39 results about "Web news" patented technology

The invention discloses a microblog hot topic detection method based on multi-information fusion. Firstly, influences of a blogger is calculated according to social relations of a microblog promulgator, and a sum of weights of a feature word in all microblogs in a given time period is further calculated based on influence information of the blogger and frequency information of the feature word; then a sudden feature word is detected according to information of the sum of the weight of the feature word along with time changes, microblog data are expanded by introducing Web news linguistic data to calculate an incidence relation valve between sudden features, and an association graph of the sudden feature word can be further constructed; finally, the association graph of the sudden feature word is divided, each strong connected subgraph indicates a topic, and accordingly detection on a microblog hot topic can be achieved. The microblog hot topic detection method comprehensively uses microblog feature word information, blogger social relation information and relative Web news document information to detect the microblog hot topic, and improves detection efficiency of the microblog hot topic.

The invention provides a tool for quantitatively analyzing internet news influence based on information retrieval and a method thereof. The tool comprises a webpage content duplication determine module, a relevant information extracting module, a news reprint rate calculating module, a newsletter source website influence determine module, a news recovery rate calculating module and a news influence calculating module, wherein, the webpage content duplication determine module determines whether a webpage is an approximate webpage or not; the relevant information extracting module extracts relevant information in the webpage; the news reprint rate calculating module reprints a website technorati authority value; a news source website and the news reprint rate are found out; the newsletter source website influence determine module determines the popularity index of the news source website to acquire a newsletter source website influence factor; the news response rate calculating module determines a network news recovery rate; and the news influence calculating module calculates the news influence value. The method comprises the following steps: determining that whether the webpage is the approximate webpage or not; extracting news webpage reprint information and recovery information; calculating the news reprint rate; calculating the recovery rate of the news; calculating the technorati authority of the news source website; calculating the influence factor of the news source website; and calculating the news influence. The quantitative calculating result is combined with qualitative analysis of a user to help the user determine the news influence.

An embodiment of the invention provides a system and method for managing computing resources in a cloud computing environment, wherein the system includes a data analysis module connected to a processor. The data analysis module predicts the demand for the computing resources, wherein the demand is predicted based on internet user activity. The internet user activity includes: the number of web pages that include at least one select term a predetermined amount of times, the number of times the web pages that include the at least one select term a predetermined amount of times are viewed, and the number of times the select term is entered into an internet search engine. The processor increases resource capacity in the cloud computing environment when the predicted demand is above a first threshold. The processor decreases resource capacity in the cloud computing environment when the predicted demand is below a second threshold.

The invention discloses a negative news recognition method based on emotion calculation and a multi-head attention mechanism, relates to the technical field of network public opinion monitoring, and solves the technical problem of how to solve the problem that objective negative news is difficult to identify, and the negative news recognition method comprises the following specific steps: (1) collecting and preprocessing network news text data; (2) establishing and expanding a negative emotion seed word bank and calculating the susceptibility; (3) performing vectorization representation, and determining the input of a discrimination model; (4) establishing a negative news discrimination model; and (5) carrying out negative news recognition. According to the method, the problem that the negative news is difficult to identify is effectively solved, and a good effect is achieved in the aspects of recognition accuracy and effectiveness of the negative news text.

The invention relates to an online Web news content extraction method. The method comprises the following steps: obtaining a DOM tree of an extracted Web news webpage; traversing the DOM tree to construct a text node information sequence and a label path information sequence; calculating a label path characteristic value sequence; fusing the label path characteristic value sequence by a weighted DS evidence theory to obtain a label path comprehensive characteristic value sequence; constructing a text node comprehensive characteristic value sequence; and extracting the text content of the Web news webpage according to the text node comprehensive characteristic value sequence. The invention furthermore discloses an online Web news content extraction system which consists of an analysis module, a calculation module, a fusion module and an extraction module. The label path characteristics are not based on templates of the webpages and have diversity, and the whole extraction process only comprises simple mathematical elements, so that massive heterogeneous Web news webpages can be effectively extracted online.

The invention discloses an online public opinion monitoring and early warning method which comprises the following steps: directionally collecting data of online news, forums and social media disclosed in the Internet by utilizing an online public opinion data acquisition module; cleaning, converting and processing the collected data through a data processing module, and converting the unstructured data into semi-structured or structured data; carrying out natural language processing on the processed data and carrying out data mining by using an artificial intelligence technology through a network public opinion data analysis module to discover and identify public opinion hotspots, sensitive topics and/or risk topics; visually displaying the public opinion monitoring analysis result through a visualization module, and outputting a public opinion analysis result chart and/or a public opinion analysis report; sending early warning information to a user through a network public opinion real-time early warning module according to an early warning rule set by the user. The accuracy of public opinion monitoring can be improved.

The invention discloses an automatic prediction method of network news expression distribution. The network news expression distribution indicates that multiple network users vote different news expressions to form the news expression distribution after browsing a piece of network news. The invention comprises the following steps: collecting network news data comprising network news samples and news expression user vote data of each sample; utilizing the collected data to carry out keyword selection and extracting the characteristics of each new sample; utilizing a probability label propagation algorithm to modify the distribution of the vote data; and automatically predicting the news expression distribution of a new network news sample by utilizing a machine learning algorithm based on instances. The automatic prediction method of the invention can be applied to predict the news expression analysis of network news and the news expression distribution of a new network news sample.

The invention discloses a social network false news detection method and device. The method comprises the following steps: obtaining character elements and image elements of social network news data; carrying out text feature extraction on the preprocessed news character elements by adopting a Bi-LSTM neural network model; performing visual feature extraction on the news image elements by adopting a VGG-19 neural network model, and performing post-processing; performing mutual fusion on the text features and the image features by using a dot product type attention mechanism to sequentially obtain corrected text features and corrected visual features; performing time-level self-attention fusion on the corrected text features at different moments to obtain ultimate text features; carrying out space-level self-attention fusion on the corrected visual features of different positions to obtain ultimate visual features; and after the ultimate text features and the ultimate visual features are spliced, a full connection layer neural network model with softmax is input, so that false news can be detected in time.

The invention discloses a text generation method and device based on dynamic weights and a storage medium, and the method comprises the steps of constructing a hot word library which stores hot search words obtained every a first preset time period and a hot search weight of each hot search word, and the hot search weight of each hot search word is generated according to the frequency parameter of each hot search word and the construction time of each hot search word; receiving any input text; extracting a plurality of subject terms from any text; matching the subject terms with a hot word library to obtain a hot search weight of each subject term; and generating an updated text according to the subject term containing the hot search weight. According to the dynamically weighted news event multi-topic automatic generation method, the corresponding weight is dynamically given to the hot word according to the hot word constructed by incremental automatic search in combination with the frequency information of the word and the hot search time information, and the invention is used for processing quickly updated real-time network news data.

The invention relates to a news-oriented Chinese food safety news event information automatic extraction method comprising the following steps: S1, collecting network news to construct a food news corpus; arranging the domestic geographic information knowledge base and constructing the geographic information knowledge base; collecting the food category information from the network and constructingthe food category knowledge base; S2, performing text classifying of the news corpus of the food news corpus by using the text classification model so as to obtain the food safety news corpus; S3 firstly extracting the event occurrence time in the food safety news corpus; secondly, extracting the event occurrence place in the food safety news corpus by utilizing the geographic information knowledge base; finally, extracting the food name and the category information involved in the food safety news corpus by utilizing the food category knowledge base; S4, counting the extracted time, place and food information; and S5, displaying the time, place and food statistical information by using the visualization technology. The information of three dimensions of the time, the place and the food category of the food safety event can be accurately extracted from the network news, statistics on the information can be performed and finally the statistical result can be visually displayed.

The invention discloses a method and system for automatically cleaning network news data based on webpage label distribution characteristics, and the method comprises the steps of crawling the networknews data through an offline crawler system, namely carrying out tree node analysis on the crawled offline news data, and extracting the attribute information, such as label name, attribute, text, link, etc., in a node; predicting the information of other node blocks through a current node by adopting an idea based on n-gram2vec, obtaining the word embedding information of the label through training, constructing an intelligent model discrimination system based on the pre-trained word embedding information, and determining the leaving of tiled nodes, namely dividing an intelligent model intoa text discrimination model and a picture discrimination model according to the type of an article label, training the two types of models by adopting different feature engineering, finally predicting, and combining the results of the two types of models according to the previous node sequence.

The invention discloses a Web news retrieval method based on event analysis. The Web news retrieval method based on the event analysis comprises the following steps of: setting two types of input frames of an event term Qe and a bound term Qc, and obtaining a query term Q={Qe, Qc}; selecting the title (T), the first paragraph (FP) and the last paragraph (LP) of Web news di to represent di, wherein di={T, FP, LP}; calculating the weight of each part of characteristic term of the di={T, FP, LP}; setting the weight of the characteristic term in the query term to be one; calculating the relevancy R(Q, di) of the query term Q and the news di; and according to the relevancy R(Q, di), carrying out descending order, and outputting a retrieval result. According to the method, the functions of different terms in a query content are distinguished, a computation method for the relevancy of the event query term and the Web news is proposed by combining with an event movement element, the importance of the Web news title and distance between the event term and the bound term, and the retrieval accuracy of the event-type Web news can be obviously improved.

The invention relates to the technical field of artificial intelligence and text analysis, in particular to a news public opinion text processing method applying artificial intelligence, a server and a medium. The method can map a target network news text and each news reference text to the same standard specification for comparative analysis, determines fragment set correlation between a first standard network news text set and each second standard network news text set according to a correlation coefficient between standard public opinion text fragments stored in advance, so as to obtain news text correlation between the target network news text and each news reference text, therefore, the process of determining the news text relevancy between the target network news text and each news reference text is effectively simplified, the news public opinion text analysis efficiency is improved, and server crash caused by processing a large number of news public opinion texts in some time periods is avoided.

The invention relates to a multi-scale visualization method for news, which is used for realizing map-oriented network news multi-scale expression and comprises the following steps: 1) crawling different news events within a certain time period, acquiring news text contents, titles and publishing time, and storing news sources; 2) geocoding the crawled news to obtain a news place and corresponding latitude and longitude coordinates; 3) performing scale division on the news according to place names involved in the news; 4) performing span calculation on the news according to the longitude and latitude of the related place in the news; and 5) constructing a map visualization model according to the scale and span of the news, thereby displaying the news with the corresponding scale and span according to different map scaling scales. According to the method, the scale of the news and the scale of the map are combined, multi-scale presentation of the news data is achieved, the browsing dimension of news information is increased, news browsing and reading are more convenient and efficient, and the news multi-scale visualization technology under the space-time dimension is achieved.