CN112650923A - Public opinion processing method and device for news events, storage medium and computer equipment - Google Patents
Public opinion processing method and device for news events, storage medium and computer equipment Download PDFInfo
- Publication number
- CN112650923A CN112650923A CN202011526767.XA CN202011526767A CN112650923A CN 112650923 A CN112650923 A CN 112650923A CN 202011526767 A CN202011526767 A CN 202011526767A CN 112650923 A CN112650923 A CN 112650923A
- Authority
- CN
- China
- Prior art keywords
- news
- public opinion
- classification
- text
- information
- 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.)
- Pending
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
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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
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)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Medical Informatics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a public opinion processing method and device of news events, a storage medium and computer equipment, relates to the field of artificial intelligence, and mainly aims to solve the problem of low public opinion processing efficiency of the existing news events. The method comprises the following steps: acquiring collected news public opinion information; carrying out first classification processing on the news public opinion information according to a trained first text classification model; extracting a second text classification model which is matched with the first classification mark obtained by the first classification processing and is trained, and performing second classification processing on the news public opinion information matched with the first classification mark according to the second text classification model; and extracting news event content from the news public opinion information of which the second classification mark is determined by the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a public opinion processing method and device for news events, a storage medium and computer equipment.
Background
With the rapid development of public opinion systems, more and more multimedia enterprises utilize the public opinion systems to obtain a great amount of public opinion information with leading features, such as financing relations of competitive companies, upstream and downstream related companies, and operation situations of opponents or related enterprises.
At present, the method for acquiring data by the existing public opinion system generally utilizes the content in the network news to directly collect the public opinion information, and performs manual analysis processing on the crawled public opinion information, thereby mining useful information aiming at news content, however, the news content scattered on each internet website is too scattered, different types of enterprises cannot meet the information required by self needs to be quickly found, personnel with strong professional classification can be required to perform long-time and high-intensity data arrangement and analysis processing, and the content required by the enterprises can be effectively obtained, so that the timeliness of the collected public opinion information containing news can be influenced after long-time processing, the influence of the public opinion information on the enterprises is influenced, a large amount of human resources are consumed, and the processing efficiency of the enterprises on the public opinion information containing news content is reduced.
Disclosure of Invention
In view of the above, the present invention provides a method and an apparatus for processing public sentiment of news event, a storage medium, and a computer device, and mainly aims to solve the problem of low efficiency in processing public sentiment of the existing news event.
According to an aspect of the present invention, there is provided a public opinion processing method for a news event, including:
acquiring collected news public opinion information, wherein the news public opinion information is information containing text contents of each news event;
carrying out first classification processing on the news public opinion information according to a trained first text classification model;
extracting a second text classification model which is matched with a first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking training sample sets with different public opinion demands in the training process of the first text classification model;
and extracting news event content from the news public opinion information of which the second classification mark is determined by the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
Further, the method further comprises:
constructing a three-layer convolutional neural network model, and extracting characteristic information from news text contents which are respectively subjected to first classification marking in a training sample set based on three preset kernel characteristic values;
performing feature screening on the feature information based on the pooling layer, splicing screened feature vectors, and training the three-layer convolutional neural network model by using the training sample set;
and optimizing the three-layer convolutional neural network model in the training process by using an adam optimizer until the three-layer convolutional neural network model is trained, so as to obtain a first text classification model.
Further, before the building of the three-layer convolutional neural network model, the method further includes:
acquiring news text contents to be marked, and determining public opinion demands;
determining a K value in a K-means cluster according to the public opinion demand, clustering the news text content, and extracting characteristic words of which the occurrence times of text words in different clustered clusters exceed a preset threshold value after clustering to serve as first classified mark content;
and performing first label classification on different clustering clusters after clustering is completed based on the first classification label content.
Further, the method further comprises:
constructing a two-layer convolutional neural network model, and extracting characteristic information from news text contents of second classification marks which belong to the first classification marks in a training sample set based on two preset kernel characteristic values, wherein different first classification marks are matched with at least one different second classification mark;
performing feature screening on the feature information based on the pooling layer, splicing screened feature vectors, and training the two-layer convolutional neural network model by using the training sample set;
and optimizing the two-layer convolutional neural network model in the training process by using an adam optimizer until the two-layer convolutional neural network model is trained, so as to obtain a second text classification model.
Further, before the acquiring of the collected news public opinion information, the method further includes:
receiving a public opinion keyword which is input, wherein the public opinion keyword is associated with the public opinion demand;
searching matched news public opinion information from a public opinion information base collected in a preset time interval according to the public opinion keywords;
and when the number of the searched news public opinion information matched with the public opinion keywords exceeds a preset threshold value, determining the news public opinion information as the collected news public opinion information.
Further, said mapping the news event content in combination with the first classification label, the second classification label, to a corresponding node in a case graph matching the public opinion demand comprises;
defining events of a affair map based on the public opinion keywords, and extracting event content from the news event content;
extracting the event relation in the event content according to the time sequence, the causal relation and the superior-inferior relation;
establishing each node of the affair relation and the node relation according to the time sequence, the causal relation and the superior-inferior relation, and constructing an affair map;
writing the news event content labeled with the first classification mark and the second classification mark into a node corresponding to the event content in the event graph.
Further, the method further comprises:
and when receiving a query request for any node in the event graph, counting the event content of each node having a node relation with the node according to a preset query level, and outputting.
According to another aspect of the present invention, there is provided a public opinion processing apparatus for a news event, comprising:
the acquisition module is used for acquiring the collected news public opinion information, wherein the news public opinion information is information containing the text content of each news event;
the first processing module is used for carrying out first classification processing on the news public opinion information according to a trained first text classification model;
the second processing module is used for extracting a second text classification model which is matched with the first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking from training sample sets with different public opinion demands in the first text classification model training process;
and the output module is used for extracting news event content from the news public opinion information of which the second classification mark is determined in the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
Further, the apparatus further comprises:
the building module is used for building a three-layer convolutional neural network model and extracting characteristic information from news text contents which are respectively marked by first classification in a training sample set on the basis of three preset kernel characteristic values;
the training module is used for carrying out feature screening on the feature information based on the pooling layer, splicing the screened feature vectors and training the three-layer convolutional neural network model by utilizing the training sample set;
and the optimization module is used for optimizing the three-layer convolutional neural network model in the training process by using an adam optimizer until the three-layer convolutional neural network model is trained to obtain a first text classification model.
Further, the apparatus further comprises:
the first determining module is used for acquiring news text contents to be marked and determining public opinion demands;
the clustering module is used for determining a K value in a K-means cluster according to the public sentiment requirement, clustering the news text content, and extracting characteristic words of which the occurrence times of text words in different clustered clusters exceed a preset threshold value after clustering to serve as first classified mark content;
and the marking module is used for carrying out first mark classification on different clustering clusters after clustering based on the first classification mark content.
Further, the building module is further configured to build a two-layer convolutional neural network model, and extract feature information from news text contents of second classification marks belonging to the first classification marks in a training sample set based on two preset kernel feature values, wherein different first classification marks are matched with at least one different second classification mark;
the training module is further used for carrying out feature screening on the feature information based on the pooling layer, splicing the screened feature vectors, and training the two-layer convolutional neural network model by using the training sample set;
the optimization module is further used for optimizing the two-layer convolutional neural network model in the training process by using an adam optimizer until the two-layer convolutional neural network model training is completed, so that a second text classification model is obtained.
Further, the apparatus further comprises:
the public opinion keyword recording module is used for recording public opinion keywords, and the public opinion keywords are related to the public opinion demand;
the searching module is used for searching matched news public opinion information from a public opinion information base collected in a preset time interval according to the public opinion keywords;
and the second determining module is used for determining the news public opinion information as the collected news public opinion information when the number of the searched news public opinion information matched with the public opinion keywords exceeds a preset threshold value.
Further, the output module includes:
the extraction unit is used for defining events of a affair map based on the public sentiment keywords and extracting event contents from the news event contents;
the computing unit is used for extracting the event relation in the event content according to the time sequence, the causal relation and the superior-inferior relation;
the construction unit is used for establishing each node and node relation of the affair relation according to the time sequence, the causal relation and the superior-inferior relation, and constructing an affair map;
and the writing unit is used for writing the news event content marked with the first classification mark and the second classification mark into a node corresponding to the event content in the event graph.
Further, the apparatus further comprises:
and the counting module is used for counting the event content of each node having a node relation with the node according to a preset query level and outputting the event content when receiving a query request for any node in the event graph.
According to another aspect of the present invention, there is provided a storage medium having at least one executable instruction stored therein, the executable instruction causing a processor to perform an operation corresponding to the above-mentioned public opinion processing method for news events.
According to still another aspect of the present invention, there is provided a computer apparatus including: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;
the memory is used for storing at least one executable instruction, and the executable instruction enables the processor to execute the operation corresponding to the public opinion processing method of the news event.
By the technical scheme, the technical scheme provided by the embodiment of the invention at least has the following advantages:
the invention provides a public opinion processing method and device for news events, a storage medium and computer equipment. Compared with the prior art, the embodiment of the invention acquires the collected news public opinion information, wherein the news public opinion information is information containing the text content of each news event; carrying out first classification processing on the news public opinion information according to a trained first text classification model; extracting a second text classification model which is matched with a first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking training sample sets with different public opinion demands in the training process of the first text classification model; follow the second classification processing confirms the second classification mark draw news event content in the news public opinion information, and combine first classification mark the second classification mark will news event content map to with corresponding node in the affairs atlas that the public opinion demand matches, export, satisfy different enterprise users and carry out the demand that accurate news public opinion was handled, greatly reduced manpower resources consumption to the processing high efficiency of news public opinion has been improved, thereby the public opinion processing efficiency of news event has been improved.
The foregoing description is only an overview of the technical solutions of the present invention, and the embodiments of the present invention are described below in order to make the technical means of the present invention more clearly understood and to make the above and other objects, features, and advantages of the present invention more clearly understandable.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
fig. 1 is a flowchart illustrating a public opinion processing method for a news event according to an embodiment of the present invention;
fig. 2 is a block diagram illustrating a public opinion processing device for news events according to an embodiment of the present invention;
fig. 3 shows a schematic structural diagram of a computer device according to an embodiment of the present invention.
Detailed Description
Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.
The embodiment of the invention provides a public opinion processing method of news events, as shown in fig. 1, the method comprises the following steps:
101. and acquiring the collected news public opinion information.
The news public opinion information is information including text contents of each news event, and may include the whole text of the news event, or may include partial text of the news event, and is collected by the public opinion system, for example, the news events of each complete text content, such as the whole report text contents of the last half year of the technical development of the enterprise a, are collected from different news internet websites by the public opinion system, and the embodiment of the present invention is not particularly limited.
It should be noted that, the acquired news public opinion information in step 101 is acquired after the news public opinion information acquired based on the public opinion system is stored, and may be in real time or at preset time intervals, so as to process the news public opinion information.
102. And carrying out first classification processing on the news public opinion information according to the trained first text classification model.
The first text classification model may be any machine learning model with a classification function, such as a neural network model and a support vector machine model, which is not limited in the embodiments of the present invention. Specifically, the first text classification model is based on classification of text features in the news public opinion information, so that before the first classification processing, natural language processing needs to be performed on the news public opinion information, and after the news public opinion information is converted into word vectors, classification is performed based on the trained first text classification model.
It should be noted that after the first classification, the obtained classification may be classified and labeled at the same time, for example, the classification performed according to each news word in the news event may include a business class label, a risk class label, a competition class label, and the like, and of course, the labeled classification label is obtained when the first text classification model is trained and the training sample set is labeled, so that the obtained news and public opinion information is subjected to the first classification, thereby limiting the first classification in the embodiment of the present invention to be a classification process according to a large range.
103. And extracting a second text classification model which is matched with the first classification mark obtained by the first classification processing and is trained, and performing second classification processing on the news public opinion information matched with the first classification mark according to the second text classification model.
In the embodiment of the present invention, after the first classification processing is completed, the first classification mark may be obtained, such as a business class mark, a risk class mark, a competition class mark, and the like in step 102. In the embodiment of the invention, different classification labels are matched with the second text classification model in advance for training, so that the detailed classification in a small range is carried out again under the classification of a large-range program. Specifically, the second text classification model may be any one machine learning model, and may be the same as or different from the first text classification model, for example, the second text classification model may be a neural network model, a support vector machine model, and the like, and the embodiment of the present invention is not limited specifically. In addition, when the second text classification model completes training, training is carried out based on training sample sets corresponding to different first classification marks, so that the corresponding second text classification model can be matched based on the first classification mark, and news public opinion information marked as the first classification mark is subjected to second classification processing based on the matched second text classification model.
It should be noted that, in order to further improve the labeling capability of the first text classification model, the first classification label is automatically labeled and determined from training sample sets with different public opinion demands in the first text classification model training process, that is, a business machine label, a risk label and a competition label are automatically labeled from the training sample sets based on different public opinion demands, that is, after a clustering feature is determined based on the public opinion demands by a clustering algorithm, a news event sample in the training sample sets is labeled, which is not specifically limited in the embodiment of the present invention.
104. And extracting news event content from the news public opinion information of which the second classification mark is determined by the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
In the embodiment of the present invention, news public opinion information of the second classification mark is obtained by classifying again based on the first classification mark, that is, one first classification mark may classify a plurality of second classification marks, for example, the second classification mark corresponding to the business machine mark as the first classification mark includes a new technology class related to an enterprise, an enterprise co-purchase news class, a new market new sales discovery class, and the like, which is not specifically limited in the embodiment of the present invention. Therefore, the finally obtained news public opinion information can have 2 classification marks, and news event content is extracted and mapped to nodes in the event graph for output.
It should be noted that the case map is a case logic knowledge base, and describes evolution rules and patterns between events. Structurally, the event graph is a directed cyclic graph, wherein nodes represent events, and directed edges represent sequential, causal, conditional, upper and lower level and other event logical relations among the events. Therefore, in order to accurately and efficiently present the event content of the news event corresponding to different classification identifications to the user so as to achieve the purpose of news early warning, the event content is mapped to the node corresponding to the event map in a logical relationship mode.
For the embodiment of the invention, in order to improve the accuracy of classification of news public opinion information, a training method for a first text classification model is further defined, and the method further comprises the following steps: constructing a three-layer convolutional neural network model, and extracting characteristic information from news text contents which are respectively subjected to first classification marking in a training sample set based on three preset kernel characteristic values; performing feature screening on the feature information based on the pooling layer, splicing screened feature vectors, and training the three-layer convolutional neural network model by using the training sample set; and optimizing the three-layer convolutional neural network model in the training process by using an adam optimizer until the three-layer convolutional neural network model is trained, so as to obtain a first text classification model.
In the embodiment of the invention, a convolutional neural network model is selected as a first text classification model to be trained, a three-layer convolutional neural network model is constructed, the characteristics are extracted from a training sample news text by utilizing kernel characteristic extraction, the sizes of kernels are preferably set to be 2, 3 and 4 respectively, wherein the model comprises an input layer, a convolutional layer, a pooling layer, a fully-connected layer dense and an output layer with three dimensions; feature screening is carried out on the features extracted by the three convolution kernels through maxPooling, and feature vectors are spliced, wherein maxPooling extracts a plurality of feature values from one Filter convolution layer, only the largest pooled Poolling layer is obtained as a reserved value, other feature values are discarded completely, the maximum value represents that only the strongest of the features is reserved,discarding other weak such features; then, through a dense layer, performing three classifications on the activation function softmax; finally, performing model optimization on the convolutional neural network model by using an adam optimizer, and setting the optimized learning rate to be 0.0005, including: the cumulative quantity and the squared cumulative quantity V of the initialization gradientd/omega＝0；Sd/omega＝0；Vdb＝0；Sdb0; in the t iteration, calculating d/omega and db by using a mini-batch gradient descent method; calculating a Momentum exponentially weighted average; updating by using an RMSprop algorithm; calculating deviation correction of Momentum and RMSprop; and updating the weight until the model iterative training is completed.
For the embodiment of the invention, in order to meet the requirements of different enterprises for acquiring public opinion information and realize the realizability of automatic marking, before the three-layer convolutional neural network model is constructed, the method further comprises the following steps: acquiring news text contents to be marked, and determining public opinion demands; determining a K value in a K-means cluster according to the public opinion demand, clustering the news text content, and extracting characteristic words of which the occurrence times of text words in different clustered clusters exceed a preset threshold value after clustering to serve as first classified mark content; and performing first label classification on different clustering clusters after clustering is completed based on the first classification label content.
The news text content to be marked is stored in the training sample set and serves as the text content to be trained, the public opinion demand is the number of classifications directly input by the user, for example, 2 numbers of classifications of competition and xx enterprises, and the k value is determined, so that the marking is automatically realized by combining the public opinion demand. After clustering is completed, different clustering clusters are obtained, that is, text contents of each class which are determined as a classification can be determined, and further determination of a mark corresponding to the text contents is required, so that the mark is determined as a mark content based on that the occurrence frequency of text words exceeds a preset feature word of a threshold, for example, in 2 clustering clusters, if a word "risk" exceeds a preset threshold 7 of feature words such as "risk insurance", "risk investment", and the like, the risk is determined as a classification mark of the class, and the embodiment of the present invention is not particularly limited.
It should be noted that the specific method for clustering by using k-means includes: determining a k value according to public opinion demands; randomly selecting k data points from the data set as a centroid; calculating the distance (such as Euclidean distance) between each point in the data set and each centroid, and dividing the point to which the centroid belongs when the point is close to which centroid; after all data are grouped together, there are k groups. Then re-computing the centroid of each set; if the distance between the newly calculated centroid and the original centroid is smaller than a certain set threshold value, the clustering can be considered to reach an expected result, and the algorithm is terminated; if the distance between the new centroid and the original centroid is changed greatly, the step of calculating the centroid distance value needs to be iterated again until the requirement is met.
For the embodiment of the invention, in order to improve the accuracy of classification of news public opinion information, a training method for a first text classification model is further defined, and the method further comprises the following steps: constructing a two-layer convolutional neural network model, and extracting characteristic information from news text contents of second classification marks which belong to the first classification marks in a training sample set based on two preset kernel characteristic values, wherein different first classification marks are matched with at least one different second classification mark; performing feature screening on the feature information based on the pooling layer, splicing screened feature vectors, and training the two-layer convolutional neural network model by using the training sample set; and optimizing the two-layer convolutional neural network model in the training process by using an adam optimizer until the two-layer convolutional neural network model is trained, so as to obtain a second text classification model.
In the embodiment of the invention, a convolutional neural network model is selected as a second text classification model to be trained, a two-layer convolutional neural network model is constructed, the characteristics are extracted from a training sample news text by utilizing kernel characteristic extraction, the sizes of kernels are preferably set to be 2 and 3 respectively, and the model comprises an input layer, a convolutional layer, a pooling layer, a fully-connected layer dense and an output layer with two dimensions; performing feature screening on the features extracted by the two convolution kernels through maxporoling, and splicing feature vectors; then, through a dense layer, the droout rate of each nerve unit is 0.2, and the activation function is softmax to carry out three classifications; and finally, performing model optimization on the CNN model by using an adam optimizer, setting the optimized learning rate to be 0.0001, and setting the optimized learning rate to be the same as the optimization method in the training process of the first text classification model, which is not repeated here.
For the embodiment of the invention, in order to improve the accuracy of obtaining news public opinion information and thus improve the public opinion processing efficiency, before obtaining the collected news public opinion information, the method further comprises: receiving a public opinion keyword which is input, wherein the public opinion keyword is associated with the public opinion demand; searching matched news public opinion information from a public opinion information base collected in a preset time interval according to the public opinion keywords; and when the number of the searched news public opinion information matched with the public opinion keywords exceeds a preset threshold value, determining the news public opinion information as the collected news public opinion information.
Specifically, the public sentiment keyword can be directly input, and is associated with the public sentiment demand, for example, the number of the input public sentiment demand is 3, and the number of the keywords can be more than 3, so that the accuracy of public sentiment processing is improved. The public opinion system stores collected news public opinion information in a public opinion information base, when public opinion processing is needed, the news public opinion information in the public opinion information base is collected according to a preset time interval, matched news public opinion information is searched by combining public opinion keywords, and for example, all the news public opinion information containing word risks is searched based on keyword risks. Further, if the number of the found matched public opinion keywords in the news public opinion information exceeds a preset threshold, the collected news public opinion information is determined, for example, if the number of the matched news public opinion information 1 includes 5 risky words, and if the number of the matched news public opinion keywords exceeds a preset threshold 3, the news public opinion information 1 is used as the collected news public opinion information.
For the embodiment of the present invention, in order to further define a mapping method of a case map, so as to improve public opinion processing efficiency, the mapping the news event content to a corresponding node in the case map matching the public opinion demand by combining the first classification mark and the second classification mark comprises: defining events of a affair map based on the public opinion keywords, and extracting event content from the news event content; extracting the event relation in the event content according to the time sequence, the causal relation and the superior-inferior relation; establishing each node of the affair relation and the node relation according to the time sequence, the causal relation and the superior-inferior relation, and constructing an affair map; writing the news event content labeled with the first classification mark and the second classification mark into a node corresponding to the event content in the event graph.
Specifically, the event of the event graph is defined based on the public sentiment keywords, that is, the term of the first-layer event needing to construct the event graph is determined, for example, the public sentiment keywords are competition, the event of the event graph is determined to be competition, that is, the constructed event graph is the event content corresponding to the relevant news public sentiment information surrounding the competition. The event content may be extracted from the news event content through a natural language processing technology, where the event content includes a core of the whole news event, for example, the news event content is a malicious competitive event for apple sales in some three business macros, and in the extraction process, the event content is extracted as apple sales malicious competition in the form of a subject, a verb, and a predicate, and the embodiment of the present invention is not particularly limited. In addition, since the collection of different news public opinion information is based on time, whether the information is in a website, whether the information is in a special column, and the like, the distribution determines the event relationship according to the time sequence (collection time), the causal relationship (whether the information is collected in a website), and the upper-lower relationship (whether the information is collected in a special column with the upper-lower relationship), that is, the relationship among the nodes in the event map, and each node is used for storing one piece of news public opinion information. When the event graph is constructed, the event graph is used as each layer of network nodes according to the time sequence, and the nodes for storing the event contents of the news public opinion information are connected according to the causal relationship and the upper and lower position relationship in sequence.
For the embodiment of the invention, in order to facilitate the timely acquisition of the public opinion processing result and improve the early warning effect on news public opinions, the method further comprises the following steps: and when receiving a query request for any node in the event graph, counting the event content of each node having a node relation with the node according to a preset query level, and outputting.
The user can trigger a query request for any node in the event picture based on the mouse, and when the current end receives the query request, the event content of each node having a node relation with the queried node is counted according to a preset query level and output. The preset query level is a level capable of querying the previous node and the next node based on the nodes, preferably 2 nodes are located upwards and 2 nodes are located downwards, and then event contents in the nodes are counted and output. In addition, the node relationship with the node is a node that can be searched by upper and lower levels, and the embodiment of the present invention is not particularly limited.
The embodiment of the invention provides a public opinion processing method for news events, which is characterized in that collected news public opinion information is obtained, and the news public opinion information is information containing text contents of each news event; carrying out first classification processing on the news public opinion information according to a trained first text classification model; extracting a second text classification model which is matched with a first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking training sample sets with different public opinion demands in the training process of the first text classification model; follow the second classification processing confirms the second classification mark draw news event content in the news public opinion information, and combine first classification mark the second classification mark will news event content map to with corresponding node in the affairs atlas that the public opinion demand matches, export, satisfy different enterprise users and carry out the demand that accurate news public opinion was handled, greatly reduced manpower resources consumption to the processing high efficiency of news public opinion has been improved, thereby the public opinion processing efficiency of news event has been improved.
Further, as an implementation of the method shown in fig. 1, an embodiment of the present invention provides a public opinion processing apparatus for a news event, as shown in fig. 2, the apparatus includes:
the acquisition module 21 is configured to acquire collected news public opinion information, where the news public opinion information is information including text of each news event;
the first processing module 22 is configured to perform first classification processing on the news public opinion information according to the trained first text classification model;
a second processing module 23, configured to extract a second text classification model that matches the first classification label obtained by the first classification processing and has been trained, and perform a second classification processing on news public sentiment information that matches the first classification label according to the second text classification model, where the first classification label is determined by automatically labeling a training sample set of different public sentiment requirements in a training process of the first text classification model;
and an output module 24, configured to extract news event content from the news public opinion information with the second classification processing determined second classification marker, and map the news event content to a corresponding node in a case map matched with the public opinion demand by combining the first classification marker and the second classification marker, so as to output the news event content.
Further, the apparatus further comprises:
the building module is used for building a three-layer convolutional neural network model and extracting characteristic information from news text contents which are respectively marked by first classification in a training sample set on the basis of three preset kernel characteristic values;
the training module is used for carrying out feature screening on the feature information based on the pooling layer, splicing the screened feature vectors and training the three-layer convolutional neural network model by utilizing the training sample set;
and the optimization module is used for optimizing the three-layer convolutional neural network model in the training process by using an adam optimizer until the three-layer convolutional neural network model is trained to obtain a first text classification model.
Further, the apparatus further comprises:
the first determining module is used for acquiring news text contents to be marked and determining public opinion demands;
the clustering module is used for determining a K value in a K-means cluster according to the public sentiment requirement, clustering the news text content, and extracting characteristic words of which the occurrence times of text words in different clustered clusters exceed a preset threshold value after clustering to serve as first classified mark content;
and the marking module is used for carrying out first mark classification on different clustering clusters after clustering based on the first classification mark content.
Further, the building module is further configured to build a two-layer convolutional neural network model, and extract feature information from news text contents of second classification marks belonging to the first classification marks in a training sample set based on two preset kernel feature values, wherein different first classification marks are matched with at least one different second classification mark;
the training module is further used for carrying out feature screening on the feature information based on the pooling layer, splicing the screened feature vectors, and training the two-layer convolutional neural network model by using the training sample set;
the optimization module is further used for optimizing the two-layer convolutional neural network model in the training process by using an adam optimizer until the two-layer convolutional neural network model training is completed, so that a second text classification model is obtained.
Further, the apparatus further comprises:
the public opinion keyword recording module is used for recording public opinion keywords, and the public opinion keywords are related to the public opinion demand;
the searching module is used for searching matched news public opinion information from a public opinion information base collected in a preset time interval according to the public opinion keywords;
and the second determining module is used for determining the news public opinion information as the collected news public opinion information when the number of the searched news public opinion information matched with the public opinion keywords exceeds a preset threshold value.
Further, the output module includes:
the extraction unit is used for defining events of a affair map based on the public sentiment keywords and extracting event contents from the news event contents;
the computing unit is used for extracting the event relation in the event content according to the time sequence, the causal relation and the superior-inferior relation;
the construction unit is used for establishing each node and node relation of the affair relation according to the time sequence, the causal relation and the superior-inferior relation, and constructing an affair map;
and the writing unit is used for writing the news event content marked with the first classification mark and the second classification mark into a node corresponding to the event content in the event graph.
Further, the apparatus further comprises:
and the counting module is used for counting the event content of each node having a node relation with the node according to a preset query level and outputting the event content when receiving a query request for any node in the event graph.
The embodiment of the invention provides a public opinion processing device for news events, which is characterized in that collected news public opinion information is obtained, and the news public opinion information is information containing text contents of each news event; carrying out first classification processing on the news public opinion information according to a trained first text classification model; extracting a second text classification model which is matched with a first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking training sample sets with different public opinion demands in the training process of the first text classification model; follow the second classification processing confirms the second classification mark draw news event content in the news public opinion information, and combine first classification mark the second classification mark will news event content map to with corresponding node in the affairs atlas that the public opinion demand matches, export, satisfy different enterprise users and carry out the demand that accurate news public opinion was handled, greatly reduced manpower resources consumption to the processing high efficiency of news public opinion has been improved, thereby the public opinion processing efficiency of news event has been improved.
According to an embodiment of the present invention, a storage medium is provided, where the storage medium stores at least one executable instruction, and the computer executable instruction can execute the public opinion processing method of news events in any of the above method embodiments.
Fig. 3 is a schematic structural diagram of a computer device according to an embodiment of the present invention, and the specific embodiment of the present invention does not limit the specific implementation of the computer device.
As shown in fig. 3, the computer apparatus may include: a processor (processor)302, a communication Interface 304, a memory 306, and a communication bus 308.
Wherein: the processor 302, communication interface 304, and memory 306 communicate with each other via a communication bus 308.
A communication interface 304 for communicating with network elements of other devices, such as clients or other servers.
The processor 302 is configured to execute the program 310, and may specifically execute the relevant steps in the above-mentioned public opinion processing method embodiment of the news event.
In particular, program 310 may include program code comprising computer operating instructions.
The processor 302 may be a central processing unit CPU, or an Application Specific Integrated Circuit (ASIC), or one or more Integrated circuits configured to implement an embodiment of the present invention. The computer device includes one or more processors, which may be the same type of processor, such as one or more CPUs; or may be different types of processors such as one or more CPUs and one or more ASICs.
And a memory 306 for storing a program 310. Memory 306 may comprise high-speed RAM memory and may also include non-volatile memory, such as at least one disk memory. The memory may be non-volatile or volatile.
The program 310 may specifically be configured to cause the processor 302 to perform the following operations:
acquiring collected news public opinion information, wherein the news public opinion information is information containing text contents of each news event;
carrying out first classification processing on the news public opinion information according to a trained first text classification model;
extracting a second text classification model which is matched with a first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking training sample sets with different public opinion demands in the training process of the first text classification model;
and extracting news event content from the news public opinion information of which the second classification mark is determined by the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented by a general purpose computing device, they may be centralized on a single computing device or distributed across a network of multiple computing devices, and alternatively, they may be implemented by program code executable by a computing device, such that they may be stored in a storage device and executed by a computing device, and in some cases, the steps shown or described may be performed in an order different than that described herein, or they may be separately fabricated into individual integrated circuit modules, or multiple ones of them may be fabricated into a single integrated circuit module. Thus, the present invention is not limited to any specific combination of hardware and software.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A public opinion processing method of news events is characterized by comprising the following steps:
acquiring collected news public opinion information, wherein the news public opinion information is information containing text contents of each news event;
carrying out first classification processing on the news public opinion information according to a trained first text classification model;
extracting a second text classification model which is matched with a first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking training sample sets with different public opinion demands in the training process of the first text classification model;
and extracting news event content from the news public opinion information of which the second classification mark is determined by the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
2. The method of claim 1, further comprising:
constructing a three-layer convolutional neural network model, and extracting characteristic information from news text contents which are respectively subjected to first classification marking in a training sample set based on three preset kernel characteristic values;
performing feature screening on the feature information based on the pooling layer, splicing screened feature vectors, and training the three-layer convolutional neural network model by using the training sample set;
and optimizing the three-layer convolutional neural network model in the training process by using an adam optimizer until the three-layer convolutional neural network model is trained, so as to obtain a first text classification model.
3. The method of claim 2, wherein prior to constructing the three-layer convolutional neural network model, the method further comprises:
acquiring news text contents to be marked, and determining public opinion demands;
determining a K value in a K-means cluster according to the public opinion demand, clustering the news text content, and extracting characteristic words of which the occurrence times of text words in different clustered clusters exceed a preset threshold value after clustering to serve as first classified mark content;
and performing first label classification on different clustering clusters after clustering is completed based on the first classification label content.
4. The method of claim 1, further comprising:
constructing a two-layer convolutional neural network model, and extracting characteristic information from news text contents of second classification marks which belong to the first classification marks in a training sample set based on two preset kernel characteristic values, wherein different first classification marks are matched with at least one different second classification mark;
performing feature screening on the feature information based on the pooling layer, splicing screened feature vectors, and training the two-layer convolutional neural network model by using the training sample set;
and optimizing the two-layer convolutional neural network model in the training process by using an adam optimizer until the two-layer convolutional neural network model is trained, so as to obtain a second text classification model.
5. The method of claim 1, wherein before obtaining the collected news public opinion information, the method further comprises:
receiving a public opinion keyword which is input, wherein the public opinion keyword is associated with the public opinion demand;
searching matched news public opinion information from a public opinion information base collected in a preset time interval according to the public opinion keywords;
and when the number of the searched news public opinion information matched with the public opinion keywords exceeds a preset threshold value, determining the news public opinion information as the collected news public opinion information.
6. The method of any one of claims 1-5, wherein the mapping the news event content in combination with the first classification label and the second classification label to corresponding nodes in a case graph matching the public sentiment needs comprises;
defining events of a affair map based on the public opinion keywords, and extracting event content from the news event content;
extracting the event relation in the event content according to the time sequence, the causal relation and the superior-inferior relation;
establishing each node of the affair relation and the node relation according to the time sequence, the causal relation and the superior-inferior relation, and constructing an affair map;
writing the news event content labeled with the first classification mark and the second classification mark into a node corresponding to the event content in the event graph.
7. The method of claim 6, further comprising:
and when receiving a query request for any node in the event graph, counting the event content of each node having a node relation with the node according to a preset query level, and outputting.
8. The utility model provides a public opinion processing apparatus of news incident which characterized in that includes:
the acquisition module is used for acquiring the collected news public opinion information, wherein the news public opinion information is information containing the text content of each news event;
the first processing module is used for carrying out first classification processing on the news public opinion information according to a trained first text classification model;
the second processing module is used for extracting a second text classification model which is matched with the first classification mark obtained by the first classification processing and is trained, and performing second classification processing on news public opinion information matched with the first classification mark according to the second text classification model, wherein the first classification mark is determined by automatically marking from training sample sets with different public opinion demands in the first text classification model training process;
and the output module is used for extracting news event content from the news public opinion information of which the second classification mark is determined in the second classification processing, and mapping the news event content to a corresponding node in a affair map matched with the public opinion demand by combining the first classification mark and the second classification mark for outputting.
9. A storage medium having at least one executable instruction stored therein, the executable instruction causing a processor to perform an operation corresponding to the public opinion processing method of a news event according to any one of claims 1 to 7.
10. A computer device, comprising: the system comprises a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface complete mutual communication through the communication bus;
the memory is used for storing at least one executable instruction, and the executable instruction causes the processor to execute the operation corresponding to the public opinion processing method of the news event according to any one of claims 1-7.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011526767.XA CN112650923A (en) | 2020-12-22 | 2020-12-22 | Public opinion processing method and device for news events, storage medium and computer equipment |
| PCT/CN2021/124890 WO2022134794A1 (en) | 2020-12-22 | 2021-10-20 | Method and apparatus for processing public opinions about news event, storage medium, and computer device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011526767.XA CN112650923A (en) | 2020-12-22 | 2020-12-22 | Public opinion processing method and device for news events, storage medium and computer equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112650923A true CN112650923A (en) | 2021-04-13 |
Family
ID=75358945
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011526767.XA Pending CN112650923A (en) | 2020-12-22 | 2020-12-22 | Public opinion processing method and device for news events, storage medium and computer equipment |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112650923A (en) |
| WO (1) | WO2022134794A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022134794A1 (en) * | 2020-12-22 | 2022-06-30 | 深圳壹账通智能科技有限公司 | Method and apparatus for processing public opinions about news event, storage medium, and computer device |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111090744A (en) * | 2019-12-17 | 2020-05-01 | 中科鼎富（北京）科技发展有限公司 | Stock market operation risk information mining method and device |
| CN114880491B (en) * | 2022-07-08 | 2022-09-23 | 云孚科技(北京)有限公司 | Method and system for automatically constructing case map |
| CN114969382B (en) * | 2022-07-19 | 2022-10-21 | 国网浙江省电力有限公司信息通信分公司 | Entity generation method based on event chain inference of event graph |
| CN115358896B (en) * | 2022-10-20 | 2023-02-03 | 四川大学华西医院 | Method, device, equipment and medium for constructing criminal evolution network by using massive documents |
| CN115827989B (en) * | 2023-02-16 | 2023-04-28 | 杭州金诚信息安全科技有限公司 | Network public opinion artificial intelligent early warning system and method in big data environment |
| CN116522013B (en) * | 2023-06-29 | 2023-09-05 | 乐麦信息技术(杭州)有限公司 | Public opinion analysis method and system based on social network platform |
| CN117312634B (en) * | 2023-11-29 | 2024-02-20 | 大文传媒集团(山东)有限公司 | Artificial intelligence data integration and propagation processing system |
| CN117649117B (en) * | 2024-01-30 | 2024-05-07 | 浙江数洋科技有限公司 | Treatment scheme determining method and device and computer equipment |
| CN118396092A (en) * | 2024-05-06 | 2024-07-26 | 苏州日报社 | Knowledge graph construction method of news data based on artificial intelligence |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11651016B2 (en) * | 2018-08-09 | 2023-05-16 | Walmart Apollo, Llc | System and method for electronic text classification |
| CN110489550A (en) * | 2019-07-16 | 2019-11-22 | 招联消费金融有限公司 | File classification method, device and computer equipment based on combination neural net |
| CN111161726B (en) * | 2019-12-24 | 2023-11-03 | 广州索答信息科技有限公司 | Intelligent voice interaction method, device, medium and system |
| CN112650923A (en) * | 2020-12-22 | 2021-04-13 | 深圳壹账通智能科技有限公司 | Public opinion processing method and device for news events, storage medium and computer equipment |
-
2020
- 2020-12-22 CN CN202011526767.XA patent/CN112650923A/en active Pending
-
2021
- 2021-10-20 WO PCT/CN2021/124890 patent/WO2022134794A1/en active Application Filing
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022134794A1 (en) * | 2020-12-22 | 2022-06-30 | 深圳壹账通智能科技有限公司 | Method and apparatus for processing public opinions about news event, storage medium, and computer device |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2022134794A1 (en) | 2022-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112650923A (en) | Public opinion processing method and device for news events, storage medium and computer equipment | |
| US11238310B2 (en) | Training data acquisition method and device, server and storage medium | |
| CN114238573B (en) | Text countercheck sample-based information pushing method and device | |
| CN110968684B (en) | Information processing method, device, equipment and storage medium | |
| CN110532451A (en) | Search method and device for policy text, storage medium, electronic device | |
| CN109886294A (en) | Knowledge fusion method, apparatus, computer equipment and storage medium | |
| CN110516074B (en) | Website theme classification method and device based on deep learning | |
| CN111914159B (en) | Information recommendation method and terminal | |
| WO2022048363A1 (en) | Website classification method and apparatus, computer device, and storage medium | |
| CN102508859A (en) | Advertisement classification method and device based on webpage characteristic | |
| CN107844533A (en) | A kind of intelligent Answer System and analysis method | |
| CN110633366A (en) | Short text classification method, device and storage medium | |
| CN113254649B (en) | Training method of sensitive content recognition model, text recognition method and related device | |
| CN110399895A (en) | The method and apparatus of image recognition | |
| CN111767725A (en) | Data processing method and device based on emotion polarity analysis model | |
| CN113626607B (en) | Abnormal work order identification method and device, electronic equipment and readable storage medium | |
| CN113051914A (en) | Enterprise hidden label extraction method and device based on multi-feature dynamic portrait | |
| CN108647322A (en) | The method that word-based net identifies a large amount of Web text messages similarities | |
| CN111680506A (en) | External key mapping method and device of database table, electronic equipment and storage medium | |
| CN114416998A (en) | Text label identification method and device, electronic equipment and storage medium | |
| CN112100372A (en) | Head news prediction classification method | |
| Qi et al. | Patent analytic citation-based vsm: Challenges and applications | |
| CN110598787B (en) | Software bug classification method based on self-defined step length learning | |
| CN116541166A (en) | Super-computing power scheduling server and resource management method | |
| CN114398891B (en) | Method for generating KPI curve and marking wave band characteristics based on log keywords |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 40049935 Country of ref document: HK |
|
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |