and the terminal 102 communicates with the server 104 through a network. The method may be applied to either the terminal 102 or the server 104. The terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers. The server 104 may be implemented as a stand-alone server or as a server cluster of multiple servers.
In one embodiment, as shown in fig. 2, there is provided an entity name recognition method, and this embodiment is illustrated by taking the application of the method to the terminal in fig. 1 as an example, and includes the following steps:
The text to be identified is input into the terminal, and the text of the entity name needs to be extracted. Words are word and phrase combinations, including words, phrases and whole vocabulary. Word vectors are vectors that are mapped to real numbers by words or phrases.
Specifically, the terminal acquires an input text to be recognized, recognizes each word in the text to be recognized, and converts each recognized word to obtain a word vector corresponding to each word. Wherein the word vector may be a fixed-length dense vector.
In one embodiment, the text to be recognized is chinese text, and the text to be recognized may be text in any language.
In one embodiment, the terminal presets the vector dimension of the word vector, and when the dimension of the word vector obtained by conversion does not reach the preset vector dimension, the dimension of the word vector reaches the preset vector dimension by supplementing 0 at the end of the word vector obtained by conversion.
The first two-way long and short memory neural network is a two-way long and short memory neural network model for extracting vector features of word vectors. The long and short memory neural network is a time recurrent neural network and is suitable for processing and predicting important events with very long intervals and delays in a time sequence; bi-directional Long-Short Term Memory Recurrent Neural Network is a variant of Long-short memory neural network (LSTM) and consists of two one-way Long-short memory neural networks. At each time t, the input is simultaneously provided to the two opposite long and short memory neural networks, and the output is determined by the two long and short memory neural networks. The vector feature is feature data output by the first two-way long and short memory neural network, and can reflect the feature of the word vector.
Specifically, after converting each word in the text to be recognized into a word vector, the terminal inputs the obtained word vector into a trained first two-way long and short memory neural network. The first two-way long and short memory neural network processes each input word vector and extracts vector features of each word vector.
And 206, screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set.
The noun screening model screens words possibly forming entity names from texts to be identified according to the extracted vector features. Candidate entity names consist of a number of words, which are possible entity names in the text to be identified. A name candidate set is a set of all candidate entity names.
Specifically, the words and the word vectors obtained through word conversion are in one-to-one correspondence, and the word vectors are in one-to-one correspondence with the extracted vector features from the word vectors, so that the words and the extracted vector features are in one-to-one correspondence. The terminal inputs the extracted vector features into a noun screening model, the noun screening model classifies words through the vector features, and combines words possibly forming entity names to obtain candidate entity names, and all candidate entity names in the text to be identified form a name candidate set.
Specifically, after the terminal obtains a name candidate set through a noun screening model, extracting each candidate entity name in the name candidate set, triggering a vector conversion instruction, and converting words in each extracted candidate entity name into word vectors according to the vector conversion instruction.
And 210, identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
The second bidirectional long and short memory neural network is a model for screening entity names from the candidate entity names. The entity name is the name of the entity.
Specifically, after the terminal converts each candidate entity name in the name candidate set into a word vector, the word vectors corresponding to the candidate entity names are input into a trained second bidirectional long and short memory neural network. And the second bidirectional long and short memory neural network processes the word vector, identifies the entity names in the candidate entity names, and outputs the identified entity names.
In one embodiment, the entity name is a company name, which may also include, but is not limited to, various personal names, place names, organization names.
In this embodiment, word vectors corresponding to each word in a text to be recognized are obtained, the obtained word vectors are input into a first two-way long and short memory neural network, vector features of the word vectors are obtained through the first two-way long and short memory neural network, the words are labeled according to the vector features through a noun screening model, candidate entity names are screened from the text to be recognized, and a name candidate set is obtained, namely, a set of all words which are possibly entity names in the text to be recognized; and extracting word vectors corresponding to the candidate recognition name candidate entity names in the name candidate set, inputting the extracted word vectors into a second bidirectional long and short memory neural network for recognition again, and recognizing the entity names in the name candidate set according to the extracted word vectors, so that the accuracy of entity name recognition is improved through the processing of the two-way long and short memory neural network.
As shown in fig. 3, in one embodiment, step 202 specifically further includes a step of acquiring a word vector, which specifically includes the steps of:
Specifically, the terminal acquires a text to be identified input by a user, and receives a triggered name extraction instruction, wherein the name extraction instruction is used for indicating the terminal to identify an entity name from the text to be identified.
In one embodiment, the terminal receives a triggered text file selection instruction, extracts a text file identifier from the text file selection instruction, queries a text file corresponding to the text file identifier from a file library, extracts the text file from the file library, and takes the text in the text file as a text to be identified.
In one embodiment, the terminal receives a picture input by a user, and recognizes characters in the picture to obtain a text to be recognized. For example, the terminal may recognize the text in the picture by OCR (Optical Character Recognition ).
And step 304, inputting the text to be recognized into a vector conversion model, and carrying out word division on the text to be recognized of the file to be recognized to obtain a plurality of words.
The vector conversion model is a model for converting words in the text to be recognized into word vectors.
Specifically, a trained vector conversion model is pre-stored in the terminal. After the terminal acquires the text to be recognized, inputting the text to be recognized into a vector conversion model, recognizing each word in the text to be recognized through the vector conversion model, and dividing the text to be recognized to obtain a plurality of words.
And 306, carrying out vector conversion on each word to obtain a word vector corresponding to each word.
Specifically, after dividing a text to be recognized into a plurality of words through a vector conversion model, the terminal respectively carries out vector conversion on each word to obtain word vectors corresponding to each word. The vector transformation model may be at least one of a Continuous Bag of Words model (CBOW) and a Skip-gram model in Word2vec (Word to vector) to generate a related model of Word vectors, which may be used to represent Word-to-Word relationships.
In one embodiment, the structure of the vector conversion model is shown in FIG. 4. The vector conversion model may be a continuous bag of words model including an input layer, a hidden layer, and an output layer. The terminal needs to obtain a continuous word bag model through training. The method comprises the steps that a training text is divided into a plurality of words by a terminal, each word is respectively converted into one-hot vectors according to an established corpus, the one-hot vectors are sparse vectors, only one of the plurality of dimensions is 1, and the rest of the plurality of dimensions are 0. The terminal selects a central word from the training text, takes a one-hot vector corresponding to the central word as expected output of an initial continuous word bag model, takes a context word of the central word, namely a word except the central word in the training text, and a corresponding one-hot vector as model input, trains the initial continuous word bag model, and obtains the continuous word bag model after training is finished. C in fig. 4 represents the number of context words in the training text; v represents the dimension of the one-hot vector, i.e., the total number of words in the corpus; n represents the number of hidden layer nodes, and the numerical value of N is the dimension of the word vector corresponding to each word in the text to be recognized.
For example, the training text is "today weather is clear", the terminal divides the training text into three words of "today", "weather" and "clear", and selects weather as a central word. The terminal takes the one-hot vector obtained by converting today and clear as the input of an initial continuous word bag model, and the value of C is 2; the terminal takes a one-hot vector corresponding to weather as expected output of an initial continuous word bag model; assuming 1000 words in the corpus, the value of V is 1000; when the value of N is 100, the dimension of the word vector corresponding to each word in the text to be recognized is 100.
In this embodiment, a text to be recognized is obtained and input into a vector conversion model, the text to be recognized is word-divided into a plurality of words by the vector conversion model, and then each word is vector-converted to obtain a word vector corresponding to each word, so that the processing from natural language to data is completed, and the processing speed of a terminal to the text to be recognized is improved.
As shown in fig. 5, in one embodiment, step 204 specifically further includes a step of obtaining a vector feature, where the step specifically includes the steps of:
The forward reference layer and the backward reference layer are one-way long and short memory neural networks in the two-way long and short memory neural network.
Specifically, the two-way long and short memory neural network consists of a forward reference layer and a backward reference layer. For each word vector, the terminal inputs into a forward reference layer and a backward reference layer in the two-way long and short memory neural network respectively. The terminal inputs each word vector into the forward reference layer in the reverse order of the input backward reference layer.
The forward reference features are feature data of word vectors extracted by the forward reference layer; the backward reference feature is feature data of a word vector extracted by the backward reference layer.
Specifically, for each input word vector, a forward reference layer in the first two-way long and short memory neural network processes the word vector, extracts feature data of the word vector, and obtains forward reference features; the backward reference layer processes the word vector to obtain backward reference characteristics. The terminal obtains a forward reference feature and a backward reference feature corresponding to each word vector respectively.
At step 506, vector features are generated from the forward reference features and the backward reference features.
Specifically, for each word vector, after the terminal obtains the forward reference feature and the backward reference feature corresponding to the word vector, the terminal calculates according to the forward reference feature and the backward reference feature to obtain vector features, and outputs the calculated vector features as a first two-way long and short memory neural network for each word vector.
In this embodiment, the obtained word vectors are respectively input to a forward reference layer and a backward reference layer in the first two-way long and short memory neural network, and the forward reference features and the backward reference features output by the forward reference layer and the backward reference layer are obtained, and are mutually referred to each other, so that vector features are generated according to the forward reference features and the backward reference features, and the accuracy of generating the vector features is improved.
As shown in fig. 6, in one embodiment, step 206 specifically further includes a step of obtaining a candidate set of names, which specifically includes the steps of:
The labeling result is obtained by labeling each word by the noun screening model.
Specifically, after obtaining the vector features output by the first two-way long and short memory neural network, the terminal inputs the vector features into a noun screening model. The noun screening model analyzes and processes the input vector features, marks words corresponding to the vector features according to the processing results, and obtains marking results of the words.
In one embodiment, the noun screening model may be a conditional random field (Conditional Random Fields, abbreviated CRF). The conditional random field is a discriminant probability model, which is a type of random field, and is commonly used for labeling or analyzing sequence data, such as natural language characters or biological sequences.
And step 604, according to the labeling result, the words labeled as name components are screened from the text to be identified.
Specifically, the noun screening model labels each term as a name component and a non-name component. And after the labeling is completed, the terminal screens the words labeled as name components from the text to be identified according to the labeling result.
Specifically, the name component in the labeling result may be further divided into a name header component and a name middle component. The name header component indicates that the term is the first term in the possible entity name and the name middle component indicates that the term is a component of a non-first term in the possible entity name. The noun screening model combines the words according to the labeling result of the screened words to obtain a plurality of candidate entity names, and obtains a name candidate set according to the plurality of candidate entity names.
For example, the text to be identified "Inactive XX software company" is divided into the words "Inactive", "XX", "software" and "company", the noun screening model labels "Inactive" as O (other), i.e., non-name components, and "XX", "software" and "company" as name components, where the noun screening model labels "XX" as B-ORG, i.e., name header components, and "software" and "company" as I-ORG, i.e., name middle components. The noun screening model screens XX, software and company marked as name components from the text to be identified, combines the XX, the software and the company according to marking results to obtain candidate entity names XX software company, and supposedly simultaneously exists Chinese tennis association in the text to be identified to obtain candidate entity names Chinese tennis association. The candidate entity names "XX software company" and "chinese tennis association" constitute a name candidate set.
In the embodiment, the vector features are input into a noun screening model to obtain labeling results of each word, and the labeling results identify whether the word is a component part of a possible entity name; according to the labeling result, words labeled as name components can be accurately screened from the text to be identified, then a plurality of candidate entity names are determined according to the screened words, a name candidate set is obtained, and accuracy of obtaining the candidate entity names is improved.
As shown in fig. 7, in one embodiment, step 210 specifically further includes a step of identifying an entity name, which specifically includes the steps of:
Specifically, for each candidate entity name, after obtaining word vectors corresponding to each word in the candidate entity name through a vector conversion model, the terminal inputs each word vector into a second bidirectional long and short memory neural network. The second bidirectional long and short memory neural network is also composed of a forward reference layer and a backward reference layer, and the terminal inputs word vectors into the forward reference layer and the backward reference layer respectively.
And step 704, acquiring a second bidirectional long and short memory neural network, and outputting a recognition result corresponding to each candidate entity name.
Specifically, the forward reference layer and the backward reference layer of the second bidirectional long and short memory neural network respectively process each word vector to obtain a vector processing result of each word vector. The second bidirectional long and short memory neural network obtains each vector processing result of each word vector in the candidate entity name by the forward reference layer and each vector processing result of each word vector in the candidate entity name by the backward reference layer, generates a recognition result corresponding to the candidate entity name according to the obtained vector processing result, and outputs the recognition result.
In one embodiment, the recognition result is an output result of an activation function between a hidden layer and an output layer of the second bidirectional long and short memory neural network. When the candidate entity name is the entity name, the activation function is in an activated state, and the output result is 1; when the candidate entity name is not the entity name, the activation function is in an inactive state, and the output result is 0.
And step 706, screening the identified entity names from the name candidate set according to the identification result.
Specifically, after the terminal obtains the recognition result of the second bidirectional long and short memory neural network, selecting a candidate entity name with the recognition result being the entity name from the name candidate set according to the recognition result, and obtaining the entity name in the text to be recognized.
For example, the entity name to be identified is a Chinese company name, and the candidate set of names includes "XX software company" and "Chinese tennis Association". The second bidirectional long and short memory neural network outputs 1 as the identification result of XX software company and 0 as the identification result of Chinese tennis Association. And taking the XX software company with the terminal screening identification result of 1 as an entity name in the text to be identified.
In this embodiment, a word vector corresponding to each candidate entity name is input into the second bidirectional long and short memory neural network, the candidate entity names are further identified, the second bidirectional long and short memory neural network is obtained, an identification result corresponding to each candidate entity name is output, whether the candidate entity name is a required entity name is identified by the identification result, the identified entity names are screened from the name candidate set according to the identification result, and accuracy of identifying the entity names is improved through second identification.
FIG. 8 is a diagram of entity name identification in one embodiment. Specifically, referring to fig. 8, the identification of the entity name may be accomplished by an entity name detection model and an entity name determination model. The entity name detection model is used for screening candidate entity names from the text to be identified to obtain a name candidate set; the entity name determination model is used to identify an entity name from among the candidate entity names.
The structure of the entity name detection model is shown in fig. 9, and the entity name detection model consists of a vector conversion model, a first two-way long and short memory neural network and a noun screening model. The vector conversion model divides the text to be identified "job-entering XX software company" into words "job-entering", "XX", "software" and "company", and converts each word into a word vector. The terminal inputs word vectors into a forward reference layer and a backward reference layer in a first two-way long and short memory neural network respectively, and the first two-way long and short memory neural network generates vector features corresponding to the word vectors respectively according to the forward reference features and the backward reference features output by the forward reference layer and the backward reference layer and inputs the vector features into a noun screening model. The noun screening model marks each word according to vector features and outputs marking results, marks 'job in' as O, namely a non-name component part, marks 'XX' as B-ORG, namely a name head component part, marks 'software' and 'company' as I-ORG, namely a name middle component part, and screens candidate name entities as 'XX software company'.
In one embodiment, the terminal obtains an entity name detection model through training, the terminal obtains training texts and manually marked candidate entity names, takes the training texts as input, manually marked candidate entity names as expected output, trains an initial entity name detection model, and adjusts parameters of each model in the initial entity name detection model to obtain the entity name detection model.
The structure of the entity name determining model is shown in fig. 10, and the entity name determining model consists of a vector conversion model and a second bidirectional long and short memory neural network. The vector conversion model divides the candidate name entity "XX software company" into words "XX", "software" and "company", and converts each word into a word vector. The terminal respectively inputs the word vectors into a forward reference layer and a backward reference layer in a second bidirectional long and short memory neural network, obtains vector processing results of the word vectors corresponding to the forward reference layer pair XX, the software and the company, and vector processing results of the word vectors corresponding to the backward reference layer pair XX, the software and the company, and outputs a recognition result 1 according to each reference result to indicate that the XX software company is an entity name.
In one embodiment, the terminal is trained to obtain an entity name determination model, and the terminal obtains a name candidate set and manually marked entity names, wherein the name candidate set comprises a plurality of candidate entity names. The terminal takes a plurality of candidate entity names in the name candidate set as input, takes the manually marked entity names as expected output, trains an initial entity name determination model, adjusts parameters of each model in the initial entity name determination model, and obtains the entity name determination model.
It should be understood that, although the steps in the flowcharts of fig. 2-3 and 5-7 are shown in order as indicated by the arrows, these steps are not necessarily performed in order as indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least some of the steps of FIGS. 2-3 and 5-7 may include multiple sub-steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the sub-steps or stages are performed necessarily occur in sequence, but may be performed alternately or alternately with other steps or at least a portion of the sub-steps or stages of other steps.
In one embodiment, as shown in fig. 11, there is provided an entity name recognition apparatus 1100, including: a vector acquisition module 1102, a feature derivation module 1104, a set derivation module 1106, a vector extraction module 1108, and a name recognition module 1110, wherein:
the vector obtaining module 1102 is configured to obtain word vectors corresponding to each word in the text to be identified.
The feature obtaining module 1104 is configured to input the obtained word vector into the first two-way long and short memory neural network, and obtain a vector feature output by the first two-way long and short memory neural network.
The collection obtaining module 1106 is configured to obtain a name candidate collection by screening candidate entity names from the text to be identified according to the vector features through the noun screening model.
The vector extraction module 1108 is configured to extract word vectors corresponding to the candidate entity names in the name candidate set.
The name recognition module 1110 is configured to recognize, through a second bidirectional long and short memory neural network, an entity name in the candidate set of names according to the extracted word vector.
In this embodiment, word vectors corresponding to each word in a text to be recognized are obtained, the obtained word vectors are input into a first two-way long and short memory neural network, vector features of the word vectors are obtained through the first two-way long and short memory neural network, the words are labeled according to the vector features through a noun screening model, candidate entity names are screened from the text to be recognized, and a name candidate set is obtained, namely, a set of all words which are possibly entity names in the text to be recognized; and extracting word vectors corresponding to the candidate recognition name candidate entity names in the name candidate set, inputting the extracted word vectors into a second bidirectional long and short memory neural network for recognition again, and recognizing the entity names in the name candidate set according to the extracted word vectors, so that the accuracy of entity name recognition is improved through the processing of the two-way long and short memory neural network.
In one embodiment, the vector acquisition module 1102 specifically includes: text acquisition module, word acquisition module and vector acquisition module, wherein:
the text acquisition module is used for acquiring a text to be identified;
the word obtaining module is used for inputting the text to be recognized into the vector conversion model, and carrying out word division on the text to be recognized to obtain a plurality of words;
the vector obtaining module is used for carrying out vector conversion on each word to obtain a word vector corresponding to each word.
In this embodiment, a text to be recognized is obtained and input into a vector conversion model, the text to be recognized is word-divided into a plurality of words by the vector conversion model, and then each word is vector-converted to obtain a word vector corresponding to each word, so that the processing from natural language to data is completed, and the processing speed of a terminal to the text to be recognized is improved.
In one embodiment, the feature obtaining module 1104 is configured to input the obtained word vectors to a forward reference layer and a backward reference layer in the first two-way long and short memory neural network, respectively; acquiring forward reference features and backward reference features output by a forward reference layer and a backward reference layer; vector features are generated from the forward reference features and the backward reference features.
In this embodiment, the obtained word vectors are respectively input to a forward reference layer and a backward reference layer in the first two-way long and short memory neural network, and the forward reference features and the backward reference features output by the forward reference layer and the backward reference layer are obtained, and are mutually referred to each other, so that vector features are generated according to the forward reference features and the backward reference features, and the accuracy of generating the vector features is improved.
In one embodiment, the collection obtaining module 1106 is configured to input the vector features into a noun screening model to obtain labeling results of each word; according to the labeling result, screening words labeled as name components from the text to be identified; and determining a plurality of candidate entity names according to the screened words to obtain a name candidate set.
In the embodiment, the vector features are input into a noun screening model to obtain labeling results of each word, and the labeling results identify whether the word is a component part of a possible entity name; according to the labeling result, words labeled as name components can be accurately screened from the text to be identified, then a plurality of candidate entity names are determined according to the screened words, a name candidate set is obtained, and accuracy of obtaining the candidate entity names is improved.
In one embodiment, the name recognition module 1110 specifically includes: vector input module, result acquisition module and name screening module, wherein:
the vector input module is used for inputting the word vector corresponding to each candidate entity name into the second bidirectional long and short memory neural network;
the result acquisition module is used for acquiring the identification result corresponding to each candidate entity name output by the second bidirectional long and short memory neural network;
and the name screening module is used for screening the identified entity names from the name candidate set according to the identification result.
In this embodiment, a word vector corresponding to each candidate entity name is input into the second bidirectional long and short memory neural network, the candidate entity names are further identified, the second bidirectional long and short memory neural network is obtained, an identification result corresponding to each candidate entity name is output, whether the candidate entity name is a required entity name is identified by the identification result, the identified entity names are screened from the name candidate set according to the identification result, and accuracy of identifying the entity names is improved through second identification.
For specific limitations of the entity name recognition apparatus, reference may be made to the above limitation of the entity name recognition method, and no further description is given here. The above-described respective modules in the entity name recognition apparatus may be implemented in whole or in part by software, hardware, and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.
In one embodiment, a computer device is provided, which may be a terminal, and the internal structure thereof may be as shown in fig. 12. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method of entity name identification. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
It will be appreciated by those skilled in the art that the structure shown in fig. 12 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer device is provided comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of when executing the computer program: acquiring word vectors corresponding to words in the text to be recognized; inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network; screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set; extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively; and identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
In one embodiment, obtaining word vectors corresponding to words in the text to be recognized includes: acquiring a text to be identified; inputting a text to be recognized into a vector conversion model, and carrying out word division on the text to be recognized to obtain a plurality of words; and carrying out vector conversion on each word to obtain a word vector corresponding to each word.
In one embodiment, inputting the obtained word vector into the first two-way long and short memory neural network, and obtaining the vector features output by the first two-way long and short memory neural network includes: respectively inputting the obtained word vectors to a forward reference layer and a backward reference layer in the first two-way long and short memory neural network; acquiring forward reference features and backward reference features output by a forward reference layer and a backward reference layer; vector features are generated from the forward reference features and the backward reference features.
In one embodiment, screening candidate entity names from the text to be identified according to vector features by a noun screening model to obtain a name candidate set comprises: inputting the vector features into a noun screening model to obtain labeling results of each word; according to the labeling result, screening words labeled as name components from the text to be identified; and determining a plurality of candidate entity names according to the screened words to obtain a name candidate set.
In one embodiment, identifying, via the second bidirectional long and short memory neural network, the entity names in the candidate set of names from the extracted word vector comprises: inputting word vectors corresponding to the names of each candidate entity into a second bidirectional long and short memory neural network; acquiring a second bidirectional long and short memory neural network, and outputting an identification result corresponding to each candidate entity name; and screening the identified entity names from the name candidate set according to the identification result.
In this embodiment, word vectors corresponding to each word in a text to be recognized are obtained, the obtained word vectors are input into a first two-way long and short memory neural network, vector features of the word vectors are obtained through the first two-way long and short memory neural network, the words are labeled according to the vector features through a noun screening model, candidate entity names are screened from the text to be recognized, and a name candidate set is obtained, namely, a set of all words which are possibly entity names in the text to be recognized; and extracting word vectors corresponding to the candidate recognition name candidate entity names in the name candidate set, inputting the extracted word vectors into a second bidirectional long and short memory neural network for recognition again, and recognizing the entity names in the name candidate set according to the extracted word vectors, so that the accuracy of entity name recognition is improved through the processing of the two-way long and short memory neural network.
In one embodiment, a computer readable storage medium is provided having a computer program stored thereon, which when executed by a processor, performs the steps of: acquiring word vectors corresponding to words in the text to be recognized; inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network; screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set; extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively; and identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network.
In one embodiment, obtaining word vectors corresponding to words in the text to be recognized includes: acquiring a text to be identified; inputting a text to be recognized into a vector conversion model, and carrying out word division on the text to be recognized to obtain a plurality of words; and carrying out vector conversion on each word to obtain a word vector corresponding to each word.
In one embodiment, inputting the obtained word vector into the first two-way long and short memory neural network, and obtaining the vector features output by the first two-way long and short memory neural network includes: respectively inputting the obtained word vectors to a forward reference layer and a backward reference layer in the first two-way long and short memory neural network; acquiring forward reference features and backward reference features output by a forward reference layer and a backward reference layer; vector features are generated from the forward reference features and the backward reference features.
In one embodiment, screening candidate entity names from the text to be identified according to vector features by a noun screening model to obtain a name candidate set comprises: inputting the vector features into a noun screening model to obtain labeling results of each word; according to the labeling result, screening words labeled as name components from the text to be identified; and determining a plurality of candidate entity names according to the screened words to obtain a name candidate set.
In one embodiment, identifying, via the second bidirectional long and short memory neural network, the entity names in the candidate set of names from the extracted word vector comprises: inputting word vectors corresponding to the names of each candidate entity into a second bidirectional long and short memory neural network; acquiring a second bidirectional long and short memory neural network, and outputting an identification result corresponding to each candidate entity name; and screening the identified entity names from the name candidate set according to the identification result.
In this embodiment, word vectors corresponding to each word in a text to be recognized are obtained, the obtained word vectors are input into a first two-way long and short memory neural network, vector features of the word vectors are obtained through the first two-way long and short memory neural network, the words are labeled according to the vector features through a noun screening model, candidate entity names are screened from the text to be recognized, and a name candidate set is obtained, namely, a set of all words which are possibly entity names in the text to be recognized; and extracting word vectors corresponding to the candidate recognition name candidate entity names in the name candidate set, inputting the extracted word vectors into a second bidirectional long and short memory neural network for recognition again, and recognizing the entity names in the name candidate set according to the extracted word vectors, so that the accuracy of entity name recognition is improved through the processing of the two-way long and short memory neural network.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the various embodiments provided herein may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double Data Rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The above examples merely represent a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (10)
1. A method of entity name identification, the method comprising:
acquiring word vectors corresponding to words in the text to be recognized;
inputting the obtained word vector into a first two-way long and short memory neural network to obtain vector characteristics output by the first two-way long and short memory neural network; the first two-way long and short memory neural network is a two-way long and short memory neural network model for extracting vector features of the word vectors;
Screening candidate entity names from the text to be identified according to the vector features by a noun screening model to obtain a name candidate set;
extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively;
identifying entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network; the second bidirectional long and short memory neural network is a model for screening entity names from the candidate entity names.
2. The method of claim 1, wherein the obtaining word vectors corresponding to words in the text to be recognized includes:
acquiring a text to be identified;
inputting the text to be recognized into a vector conversion model, and carrying out word division on the text to be recognized to obtain a plurality of words;
and carrying out vector conversion on each word to obtain a word vector corresponding to each word.
3. The method of claim 1, wherein inputting the obtained word vector into a first two-way long and short memory neural network, and obtaining the vector features output by the first two-way long and short memory neural network comprises:
respectively inputting the obtained word vectors to a forward reference layer and a backward reference layer in the first two-way long and short memory neural network;
Acquiring forward reference features and backward reference features output by the forward reference layer and the backward reference layer;
vector features are generated from the forward reference features and the backward reference features.
4. The method of claim 1, wherein the screening candidate entity names from the text to be identified according to the vector features by the noun screening model to obtain a candidate set of names comprises:
inputting the vector features into a noun screening model to obtain labeling results of each word;
according to the labeling result, screening words labeled as name components from the text to be identified;
and determining a plurality of candidate entity names according to the screened words to obtain a name candidate set.
5. The method of claim 1, wherein the identifying, via the second bidirectional long and short memory neural network, the entity name in the candidate set of names from the extracted word vector comprises:
inputting word vectors corresponding to the names of each candidate entity into a second bidirectional long and short memory neural network;
acquiring the second bidirectional long and short memory neural network, and outputting an identification result corresponding to each candidate entity name;
And screening the identified entity names from the name candidate set according to the identification result.
6. An entity name recognition apparatus, the apparatus comprising:
the vector acquisition module is used for acquiring word vectors corresponding to each word in the text to be identified;
the feature obtaining module is used for inputting the obtained word vector into a first two-way long and short memory neural network to obtain the vector feature output by the first two-way long and short memory neural network; the first two-way long and short memory neural network is a two-way long and short memory neural network model for extracting vector features of the word vectors;
the collection obtaining module is used for screening candidate entity names from the text to be identified according to the vector characteristics through a noun screening model to obtain a name candidate collection;
the vector extraction module is used for extracting word vectors corresponding to the names of the candidate entities in the name candidate set respectively;
the name recognition module is used for recognizing entity names in the name candidate set according to the extracted word vector through a second bidirectional long and short memory neural network; the second bidirectional long and short memory neural network is a model for screening entity names from the candidate entity names.
7. The apparatus of claim 6, wherein the vector acquisition module comprises:
the text acquisition module is used for acquiring a text to be identified;
the word obtaining module is used for inputting the text to be recognized into a vector conversion model, and dividing the text to be recognized into words to obtain a plurality of words;
the vector obtaining module is used for carrying out vector conversion on each word to obtain a word vector corresponding to each word.
8. The apparatus of claim 6, wherein the name recognition module comprises:
the vector input module is used for inputting the word vector corresponding to each candidate entity name into the second bidirectional long and short memory neural network;
the result acquisition module is used for acquiring the identification result corresponding to each candidate entity name output by the second bidirectional long and short memory neural network;
and the name screening module is used for screening the identified entity names from the name candidate set according to the identification result.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1 to 5 when the computer program is executed by the processor.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811592664.6A CN109753653B (en) | 2018-12-25 | 2018-12-25 | Entity name recognition method, entity name recognition device, computer equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811592664.6A CN109753653B (en) | 2018-12-25 | 2018-12-25 | Entity name recognition method, entity name recognition device, computer equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109753653A CN109753653A (en) | 2019-05-14 |
| CN109753653B true CN109753653B (en) | 2023-07-11 |
Family
ID=66403904
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811592664.6A Active CN109753653B (en) | 2018-12-25 | 2018-12-25 | Entity name recognition method, entity name recognition device, computer equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109753653B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110532350B (en) * | 2019-06-14 | 2024-05-03 | 平安科技（深圳）有限公司 | Text information online identification processing method and device based on jurisdiction |
| CN110688853B (en) * | 2019-08-12 | 2022-09-30 | 平安科技（深圳）有限公司 | Sequence labeling method and device, computer equipment and storage medium |
| CN110457706B (en) * | 2019-08-15 | 2023-08-22 | 腾讯科技（深圳）有限公司 | Point-of-interest name selection model training method, using method, device and storage medium |
| CN110909548B (en) * | 2019-10-10 | 2024-03-12 | 平安科技（深圳）有限公司 | Chinese named entity recognition method, device and computer-readable storage medium |
| CN111104791B (en) * | 2019-11-14 | 2024-02-20 | 北京金堤科技有限公司 | Industry information acquisition method and device, electronic equipment and medium |
| CN111145914B (en) * | 2019-12-30 | 2023-08-04 | 四川大学华西医院 | Method and device for determining text entity of lung cancer clinical disease seed bank |
| CN112668336B (en) * | 2020-05-14 | 2024-01-09 | 北京金山数字娱乐科技有限公司 | A word processing method based on task model |
| CN112329469B (en) * | 2020-11-05 | 2023-12-19 | 新华智云科技有限公司 | Administrative region entity identification method and system |
| CN113342964B (en) * | 2021-06-03 | 2022-04-19 | 云南大学 | Recommendation type determination method and system based on mobile service |
| CN116720520B (en) * | 2023-08-07 | 2023-11-03 | 烟台云朵软件有限公司 | Text data-oriented alias entity rapid identification method and system |
| CN120071214B (en) * | 2025-01-21 | 2026-02-06 | 深圳市塔普智能科技有限公司 | Video object segmentation method, device, equipment and storage medium |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11069335B2 (en) * | 2016-10-04 | 2021-07-20 | Cerence Operating Company | Speech synthesis using one or more recurrent neural networks |
| AU2017272141A1 (en) * | 2016-12-19 | 2018-07-05 | Accenture Global Solutions Limited | Duplicate and similar bug report detection and retrieval using neural networks |
| CN106682220A (en) * | 2017-01-04 | 2017-05-17 | 华南理工大学 | Online traditional Chinese medicine text named entity identifying method based on deep learning |
| JP6820764B2 (en) * | 2017-02-28 | 2021-01-27 | 日本放送協会 | Acoustic model learning device and acoustic model learning program |
| CN107644014A (en) * | 2017-09-25 | 2018-01-30 | 南京安链数据科技有限公司 | A kind of name entity recognition method based on two-way LSTM and CRF |
| CN107992542A (en) * | 2017-11-27 | 2018-05-04 | 中山大学 | A kind of similar article based on topic model recommends method |
| CN108536679B (en) * | 2018-04-13 | 2022-05-20 | 腾讯科技(成都)有限公司 | Named entity recognition method, device, equipment and computer readable storage medium |
| CN108920445B (en) * | 2018-04-23 | 2022-06-17 | 华中科技大学鄂州工业技术研究院 | Named entity identification method and device based on Bi-LSTM-CRF model |
| CN108664595B (en) * | 2018-05-08 | 2020-10-16 | 和美（深圳）信息技术股份有限公司 | Domain knowledge base construction method and device, computer equipment and storage medium |
-
2018
- 2018-12-25 CN CN201811592664.6A patent/CN109753653B/en active Active
Non-Patent Citations (1)
| Title |
|---|
| Bi-directional LSTM Recurrent Neural Network for Chinese Word Segmentation;Yushi Yao et al.;《ICONIP 2016》;第345-353页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109753653A (en) | 2019-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109753653B (en) | Entity name recognition method, entity name recognition device, computer equipment and storage medium | |
| CN111444723B (en) | Information extraction method, computer device, and storage medium | |
| CN110444198B (en) | Retrieval method, retrieval device, computer equipment and storage medium | |
| CN111666401B (en) | Document recommendation method, device, computer equipment and medium based on graph structure | |
| CN110334179B (en) | Question-answer processing method, device, computer equipment and storage medium | |
| CN111914825B (en) | Text recognition methods, devices and electronic equipment | |
| CN111461301B (en) | Serialized data processing method and device, text processing method and device | |
| CN111459977B (en) | Transformation of natural language queries | |
| CN109165384A (en) | A kind of name entity recognition method and device | |
| CN109858010A (en) | Field new word identification method, device, computer equipment and storage medium | |
| CN111783471A (en) | Semantic recognition method, device, equipment and storage medium of natural language | |
| JP6693582B2 (en) | Document abstract generation method, device, electronic device, and computer-readable storage medium | |
| US20210390370A1 (en) | Data processing method and apparatus, storage medium and electronic device | |
| CN110427612B (en) | Entity disambiguation method, device, equipment and storage medium based on multiple languages | |
| CN111737548A (en) | Click verification code identification method and device, computer equipment and storage medium | |
| CN112559725A (en) | Text matching method, device, terminal and storage medium | |
| CN109063155B (en) | Language model parameter determination method and device and computer equipment | |
| CN114462397B (en) | Language identification model training method, language identification method, device and electronic equipment | |
| CN115883919A (en) | Video processing method, video processing device, electronic equipment and storage medium | |
| CN118133845B (en) | Fusion method, device, equipment and storage medium for multi-channel semantic understanding | |
| CN115661846A (en) | Data processing method, device, electronic device and storage medium | |
| US12608544B2 (en) | System and method for automatic language detection for handwritten text | |
| CN114036260B (en) | Sensitive word determination method, device, equipment, storage medium and program product | |
| CN112541363B (en) | Method, device and server for identifying text data in target language | |
| CN116403225A (en) | Internet automatic recognition method and system for a special font |
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 | ||
| TG01 | Patent term adjustment | ||
| TG01 | Patent term adjustment |