CN116127321A - Training method, pushing method and system for ship news pushing model - Google Patents
Training method, pushing method and system for ship news pushing model Download PDFInfo
- Publication number
- CN116127321A CN116127321A CN202310133118.0A CN202310133118A CN116127321A CN 116127321 A CN116127321 A CN 116127321A CN 202310133118 A CN202310133118 A CN 202310133118A CN 116127321 A CN116127321 A CN 116127321A
- Authority
- CN
- China
- Prior art keywords
- news
- ship
- data set
- pushing
- training
- 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
- 238000012549 training Methods 0.000 title claims abstract description 114
- 238000000034 method Methods 0.000 title claims abstract description 60
- 238000012545 processing Methods 0.000 claims abstract description 45
- 238000004364 calculation method Methods 0.000 claims abstract description 30
- 239000013598 vector Substances 0.000 claims description 17
- 238000011176 pooling Methods 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000005070 sampling Methods 0.000 claims description 8
- 230000009193 crawling Effects 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 5
- 239000013604 expression vector Substances 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims 1
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 27
- 239000011159 matrix material Substances 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000006467 substitution reaction Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000008358 core component Substances 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
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
- G06F16/951—Indexing; Web crawling techniques
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Electrically Operated Instructional Devices (AREA)
Abstract
The invention discloses a training method, a pushing method and a system of a ship news pushing model, wherein the training method comprises the steps of acquiring a ship news data set; carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts; and inputting the training data set into a ship news pushing model for training treatment to obtain a trained ship news pushing model. The embodiment of the invention can solve the difficulty that proper news is difficult to push when the ship news is read for the first time by using the ship news push model, improves the efficiency of ship news push, and can be widely applied to the technical field of artificial intelligence.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a training method, a pushing method and a pushing system of a ship news pushing model.
Background
In the context of the internet and the big data age, as the marine industry has grown faster in recent years, and personal news reading habits have also changed, the transfer from traditional media, such as paper newspapers or broadcast television newscasts, to network media platforms, etc. has also been made. Although the ship news can be acquired on a large number of network platforms, in the Internet age, the pushing of the ship news is often quite complicated, and if the needed ship news information cannot be acquired in time, decision judgment can be deviated. In the related art, the ship news pushing method pushes the ship news according to the preference interests of the user, but it is difficult to push the proper related ship news when the user browses the ship news for the first time. In addition, information such as comments of other users on the ship news and click rate of the ship news cannot be obtained, so that importance of the ship news is difficult to know from a large amount of ship news information, and the ship news pushing efficiency is low. In view of the foregoing, there is a need for solving the technical problems in the related art.
Disclosure of Invention
In view of the above, the embodiment of the invention provides a training method, a pushing method and a system for a ship news pushing model, so as to improve the ship news pushing efficiency.
In one aspect, the invention provides a training method of a ship news push model, which comprises the following steps:
acquiring a ship news data set;
carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts;
and inputting the training data set into a ship news pushing model for training treatment to obtain a trained ship news pushing model.
Optionally, the acquiring the ship news data set includes:
carrying out news crawling processing on a ship website to obtain a website news data set;
and denoising and sampling the website news data set to obtain the ship news data set.
Optionally, the performing news text score calculation on the ship news data set to obtain a training data set includes: carrying out news text calculation processing on each news text in the ship news data set to obtain a news text score corresponding to the news text, wherein the method comprises the following steps:
performing weight calculation processing on the news text to obtain text weight;
carrying out relevance score calculation processing on the news text to obtain a relevance score;
and carrying out scoring calculation processing on the news text according to a scoring formula by combining the text weight and the relevance score to obtain a news text score.
The training data set is input into a ship news pushing model for training treatment, and a trained ship news pushing model is obtained, which comprises the following steps:
inputting the training data set into a ship news pushing model to obtain a ship news pushing prediction result;
determining a training loss value according to the ship news pushing prediction result and the news text score;
and updating parameters of the ship news pushing model according to the loss value to obtain a trained ship news pushing model.
Optionally, the inputting the training data set into the ship news pushing model to obtain a ship news pushing prediction result includes:
performing corpus preprocessing and splicing processing on the training data set to obtain an embedded layer representation set;
performing convolution operation processing on the embedded layer representation set to obtain a convolution characteristic representation set;
pooling and connecting the convolution characteristic representation set to obtain a characteristic representation vector set;
and carrying out multi-head attention processing and full connection processing on the characteristic expression vector set to obtain a ship news pushing prediction result.
Optionally, the ship news push model includes an embedded layer, a convolution layer, a global average pooling layer, a connection layer, a multi-headed self-attention mechanism layer, and a full connection layer.
On the other hand, the embodiment of the invention also provides a pushing method of the ship news pushing model, which comprises the following steps:
acquiring ship news to be pushed;
and inputting the ship news to be pushed into the ship news pushing model obtained by the training method of the ship news pushing model to obtain a ship news pushing result.
On the other hand, the embodiment of the invention also provides a training system of the ship news push model, which comprises the following steps:
the first module is used for acquiring a ship news data set;
the second module is used for carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts;
and the third module is used for inputting the training data set into the ship news pushing model for training treatment to obtain a trained ship news pushing model.
Optionally, the first module includes:
the first unit is used for carrying out news crawling processing on the ship website to obtain a website news data set;
and the second unit is used for denoising and sampling the website news data set to obtain a ship news data set.
Optionally, the third module includes:
the third unit is used for inputting the training data set into the ship news pushing model to obtain a ship news pushing prediction result;
a fourth unit, configured to determine a training loss value according to the ship news push prediction result and the news text score;
and a fifth unit, configured to update parameters of the ship news pushing model according to the loss value, so as to obtain a trained ship news pushing model.
Compared with the prior art, the technical scheme provided by the invention has the following technical effects: according to the embodiment of the invention, the ship news data set is acquired; carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts; and inputting the training data set into a ship news pushing model for training treatment to obtain a trained ship news pushing model. According to the embodiment of the invention, the difficulty that proper news is difficult to push when the ship news is read for the first time can be solved by using the ship news push model, and news push can be performed based on the importance of the news text, so that the efficiency of ship news push is improved.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a flowchart of a training method of a ship news push model provided in an embodiment of the present application;
fig. 2 is a schematic structural diagram of a ship news pushing model according to an embodiment of the present application;
fig. 3 is a frame diagram of a news recommending method in the related art.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
In the related art, a news recommendation system generally consists of a technology of filtering information and generating recommendations, and the recommendation system is generally classified into collaborative filtering, content-based and hybrid methods according to an underlying algorithm. The collaborative filtering system recommends items that the user liked in the past that are similar to the current user preferences. In content-based algorithms, the recommendation depends only on the user's past scoring of the items, which means that the recommended items will have similar characteristics to items that the current user liked in the past. The hybrid model incorporates one or more recommendation methods to alleviate the weaknesses of a single technology. But the above method makes it difficult to push proper news when the user browses news for the first time. Today there are a lot of news releases each day and articles are updated continuously. Such large amounts of data, distributed over a short period of time, combined with unstructured formats of news articles, require more complex analysis and more computationally intensive computations. Furthermore, over time, the interests of the user may also change as individuals exhibit short-term and long-term preferences. Another related challenge is that users rarely provide explicit feedback on likes and ratings.
In view of this, the embodiment of the present application provides a training method for a ship news push model, which may be applied to a terminal, a server, software running in a terminal or a server, and the like. The terminal may be, but is not limited to, a tablet computer, a notebook computer, a desktop computer, etc. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDNs, basic cloud computing services such as big data and artificial intelligent platforms. Referring to fig. 1, the method mainly includes the steps of:
s101, acquiring a ship news data set;
s102, carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts;
and S103, inputting the training data set into a ship news pushing model for training treatment, and obtaining a trained ship news pushing model.
In the embodiment of the invention, in order to solve the problem that news pushed to a user is difficult to accurately according to news content in the related technology, a ship news push model is provided for realizing ship news recommendation. And training the ship news pushing model by establishing a ship news pushing module, and carrying out ship news pushing by the ship news pushing model after training. According to the embodiment of the invention, a ship news data set is firstly required to be acquired, the importance of news texts in the data set is calculated for the ship news data set, news text score calculation processing is carried out, a training data set is obtained, and the training data set consists of a plurality of news texts in the ship news data set and news text scores corresponding to the news texts. Wherein the news text score corresponding to the news text corresponds to the label of the news text. And finally, inputting the training data set into a ship news pushing model for training treatment to obtain a trained ship news pushing model.
Further as a preferred embodiment, the acquiring a ship news data set includes:
carrying out news crawling processing on a ship website to obtain a website news data set;
and denoising and sampling the website news data set to obtain the ship news data set.
In the embodiment of the invention, the acquisition method of the data set can be acquired by a manual method, and news on a website can be acquired by a web crawler method. According to the embodiment of the invention, news of a plurality of ship websites is collected and used as a data set of a ship news pushing model, news crawling processing is carried out on the plurality of ship websites, denoising and sampling processing are carried out on the crawled data to obtain a ship news data set, the denoising method can adopt methods such as mean value filtering and singular value decomposition, and the sampling method can adopt random extraction and the like.
Further as a preferred embodiment, the performing news text score calculation processing on the ship news data set to obtain a training data set includes: carrying out news text calculation processing on each news text in the ship news data set to obtain a news text score corresponding to the news text, wherein the method comprises the following steps:
performing weight calculation processing on the news text to obtain text weight;
carrying out relevance score calculation processing on the news text to obtain a relevance score;
and carrying out scoring calculation processing on the news text according to a scoring formula by combining the text weight and the relevance score to obtain a news text score.
In the embodiment of the invention, in the process of carrying out news text score calculation processing on the ship news data set to obtain the training data set, the news text score of each news text in the ship news data set needs to be calculated, wherein the method comprises the following steps: firstly, carrying out weight calculation processing on news texts to obtain text weights, wherein the weight calculation is to calculate the inverse document probability (IDF) of the news texts, which is a measure of the general importance of words. The weight calculation mode is as follows:
where N represents the number of all news texts and N (qi) represents the number of words qi contained in each news text.
And secondly, calculating the relevance score of the news text to obtain the relevance score, wherein the calculation method of the relevance score is as follows:
wherein k is 1 ,k 2 B represents a regulatory factor. f (f) i Representation word q i Frequency of occurrence in document d, qf i Representation word q i Frequency in sentences. dl denotes the length of the news text d, and avgdl is the average length of all documents in all the news texts d.
The third step is to calculate the Score of the news text d according to the scoring formula, the scoring formula Score (Q, d) is as follows:
wherein W is i Represents text weight, n represents total number of news text, R (q i D) represents a relevance score.
The news text score calculated by the embodiment of the invention can be approximately equal to the importance of the news, and the news text in the training data set can be marked by calculating the importance of the news so as to facilitate the training of a subsequent ship news push model.
Further as a preferred embodiment, the training data set is input into a ship news pushing model for training, to obtain a trained ship news pushing model, which includes:
inputting the training data set into a ship news pushing model to obtain a ship news pushing prediction result;
determining a training loss value according to the ship news pushing prediction result and the news text score;
and updating parameters of the ship news pushing model according to the loss value to obtain a trained ship news pushing model.
In the embodiment of the invention, after the training data set is obtained after the news text score is calculated on the news text, the training data set can be input into the initialized ship news push model for training. Specifically, after data in the training data set is input into the initialized ship news pushing model, a pushing result output by the model, namely, a ship news pushing prediction result, can be obtained, and the accuracy of model prediction can be evaluated and identified according to the ship news pushing prediction result and the label, namely, the news text score, so that parameters of the model are updated. For the ship news push model, the accuracy of the model prediction result can be measured by a Loss Function (Loss Function), wherein the Loss Function is defined on single training data and is used for measuring the prediction error of one training data, and particularly determining the Loss value of the training data through the label of the single training data and the prediction result of the model on the training data. In actual training, one training data set has a lot of training data, so that a Cost Function (Cost Function) is generally adopted to measure the overall error of the training data set, and the Cost Function is defined on the whole training data set and is used for calculating the average value of the prediction errors of all the training data, so that the prediction effect of the model can be better measured. For a general machine learning model, based on the cost function, a regular term for measuring the complexity of the model can be used as a training objective function, and based on the objective function, the loss value of the whole training data set can be obtained. There are many kinds of common loss functions, such as 0-1 loss function, square loss function, absolute loss function, logarithmic loss function, cross entropy loss function, etc., which can be used as the loss function of the machine learning model, and will not be described in detail herein. In the embodiment of the application, one loss function can be selected to determine the loss value of training. Based on the trained loss value, updating the parameters of the model by adopting a back propagation algorithm, and iterating for several rounds to obtain the trained ship news push model. Specifically, the number of iteration rounds may be preset, or training may be considered complete when the test set meets the accuracy requirements.
Further as a preferred embodiment, the inputting the training data set into the ship news pushing model to obtain a ship news pushing prediction result includes:
performing corpus preprocessing and splicing processing on the training data set to obtain an embedded layer representation set;
performing convolution operation processing on the embedded layer representation set to obtain a convolution characteristic representation set;
pooling and connecting the convolution characteristic representation set to obtain a characteristic representation vector set;
and carrying out multi-head attention processing and full connection processing on the characteristic expression vector set to obtain a ship news pushing prediction result.
In the embodiment of the invention, a ship news push model comprises an embedding layer, a convolution layer, a global average pooling layer, a connecting layer, a multi-head self-attention mechanism layer and a full connecting layer, wherein the embedding layer carries out corpus preprocessing on news texts to obtain word vector embedded representations, part-of-speech vector embedded representations and entity vector embedded representations, and then the word vector embedded representations, the part-of-speech vector embedded representations and the entity vector embedded representations are spliced to obtain embedded layer representations; the convolution layer is to input the representation of the embedded layer into a convolution neural network, and obtain a convolution characteristic representation through convolution operation; the global average pooling layer pools the convolution characteristic representation and then obtains a characteristic representation vector after passing through the connecting layer; the multi-head self-attention mechanism layer enhances the characterization of news by capturing the interaction of the news; and finally, obtaining output through the full connection layer. Wherein the input of the embedded layer is the entire news text. The news text will be converted into a matrix representation at the embedded layer. After the matrix is obtained, the matrix is input into a convolution layer for convolution operation, an activation function used by the convolution layer in the embodiment of the invention is a ReLu function, and the output of the convolution layer is subjected to Pooling operation through a global average Pooling layer. And then a combined matrix is obtained after passing through the connecting layer. Taking the output matrix as the input of a multi-head self-attention mechanism layer, and expressing the ith news learned by the kth attention head in the multi-head self-attention mechanism layer as follows:
in the method, in the process of the invention,and->Is a parameter of the kth news self-attention header, M is the number of candidate news, e i And e j Representing the output of the ith and jth news from the link layer, respectively,/for>Representing the relative importance of the ith and jth news interactions. The multiple head representation of the ith news is to stitch the outputs of the h independent self-attention representations, i.e
In the embodiment of the invention, the output ship news push prediction result is finally generated through full connection, and is used for measuring the ship news push prediction result y and the true value according to the Mean Absolute Error (MAE)Errors between them. The Mean Absolute Error (MAE) calculation method is as follows:
where n is the number of news texts.
According to the ship news pushing model based on the multi-head self-attention mechanism, which is provided by the embodiment of the invention, the ship news can be pushed to a user for reading according to the importance of the news text, and the model is more accurately predicted by using the convolution layer and the multi-head self-attention mechanism layer, so that the user can accurately grasp the development dynamics of the ship industry, and convenience is brought to the user.
Further as a preferred embodiment, the ship news push model includes an embedded layer, a convolution layer, a global average pooling layer, a connection layer, a multi-headed self-attention mechanism layer, and a fully connected layer.
Referring to fig. 2, fig. 2 is a schematic structural diagram of a ship news push model provided in an embodiment of the present application, where in the embodiment of the present invention, a news text is subjected to corpus preprocessing by an embedding layer to obtain a word vector embedded representation, a part-of-speech vector embedded representation and an entity vector embedded representation, and then the word vector embedded representation, the part-of-speech vector embedded representation and the entity vector embedded representation are spliced to obtain an embedding layer representation; the convolution layer is to input the representation of the embedded layer into a convolution neural network, and obtain a convolution characteristic representation through convolution operation; the global average pooling layer pools the convolution characteristic representation and then obtains a characteristic representation vector after passing through the connecting layer; the multi-head self-attention mechanism layer enhances the characterization of news by capturing the interaction of the news; and finally obtaining an output result through the full connection layer.
On the other hand, the embodiment of the invention also provides a pushing method of the ship news pushing model, which comprises the following steps:
acquiring ship news to be pushed;
and inputting the ship news to be pushed into the ship news pushing model obtained by the training method of the ship news pushing model to obtain a ship news pushing result.
It can be understood that the content in the training method embodiment of the ship news push model is applicable to the pushing method embodiment of the ship news push model, the functions specifically realized by the pushing method embodiment of the ship news push model are the same as those of the training method embodiment of the ship news push model, and the beneficial effects achieved by the pushing method embodiment of the ship news push model are the same as those achieved by the training method embodiment of the ship news push model.
On the other hand, the embodiment of the invention also provides a training system of the ship news push model, which comprises the following steps:
the first module is used for acquiring a ship news data set;
the second module is used for carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts;
and the third module is used for inputting the training data set into the ship news pushing model for training treatment to obtain a trained ship news pushing model.
Further as a preferred embodiment, the first module includes:
the first unit is used for carrying out news crawling processing on the ship website to obtain a website news data set;
and the second unit is used for denoising and sampling the website news data set to obtain a ship news data set.
Further as a preferred embodiment, the third module includes:
the third unit is used for inputting the training data set into the ship news pushing model to obtain a ship news pushing prediction result;
a fourth unit, configured to determine a training loss value according to the ship news push prediction result and the news text score;
and a fifth unit, configured to update parameters of the ship news pushing model according to the loss value, so as to obtain a trained ship news pushing model.
It can be understood that the content in the training method embodiment of the ship news push model is applicable to the training system embodiment of the ship news push model, the specific functions of the training system embodiment of the ship news push model are the same as those of the training method embodiment of the ship news push model, and the beneficial effects achieved by the training system embodiment of the ship news push model are the same as those achieved by the training method embodiment of the ship news push model.
In the related art, a general framework of the news recommending method is generally shown in fig. 3. The core components of the related art news recommendation method include a news encoder, a user encoder, and a click prediction module. Wherein the news encoder is operative to learn news embeddings from text, the user encoder is operative to learn user embeddings from click news embeddings, and the click prediction module is operative to calculate a personalized click score for the news rank based on a correlation between the user embeddings and candidate news embeddings. Suppose the user has T historical click news, denoted as [ D ] 1 ,D 2 ,…,D T ]. News encoder provides these click news and each candidate news D to the user c Processing to obtain their embeddings, denoted as [ h ] 1 ,h 2 ,…h T ]. The user encoder takes as input the embedded sequence of these clicked news and outputs the user's embedded u that summarizes the information of interest to the user. The click prediction module embeds u and h by the user c For input, a click score is calculated from the correlation. Related news is pushed by the score. The news pushing method in the related art mostly pushes news according to interest preference of a user, but it is difficult to push proper related news when the user browses news for the first time.
In summary, the embodiment of the invention has the following advantages: compared with other models related to news recommendation, the embodiment of the invention can predict the importance ranking of the news text in a plurality of news according to the content of the news text without depending on the past reading interest preference of the user, the click rate, comments and the like of the news, and can play a good role in the field with stronger professionals such as the ship industry.
In some alternative embodiments, the functions/acts noted in the block diagrams may occur out of the order noted in the operational illustrations. For example, two blocks shown in succession may in fact be executed substantially concurrently or the blocks may sometimes be executed in the reverse order, depending upon the functionality/acts involved. Furthermore, the embodiments presented and described in the flowcharts of the present invention are provided by way of example in order to provide a more thorough understanding of the technology. The disclosed methods are not limited to the operations and logic flows presented herein. Alternative embodiments are contemplated in which the order of various operations is changed, and in which sub-operations described as part of a larger operation are performed independently.
Furthermore, while the invention is described in the context of functional modules, it should be appreciated that, unless otherwise indicated, one or more of the described functions and/or features may be integrated in a single physical device and/or software module or one or more functions and/or features may be implemented in separate physical devices or software modules. It will also be appreciated that a detailed discussion of the actual implementation of each module is not necessary to an understanding of the present invention. Rather, the actual implementation of the various functional modules in the apparatus disclosed herein will be apparent to those skilled in the art from consideration of their attributes, functions and internal relationships. Accordingly, one of ordinary skill in the art can implement the invention as set forth in the claims without undue experimentation. It is also to be understood that the specific concepts disclosed are merely illustrative and are not intended to be limiting upon the scope of the invention, which is to be defined in the appended claims and their full scope of equivalents.
Logic and/or steps represented in the flowcharts or otherwise described herein, e.g., a ordered listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. For the purposes of this description, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that: many changes, modifications, substitutions and variations may be made to the embodiments without departing from the spirit and principles of the invention, the scope of which is defined by the claims and their equivalents.
While the preferred embodiment of the present invention has been described in detail, the present invention is not limited to the embodiments described above, and those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of the present invention, and these equivalent modifications or substitutions are included in the scope of the present invention as defined in the appended claims.
Claims (10)
1. A training method for a ship news push model, the method comprising:
acquiring a ship news data set;
carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts;
and inputting the training data set into a ship news pushing model for training treatment to obtain a trained ship news pushing model.
2. The method of claim 1, wherein the acquiring a ship news dataset comprises:
carrying out news crawling processing on a ship website to obtain a website news data set;
and denoising and sampling the website news data set to obtain the ship news data set.
3. The method of claim 1, wherein the performing news text score calculation on the ship news data set to obtain a training data set comprises: carrying out news text calculation processing on each news text in the ship news data set to obtain a news text score corresponding to the news text, wherein the method comprises the following steps:
performing weight calculation processing on the news text to obtain text weight;
carrying out relevance score calculation processing on the news text to obtain a relevance score;
and carrying out scoring calculation processing on the news text according to a scoring formula by combining the text weight and the relevance score to obtain a news text score.
4. The method according to claim 1, wherein the training data set input to the ship news push model to obtain a trained ship news push model comprises:
inputting the training data set into a ship news pushing model to obtain a ship news pushing prediction result;
determining a training loss value according to the ship news pushing prediction result and the news text score;
and updating parameters of the ship news pushing model according to the loss value to obtain a trained ship news pushing model.
5. The method of claim 4, wherein inputting the training data set into a ship news push model to obtain a ship news push prediction result comprises:
performing corpus preprocessing and splicing processing on the training data set to obtain an embedded layer representation set;
performing convolution operation processing on the embedded layer representation set to obtain a convolution characteristic representation set;
pooling and connecting the convolution characteristic representation set to obtain a characteristic representation vector set;
and carrying out multi-head attention processing and full connection processing on the characteristic expression vector set to obtain a ship news pushing prediction result.
6. The method of any of claims 1 to 5, wherein the marine news push model comprises an embedded layer, a convolution layer, a global averaging pooling layer, a connection layer, a multi-headed self-attention mechanism layer, and a fully connected layer.
7. A pushing method of a ship news pushing model, the method comprising:
acquiring ship news to be pushed;
inputting the ship news to be pushed into the ship news pushing model obtained by the training method of the ship news pushing model according to any one of claims 1-6, and obtaining a ship news pushing result.
8. A training system for a marine news push model, the training system comprising:
the first module is used for acquiring a ship news data set;
the second module is used for carrying out news text score calculation processing on the ship news data set to obtain a training data set, wherein the training data set consists of a plurality of news texts and news text scores corresponding to the news texts;
and the third module is used for inputting the training data set into the ship news pushing model for training treatment to obtain a trained ship news pushing model.
9. The training system of claim 8, wherein the first module comprises:
the first unit is used for carrying out news crawling processing on the ship website to obtain a website news data set;
and the second unit is used for denoising and sampling the website news data set to obtain a ship news data set.
10. The training system of claim 8, wherein the third module comprises:
the third unit is used for inputting the training data set into the ship news pushing model to obtain a ship news pushing prediction result;
a fourth unit, configured to determine a training loss value according to the ship news push prediction result and the news text score;
and a fifth unit, configured to update parameters of the ship news pushing model according to the loss value, so as to obtain a trained ship news pushing model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310133118.0A CN116127321A (en) | 2023-02-16 | 2023-02-16 | Training method, pushing method and system for ship news pushing model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202310133118.0A CN116127321A (en) | 2023-02-16 | 2023-02-16 | Training method, pushing method and system for ship news pushing model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN116127321A true CN116127321A (en) | 2023-05-16 |
Family
ID=86304527
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202310133118.0A Pending CN116127321A (en) | 2023-02-16 | 2023-02-16 | Training method, pushing method and system for ship news pushing model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN116127321A (en) |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019105432A1 (en) * | 2017-11-29 | 2019-06-06 | 腾讯科技(深圳)有限公司 | Text recommendation method and apparatus, and electronic device |
CN110413863A (en) * | 2019-08-01 | 2019-11-05 | 信雅达系统工程股份有限公司 | A kind of public sentiment news duplicate removal and method for pushing based on deep learning |
US20200089711A1 (en) * | 2018-09-17 | 2020-03-19 | Yandex Europe Ag | Method and system for generating push notifications related to digital news |
CN111460817A (en) * | 2020-03-30 | 2020-07-28 | 中南大学 | Method and system for recommending criminal legal document related law provision |
CN112712901A (en) * | 2021-01-14 | 2021-04-27 | 西京学院 | Grammar quantum based long-short time memory model and medicine interaction extraction method |
CN114154503A (en) * | 2021-12-02 | 2022-03-08 | 四川启睿克科技有限公司 | Sensitive data type identification method |
CN114298053A (en) * | 2022-03-10 | 2022-04-08 | 中国科学院自动化研究所 | Event joint extraction system based on feature and attention mechanism fusion |
CN114706972A (en) * | 2022-03-21 | 2022-07-05 | 北京理工大学 | Unsupervised scientific and technical information abstract automatic generation method based on multi-sentence compression |
CN115146175A (en) * | 2022-07-26 | 2022-10-04 | 中国长江三峡集团有限公司 | Optimization method and device for news recommendation algorithm |
-
2023
- 2023-02-16 CN CN202310133118.0A patent/CN116127321A/en active Pending
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2019105432A1 (en) * | 2017-11-29 | 2019-06-06 | 腾讯科技(深圳)有限公司 | Text recommendation method and apparatus, and electronic device |
US20200089711A1 (en) * | 2018-09-17 | 2020-03-19 | Yandex Europe Ag | Method and system for generating push notifications related to digital news |
CN110413863A (en) * | 2019-08-01 | 2019-11-05 | 信雅达系统工程股份有限公司 | A kind of public sentiment news duplicate removal and method for pushing based on deep learning |
CN111460817A (en) * | 2020-03-30 | 2020-07-28 | 中南大学 | Method and system for recommending criminal legal document related law provision |
CN112712901A (en) * | 2021-01-14 | 2021-04-27 | 西京学院 | Grammar quantum based long-short time memory model and medicine interaction extraction method |
CN114154503A (en) * | 2021-12-02 | 2022-03-08 | 四川启睿克科技有限公司 | Sensitive data type identification method |
CN114298053A (en) * | 2022-03-10 | 2022-04-08 | 中国科学院自动化研究所 | Event joint extraction system based on feature and attention mechanism fusion |
CN114706972A (en) * | 2022-03-21 | 2022-07-05 | 北京理工大学 | Unsupervised scientific and technical information abstract automatic generation method based on multi-sentence compression |
CN115146175A (en) * | 2022-07-26 | 2022-10-04 | 中国长江三峡集团有限公司 | Optimization method and device for news recommendation algorithm |
Non-Patent Citations (1)
Title |
---|
曹玉婵 著: "Python与神经网络实战研究", 30 June 2022, 青岛:中国海洋大学出版社, pages: 182 - 184 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ghasemi et al. | Neural text similarity of user reviews for improving collaborative filtering recommender systems | |
CN107341145B (en) | A kind of user feeling analysis method based on deep learning | |
CN112231569B (en) | News recommendation method, device, computer equipment and storage medium | |
US20130204833A1 (en) | Personalized recommendation of user comments | |
US9141966B2 (en) | Opinion aggregation system | |
Li et al. | Deep cross-platform product matching in e-commerce | |
US20090281975A1 (en) | Recommending similar content identified with a neural network | |
CN110348968A (en) | A kind of recommender system and method analyzed based on user and project coupled relation | |
CN113343125A (en) | Academic-precision-recommendation-oriented heterogeneous scientific research information integration method and system | |
CN112948676A (en) | Training method of text feature extraction model, and text recommendation method and device | |
Wu et al. | Optimization matrix factorization recommendation algorithm based on rating centrality | |
CN114385930A (en) | Interest point recommendation method and system | |
Liu et al. | A multi-task dual attention deep recommendation model using ratings and review helpfulness | |
CN109086463A (en) | A kind of Ask-Answer Community label recommendation method based on region convolutional neural networks | |
Wang et al. | Using semantic relations for content-based recommender systems in cultural heritage | |
Yin et al. | An efficient recommendation algorithm based on heterogeneous information network | |
CN114117233A (en) | Conversation news recommendation method and system based on user implicit feedback | |
Kharrat et al. | Recommendation system based contextual analysis of Facebook comment | |
Ng et al. | Personalized book recommendation based on a deep learning model and metadata | |
CN116127321A (en) | Training method, pushing method and system for ship news pushing model | |
Zeng et al. | RACMF: robust attention convolutional matrix factorization for rating prediction | |
Iftikhar et al. | Amazon products reviews classification based on machine learning, deep learning methods and BERT | |
Wang et al. | Predicting best answerers for new questions: An approach leveraging convolution neural networks in community question answering | |
Alatrash et al. | Collaborative filtering integrated fine-grained sentiment for hybrid recommender system | |
Sowbhagya et al. | User profiling for web personalization |
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 |