CN106815192B - Model training method and device, and sentence emotion recognition method and device - Google Patents
Model training method and device, and sentence emotion recognition method and device Download PDFInfo
- Publication number
- CN106815192B CN106815192B CN201510850126.2A CN201510850126A CN106815192B CN 106815192 B CN106815192 B CN 106815192B CN 201510850126 A CN201510850126 A CN 201510850126A CN 106815192 B CN106815192 B CN 106815192B
- Authority
- CN
- China
- Prior art keywords
- word
- sentence
- vector
- statement
- emotion
- 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.)
- Expired - Fee Related
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/20—Natural language analysis
- G06F40/279—Recognition of textual entities
- G06F40/289—Phrasal analysis, e.g. finite state techniques or chunking
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Machine Translation (AREA)
Abstract
本申请公开了一种模型训练方法及装置和语句情感识别方法及装置。其中，该模型训练方法包括：获取带有情感标记的文本信息，其中，文本信息包括按照预先定义的情感类型标记的多条语句，每条语句携带有与其对应的情感标记；确定文本信息中每条语句对应的语句向量，其中，语句向量为用于唯一表示对应的语句的多维数组；将每条语句对应的情感标记及其对应的语句向量输入到循环神经网络，训练得到神经网络模型，其中，神经网络模型用于识别语句中的情感类型。本申请解决了现有技术中对语句中所包含的情感识别准确性低的技术问题。
The present application discloses a model training method and device, and a sentence emotion recognition method and device. Wherein, the model training method includes: acquiring text information with emotional tags, wherein the text information includes a plurality of sentences marked according to a predefined emotional type, and each sentence carries a corresponding emotional tag; determining each sentence in the text information sentence vector corresponding to each sentence, where the sentence vector is a multi-dimensional array used to uniquely represent the corresponding sentence; input the emotion tag corresponding to each sentence and its corresponding sentence vector into the recurrent neural network, and train the neural network model, where , a neural network model is used to identify the type of emotion in a sentence. The present application solves the technical problem of low accuracy of emotion recognition contained in sentences in the prior art.
Description
Technical Field
The application relates to the field of text processing, in particular to a model training method and device and a sentence emotion recognition method and device.
Background
The user's emotion is usually contained in the sentence input by the user, and the emotion of the sentence is mainly recognized according to the emotion keyword or punctuation mark, emoticon and the like in the sentence in the prior art. However, when the analyzed sentence has no emotion-like words or emotion that is not well recognized by the machine, the recognition method has a low accuracy, such as: do you not sing? Whether the sentence is sad, angry or happy emotion needs to be understood in conjunction with the context, in "do you not sing? Not useful! "," do you not sing a song? Unfortunately "," do you not sing? Do not like you, i sing, we play. "respectively indicates angry, sadness, and happy emotion. However, the words of the sentence do not contain clear emotional tendency, so that the emotion of the sentence cannot be recognized in the prior art.
In view of the above problems, no effective solution has been proposed.
Disclosure of Invention
The embodiment of the application provides a model training method and device and a sentence emotion recognition method and device, so as to at least solve the technical problem that in the prior art, the emotion recognition accuracy included in a sentence is low.
According to an aspect of an embodiment of the present application, there is provided a model training method, including: acquiring text information with emotion marks, wherein the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence; determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement; and inputting the emotion mark corresponding to each statement and the statement vector corresponding to the emotion mark into a cyclic neural network, and training to obtain a neural network model, wherein the neural network model is used for identifying the emotion type in the statement.
Further, determining a sentence vector corresponding to each sentence in the text information includes: performing word segmentation processing on each sentence in the text information; searching a word vector corresponding to each word after each sentence is participated, wherein the word vector is a multidimensional array for uniquely representing the corresponding word; and determining the sentence vector of the corresponding sentence according to the word vector corresponding to each word in each sentence.
Further, before searching for a word vector corresponding to each word after each sentence is participated, the model training method further includes: acquiring text information with preset data volume to obtain a text information set; generating a word vector of each word in the text information set by using a machine learning mode to obtain a word vector set; searching for the word vector corresponding to each word after each sentence is participated comprises: and searching a word vector corresponding to each word after word segmentation from the word vector set.
Further, determining the sentence vector of the corresponding sentence according to the word vector corresponding to each word in each sentence includes: for each statement, adding word vectors corresponding to all words contained in the statement to obtain a word vector sum, wherein the word vector sum is a vector which is obtained by adding numerical values of corresponding positions in the word vectors and has the same number of dimensions as the word vectors; and dividing the word vector sum by the number of words to obtain the sentence vector, wherein the word vector sum divided by the number of words refers to a numerical value of each position in the word vector sum divided by the number of words.
According to another aspect of the embodiments of the present application, there is also provided a sentence emotion recognition method, including: determining a sentence vector corresponding to each sentence in the text to be tested; and inputting a statement vector corresponding to each statement in the text to be detected into a neural network model, and identifying the emotion type of each statement in the text to be detected by using the neural network model.
According to another aspect of the embodiments of the present application, there is also provided a model training apparatus, including: the system comprises an acquisition unit, a judgment unit and a display unit, wherein the acquisition unit is used for acquiring text information with emotion marks, the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence; the determining unit is used for determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement; and the training unit is used for inputting the emotion marks corresponding to each statement and the statement vectors corresponding to the emotion marks into a cyclic neural network and training to obtain a neural network model, wherein the neural network model is used for identifying the emotion types in the statements.
Further, the determining unit includes: the word segmentation module is used for carrying out word segmentation processing on each sentence in the text information; the query module is used for searching a word vector corresponding to each word after each sentence is subjected to word segmentation, wherein the word vector is a multidimensional array used for uniquely representing the corresponding word; and the determining module is used for determining the sentence vector of the corresponding sentence according to the word vector corresponding to each word in each sentence.
Further, the model training apparatus further includes: the acquisition unit is used for acquiring text information with preset data volume before searching for a word vector corresponding to each word after each sentence is participled to obtain a text information set; the generating unit is used for generating a word vector of each word in the text information set by utilizing a machine learning mode to obtain a word vector set; the query module is specifically configured to find a word vector corresponding to each word after the word segmentation from the word vector set.
Further, the determining module includes: the first calculation submodule is used for adding word vectors corresponding to all words contained in each statement to obtain a word vector sum, wherein the word vector sum is a vector which is obtained by adding numerical values of corresponding positions in the word vectors and has the same number of dimensions as the word vectors; and the second calculation submodule is used for dividing the word vector sum by the number of the words to obtain the statement vector, wherein the word vector sum divided by the number of the words refers to a numerical value of each position in the word vector sum divided by the number of the words.
According to another aspect of the embodiments of the present application, there is also provided a sentence emotion recognition apparatus, including: the vector determining unit is used for determining statement vectors corresponding to each statement in the text to be tested; and the identification unit is used for inputting the statement vector corresponding to each statement in the text to be detected into the neural network model, and identifying the emotion type of each statement in the text to be detected by using the neural network model.
According to the embodiment of the application, text information with emotion marks is obtained, wherein the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence; determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement; the emotion marks corresponding to each sentence and the sentence vectors corresponding to the emotion marks are input into a cyclic neural network, and a neural network model is obtained through training, so that the emotion types in the sentences can be conveniently recognized through the neural network model, the technical problem that the emotion recognition accuracy contained in the sentences is low in the prior art is solved, and the effect of improving the emotion recognition accuracy in the sentences is achieved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a flow diagram of a model training method according to an embodiment of the present application;
FIG. 2 is a flow chart of a sentence emotion recognition method according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a model training apparatus according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a sentence emotion recognition apparatus according to an embodiment of the present application.
Detailed Description
In order to make the technical solutions better understood by those skilled in the art, 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 partial embodiments of the present application, but not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It should be noted that the terms "first," "second," and the like in the description and claims of this application and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the application described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
In accordance with an embodiment of the present application, there is provided a method embodiment of a model training method, it should be noted that the steps illustrated in the flowchart of the figure may be performed in a computer system, such as a set of computer-executable instructions, and that while a logical order is illustrated in the flowchart, in some cases the steps illustrated or described may be performed in an order different than here.
Fig. 1 is a flowchart of a model training method according to an embodiment of the present application, as shown in fig. 1, the method includes the following steps:
step S102, acquiring text information with emotion marks, wherein the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence.
The text information with emotion marks in the embodiment of the application can be text information formed by manually marking the emotion of each sentence in a pre-collected text information sample. The emotion classification of the statement can be set as required, and the emotion classification can be a customized standard classification, such as if the statement emotion is divided into three types: positive, neutral, and negative, respectively, are labeled with 1,0, -1 for emotion correspondence, or are divided into five types: happiness, sadness, anger, fear, disgust, etc.
Step S104, determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement.
After the text information with the emotion marks is obtained, statement vectors corresponding to each statement in the text are determined, the statement vectors corresponding to each statement can be statement vectors generated according to a predefined rule, and the statement vectors corresponding to different statements are different. Because each sentence in the text information carries the corresponding emotion mark, the sentence vector corresponding to each sentence also corresponds to the same emotion mark as the sentence.
And S106, inputting the emotion marks corresponding to each statement and the statement vectors corresponding to the emotion marks into a cyclic neural network, and training to obtain a neural network model, wherein the neural network model is used for identifying the emotion types in the statements.
In this embodiment, after determining a sentence vector corresponding to each sentence of the text information, the sentence vectors carrying emotion marks are sequentially input into the recurrent neural network to be trained, so as to obtain the neural network model.
By inputting the statement vector corresponding to the statement in the text into the recurrent neural network, the machine can memorize the words and their combinations and their corresponding emotion marks in the statement, and memorize these words and their combinations and their corresponding emotion marks with the parameters in the neural network model (the parameters in the neural network model are determined, mostly as a matrix).
When the neural network model is used for emotion recognition of sentences, the emotion types of the sentences in the text can be effectively and quickly recognized according to the context of the sentences, and compared with the mode that the sentences are recognized through emotion keywords and the like in the prior art, the accuracy of the sentence emotion recognition through the neural network model is greatly improved.
According to the embodiment of the application, text information with emotion marks is obtained, wherein the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence; determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement; the emotion marks corresponding to each sentence and the sentence vectors corresponding to the emotion marks are input into a cyclic neural network, and a neural network model is obtained through training, so that the emotion types in the sentences can be conveniently recognized through the neural network model, the technical problem that the emotion recognition accuracy contained in the sentences is low in the prior art is solved, and the effect of improving the emotion recognition accuracy in the sentences is achieved.
Preferably, determining a sentence vector corresponding to each sentence in the text information includes: performing word segmentation processing on each sentence in the text information; searching a word vector corresponding to each word after each sentence is participated, wherein the word vector is a multidimensional array for uniquely representing the corresponding word; and determining the sentence vector of the corresponding sentence according to the word vector corresponding to each word in each sentence.
In this embodiment, for determining the sentence vectors, word segmentation processing may be performed on each sentence, and then the word vector corresponding to each word in each sentence is queried. Wherein the word segmentation result is to arrange words in order according to the original sentence, including punctuation marks such as "will you not sing? "do the participles" do you not sing a song? "or" do you not sing? ". The word vector of each term may be obtained by searching the word vector corresponding to each term from a pre-generated word vector set, or may be generated by using a pre-generated rule.
Generating a statement vector corresponding to a statement according to a word vector corresponding to a word in the statement may be to add word vectors corresponding to each word in the statement to obtain a statement vector corresponding to the statement, that is, to use a value obtained by adding a value at a corresponding position of each word vector as a value at the corresponding position of the statement vector; or adding the word vectors corresponding to each word in the sentence and dividing the added word vectors by the number of the word vectors, namely adding the numerical value at the corresponding position of each word vector and dividing the numerical value by the number of the word vectors to obtain a value serving as the numerical value at the corresponding position of the sentence vector.
Further, before searching for a word vector corresponding to each word after each sentence is participated, the model training method further includes: acquiring text information with preset data volume to obtain a text information set; generating a word vector of each word in the text information set by using a machine learning mode to obtain a word vector set; searching for the word vector corresponding to each word after each sentence is participated comprises: and searching a word vector corresponding to each word after word segmentation from the word vector set.
In this embodiment, a large amount of text information is collected as a training text set, and word vectors are generated by using a machine learning method to obtain a word vector set of texts related to each service type. The word vector is generated by a machine learning method, and can be obtained by a method of performing word vector training by using Google word2vec and the like.
By means of the pre-generated word vector set, when the statement vector is generated according to the word vector, the word vector of the corresponding word can be directly inquired from the word vector set, and therefore the determining efficiency of the statement vector is improved.
Further, determining the sentence vector of the corresponding sentence according to the word vector corresponding to each word in each sentence includes: for each statement, adding word vectors corresponding to all words contained in the statement to obtain a word vector sum, wherein the word vector sum is a vector which is obtained by adding numerical values of corresponding positions in the word vectors and has the same number of dimensions as the word vectors; and dividing the word vector sum by the number of the words to obtain a statement vector, wherein the word vector sum divided by the number of the words refers to a numerical value of each position in the word vector sum divided by the number of the words.
In this embodiment, the word vectors corresponding to each word in the sentence are added and then divided by the number of the word vectors, that is, the numerical value at the corresponding position of each word vector is added and then divided by the number of the word vectors, and the obtained value is used as the numerical value at the corresponding position of the sentence vector.
For example, for "do you not sing a song? "sentence vector, which is segmented into words to get" do you not sing? "assume where" you "correspond to a word vector of [1,0,0,0,0, … … ]," none "corresponds to a word vector of [0,1,0,0,0, … … ]," sing "corresponds to a word vector of [0,0,1,0,0, … … ],"? "the corresponding word vector is [0,0,0,0,1, … … ], then the calculated statement vector for the statement is [0.25,0.25,0.25,0,0.25, … … ].
An optional mode of the model training method in the embodiment of the present application includes:
step one, collecting a large amount of text information as a training text set.
And step two, performing word segmentation on the training text set, and generating a word vector of each word in the text by using a machine learning method to obtain a word vector set. The word vector generated by the machine learning method can be obtained by using Google word2vec to perform word vector training and the like.
And step three, collecting new text information, performing word segmentation on the text information, and performing manual labeling on emotion classification of each sentence in the text information to serve as a sentence emotion training set. The emotion classification can be a custom standard classification, for example, if the sentence emotion is classified into three types: positive, neutral, and negative, respectively, are labeled with 1,0, -1 for emotion correspondence, or are divided into five types: happiness, sadness, anger, fear and disgust.
And step four, obtaining a statement vector with emotion marks corresponding to each statement in the statement emotion training set according to the word vector set, and using the statement vector as the statement vector training set. The statement vector may be calculated by: and adding the word vectors corresponding to each word in the sentence and dividing the sum by the number of the word vectors.
And step five, training the statement vector training set obtained in the step four by using an RNN (recurrent neural network) model to obtain an RNN training model. The machine can memorize words and their combination forms in the sentences and emotion labels of the sentences by using the sentence vectors as the input content of the RNN (recurrent neural network) initial model, and memorize the combinations by using the parameters in the model.
In the embodiment of the application, after the sentences and the emotions are represented by data, machine learning is carried out, the emotion types of the text sentences can be effectively and quickly identified based on context semantics, the problem that in the prior art, the sentence emotion classification accuracy is low by identifying the sentences according to the semantics of the sentences is solved, and the effect of improving the sentence emotion classification accuracy is achieved.
According to the embodiment of the application, a sentence emotion recognition method is further provided, and the sentence emotion recognition method can be used for recognizing the service type of the sentence to be detected through the neural network model obtained through the model training method of the embodiment of the application. As shown in fig. 2, the sentence emotion recognition method includes:
step S202, determining a sentence vector corresponding to each sentence in the text to be tested.
For the determination of the statement vector corresponding to the statement to be detected, reference may be made to the determination manner of the statement vector corresponding to the statement in the collected text information with emotion marks in the above embodiments of the present application, which is not described herein again.
Step S204, inputting the statement vector corresponding to each statement in the text to be detected into the neural network model, and identifying the emotion type of each statement in the text to be detected by using the neural network model.
The neural network model of the embodiment of the present application is a neural network model obtained by training the model training method of the above embodiment of the present application. Inputting the statement vector of the text to be tested into the RNN training model to obtain the emotion type of each statement, wherein the mark is 1 to represent positive, the mark is 0 to represent neutral, the mark is-1 to represent negative, and the mark depends on the definition mode of the mark of the emotion type.
Because the machine can memorize words and words in the sentences and combinations thereof and corresponding emotion marks by inputting the sentence vectors corresponding to the sentences in the text into the cyclic neural network and memorize the words and the combinations thereof and the corresponding emotion marks by using parameters (most of which are determined by parameters in the neural network model) in the neural network model, when the neural network model is used for carrying out emotion recognition on the sentences, the emotion types of the sentences in the text can be effectively and quickly recognized according to the context of the sentences, and compared with the mode of recognizing the emotion of the sentences by using emotion keywords and the like in the prior art, the accuracy of recognizing the emotion of the sentences by the neural network model is greatly improved.
An embodiment of the present application further provides a model training apparatus, which may be used to execute the model training method according to the embodiment of the present application, and as shown in fig. 3, the apparatus includes: an acquisition unit 301, a determination unit 303 and a training unit 305.
The acquiring unit 301 is configured to acquire text information with emotion marks, where the text information includes a plurality of sentences marked according to predefined emotion types, and each sentence carries an emotion mark corresponding to the sentence.
The text information with emotion marks in the embodiment of the application can be text information formed by manually marking the emotion of each sentence in a pre-collected text information sample. The emotion classification of the statement can be set as required, and the emotion classification can be a customized standard classification, such as if the statement emotion is divided into three types: positive, neutral, and negative, respectively, are labeled with 1,0, -1 for emotion correspondence, or are divided into five types: happiness, sadness, anger, fear, disgust, etc.
The determining unit 303 is configured to determine a statement vector corresponding to each statement in the text information, where the statement vector is a multidimensional array for uniquely representing the corresponding statement.
After the text information with the emotion marks is obtained, statement vectors corresponding to each statement in the text are determined, the statement vectors corresponding to each statement can be statement vectors generated according to a predefined rule, and the statement vectors corresponding to different statements are different. Because each sentence in the text information carries the corresponding emotion mark, the sentence vector corresponding to each sentence also corresponds to the same emotion mark as the sentence.
The training unit 305 is configured to input the emotion mark corresponding to each sentence and the sentence vector corresponding to the emotion mark into a cyclic neural network, and train to obtain a neural network model, where the neural network model is used to identify an emotion type in the sentence.
In this embodiment, after determining a sentence vector corresponding to each sentence of the text information, the sentence vectors carrying emotion marks are sequentially input into the recurrent neural network to be trained, so as to obtain the neural network model.
By inputting the statement vector corresponding to the statement in the text into the recurrent neural network, the machine can memorize the words and their combinations and their corresponding emotion marks in the statement, and memorize these words and their combinations and their corresponding emotion marks with the parameters in the neural network model (the parameters in the neural network model are determined, mostly as a matrix).
When the neural network model is used for emotion recognition of sentences, the emotion types of the sentences in the text can be effectively and quickly recognized according to the context of the sentences, and compared with the mode that the sentences are recognized through emotion keywords and the like in the prior art, the accuracy of the sentence emotion recognition through the neural network model is greatly improved.
According to the embodiment of the application, text information with emotion marks is obtained, wherein the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence; determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement; the emotion marks corresponding to each sentence and the sentence vectors corresponding to the emotion marks are input into a cyclic neural network, and a neural network model is obtained through training, so that the emotion types in the sentences can be conveniently recognized through the neural network model, the technical problem that the emotion recognition accuracy contained in the sentences is low in the prior art is solved, and the effect of improving the emotion recognition accuracy in the sentences is achieved.
Preferably, the determination unit includes: the word segmentation module is used for carrying out word segmentation processing on each sentence in the text information; the query module is used for searching a word vector corresponding to each word after each sentence is subjected to word segmentation, wherein the word vector is a multidimensional array used for uniquely representing the corresponding word; and the determining module is used for determining the sentence vector of the corresponding sentence according to the word vector corresponding to each word in each sentence.
In this embodiment, for determining the sentence vectors, word segmentation processing may be performed on each sentence, and then the word vector corresponding to each word in each sentence is queried. Wherein the word segmentation result is to arrange words in order according to the original sentence, including punctuation marks such as "will you not sing? "do the participles" do you not sing a song? "or" do you not sing? ". The word vector of each term may be obtained by searching the word vector corresponding to each term from a pre-generated word vector set, or may be generated by using a pre-generated rule.
Generating a statement vector corresponding to a statement according to a word vector corresponding to a word in the statement may be to add word vectors corresponding to each word in the statement to obtain a statement vector corresponding to the statement, that is, to use a value obtained by adding a value at a corresponding position of each word vector as a value at the corresponding position of the statement vector; or adding the word vectors corresponding to each word in the sentence and dividing the added word vectors by the number of the word vectors, namely adding the numerical value at the corresponding position of each word vector and dividing the numerical value by the number of the word vectors to obtain a value serving as the numerical value at the corresponding position of the sentence vector.
Further, the model training device further comprises: the acquisition unit is used for acquiring text information with preset data volume before searching for a word vector corresponding to each word after each sentence is participled to obtain a text information set; the generating unit is used for generating a word vector of each word in the text information set by utilizing a machine learning mode to obtain a word vector set; the query module is specifically configured to find a word vector corresponding to each word after word segmentation from the word vector set.
In this embodiment, a large amount of text information is collected as a training text set, and word vectors are generated by using a machine learning method to obtain a word vector set of texts related to each service type. The word vector is generated by a machine learning method, and can be obtained by a method of performing word vector training by using Google word2vec and the like.
By means of the pre-generated word vector set, when the statement vector is generated according to the word vector, the word vector of the corresponding word can be directly inquired from the word vector set, and therefore the determining efficiency of the statement vector is improved.
Further, the determining module includes: the first calculation submodule is used for adding word vectors corresponding to all words contained in each statement to obtain a word vector sum, wherein the word vector sum is a vector which is obtained by adding numerical values of corresponding positions in the word vectors and has the same number of dimensions as the word vectors; and the second calculation submodule is used for dividing the word vector sum by the number of the words to obtain a statement vector, wherein the word vector sum divided by the number of the words refers to a numerical value of each position in the word vector sum divided by the number of the words.
In this embodiment, the word vectors corresponding to each word in the sentence are added and then divided by the number of the word vectors, that is, the numerical value at the corresponding position of each word vector is added and then divided by the number of the word vectors, and the obtained value is used as the numerical value at the corresponding position of the sentence vector.
For example, for "do you not sing a song? "sentence vector, which is segmented into words to get" do you not sing? "assume where" you "correspond to a word vector of [1,0,0,0,0, … … ]," none "corresponds to a word vector of [0,1,0,0,0, … … ]," sing "corresponds to a word vector of [0,0,1,0,0, … … ],"? "the corresponding word vector is [0,0,0,0,1, … … ], then the calculated statement vector for the statement is [0.25,0.25,0.25,0,0.25, … … ].
The model training apparatus includes a processor and a memory, the acquiring unit 301, the determining unit 303, the training unit 305, and the like are stored in the memory as program units, and the processor executes the program units stored in the memory.
The processor comprises a kernel, and the kernel calls the corresponding program unit from the memory. The kernel can be set to be one or more than one, and the neural network model is obtained through training by adjusting kernel parameters.
The memory may include volatile memory in a computer readable medium, Random Access Memory (RAM) and/or nonvolatile memory such as Read Only Memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.
The present application further provides an embodiment of a computer program product, which, when being executed on a data processing device, is adapted to carry out program code for initializing the following method steps: acquiring text information with emotion marks, wherein the text information comprises a plurality of sentences marked according to predefined emotion types, and each sentence carries the emotion mark corresponding to the sentence; determining a statement vector corresponding to each statement in the text information, wherein the statement vector is a multidimensional array used for uniquely representing the corresponding statement; and inputting the emotion mark corresponding to each statement and the statement vector corresponding to the emotion mark into a cyclic neural network, and training to obtain a neural network model, wherein the neural network model is used for identifying the emotion type in the statement.
According to the embodiment of the application, a sentence emotion recognition device is further provided, and the wrongly written word recognition device can be used for executing the sentence emotion recognition method provided by the embodiment of the application. As shown in fig. 4, the sentence emotion recognition apparatus includes: a vector determination unit 401 and a recognition unit 403.
The vector determining unit 401 is configured to determine a sentence vector corresponding to each sentence in the text to be tested.
For the determination of the statement vector corresponding to the statement to be detected, reference may be made to the determination manner of the statement vector corresponding to the statement in the collected text information with emotion marks in the above embodiments of the present application, which is not described herein again.
The identifying unit 403 is configured to input a statement vector corresponding to each statement in the text to be detected into the neural network model, and identify an emotion type of each statement in the text to be detected by using the neural network model.
The neural network model of the embodiment of the present application is a neural network model obtained by training the model training method of the above embodiment of the present application. Inputting the statement vector of the text to be tested into the RNN training model to obtain the emotion type of each statement, wherein the mark is 1 to represent positive, the mark is 0 to represent neutral, the mark is-1 to represent negative, and the mark depends on the definition mode of the mark of the emotion type.
Because the machine can memorize words and words in the sentences and combinations thereof and corresponding emotion marks by inputting the sentence vectors corresponding to the sentences in the text into the cyclic neural network and memorize the words and the combinations thereof and the corresponding emotion marks by using parameters (most of which are determined by parameters in the neural network model) in the neural network model, when the neural network model is used for carrying out emotion recognition on the sentences, the emotion types of the sentences in the text can be effectively and quickly recognized according to the context of the sentences, and compared with the mode of recognizing the emotion of the sentences by using emotion keywords and the like in the prior art, the accuracy of recognizing the emotion of the sentences by the neural network model is greatly improved.
The sentence emotion recognition apparatus includes a processor and a memory, and the vector determination unit 401, the recognition unit 403, and the like are stored in the memory as program units, and the processor executes the program units stored in the memory.
The processor comprises a kernel, and the kernel calls the corresponding program unit from the memory. The kernel can be set to be one or more, and the emotion type of the statement in the text to be tested is identified by adjusting the kernel parameters.
The memory may include volatile memory in a computer readable medium, Random Access Memory (RAM) and/or nonvolatile memory such as Read Only Memory (ROM) or flash memory (flash RAM), and the memory includes at least one memory chip.
The present application further provides an embodiment of a computer program product, which, when being executed on a data processing device, is adapted to carry out program code for initializing the following method steps: determining a sentence vector corresponding to each sentence in the text to be tested; and inputting the statement vector corresponding to each statement in the text to be detected into the neural network model, and identifying the emotion type of each statement in the text to be detected by using the neural network model.
The above-mentioned serial numbers of the embodiments of the present application are merely for description and do not represent the merits of the embodiments.
In the above embodiments of the present application, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.
In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The integrated unit, if implemented in the form of a software functional unit 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 may be substantially implemented or contributed to by the prior art, or all or part of the technical solution may be embodied in a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.
The foregoing is only a preferred embodiment of the present application and it should be noted that those skilled in the art can make several improvements and modifications without departing from the principle of the present application, and these improvements and modifications should also be considered as the protection scope of the present application.
Claims (6)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510850126.2A CN106815192B (en) | 2015-11-27 | 2015-11-27 | Model training method and device, and sentence emotion recognition method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510850126.2A CN106815192B (en) | 2015-11-27 | 2015-11-27 | Model training method and device, and sentence emotion recognition method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106815192A CN106815192A (en) | 2017-06-09 |
| CN106815192B true CN106815192B (en) | 2020-04-21 |
Family
ID=59103582
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510850126.2A Expired - Fee Related CN106815192B (en) | 2015-11-27 | 2015-11-27 | Model training method and device, and sentence emotion recognition method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106815192B (en) |
Families Citing this family (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109388793B (en) * | 2017-08-03 | 2023-04-07 | 阿里巴巴集团控股有限公司 | Entity marking method, intention identification method, corresponding device and computer storage medium |
| CN107967258B (en) * | 2017-11-23 | 2021-09-17 | 广州艾媒数聚信息咨询股份有限公司 | Method and system for emotion analysis of text information |
| CN108170683B (en) * | 2018-01-22 | 2019-05-07 | 北京百度网讯科技有限公司 | Method and apparatus for obtaining information |
| CN108304387B (en) * | 2018-03-09 | 2021-06-15 | 联想(北京)有限公司 | Method, device, server group and storage medium for recognizing noise words in text |
| CN108536784B (en) * | 2018-03-29 | 2021-08-24 | 阿里巴巴（中国）有限公司 | Comment information sentiment analysis method and device, computer storage medium and server |
| CN108984522B (en) * | 2018-06-21 | 2022-12-23 | 北京亿家老小科技有限公司 | Intelligent nursing system |
| CN109147826B (en) * | 2018-08-22 | 2022-12-27 | 平安科技（深圳）有限公司 | Music emotion recognition method and device, computer equipment and computer storage medium |
| CN109461039A (en) * | 2018-08-28 | 2019-03-12 | 厦门快商通信息技术有限公司 | A kind of text handling method and intelligent customer service method |
| CN111291570B (en) * | 2018-12-07 | 2022-07-05 | 北京国双科技有限公司 | Method and device for realizing element identification in judicial documents |
| CN109783800B (en) * | 2018-12-13 | 2024-04-12 | 北京百度网讯科技有限公司 | Method, device, equipment and storage medium for acquiring emotional keywords |
| CN111435410B (en) * | 2019-01-14 | 2023-04-14 | 阿里巴巴集团控股有限公司 | Relation extraction method and device for medical text |
| US11132511B2 (en) * | 2019-02-05 | 2021-09-28 | International Business Machines Corporation | System for fine-grained affective states understanding and prediction |
| CN110287461B (en) * | 2019-05-24 | 2023-04-18 | 北京百度网讯科技有限公司 | Text conversion method, device and storage medium |
| CN112445898B (en) * | 2019-08-16 | 2024-06-14 | 阿里巴巴集团控股有限公司 | Dialogue emotion analysis method and device, storage medium and processor |
| CN110727695B (en) * | 2019-09-29 | 2022-05-03 | 浙江大学 | Natural language query analysis method for novel power supply urban rail train data operation and maintenance |
| CN111274807B (en) * | 2020-02-03 | 2022-05-10 | 华为技术有限公司 | Text information processing method and device, computer equipment and readable storage medium |
| CN111666761B (en) * | 2020-05-13 | 2022-12-09 | 北京大学 | Fine-grained sentiment analysis model training method and device |
| CN112732875A (en) * | 2021-01-20 | 2021-04-30 | 珠海格力电器股份有限公司 | Method and device for determining corpus data tags |
| CN113220889A (en) * | 2021-06-01 | 2021-08-06 | 北京读我科技有限公司 | Semantic classification method and device, storage medium and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573046A (en) * | 2015-01-20 | 2015-04-29 | 成都品果科技有限公司 | Comment analyzing method and system based on term vector |
| CN104731770A (en) * | 2015-03-23 | 2015-06-24 | 中国科学技术大学苏州研究院 | Chinese microblog emotion analysis method based on rules and statistical model |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| RU2571373C2 (en) * | 2014-03-31 | 2015-12-20 | Общество с ограниченной ответственностью "Аби ИнфоПоиск" | Method of analysing text data tonality |
-
2015
- 2015-11-27 CN CN201510850126.2A patent/CN106815192B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104573046A (en) * | 2015-01-20 | 2015-04-29 | 成都品果科技有限公司 | Comment analyzing method and system based on term vector |
| CN104731770A (en) * | 2015-03-23 | 2015-06-24 | 中国科学技术大学苏州研究院 | Chinese microblog emotion analysis method based on rules and statistical model |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106815192A (en) | 2017-06-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106815192B (en) | Model training method and device, and sentence emotion recognition method and device | |
| CN114416999B (en) | Hierarchical classifiers for image and text data | |
| CN106815194A (en) | Model training method and device and keyword recognition method and device | |
| CN111666761B (en) | Fine-grained sentiment analysis model training method and device | |
| JP6850806B2 (en) | Annotation system for extracting attributes from electronic data structures | |
| KR101561464B1 (en) | Collected data sentiment analysis method and apparatus | |
| CN111125354A (en) | Text classification method and device | |
| KR102410715B1 (en) | Apparatus and method for analyzing sentiment of text data based on machine learning | |
| CN116882414B (en) | Automatic comment generation method and related device based on large-scale language model | |
| CN106815198A (en) | The recognition methods of model training method and device and sentence type of service and device | |
| CN104915420B (en) | Knowledge base data processing method and system | |
| US9652997B2 (en) | Method and apparatus for building emotion basis lexeme information on an emotion lexicon comprising calculation of an emotion strength for each lexeme | |
| CN107515849A (en) | It is a kind of into word judgment model generating method, new word discovery method and device | |
| CN110647504B (en) | Method and device for searching judicial documents | |
| CN111159400B (en) | Product comment emotion classification method and system | |
| CN110321549B (en) | New concept mining method based on sequential learning, relation mining and time sequence analysis | |
| CN104408036B (en) | It is associated with recognition methods and the device of topic | |
| CN114722827B (en) | Model training method, device and equipment for task processing model and storage medium | |
| CN109408175B (en) | Real-time interaction method and system in general high-performance deep learning calculation engine | |
| JP4600045B2 (en) | Opinion extraction learning device and opinion extraction classification device | |
| CN111488452A (en) | Webpage tampering detection method, detection system and related equipment | |
| CN109033078A (en) | The recognition methods of sentence classification and device, storage medium, processor | |
| CN110019672A (en) | A kind of method for pushing of similar case, system, storage medium and processor | |
| CN110019659A (en) | The search method and device of judgement document | |
| CN113688243B (en) | Annotation methods, devices, equipment and storage media for entities in statements |
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 | ||
| CB02 | Change of applicant information |
Address after: 100083 No. 401, 4th Floor, Haitai Building, 229 North Fourth Ring Road, Haidian District, Beijing Applicant after: BEIJING GRIDSUM TECHNOLOGY Co.,Ltd. Address before: 100086 Cuigong Hotel, 76 Zhichun Road, Shuangyushu District, Haidian District, Beijing Applicant before: BEIJING GRIDSUM TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200421 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |