45 results about "Financial news" patented technology

A system and method for delivering highly customized, natural-sounding/appearing audio and/or visual content to existing player devices, including but not restricted to wired and wireless voicemail, sound-enabled PCs, and portable MP3 or DVD players. Subscribers register with existing content providers to receive alerts and information on topics they care about (e.g., portfolio updates, financial news, sports). If a user selects the audio and/or visual delivery option, the content provider passes his or her registration and preference information to the system. The content providers then pass news information to the system, which converts it to audio and/or visuals in one of two ways. For short, formulaic messages, the system concatenates spoken phrases and clauses previously recorded by human talent and stored in a multimedia library database, to create natural-seeming audio and/or visual sequences. For longer messages, the system uses human abilities entirely—i.e., a human reader records a complete text and inputs audio and/or visual files to the system. Turnaround time in both cases is minimal, and quality is high. The system then organizes customized audio and/or visual news deliveries in accordance with user preference information and customizable playlist rules, which order the selected news information by vertical, subject, paragraph, sentence, or other dimension(s). Customized news packages are then delivered as audio and/or visuals to the listeners' player devices.

The invention discloses a Chinese financial news text classification method based on a convolutional neural network. The method is mainly divided into four parts including word vector training, text preprocessing, neural network model training and news classification. Large-scale financial news corpora are used for training to obtain a broadly common financial-category word vector model through anunsupervised learning method, the word vector is effectively imported into the training of the convolutional neural network, and the statistical information of the model is increased through a methodof dynamically regulating the word vector. The adopted convolutional neural network has a simple structure, and can show excellent performances by aiming at a small sample set, so that a Chinese financial news text classification problem can be effectively solved, and effectiveness of the convolutional neural network in processing a text classification problem can be fully proved.

A method of creating a price prediction model that forecasts short-term price fluctuations in financial instruments by collecting, analyzing and classifying financial news for a financial instrument into categories. Distributions for the changes in price of the financial instrument for a set period of time and distributions for the changes in price of the financial instrument as a result of the financial news for each news category for a set period of time are then obtained. If the distributions for the changes in price of the financial instrument are statistically significantly different than the distributions for the changes in price of the financial instrument for a particular news category, and the mean for the change in price is greater or less than zero, a signal is produced indicating the trading action that should be taken for the financial instrument.

The invention provides a stock price fluctuation prediction method of a convolutional neural network combining financial news. According to the method, natural-language processing technology is utilized to extract features in relevant news, and thus an association degree of the financial news and a stock price trend is analyzed and observed. In combination with valid information of news reports, the invention provides the stock price fluctuation prediction method based on the convolutional neural network. Firstly, word segmentation is carried out on the news, main events are extracted, top 3000 words appearing most frequently are used as keywords, and a Glove model is used to indicate the same as low-dimensional dense word vectors; then news features are correlated with stock prices, timeis divided into a short time period, a medium time period and a long time period, the convolutional neural network is used to simulate short-term and long-term effects of the news events on stock price changes; and finally, up and down situations of a stock are predicted through a trained model.

The invention discloses a public opinion information extraction and knowledge base generation method based on natural language processing. The public opinion information extraction and knowledge basegeneration method comprises the following steps of 1, text preprocessing; 2, named entity recognition: identifying company institution names and names, and finishing named entity recognition by adopting a neural network-based method; 3, relationship extraction: extracting six types of relationships in the financial field by adopting a feature layer + GRU + Attention; 4, entity linking; a Jaro winker distance method is adopted, and the distance between a link entity and a target entity is calculated to judge whether the link entity and the target entity are the same entity or not, so that entity disambiguation is achieved. According to the method, an end-to-end model and a feature extraction input class model are combined, a one-stop process from financial unstructured text to structured data storage is constructed, financial news context information is fully utilized, knowledge is extracted with fewer parameters and faster training prediction speed, and good performance is achieved inthe field of financial public opinion information.

The invention discloses a knowledge network building method for financial news. The knowledge network building method for financial news includes that acquiring different financial news data from a news website or a database; using a classification technique to identify the industry label of each acquired news data, using an improved topic model to extract the topic information thereof, building upstream and downstream relationship networks between industries, and creating corresponding news knowledge bases; building corresponding knowledge sub-networks based on the news knowledge bases, wherein each knowledge sub-network comprises four layers of topological structures, to be specific, each knowledge sub-network comprises four kinds of type nodes (news, news cluster, topic cluster and topic) and two kinds of type relationships (inclusion relationship and correlation). Each piece of news would generate the knowledge sub-network thereof, and the knowledge sub-network of each piece of news would be displayed below the news content. The high timeliness of different news can be guaranteed from extracting to displaying based on the built knowledge network; the knowledge network building method for financial news brings better experience to users.

The invention relates to a Chinese semantic structure and finely segmented word bank combination based emotional analysis method. The emotional analysis method comprises: 1) inputting a to-be-tested text at least consisting of a statement into a computer system; 2) performing word segmentation processing on each statement of the to-be-tested text and marking emotional words and other words in each statement; 3) performing matching on the to-be-tested text subjected to word segmentation processing to obtain a semantic mode of each semantic unit; and 4) correspondingly converting the semantic mode of each semantic unit of the to-be-tested text to an emotional value, and performing accumulation on the emotional values of all the semantic units in the text to obtain an emotional value of the to-be-tested text. According to the method, the emotional words, connective words, transitional words and the like are segmented from a non-structured text; according to actual arrangement of the words, sentence patterns are matched to obtain the emotional values of the semantic units; and the emotional value of a sentence is comprehensively calculated according to the emotional values of the semantic units to achieve the purpose of quantizing the emotional value of a financial news comment sentence.

The invention provides a method and a system for constructing a business social network. The method comprises the following steps: identifying input business entities in financial news; identifying the business relationship among the business entities; and constructing a business social network according to the identified business entities and the business relationship among the business entities. In the invention, the mentioned business entities and the business relationship among the business entities can be automatically acquired from large-scale business news so as to construct the business social network, thus rapidly, thoroughly, accurately and intelligently completing construction of the business social network in real time, accurately reflecting the conditions of current business society, providing important basis for the business entities in information analysis and enterprise decisions, and saving much time and expense.

The present invention discloses a calculation method and system for relevance degree of an individual share and an article. The system comprises a data acquisition module, a relevance degree analysis module, an emotion analysis module, a heat analysis module, a data storage module and a data retrieval module. The method comprises: acquiring massive financial news corpus every day; performing text data mining; analyzing a relevance between an individual share and an article among the corpus that is acquired in real time; analyzing individual share emotion in the corpus that is acquired in real time; summarizing the relevance between the individual share and the article, i.e. heat of the individual share daily; enabling an investor to retrieve market condition news of a share that draws the attention of the investor, and providing the relevance degree with the article, and historical heat indicators of the emotion and share for reference, so that the method and system become an excellent retrieval tool of the investor for individual share market condition news.

A mass customizable interactive, multi-faceted system for data collection, processing, analysis, transmission, and trading in securities comprising a multidimensional database, hardware including a network of computers, software including internet browsers and software programs, wherein variables are presented to different users to allow each user to generate personalized financial product portfolio and set filters that control the presentation of information relating to their own financial news, research and/or trading.

The invention relates to the technical field of sentiment dictionary analysis, in particular to a sentiment dictionary-based sentiment analysis method for fine-grained entities in financial news. Themethod comprises the following steps: analyzing a large amount of financial news; obtaining all listed company entity sets of the news to be analyzed on the basis of existing data service-entity identification and extraction of the company; filtering to obtain a listed company sentiment sentence set only containing sentiment words from the listed company sentence set obtained in S2; traversing each word of the sentiment sentence subjected to word segmentation filtered in the step S3, and judging whether the word is a sentiment word or not; performing weighted summation on the emotion scores ofall emotion sentences obtained by each listed company in the step S4; and performing polarity division on the emotion scores. According to the sentiment dictionary-based sentiment analysis method forthe fine-grained entities in the financial news, sentiment analysis and calculation are performed on each listed company in the news by adopting a sentiment dictionary method, sentiment analysis is performed on each listed company involved in the financial news, and the sentiment polarity of each listed company of each article can be obtained.

The invention relates to an emotion analysis method and system for enterprise subjects in financial news, and the method comprises the following steps: S1, collecting news data, carrying out the modeltraining according to the collected news data, and obtaining a classification prediction model; and S2, inputting to-be-classified news data into the classification prediction model, and performing classification prediction on the emotion tags of the enterprise subjects in the to-be-classified news data. The method is designed on the basis of a more advanced text representation model BERT and a memory network model of a double-storage structure, the classification accuracy is higher, meanwhile, domain experts do not need to formulate a rule template to extract additional features, the labor cost is reduced, and maintenance is convenient.

The invention provides a hotspot mining method and system based on Internet financial information, which can help investors to track the development process of market hotspots according to the mined hot topics. The method comprises the following steps: capturing financial news from a network; extracting a feature vector of each piece of news; clustering according to the extracted feature vectors to obtain a plurality of news clusters, wherein each news cluster corresponds to one hot spot; and carrying out importance ranking on all the titles in each news cluster, and obtaining the title with the highest importance to describe the hot topic in the corresponding news cluster. The invention relates to the field of data mining.

The invention relates to a method for analyzing tendentiousness of financial news. The method comprises: identifying a company name, extracting key sentence groups and classifying the key sentence groups by using an LSTM model. The invention provides a method for analyzing tendentiousness of financial news, and the company name is identified by using methods of company name abbreviation dictionaryand encyclopedia query, and the method has good effect and expansibility. A key sentence group extraction method based on deep learning framework doc2vec text similarity matched with comprehensive feature attributes is used, and the method has good extraction effect, high accuracy and recall rate, high accuracy rate of text tendency determination, and good effect, and can preferably meet needs ofpractical application.

The invention relates to an automatic stock selection method based on big data of news. According to the technical scheme, firstly, emotional vocabularies and industry vocabularies which come from a professional Chinese Financial Dictionary are stored in a storage; internet financial news are acquired in real time through RSS and updated every half hour; news text content of the day is analyzed through a server, text analysis of the news content comprises two sub-sections: 1) emotional dimension analysis of the news content: emotional tendency of the news content is obtained through calculation; 2) industry dimension analysis of the news content: industry attention embodied by the news content is obtained through calculation. The stock ranking is calculated by use of the emotional tendency and the industry attention, and a plurality of stocks ranking in the top are selected as investment objects. Vocabularies with emotional dimension indexes and industry attention indexes in the news are filtered and counted, and more information having influence on stock prices is mined at multiple angles.

The invention provides a financial news model training method and a device, a financial news generation method and a device, equipment and a medium, and belongs to the technical field of data processing. The problem that existing financial news is not high in readability is solved. The financial news generation method comprises the following steps: acquiring real-time announcement information, namely acquiring the real-time announcement information by monitoring announcements published by listed companies in a database at the first time; announcement preprocessing, converting the acquired real-time announcement information in the PDF form into a text form, and cleaning unnecessary data in the acquired real-time announcement information; refining announcement information, refining the obtained real-time announcement information into a news form by applying the announcement model trained by the financial news model training method; and news pushing, pushing the news to a client. The method has the advantages of high readability and the like.

The invention relates to an emotion discrimination method based on fine-grained annotation data. The method comprises the following steps: collecting financial news data, dividing news data into a labeled sample set and an unlabeled sample set; training a first classifier and a second classifier through the labeled sample set and the unlabeled sample set; the first classifier can screen out the key sentences in the article; the second classifier judges the emotional tendency of the article; model parameters of the first classifier and model parameters of the second classifier are obtained respectively; and adding the data with high confidence in the classification result into the annotation sample set, selecting the most worthy annotated data C from the unannotated sample set by using an active learning theory, sending the most worthy annotated data C to a worker for annotation, and circularly training the emotion discrimination model until the classification precision is achieved, andending the training to obtain the discrimination model.

The invention discloses a financial news text emotional tendency analysis method based on a graph convolutional network. The method comprises the following steps: determining a data source to obtain financial text data; preprocessing the financial text data to obtain a clean text list; sampling the clean text list to obtain a sample list; carrying out manual labeling on the sample list; establishing a heterogeneous graph by using the clean text list; performing feature extraction on the heterogeneous image to obtain a feature matrix, a label matrix and an adjacent matrix; establishing a four-layer graph convolution network by taking the feature matrix as input, the label matrix as supervision information and the adjacent matrix as a support matrix of graph convolution operation; and obtaining the classification accuracy of the sample list and the classification result of the clean text list through iterative training. According to the method, the unlabeled data is introduced into the heterogeneous graph, learning can be carried out under the condition that no priori word embedding knowledge exists, and the dilemma that an emotion dictionary is difficult to construct and maintain in a web environment and the strong dependence on the proportion of labeled data and the word embedding effect are eliminated.

The invention discloses a data analysis and prediction method and device, and relates to the technical field of big data, and the method comprises the steps: building a financial news data set; performing unsupervised corpus pre-training on the financial news data set by adopting a BERT model to obtain a pre-training model; performing binary classification training with supervised training on the pre-training model by using the financial news data of which the data labels are established according to the return rate of the stock corresponding to each piece of financial news data to obtain a news sentiment classification model; inputting the target financial news data into a news sentiment classification model to obtain a sentiment index; establishing a stock sample data set corresponding to the financial news data set, and constructing an LSTM model for predicting a stock trend; inputting target stock data corresponding to the target financial news data into the LSTM model to obtain a trend prediction value; analyzing and predicting the target stock according to the trend prediction value and the emotion index, thus the efficiency and accuracy of stock data analysis and prediction can be improved, and the user experience is improved.

A system and software-implemented method for reporting financial market news and events. Machine-readable extracted data is reformulated in content and format to provide a more efficient display and understanding of the relevance of news and events to the end user. Audio and/or graphic indicators are added to further promote efficient understanding of the delivered news or event message.

The invention discloses a method for automatically correlating financial news to stocks. The method comprises the steps that annotation data and to-be-processed data are adopted to process original financial news information; the original financial news information is preprocessed, wherein original financial news information data is converted to be in a data form, including a multidimensional vector form, etc., facilitating subsequent deep learning; a multi-tag classification model is adopted to perform deep learning on the preprocessed financial news information; and a trained model is used to process new financial news information, and the financial news information is automatically correlated to stocks. In this way, a word segmentation mode and a deep learning mode can be combined, theoriginal financial news information is automatically correlated to the stocks needing correlation with high accuracy, and it is convenient for a user to acquire comprehensive information of the concerned stocks.

The invention relates to an event extraction method and device in public opinion monitoring in the financial field and computer equipment. The method comprises the steps of obtaining sample data froma financial news text, and preprocessing the sample data to obtain a sample set; obtaining a plurality of preset different event extraction models, and training the event extraction models in a K-foldcross validation mode according to the sample set to obtain K event extraction sub-models of each event extraction model; preprocessing the to-be-extracted text, inputting the preprocessed to-be-extracted text into K event extraction sub-models of each event extraction model, outputting a prediction text ID, a prediction event and a prediction entity corresponding to the to-be-extracted text, andconstructing a triple according to the prediction text ID, the prediction event and the prediction entity; and voting triples output by the K event extraction sub-models of each event extraction model in a voting mode to determine a real triad. By adopting the method, accumulated errors can be reduced.

A stock market prediction method based on the convolutional neural network model comprises the four steps of S1, data collection, S2, data processing, S3, neural network model training and S4, stock market trend prediction. Financial news corpora of listed companies are collected and analyzed; according to the DJIA historical data, the financial news corpus is labeled; and the marked financial corpus information is used as an input value of the convolutional neural network model, so that a predicted value of news marking is output through processing of the convolutional neural network model, and the fluctuation condition and stock market trend of the DJIA can be predicted through the predicted value of the news marking. Through cross validation, the prediction accuracy can reach 65.5%.

The invention discloses a financial news stream emergency detection method based on hierarchical clustering. The method comprises the following steps: preprocessing a text; extracting keywords and constructing a keyword co-occurrence graph; clustering the keywords by adopting a bisection K-Means algorithm, dividing the keyword co-occurrence graph into a plurality of sub-graphs, with the keyword in each sub-graph being a financial theme; identifying the financial theme to which each piece of financial news belongs through similarity calculation; constructing an undirected graph with each piece of financial news as a node, clustering the financial news by adopting the bisection K-Means algorithm, dividing the financial news node undirected graph into a plurality of sub-graphs, with the financial news in each sub-graph being a financial event; generating a story chain through similarity calculation; performing emergency detection. According to the method, event clustering is carried out on financial news through natural language processing and graph theory related technologies, the problem that related news of the same event cannot be comprehensively considered in traditional financial emergencies is solved, the financial emergencies are efficiently and accurately detected, and certain industrial value is achieved.