CN111291187A - Emotion analysis method and device, electronic equipment and storage medium - Google Patents

Emotion analysis method and device, electronic equipment and storage medium Download PDF

Info

Publication number
CN111291187A
CN111291187A CN202010074496.2A CN202010074496A CN111291187A CN 111291187 A CN111291187 A CN 111291187A CN 202010074496 A CN202010074496 A CN 202010074496A CN 111291187 A CN111291187 A CN 111291187A
Authority
CN
China
Prior art keywords
analyzed
word
words
vector
target dimension
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
Application number
CN202010074496.2A
Other languages
Chinese (zh)
Other versions
CN111291187B (en
Inventor
任鑫涛
郭豪
蔡准
孙悦
郭晓鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Trusfort Technology Co ltd
Original Assignee
Beijing Trusfort Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Trusfort Technology Co ltd filed Critical Beijing Trusfort Technology Co ltd
Priority to CN202010074496.2A priority Critical patent/CN111291187B/en
Publication of CN111291187A publication Critical patent/CN111291187A/en
Application granted granted Critical
Publication of CN111291187B publication Critical patent/CN111291187B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/35—Clustering; Classification
    • 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
    • G06F16/33—Querying
    • G06F16/3331—Query processing
    • G06F16/334—Query execution
    • G06F16/3344—Query execution using natural language analysis
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Machine Translation (AREA)

Abstract

The application provides an emotion analysis method, an emotion analysis device, electronic equipment and a storage medium, wherein the method comprises the following steps: acquiring a text to be analyzed; dividing a text to be analyzed into a plurality of words to be analyzed including target dimension words; obtaining semantic feature vectors of the words to be analyzed based on the words to be analyzed and the trained semantic extraction model; determining attention weight of target dimension words to each word to be analyzed based on each semantic feature vector, and determining a first vector of a text to be analyzed based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension words to the word to be analyzed; and determining an emotion analysis result corresponding to the target dimension word based on the first vector of the text to be analyzed and the second vector of the target dimension word. According to the scheme, the attention weight of the target dimension words to each word to be analyzed is utilized to determine the associated information between the target dimension words and the text context, so that emotion analysis can be carried out more completely and accurately.

Description

Emotion analysis method and device, electronic equipment and storage medium
Technical Field
The application relates to the technical field of computer processing, in particular to an emotion analysis method, an emotion analysis device, electronic equipment and a storage medium.
Background
With the rapid development of the internet, the number of network users increases sharply, and people generate a great amount of valuable comment information such as characters, events, products and the like in the information interaction process. For example, in the fields of e-commerce, intelligent tourism, network car booking and the like, a user can evaluate multiple dimensions such as commodity quality, service and the like after consumption, each dimension contains abundant emotional information, and the user behavior can be better understood through mining the emotional information, so that the development direction of an event is predicted.
The main flow of the related emotion analysis is generally that a single user comment is analyzed to give the emotion polarity, and then the emotion polarities of all the user comments are aggregated to obtain a final analysis result. Many user reviews do not simply express one emotional polarity. For example, the user makes a comment after a meal, "the atmosphere of the restaurant is good, the dishes taste good, or the attitude of the waiter is somewhat poor. "i.e. the user evaluates from three dimensions" environment "," taste "and" service ", respectively, and if only the overall emotional polarity of the comment is given, a large information loss or analysis deviation will occur.
Disclosure of Invention
In view of the above, an object of the present application is to provide an emotion analysis method, apparatus, electronic device and storage medium, so as to improve the integrity and accuracy of emotion analysis.
Mainly comprises the following aspects:
in a first aspect, the present application provides a sentiment analysis method, including:
acquiring a text to be analyzed;
dividing the text to be analyzed into a plurality of words to be analyzed including target dimension words;
obtaining semantic feature vectors aiming at the target dimension words and semantic feature vectors of other words to be analyzed except the target dimension words in the plurality of words to be analyzed based on the words to be analyzed and the trained recurrent neural network semantic extraction model;
determining attention weight of the target dimension word to each word to be analyzed based on the semantic feature vector of the target dimension word and the semantic feature vectors of other words to be analyzed, and determining a first vector corresponding to the text to be analyzed based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension word to the word to be analyzed;
determining an emotion analysis result corresponding to the target dimension word based on the first vector of the text to be analyzed and the second vector of the target dimension word.
In one embodiment, the obtaining semantic feature vectors for the target dimension words and semantic feature vectors of other words to be analyzed in the plurality of words to be analyzed except the target dimension words based on the respective words to be analyzed and the trained semantic extraction model of the recurrent neural network includes:
inputting each word to be analyzed into a trained word vector conversion model to obtain a word vector corresponding to each word to be analyzed;
and inputting the word vector corresponding to each word to be analyzed into a trained recurrent neural network semantic extraction model to obtain a semantic feature vector for the target dimension word and semantic feature vectors of other words to be analyzed except the target dimension word in the plurality of words to be analyzed.
In one embodiment, the determining the attention weight of the target dimension word for each of the words to be analyzed based on the semantic feature vector of the target dimension word and the semantic feature vectors of the other words to be analyzed includes:
for each word to be analyzed, determining a first product based on the semantic feature vector of the word to be analyzed and the semantic feature vector of the target dimension word; determining a second product based on the semantic feature vector of the word to be analyzed and the second vector of the target dimension word;
determining a ratio between the first product and the product-sum value; the product sum value is determined by the semantic feature vector of each word to be analyzed and the semantic feature vector of the target dimension word;
determining an attention weight of the target dimension word for each of the words to be analyzed based on the determined ratio and the second product.
In one embodiment, the determining a first vector corresponding to the text to be analyzed based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension word to the word to be analyzed includes:
for each word to be analyzed, performing product operation on the semantic feature vector of the word to be analyzed and the attention weight of the target dimension word to the word to be analyzed to obtain a third product;
and performing summation operation on the third product corresponding to each word to be analyzed to obtain a first vector corresponding to the text to be analyzed.
In one embodiment, the determining, based on the first vector of the text to be analyzed and the second vector of the target dimension word, an emotion analysis result corresponding to the target dimension word includes:
assigning a first weight and a second weight to the first vector and the second vector, respectively;
performing a product operation on the first vector and the first weight to obtain a fourth product, and performing a product operation on the second vector and the second weight to obtain a fifth product;
and performing summation operation on the fourth product and the fifth product to obtain an emotion analysis result corresponding to the target dimension word.
In one embodiment, the method further comprises: training the semantic extraction model of the recurrent neural network and the attention weight;
the semantic extraction model of the recurrent neural network and the attention weight are obtained by training based on the obtained analysis text samples and the text labeling information corresponding to the dimension words in each analysis text sample.
In a second aspect, the present application further provides an emotion analyzing apparatus, including:
the acquisition module is used for acquiring a text to be analyzed;
the dividing module is used for dividing the text to be analyzed into a plurality of words to be analyzed including target dimension words;
the generating module is used for obtaining semantic feature vectors aiming at the target dimension words and semantic feature vectors of other words to be analyzed except the target dimension words in the plurality of words to be analyzed based on the words to be analyzed and the trained recurrent neural network semantic extraction model;
the determining module is used for determining the attention weight of the target dimension word to each word to be analyzed based on the semantic feature vector of the target dimension word and the semantic feature vectors of other words to be analyzed, and determining a first vector corresponding to the text to be analyzed based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension word to the word to be analyzed;
and the analysis module is used for determining an emotion analysis result corresponding to the target dimension word based on the first vector of the text to be analyzed and the second vector of the target dimension word.
In one embodiment, the apparatus further comprises:
the training module is used for training the semantic extraction model of the recurrent neural network and the attention weight;
the semantic extraction model of the recurrent neural network and the attention weight are obtained by training based on the obtained analysis text samples and the text labeling information corresponding to the dimension words in each analysis text sample.
In a third aspect, the present application further provides an electronic device, including: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, the processor and the memory communicate via the bus when the electronic device is running, and the processor executes the machine-readable instructions to implement the steps of the emotion analysis method according to the first aspect and any of its various embodiments.
In a fourth aspect, the present application further provides a computer readable storage medium having a computer program stored thereon, which when executed by a processor, implements the steps of the sentiment analysis method according to the first aspect and any of its various embodiments.
By adopting the scheme, firstly, a text to be analyzed can be divided into a plurality of words to be analyzed including target dimension words, semantic feature vectors aiming at the target dimension words and semantic feature vectors of other words to be analyzed except the target dimension words in the plurality of words to be analyzed can be obtained based on each divided word to be analyzed and a trained semantic extraction model of the recurrent neural network, then, attention weights of the target dimension words to the words to be analyzed can be determined based on the semantic feature vectors of the target dimension words and the semantic feature vectors of the other words to be analyzed, a first vector corresponding to the text to be analyzed can be determined based on the semantic feature vectors of each word to be analyzed and the attention weights of the target word dimension words to the words to be analyzed, and finally, a first vector of the text to be analyzed and a second vector of the target dimension words can be determined based on the first vector of the text to be analyzed and the second vector of the target dimension words, and determining an emotion analysis result corresponding to the target dimension word. That is, when the emotion analysis is performed on the text to be analyzed, the attention weight of the target dimension word to each word to be analyzed is used for determining the associated information between the target dimension word and the text context, so that the emotion analysis can be performed more completely and accurately.
In order to make the aforementioned objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.
FIG. 1 is a flow chart of a sentiment analysis method provided in an embodiment of the present application;
FIG. 2 is a flow chart of another emotion analysis method provided in the first embodiment of the present application;
FIG. 3 is a flow chart of another emotion analysis method provided in the first embodiment of the present application;
FIG. 4 is a flow chart of another emotion analysis method provided in the first embodiment of the present application;
FIG. 5 is a schematic diagram illustrating an application of a sentiment analysis method provided in an embodiment of the present application;
FIG. 6 is a schematic structural diagram illustrating an emotion analyzing apparatus provided in the second embodiment of the present application;
fig. 7 shows a schematic structural diagram of an electronic device according to a third embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present application without making any creative effort, shall fall within the protection scope of the present application.
Considering emotion analysis in the related technology, a single user comment is analyzed to give emotion polarity, and then emotion polarities of all user comments are aggregated to obtain a final analysis result. However, many user comments do not simply express one emotion polarity, and if only the overall emotion polarity of the comment is given, a large information loss or analysis deviation is generated. Based on the emotion analysis method, at least one emotion analysis scheme is provided to improve the completeness and accuracy of emotion analysis.
To facilitate understanding of the present embodiment, first, an emotion analysis method applied in the embodiments of the present application is described in detail, where an execution subject of the emotion analysis method provided in the embodiments of the present application is generally an electronic device with certain computing capability, and the electronic device includes, for example: a terminal device, which may be a User Equipment (UE), a mobile device, a User terminal, a cellular phone, a cordless phone, a Personal Digital Assistant (PDA), a handheld device, a computing device, a vehicle mounted device, a wearable device, or a server or other processing device. In some possible implementations, the sentiment analysis method may be implemented by a processor calling computer-readable instructions stored in a memory.
The emotion analysis method provided in the embodiments of the present application is described below by taking an execution subject as a server.
Example one
Referring to fig. 1, a flowchart of an emotion analysis method provided in an embodiment of the present application is shown, where the method specifically includes the following steps:
s101, obtaining a text to be analyzed;
s102, dividing a text to be analyzed into a plurality of words to be analyzed including target dimension words;
s103, obtaining a semantic feature vector for the target dimension word and semantic feature vectors of other words to be analyzed except the target dimension word in the plurality of words to be analyzed based on each word to be analyzed and the trained semantic extraction model;
s104, determining the attention weight of the target dimensional words to each word to be analyzed based on the semantic feature vectors of the target dimensional words and the semantic feature vectors of other words to be analyzed, and determining a first vector corresponding to the text to be analyzed based on the semantic feature vectors of each word to be analyzed and the attention weight of the target dimensional words to the word to be analyzed;
and S105, determining an emotion analysis result corresponding to the target dimensional word based on the first vector of the text to be analyzed and the second vector of the target dimensional word.
Here, after the semantic extraction model is obtained through training, the semantic feature vector may be extracted based on the semantic extraction model, and the first vector corresponding to the text to be analyzed may be determined based on the extracted semantic feature vector, so that emotion analysis may be performed based on the first vector.
Before extracting the semantic feature vector based on the semantic extraction model, the embodiment of the application can firstly perform word segmentation processing on the text to be analyzed to obtain a plurality of words to be analyzed. In order to facilitate the completeness of emotion analysis, in the embodiment of the present application, before extracting a semantic feature vector based on a semantic extraction model, a target dimension word in a word to be analyzed needs to be identified, where the target dimension word is used to represent a dimension word that can have an emotion analysis intention in a text to be analyzed, for example, service quality, logistics speed, and the like.
One or more target dimension words can be provided. For example, for the text to be analyzed, which is "the piece of clothing has very good quality but very slow logistics", both quality and logistics can be used as target dimension words, that is, the word to be analyzed, which is intended to pay attention to the emotion analysis result, can be selected as the target dimension word.
After determining a plurality of words to be analyzed including the target dimension words, the semantic feature vectors of the words to be analyzed including the target dimension words can be extracted and obtained based on the trained semantic extraction model.
When extracting the semantic feature vectors, the embodiment of the application can firstly input each word to be analyzed into the trained word vector conversion model, and then extract the semantic feature vectors based on the trained semantic extraction model.
After obtaining each word to be analyzed, the word to be analyzed, which is a natural language, may be converted into digital information in a vector form based on a mathematical method word2vec, so as to facilitate machine identification, and this process is called encoding (Encoder). That is, a word vector obtained by converting the word vector conversion model is used to represent a word, and then the word vector is used as an input feature of the semantic extraction model.
The word vector conversion model that can be adopted in the embodiment of the present application includes two types, One is a word vector conversion model based on One-hot Representation (One-hot Representation), and the other is a word vector conversion model based on distributed Representation (DistributedRepresentation).
The former word vector conversion model uses a very long vector to represent a word, the length of the vector is the word quantity N of the dictionary, each vector only has one dimension of 1, the rest dimensions are all 0, and the position of 1 represents the position of the word in the dictionary. That is, the former word vector conversion model stores word information in a sparse manner, that is, each word is assigned with a digital identifier, and the representation form is relatively simple. The latter word vector conversion model needs to perform semantic representation according to context information, that is, words appearing in the same context have similar semantics. That is, the latter word vector conversion model stores word information in a dense manner, and the representation form is relatively complex. Considering that the former word vector conversion model based on One-hot Representation often encounters dimension disaster when solving the practical problem and cannot reveal the potential relation between vocabularies, the latter word vector conversion model based on Distributed Representation can be adopted to carry out vector Representation on the label information in the practical implementation, thereby not only avoiding the problem of dimension disaster, but also mining the correlation attributes between vocabularies, and improving the accuracy of semantic expression.
In the embodiment of the application, after the word vectors are extracted based on the word vector conversion model, the extraction of the semantic feature vectors can be performed based on the trained semantic extraction model. Considering that the extraction process of the semantic feature vector is taken as a key step of the emotion analysis method provided by the embodiment of the application, the training process of the semantic extraction model for semantic extraction can be simply described next.
After the semantic extraction model is trained, each analysis text sample and text labeling information corresponding to the dimension words in each analysis text sample need to be acquired, in the embodiment of the application, semantic labeling can be performed based on the emotional states of the dimension words, for example, when it is determined that the emotional states can be divided into positive emotions and negative emotions, the positive emotion corresponding is labeled as 1, and the negative emotion corresponding is labeled as 0; as another example, where it is determined that emotional states can be classified as positive, negative, and neutral, the corresponding labels can be 1, 0, -1. The semantic annotation is only an example, and in a specific application, the semantic annotation can be performed based on the rough classification, and the semantic annotation can be performed after the rough classification emotion is further refined, which is not limited specifically herein.
After semantic annotation, a semantic extraction model to be trained and an attention weight based on the obtained analysis text samples and text annotation information corresponding to the dimension words in each analysis text sample can be trained, and training of the relevant semantic extraction model is a process of training parameters of the semantic extraction model.
In a particular application, the semantic extraction model maps an input vector to an output vector. The embodiment of the application can adopt a special type of Recurrent Neural Networks (RNN) -Long Short Term Memory (LSTM) network to perform model training, wherein the LSTM comprises a 3-gate structure for controlling transmission and change of information, and the two methods are as follows: an input gate, an output gate and a forgetting gate. The input gate is used for controlling the proportion occupied by the input signal, the output gate is used for controlling the proportion occupied by the output signal, and the forgetting gate is used for controlling the proportion forgotten information. The three work cooperatively to control the internal operation mode of the LSTM. The model is a model which is very suitable for processing the characteristics of sequence information and receives a signal input at each moment, simultaneously outputs a signal and changes the internal parameter state of the signal. Therefore, the LSTM network is adopted in the embodiment of the application, various basic knowledge is gradually mastered through repeated iterative learning, and finally how to generate a voice feature vector meeting the requirements according to the word vector is learned.
In the emotion analysis method provided by the embodiment of the application, in the process of model training, in order to measure whether the result output by the model is matched with the pre-labeled information, various loss function representation modes can be adopted. In the embodiment of the application, the cross entropy can be used as a loss function to measure the matching degree of information, and the problem that training cannot be continued due to small gradient existing in small error can be solved mainly by considering the cross entropy loss, so that the training robustness is good.
Considering the influence of the target dimension word on the emotion analysis result of the text to be analyzed, the attention weight of the target dimension word on each word to be analyzed may be determined based on the semantic feature vector of each word to be analyzed, so as to determine the emotion analysis result corresponding to the target dimension word according to the attention weight, the emotion analysis result in the embodiment of the present application may adopt one of two emotions, namely, a positive emotion and a negative emotion, may also adopt one of three emotions, namely, a positive emotion, a negative emotion, a neutral emotion, and the like, and may also adopt one of a plurality of emotions, specifically which classification manner is adopted, and may be determined by the classification type in the training process, and no specific limitation is made herein.
Still taking the text to be analyzed, namely the piece of clothes is very good in quality but very slow in logistics, if a two-classification mode is adopted, the emotion analysis result is positive for the target dimension word of quality, and the emotion analysis result is negative for the target dimension word of logistics.
In the embodiment of the application, the emotion analysis result corresponding to the target dimension word can be determined based on the splicing result of the first vector of the text to be analyzed and the second vector of the target dimension word. Wherein the second vector related to the target dimension term may be a vector representation related to the target dimension term. The first vector relating to the text to be analyzed may then be determined based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension word to that word to be analyzed.
In the embodiment of the application, the attention weight may be determined based on the semantic feature vector of the target dimension word and the semantic feature vectors of other words to be analyzed. Considering the key role of the determination of the attention weight on the first vector corresponding to the text to be analyzed, the above process of determining the attention weight can be explained with reference to fig. 2.
S201, aiming at each word to be analyzed, determining a first product based on the semantic feature vector of the word to be analyzed and the semantic feature vector of a target dimension word; determining a second product based on the semantic feature vector of the word to be analyzed and the second vector of the target dimension word;
s202, determining a ratio of the first product to a product sum; the product sum is determined by the semantic feature vector of each word to be analyzed and the semantic feature vector of the target dimension word;
s203, determining the attention weight of the target dimension word to each word to be analyzed based on the determined ratio and the second product.
Here, for each word to be analyzed, a first product may be first determined based on the semantic feature vector of the word to be analyzed and the semantic feature vector of the target dimension word, and a second product may be determined based on the semantic feature vector of the word to be analyzed and the second vector of the target dimension word, then a ratio between the first product and the product-sum value may be determined, and finally an attention weight of the target dimension word for each word to be analyzed may be determined based on the determined ratio and the second product. The product sum value is determined by the semantic feature vector of each word to be analyzed and the semantic feature vector of the target dimension word.
That is, the embodiment of the application can determine the influence of the target dimension word on different words to be analyzed based on the weight calculation strategy, and still take the text to be analyzed, which is "the clothing quality is very good, but the logistics is very slow", as for the word to be analyzed, which is "good", the influence of the target dimension word "quality" on the target dimension word "quality" will far exceed the influence of other words on the target dimension word, so that the emotion word most relevant to the target dimension word "quality" can be determined, and the accuracy of emotion analysis is further improved. Similarly, for the words to be analyzed in the slow mode, the influence of the target dimension word logistics is far beyond the influence of other words, so that the emotion words most relevant to the target dimension word logistics can be determined.
After determining the attention weight of the target dimension word for each analysis word, a first vector corresponding to the text to be analyzed may be determined. As shown in fig. 3, the process of determining the first vector specifically includes the following steps:
s301, for each word to be analyzed, performing product operation on the semantic feature vector of the word to be analyzed and the attention weight of the target dimension word to the word to be analyzed to obtain a third product;
and S302, performing summation operation on the third products corresponding to the words to be analyzed to obtain a first vector corresponding to the text to be analyzed.
Here, for each word to be analyzed, firstly, a semantic feature vector of the word to be analyzed and a target dimension word may be multiplied by the attention weight of the word to be analyzed, and a third product corresponding to each word to be analyzed is summed to obtain a first vector corresponding to the text to be analyzed, that is, in the embodiment of the present application, a vector representation of the current text to be analyzed may be obtained by using a weighted summation manner, so that a final semantic representation may be obtained by using a concatenation result between the vector representation and a second vector of the target dimension word, and an emotion analysis result corresponding to each target dimension word may be determined by using the semantic representation.
As shown in fig. 4, the process of determining the emotion analysis result by using the final semantic representation in the embodiment of the present application includes the following steps:
s401, respectively giving a first weight and a second weight to the first vector and the second vector;
s402, performing product operation on the first vector and the first weight to obtain a fourth product, and performing product operation on the second vector and the second weight to obtain a fifth product;
and S403, performing summation operation on the fourth product and the fifth product to obtain an emotion analysis result corresponding to the target dimension word.
Here, in the embodiment of the present application, a first weight may be first assigned to a first vector of a text to be analyzed, and a second weight may be assigned to a second vector of a target dimension word, so that a fourth product may be obtained by performing a product operation on the first vector and the first weight, a fifth product may be obtained by performing a product operation on the second vector and the second weight, and then a final vector representation may be determined to obtain an emotion analysis result corresponding to the target dimension word by performing a sum operation on the fourth product and the fifth product.
In order to determine the emotion analysis result corresponding to the target dimension word, the final vector representation may be input to the softmax function to obtain a final probability output, for example, for the second classification, the probability of positive emotion corresponding to the target dimension word, which is quality, is determined to be 98%, and the emotion analysis result can be determined to be positive emotion by setting a probability threshold.
In order to further understand the process of the emotion analysis method provided in the embodiments of the present application, a detailed description may be made with reference to fig. 5 and the following formula. Here again, the piece of clothing is of very good quality, but the logistics are very slow to explain as text to be analyzed.
(1) The first input text to be analyzed can be converted into a word sequence through word segmentation, as shown in fig. 5, the clothes are very good in quality, but the logistics are very slow, and the word sequence is mapped into a word vector sequence W ═ W-1,W2,W3,W4,W5,W6,W7,W8,W9,W10},W∈Rm×nM is the length of the word vector and n is the length of the input text, and finally these word vectors can be trained with the model as part of the model parameters. The method comprises the following steps of training, wherein two dimension words, namely quality and logistics, are provided, and the emotion polarity of one dimension word is trained in each training.
(2) Each input word to be analyzed is mapped into a semantic feature vector sequence H ═ H through an LSTM network1,H2,H3,H4,H5,H6,H7,H8,H9,H10},H∈Rm×nM is the length of the hidden vector and n is the length of the input text, and finally these vectors are trained with the model as part of the model parameters.
(3) Then H1,H2,H3,H4,H5,H6,H7,H8,H9,H10Attention weight α may be derived by an attention mechanism along with its corresponding target word1,α2,α3,α4,α5,α6,α7,α8,α9,α10The attention weight is calculated by the formula
Figure BDA0002378145300000141
Wherein, Wa∈Rm×nWord vectors representing dimensional words in sentences, H ∈ Rm×nRepresenting hidden layer vectors output via LSTM, m being the length of the word vector, n being the length of the input text, HaRepresents a dimension word WaHidden layer vector, H, output over LSTMiThe ith word W representing the input textiHidden layer vector output by LSTM, a ∈ R1×1,b∈R1×1,Q∈Rn×nAs a parameter, αiRepresents the attention weight of the ith word, [ H ]T*Q*Wa]iIs represented by (H)T*Q*Wa)∈Rm×nThe ith position component of the vector.
(4) After obtaining the attention weight of the hidden layer vector, weighting the hidden layer vector by the attention weight to obtain
Figure BDA0002378145300000142
Obtaining the vector representation form (namely a first vector) of the current text input, and finally adding the vector representation form and the word vector W of the dimension worda(namely the second vector) is spliced by a splicing module to form a complete semantic representation T of the word, and the splicing mode of the splicing module is that T is equal to M1×Htotal+M2×WaWherein M is1∈Rm×m,M2∈Rm×mAs a parameter, m is the length of the word vector.
(5) Obtaining the final vector representation T of the text to be analyzed in the step 4, and then inputting the final vector representation T into the softmax function
Figure BDA0002378145300000151
And obtaining the probability output of the final model. The model may use cross entropy as a loss function, with the formula:
Figure BDA0002378145300000152
wherein, yiFor characterizing the true identity result, TiFor characterizing the model input results.
In summary, the emotion analysis method provided in the embodiment of the present application can utilize dimension word information and dimension word and context semantic information at the same time, and provides a network architecture, so that a more comprehensive and complete word meaning can be utilized by a model in terms of word information representation, that is, an attention weight generation mode based on dimension information is provided, and public opinion monitoring depends on accurate and comprehensive understanding of the word meaning, so that the related model architecture adopted in the embodiment of the present application is helpful for improving the accuracy of public opinion monitoring based on dimension words.
Example two
Based on the same application concept, the second embodiment of the present application provides an emotion analysis device corresponding to the emotion analysis method, and because the principle of solving the problem of the device in the embodiment of the present application is similar to that of the emotion analysis method in the embodiment of the present application, the implementation of the device may refer to the implementation of the method, and repeated details are not repeated.
As shown in fig. 6, an emotion analysis apparatus provided in an embodiment of the present application includes:
an obtaining module 601, configured to obtain a text to be analyzed;
a dividing module 602, configured to divide a text to be analyzed into a plurality of terms to be analyzed, where the terms include target dimension terms;
a generating module 603, configured to obtain a semantic feature vector for the target dimensional word and semantic feature vectors of other words to be analyzed, except the target dimensional word, in the plurality of words to be analyzed, based on each word to be analyzed and the trained semantic extraction model;
a determining module 604, configured to determine, based on the semantic feature vector of the target dimensional word and the semantic feature vectors of other words to be analyzed, an attention weight of the target dimensional word to each word to be analyzed, and determine, based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimensional word to the word to be analyzed, a first vector corresponding to the text to be analyzed;
and the analysis module 605 is configured to determine an emotion analysis result corresponding to the target dimension word based on the first vector of the text to be analyzed and the second vector of the target dimension word.
In one embodiment, the determining module 604 is configured to determine attention weights of the target dimension words for the respective words to be analyzed according to the following steps:
for each word to be analyzed, determining a first product based on the semantic feature vector of the word to be analyzed and the semantic feature vector of the target dimension word; determining a second product based on the semantic feature vector of the word to be analyzed and the second vector of the target dimension word;
determining a ratio between the first product and the product-sum value; the product sum is determined by the semantic feature vector of each word to be analyzed and the semantic feature vector of the target dimension word;
based on the determined ratio and the second product, an attention weight of the target dimension word for each word to be analyzed is determined.
In one embodiment, the determining module 604 is configured to determine a first vector corresponding to a text to be analyzed according to the following steps:
for each word to be analyzed, performing product operation on the semantic feature vector of the word to be analyzed and the attention weight of the target dimension word to the word to be analyzed to obtain a third product;
and carrying out summation operation on the third products corresponding to the words to be analyzed to obtain a first vector corresponding to the text to be analyzed.
In one embodiment, the analysis module 605 is configured to determine an emotion analysis result corresponding to the target dimension word according to the following steps:
assigning a first weight and a second weight to the first vector and the second vector, respectively;
performing product operation on the first vector and the first weight to obtain a fourth product, and performing product operation on the second vector and the second weight to obtain a fifth product;
and performing summation operation on the fourth product and the fifth product to obtain an emotion analysis result corresponding to the target dimension word.
In one embodiment, the above apparatus further comprises:
a training module 606 for training the semantic extraction model and the attention weight;
the semantic extraction model and the attention weight are obtained by training based on the obtained analysis text samples and the text labeling information corresponding to the dimension words in each analysis text sample.
EXAMPLE III
As shown in fig. 7, a schematic structural diagram of an electronic device provided in an embodiment of the present application is shown, where the electronic device includes: a processor 701, a memory 702 and a bus 703, wherein the memory 702 stores executable instructions, and when the apparatus is operated, the processor 701 communicates with the memory 702 through the bus 703, and the processor 701 implements the steps of the emotion analysis method according to an embodiment when executing the machine-readable instructions stored in the memory 702.
An embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by the processor 701 to perform the steps of the emotion analysis method.
Specifically, the storage medium can be a general storage medium, such as a removable disk, a hard disk, and the like, and when a computer program on the storage medium is run, the emotion analysis method can be executed, so that the problem of large information loss or analysis deviation generated at present is solved, and the effect of improving the integrity and accuracy of emotion analysis is achieved.
The computer program product of the emotion analysis method provided in the embodiment of the present application includes a computer-readable storage medium storing a program code, and instructions included in the program code may be used to execute the method in the foregoing method embodiment, and specific implementation may refer to the method embodiment, which is not described herein again.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the system and the apparatus described above may refer to the corresponding processes in the foregoing method embodiments, and are 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 application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including 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 application. 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.
The above description is only for the specific embodiments of the present application, but the scope of the present application is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present application, and shall be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims (10)

1. A sentiment analysis method, characterized in that the method comprises:
acquiring a text to be analyzed;
dividing the text to be analyzed into a plurality of words to be analyzed including target dimension words;
obtaining semantic feature vectors aiming at the target dimension words and semantic feature vectors of other words to be analyzed except the target dimension words in the plurality of words to be analyzed based on the words to be analyzed and the trained semantic extraction model;
determining attention weight of the target dimension word to each word to be analyzed based on the semantic feature vector of the target dimension word and the semantic feature vectors of other words to be analyzed, and determining a first vector corresponding to the text to be analyzed based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension word to the word to be analyzed;
determining an emotion analysis result corresponding to the target dimension word based on the first vector of the text to be analyzed and the second vector of the target dimension word.
2. The method according to claim 1, wherein the obtaining semantic feature vectors for the target dimension words and semantic feature vectors of other words to be analyzed in the plurality of words to be analyzed except the target dimension words based on the respective words to be analyzed and the trained semantic extraction model comprises:
inputting each word to be analyzed into a trained word vector conversion model to obtain a word vector corresponding to each word to be analyzed;
and inputting the word vector corresponding to each word to be analyzed into a trained semantic extraction model to obtain a semantic feature vector for the target dimension word and semantic feature vectors of other words to be analyzed except the target dimension word in the plurality of words to be analyzed.
3. The method of claim 1, wherein the determining attention weights of the target dimension words for the respective words to be analyzed based on the semantic feature vectors of the target dimension words and the semantic feature vectors of the other words to be analyzed comprises:
for each word to be analyzed, determining a first product based on the semantic feature vector of the word to be analyzed and the semantic feature vector of the target dimension word; determining a second product based on the semantic feature vector of the word to be analyzed and the second vector of the target dimension word;
determining a ratio between the first product and the product-sum value; the product sum value is determined by the semantic feature vector of each word to be analyzed and the semantic feature vector of the target dimension word;
determining an attention weight of the target dimension word for each of the words to be analyzed based on the determined ratio and the second product.
4. The method of claim 3, wherein determining the first vector corresponding to the text to be analyzed based on the semantic feature vector of each term to be analyzed and the attention weight of the target dimension term to the term to be analyzed comprises:
for each word to be analyzed, performing product operation on the semantic feature vector of the word to be analyzed and the attention weight of the target dimension word to the word to be analyzed to obtain a third product;
and performing summation operation on the third product corresponding to each word to be analyzed to obtain a first vector corresponding to the text to be analyzed.
5. The method of claim 1, wherein determining emotion analysis results corresponding to the target dimension words based on the first vector of the text to be analyzed and the second vector of the target dimension words comprises:
assigning a first weight and a second weight to the first vector and the second vector, respectively;
performing a product operation on the first vector and the first weight to obtain a fourth product, and performing a product operation on the second vector and the second weight to obtain a fifth product;
and performing summation operation on the fourth product and the fifth product to obtain an emotion analysis result corresponding to the target dimension word.
6. The method according to any one of claims 1 to 5, further comprising: training the semantic extraction model and the attention weight;
the semantic extraction model and the attention weight are obtained by training based on the obtained analysis text samples and the text labeling information corresponding to the dimension words in each analysis text sample.
7. An emotion analysis apparatus, characterized in that the apparatus comprises:
the acquisition module is used for acquiring a text to be analyzed;
the dividing module is used for dividing the text to be analyzed into a plurality of words to be analyzed including target dimension words;
the generating module is used for obtaining semantic feature vectors aiming at the target dimension words and semantic feature vectors of other words to be analyzed except the target dimension words in the plurality of words to be analyzed based on the words to be analyzed and the trained semantic extraction model;
the determining module is used for determining the attention weight of the target dimension word to each word to be analyzed based on the semantic feature vector of the target dimension word and the semantic feature vectors of other words to be analyzed, and determining a first vector corresponding to the text to be analyzed based on the semantic feature vector of each word to be analyzed and the attention weight of the target dimension word to the word to be analyzed;
and the analysis module is used for determining an emotion analysis result corresponding to the target dimension word based on the first vector of the text to be analyzed and the second vector of the target dimension word.
8. The apparatus of claim 7, further comprising:
a training module for training the semantic extraction model and the attention weight;
the semantic extraction model and the attention weight are obtained by training based on the obtained analysis text samples and the text labeling information corresponding to the dimension words in each analysis text sample.
9. An electronic device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory communicating via the bus when the electronic device is operating, the processor implementing the steps of the sentiment analysis method of any one of claims 1-6 when executing the machine-readable instructions.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when being executed by a processor, carries out the steps of the sentiment analysis method according to any one of claims 1 to 6.
CN202010074496.2A 2020-01-22 2020-01-22 Emotion analysis method and device, electronic equipment and storage medium Active CN111291187B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010074496.2A CN111291187B (en) 2020-01-22 2020-01-22 Emotion analysis method and device, electronic equipment and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010074496.2A CN111291187B (en) 2020-01-22 2020-01-22 Emotion analysis method and device, electronic equipment and storage medium

Publications (2)

Publication Number Publication Date
CN111291187A true CN111291187A (en) 2020-06-16
CN111291187B CN111291187B (en) 2023-08-08

Family

ID=71030717

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010074496.2A Active CN111291187B (en) 2020-01-22 2020-01-22 Emotion analysis method and device, electronic equipment and storage medium

Country Status (1)

Country Link
CN (1) CN111291187B (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111930941A (en) * 2020-07-31 2020-11-13 腾讯音乐娱乐科技（深圳）有限公司 Method and device for identifying abuse content and server
CN111950258A (en) * 2020-08-10 2020-11-17 深圳市慧择时代科技有限公司 Emotion classification method and device
CN112307179A (en) * 2020-10-21 2021-02-02 深圳技术大学 Text matching method, device, equipment and storage medium
CN112905790A (en) * 2021-02-04 2021-06-04 中国建设银行股份有限公司 Method, device and system for extracting qualitative indexes of supervision events
CN113469197A (en) * 2021-06-29 2021-10-01 北京达佳互联信息技术有限公司 Image-text matching method, device, equipment and storage medium

Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108052505A (en) * 2017-12-26 2018-05-18 上海智臻智能网络科技股份有限公司 Text emotion analysis method and device, storage medium, terminal
CN108170681A (en) * 2018-01-15 2018-06-15 中南大学 Text emotion analysis method, system and computer readable storage medium
CN108416032A (en) * 2018-03-12 2018-08-17 腾讯科技（深圳）有限公司 A kind of file classification method, device and storage medium
CN108959268A (en) * 2018-07-20 2018-12-07 科大讯飞股份有限公司 A kind of text emotion analysis method and device
US20190121853A1 (en) * 2017-10-25 2019-04-25 International Business Machines Corporation Facilitating automatic detection of relationships between sentences in conversations
CN109710761A (en) * 2018-12-21 2019-05-03 中国标准化研究院 The sentiment analysis method of two-way LSTM model based on attention enhancing
CN109766557A (en) * 2019-01-18 2019-05-17 河北工业大学 A kind of sentiment analysis method, apparatus, storage medium and terminal device
CN109918636A (en) * 2017-12-13 2019-06-21 上海智臻智能网络科技股份有限公司 Kansei Information Processing method, computer equipment and computer readable storage medium
CN109960793A (en) * 2017-12-25 2019-07-02 上海智臻智能网络科技股份有限公司 Opinion mining device and intelligent terminal
CN109960794A (en) * 2017-12-25 2019-07-02 上海智臻智能网络科技股份有限公司 Judge the device and intelligent terminal of text emotion
CN110032736A (en) * 2019-03-22 2019-07-19 深兰科技（上海）有限公司 A kind of text analyzing method, apparatus and storage medium
CN110222178A (en) * 2019-05-24 2019-09-10 新华三大数据技术有限公司 Text sentiment classification method, device, electronic equipment and readable storage medium storing program for executing
CN110377740A (en) * 2019-07-22 2019-10-25 腾讯科技（深圳）有限公司 Feeling polarities analysis method, device, electronic equipment and storage medium
CN110427610A (en) * 2019-06-25 2019-11-08 平安科技（深圳）有限公司 Text analyzing method, apparatus, computer installation and computer storage medium
CN110704622A (en) * 2019-09-27 2020-01-17 北京明略软件***有限公司 Text emotion classification method and device and electronic equipment

Patent Citations (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190121853A1 (en) * 2017-10-25 2019-04-25 International Business Machines Corporation Facilitating automatic detection of relationships between sentences in conversations
CN109918636A (en) * 2017-12-13 2019-06-21 上海智臻智能网络科技股份有限公司 Kansei Information Processing method, computer equipment and computer readable storage medium
CN109960794A (en) * 2017-12-25 2019-07-02 上海智臻智能网络科技股份有限公司 Judge the device and intelligent terminal of text emotion
CN109960793A (en) * 2017-12-25 2019-07-02 上海智臻智能网络科技股份有限公司 Opinion mining device and intelligent terminal
CN108052505A (en) * 2017-12-26 2018-05-18 上海智臻智能网络科技股份有限公司 Text emotion analysis method and device, storage medium, terminal
CN108170681A (en) * 2018-01-15 2018-06-15 中南大学 Text emotion analysis method, system and computer readable storage medium
CN108416032A (en) * 2018-03-12 2018-08-17 腾讯科技（深圳）有限公司 A kind of file classification method, device and storage medium
CN108959268A (en) * 2018-07-20 2018-12-07 科大讯飞股份有限公司 A kind of text emotion analysis method and device
CN109710761A (en) * 2018-12-21 2019-05-03 中国标准化研究院 The sentiment analysis method of two-way LSTM model based on attention enhancing
CN109766557A (en) * 2019-01-18 2019-05-17 河北工业大学 A kind of sentiment analysis method, apparatus, storage medium and terminal device
CN110032736A (en) * 2019-03-22 2019-07-19 深兰科技（上海）有限公司 A kind of text analyzing method, apparatus and storage medium
CN110222178A (en) * 2019-05-24 2019-09-10 新华三大数据技术有限公司 Text sentiment classification method, device, electronic equipment and readable storage medium storing program for executing
CN110427610A (en) * 2019-06-25 2019-11-08 平安科技（深圳）有限公司 Text analyzing method, apparatus, computer installation and computer storage medium
CN110377740A (en) * 2019-07-22 2019-10-25 腾讯科技（深圳）有限公司 Feeling polarities analysis method, device, electronic equipment and storage medium
CN110704622A (en) * 2019-09-27 2020-01-17 北京明略软件***有限公司 Text emotion classification method and device and electronic equipment

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
李苑等: "基于注意力机制的评论情感分析及情感词检测", 《计算机科学》, no. 01 *
翟社平等: "基于注意力机制Bi-LSTM算法的双语文本情感分析", 《计算机应用与软件》, no. 12 *
胡朝举等: "基于词向量技术和混合神经网络的情感分析", 《计算机应用研究》, no. 12 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111930941A (en) * 2020-07-31 2020-11-13 腾讯音乐娱乐科技（深圳）有限公司 Method and device for identifying abuse content and server
CN111950258A (en) * 2020-08-10 2020-11-17 深圳市慧择时代科技有限公司 Emotion classification method and device
CN112307179A (en) * 2020-10-21 2021-02-02 深圳技术大学 Text matching method, device, equipment and storage medium
CN112905790A (en) * 2021-02-04 2021-06-04 中国建设银行股份有限公司 Method, device and system for extracting qualitative indexes of supervision events
CN113469197A (en) * 2021-06-29 2021-10-01 北京达佳互联信息技术有限公司 Image-text matching method, device, equipment and storage medium
CN113469197B (en) * 2021-06-29 2024-03-22 北京达佳互联信息技术有限公司 Image-text matching method, device, equipment and storage medium

Also Published As

Publication number Publication date
CN111291187B (en) 2023-08-08

Similar Documents

Publication Publication Date Title
CN111291187B (en) Emotion analysis method and device, electronic equipment and storage medium
CN110162749B (en) Information extraction method, information extraction device, computer equipment and computer readable storage medium
CN111553162B (en) Intention recognition method and related device
CN110009430B (en) Cheating user detection method, electronic device and computer readable storage medium
WO2018033030A1 (en) Natural language library generation method and device
CN111324696A (en) Entity extraction method, entity extraction model training method, device and equipment
CN111538809B (en) Voice service quality detection method, model training method and device
EP4113357A1 (en) Method and apparatus for recognizing entity, electronic device and storage medium
CN111274822A (en) Semantic matching method, device, equipment and storage medium
CN111859967A (en) Entity identification method and device and electronic equipment
CN110826327A (en) Emotion analysis method and device, computer readable medium and electronic equipment
CN114896386A (en) Film comment semantic emotion analysis method and system based on BilSTM
CN113011689B (en) Evaluation method and device for software development workload and computing equipment
CN116245097A (en) Method for training entity recognition model, entity recognition method and corresponding device
CN110532562B (en) Neural network training method, idiom misuse detection method and device and electronic equipment
CN114330704A (en) Statement generation model updating method and device, computer equipment and storage medium
CN113821588A (en) Text processing method and device, electronic equipment and storage medium
CN116680401A (en) Document processing method, document processing device, apparatus and storage medium
CN116628147A (en) Method for training text prediction model, text prediction method and device
CN116644148A (en) Keyword recognition method and device, electronic equipment and storage medium
CN115130455A (en) Article processing method and device, electronic equipment and storage medium
CN115129863A (en) Intention recognition method, device, equipment, storage medium and computer program product
CN112818084B (en) Information interaction method, related device, equipment and computer readable medium
CN114297380A (en) Data processing method, device, equipment and storage medium
CN113869058A (en) LC-GCN (LC-GCN-based) aspect level emotion analysis method and system, storage medium and electronic equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant