CN116541517A

CN116541517A - Text information processing method, apparatus, device, software program, and storage medium

Info

Publication number: CN116541517A
Application number: CN202210089230.4A
Authority: CN
Inventors: 李涛
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2022-01-25
Filing date: 2022-01-25
Publication date: 2023-08-04

Abstract

The invention provides a text information processing method, a device, an electronic device, a software program and a storage medium, wherein the method comprises the following steps: acquiring text information to be processed; carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed; determining multi-hop information of different levels of the syntactic analysis result according to the syntactic analysis result of the text information to be processed; determining a multi-level graph neural network matched with the text information to be processed based on the multi-hop information; and processing the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed, so that the emotion state of the text information can be analyzed through diversification of the multi-level graph neural network, the complexity of emotion analysis of the text information is reduced, the accuracy of emotion analysis is improved, and the use experience of a user is improved.

Description

Text information processing method, apparatus, device, software program, and storage medium

Technical Field

The present invention relates to a text information processing technology, and more particularly, to a text information processing method, apparatus, electronic device, software program, and storage medium.

Background

In the text information processing process, because the text content field span is large, the used text information emotion classification technology is mainly based on a Long Short-Term Memory network (LSTM), but if the text information is longer, a large amount of key information can be lost by the method, so that the final emotion classification effect is poor; another common technique is to use convolutional neural network (Convolutional Neural Networks, CNN), and when CNN is used, the characteristics with different spans are extracted due to the window characteristics, so that the parallelism is better, the model is easy to train, but the relation between the front and the back of the word cannot be grasped, the position characteristics cannot be grasped, the analysis of the emotion state of the text information is influenced, and especially, the adaptability is poor in the text information processing in the financial field, and the use experience of a user is influenced.

Disclosure of Invention

In view of the above, embodiments of the present invention provide a text information processing method, apparatus, electronic device, software program, and storage medium, which can implement analysis of emotion states of text information through multiple levels of graph neural networks, reduce complexity of emotion analysis of text information, improve accuracy of emotion analysis, and improve user experience.

The technical scheme of the embodiment of the invention is realized as follows:

the embodiment of the invention provides a text information processing method, which comprises the following steps:

responding to a text information processing request, and acquiring text information to be processed;

carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed;

determining multi-hop information of different levels of the syntactic analysis result according to the syntactic analysis result of the text information to be processed;

determining a multi-level graph neural network matched with the text information to be processed based on the multi-hop information;

and processing the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed, so as to determine the emotion state of the text information to be processed through the classification result.

The embodiment of the invention also provides a text information processing device, which comprises:

the information transmission module is used for responding to the text information processing request and acquiring text information to be processed;

the information processing module is used for carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed;

The information processing module is used for determining multi-hop information of different levels of the syntactic analysis result according to the syntactic analysis result of the text information to be processed;

the information processing module is used for determining a multi-level graph neural network matched with the text information to be processed based on the multi-hop information;

the information processing module is used for processing the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed, so that the emotion state of the text information to be processed is determined through the classification result.

In the above-described arrangement, the first and second embodiments,

the information processing module is used for analyzing the text information processing request and determining a target object and a financial scene corresponding to the target object, which are included in the text information processing request;

the information processing module is used for determining historical behavior parameters of the target object and historical parameters of the financial scene in the financial scene;

the information processing module is used for carrying out data cross screening processing on the historical behavior parameters of the target object and the historical parameters of the financial scene based on the target object, and obtaining the text information to be processed, which is matched with the target object.

In the above-described arrangement, the first and second embodiments,

the information processing module is used for triggering a corresponding word segmentation library according to the identification environment of the text information to be processed;

the information processing module is used for carrying out word segmentation processing on the text information to be processed through the triggered word dictionary of the word segmentation library, extracting Chinese character texts and forming different word-level feature vectors;

the information processing module is used for carrying out dependency syntax processing on the word-level feature vector to obtain at least one dependency relationship;

the information processing module is used for analyzing the word-level feature vector according to the main-predicate relation in the at least one dependency relation to obtain a syntactic analysis result in the text to be processed.

In the above-described arrangement, the first and second embodiments,

the information processing module is used for determining the syntactic structure information of different word-level feature vectors according to the syntactic analysis result of the text information to be processed;

the information processing module is configured to determine different level multi-hop information of the syntactic analysis result according to syntactic structure information of the different word level feature vectors, where the multi-hop information at least includes: first hop information and second hop information.

In the above-described arrangement, the first and second embodiments,

the information processing module is used for determining a local map neural network matched with the text information to be processed according to multi-hop information of different levels in the multi-hop information;

the information processing module is used for determining a global graph neural network matched with the text information to be processed according to the complete sentence information in the text to be processed, wherein the multi-level graph neural network comprises the global graph neural network and at least one local graph neural network.

In the above-described arrangement, the first and second embodiments,

the information processing module is used for processing the text information to be processed through the local map neural network to obtain a first processing result;

the information processing module is used for processing the text information to be processed through the global graph neural network to obtain a second processing result;

the information processing module is used for carrying out fusion processing on the first processing result and the second processing result based on an attention mechanism to obtain a fusion processing result, and carrying out normalization processing on the fusion processing result to obtain a classification result of the text information to be processed.

The embodiment of the invention also provides electronic equipment, which comprises:

A memory for storing executable instructions;

and the processor is used for realizing the text information processing method when the executable instructions stored in the memory are operated.

The invention provides a computer readable storage medium storing executable instructions which when executed by a processor implement the text information processing method.

The embodiment of the invention has the following beneficial effects:

the method comprises the steps of responding to a text information processing request to obtain text information to be processed; carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed; determining multi-hop information of different levels of the syntactic analysis result according to the syntactic analysis result of the text information to be processed; determining a multi-level graph neural network matched with the text information to be processed based on the multi-hop information; the text information to be processed is processed through the multi-level graph neural network, so that a classification result of the text information to be processed is obtained, the emotion state of the text information to be processed is determined according to the classification result, the emotion state of the text information can be analyzed through diversification of the multi-level graph neural network, complexity of emotion analysis of the text information is reduced, accuracy of emotion analysis is improved, and user experience is improved.

Drawings

FIG. 1 is a schematic view of a usage environment of a text information processing method according to an embodiment of the present invention;

fig. 2 is a schematic diagram of a composition structure of a text information processing apparatus according to an embodiment of the present invention;

FIG. 3 is a schematic flow chart of an alternative text message processing method according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of an alternative processing procedure of multi-hop information according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of an alternative architecture of a multi-level graph neural network in accordance with an embodiment of the present invention;

FIG. 6 is a schematic flow chart of an alternative text message processing method according to an embodiment of the present invention;

fig. 7 is a schematic structural diagram of a neural network according to an embodiment of the present invention.

Detailed Description

The present invention will be further described in detail with reference to the accompanying drawings, for the purpose of making the objects, technical solutions and advantages of the present invention more apparent, and the described embodiments should not be construed as limiting the present invention, and all other embodiments obtained by those skilled in the art without making any inventive effort are within the scope of the present invention.

In the following description, reference is made to "some embodiments" which describe a subset of all possible embodiments, but it will be appreciated; "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

Before describing embodiments of the present invention in further detail, the terms and terminology involved in the embodiments of the present invention will be described, and the terms and terminology involved in the embodiments of the present invention will be used in the following explanation.

1) In response to a condition or state that is used to represent the condition or state upon which the performed operation depends, the performed operation or operations may be in real-time or with a set delay when the condition or state upon which it depends is satisfied; without being specifically described, there is no limitation in the execution sequence of the plurality of operations performed.

2) Information, various forms of information available in the internet, such as video files, multimedia information, news information, etc. presented in a client or smart device.

3) Convolutional neural network (CNN Convolutional Neural Networks) is a type of feedforward neural network (Feed forward Neural Networks) that includes convolutional computation and has a deep structure, and is one of representative algorithms of deep learning. Convolutional neural networks have the capability of token learning (representation learning) and are capable of performing a shift-invariant classification (shift-invariant classification) on input information in their hierarchical structure.

4) Model training, multi-classification learning is carried out on the image data set. The model can be constructed by adopting deep learning frameworks such as Tensor Flow and torch, and a multi-classification model is formed by using multi-layer combination of neural network layers such as CNN. The input of the model is a three-channel or original channel matrix formed by reading an image through tools such as openCV, the model is output as multi-classification probability, and the webpage category is finally output through algorithms such as softmax. During training, the model approaches to the correct trend through an objective function such as cross entropy and the like.

5) Neural Networks (NN): an artificial neural network (Artificial Neural Network, ANN), abbreviated as neural network or neural-like network, is a mathematical or computational model that mimics the structure and function of biological neural networks (the central nervous system of animals, particularly the brain) for estimating or approximating functions in the field of machine learning and cognitive sciences.

6) Fig. neural network (Graph Neural Network, GNN): a neural network acting directly on a graph structure, which is mainly aimed at processing data of a non-euclidean space structure (graph structure). Having an input order of ignore nodes; in the calculation process, the representation of the node is influenced by the neighboring nodes around the node, and the connection of the graph is unchanged; the representation of the graph structure enables graph-based reasoning. In general, a graph neural network consists of two modules: a propagation Module (Propagation Module) for transferring information between nodes in the graph and updating the state, and an Output Module (Output Module) for defining an objective function according to different tasks based on vector representations of nodes and edges of the graph. The graphic neural network has: graph convolutional neural networks (Graph Convolutional Networks, GCNs), gated graph neural networks (Gated Graph Neural Networks, GGNNs), and attention mechanism based graph attention neural networks (Graph Attention Networks, GAT).

7) Directed graph: representing the relationship between items, a directed graph may be represented by an ordered triplet (V (D), A (D), ψD), where ψD is an association function, which is an ordered pair of elements corresponding to V (D) for each element in A (D).

8) Encoder-decoder structure: network architecture commonly used in machine translation technology. The method comprises two parts of an encoder and a decoder, wherein the encoder converts input text into a series of context vectors capable of expressing input text characteristics, and the decoder receives the output result of the encoder as own input and outputs a corresponding text sequence in another language.

9) The bidirectional attention neural network model (BERT Bidirectional Encoder Representations from Transformers) is a bidirectional attention neural network model proposed by google.

10 Token). Word units, the input text, before any actual processing, needs to be split into language units such as words, punctuation marks, numbers or pure alphanumerics. These units are referred to as word units.

11 Softmax): the normalized exponential function is a generalization of the logic function. It can "compress" a K-dimensional vector containing arbitrary real numbers into another K-dimensional real vector such that each element ranges between 0,1 and the sum of all elements is 1.

12 Word segmentation): and segmenting the Chinese text by using a Chinese word segmentation tool to obtain a set of fine-grained words. Stop words: there is no contribution to the semantics of the text or words that may be ignored. Cosin similarity: the two texts are expressed as cosine similarity after the vector.

Fig. 1 is a schematic view of a usage scenario of a text information processing method provided in an embodiment of the present invention, referring to fig. 1, a terminal (including a terminal 10-1 and a terminal 10-2) is provided with a client capable of displaying software of corresponding financial text information, for example, a client or a plug-in for performing financial activities on virtual resources or entity resources or paying by virtual resources (such as Q-notes), a target object may obtain and display the financial text information through the corresponding client, and trigger a corresponding text information processing process (for example, a payment applet of the instant messaging software or a process of buying stocks by funds in the instant messaging software) in a text information processing process, and through processing the text information, an emotional state of the target object on different stocks may be obtained to determine a preference degree of the target object on different stocks); the terminal is connected to the server 200 through the network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two, and uses a wireless link to implement data transmission.

As an example, the server 200 is configured to arrange the text information processing apparatus to implement the text information processing method provided by the present invention, so as to obtain text information to be processed by responding to a text information processing request; carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed; determining multi-hop information of different levels of the syntactic analysis result according to the syntactic analysis result of the text information to be processed; determining a multi-level graph neural network matched with the text information to be processed based on the multi-hop information; and processing the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed, so as to determine the emotion state of the text information to be processed through the classification result, and executing related operations, such as purchasing behavior of stocks, emotion recognition of user conversations and the like, based on the emotion state of the text information, wherein specific behavior content is not limited in the application.

Of course, the text information processing device provided by the invention can be applied to a virtual resource or an entity resource to perform financial activities or a use environment of information interaction through entity financial resource payment environments (including but not limited to entity financial resource change environments of various types) or social software, financial text information of different data sources is usually processed in the process of performing financial activities through various types of entity financial resources or through virtual resource payment, and finally, the financial text information corresponding to a corresponding target object selected by the target object is presented on a User Interface (UI). The emotional state of the text information to be processed (for example, the emotional state of real-time price fluctuation of stocks and the emotional state of the rising and falling range of futures) formed by the target object in the current display interface can be called by other application programs, and it is to be noted that the emotional state related in the application can include at least the following contents: the emotion state can feed back the viewpoint tendency and emotion information of a target object, and has wide application prospects in the fields of topic discovery, poll, targeted advertisement delivery, after-sales service evaluation and the like.

The text information processing method provided by the embodiment of the application is realized based on artificial intelligence, wherein the artificial intelligence (Artificial Intelligence, AI) is a theory, a method, a technology and an application system which simulate, extend and expand human intelligence by using a digital computer or a machine controlled by the digital computer, sense environment, acquire knowledge and acquire an optimal result by using the knowledge. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and to produce a new intelligent machine that can react in a similar way to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision.

The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.

In the embodiments of the present application, the mainly related artificial intelligence software technology includes the above-mentioned speech processing technology, machine learning, and other directions. For example, speech recognition techniques (Automatic Speech Recognition, ASR) in Speech technology (Speech Technology) may be involved, including Speech signal preprocessing (Speech signal preprocessing), speech signal frequency domain analysis (Speech signal frequency analyzing), speech signal feature extraction (Speech signal feature extraction), speech signal feature matching/recognition (Speech signal feature matching/recognition), training of Speech (Speech training), and the like.

For example, machine Learning (ML) may be involved, which is a multi-domain interdisciplinary, involving multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It is specially studied how a computer simulates or implements learning behavior of a human to acquire new knowledge or skills, and reorganizes existing knowledge structures to continuously improve own performance. Machine learning is the core of artificial intelligence, a fundamental approach to letting computers have intelligence, which is applied throughout various areas of artificial intelligence. Machine Learning typically includes Deep Learning (Deep Learning) techniques, including artificial neural networks (artificial neural network), such as convolutional neural networks (Convolutional Neural Network, CNN), recurrent neural networks (Recurrent Neural Network, RNN), deep neural networks (Deep neural network, DNN), and the like.

The following describes in detail the structure of the text information processing apparatus according to the embodiment of the present invention, and the text information processing apparatus may be implemented in various forms, such as a dedicated terminal with a processing function of the text information processing apparatus, or may be a server provided with a processing function of the text information processing apparatus, for example, the server 200 in fig. 1. Fig. 2 is a schematic diagram of a composition structure of a text information processing apparatus according to an embodiment of the present invention, and it is understood that fig. 2 shows only an exemplary structure of the text information processing apparatus, not all the structures, and that a part of or all the structures shown in fig. 2 may be implemented as needed.

The text information processing device provided by the embodiment of the invention comprises: at least one processor 201, a memory 202, a user interface 203, and at least one network interface 204. The various components in the text information processing apparatus are coupled together by a bus system 205. It is understood that the bus system 205 is used to enable connected communications between these components. The bus system 205 includes a power bus, a control bus, and a status signal bus in addition to the data bus. But for clarity of illustration the various buses are labeled as bus system 205 in fig. 2.

The user interface 203 may include, among other things, a display, keyboard, mouse, trackball, click wheel, keys, buttons, touch pad, or touch screen, etc.

It will be appreciated that the memory 202 may be either volatile memory or nonvolatile memory, and may include both volatile and nonvolatile memory. The memory 202 in embodiments of the present invention is capable of storing data to support operation of the terminal (e.g., 10-1). Examples of such data include: any computer program, such as an operating system and application programs, for operation on the terminal (e.g., 10-1). The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, for implementing various basic services and processing hardware-based tasks. The application may comprise various applications.

In some embodiments, the text information processing apparatus provided in the embodiments of the present invention may be implemented by combining software and hardware, and by way of example, the text information processing apparatus provided in the embodiments of the present invention may be a processor in the form of a hardware decoding processor, which is programmed to perform the text information processing method provided in the embodiments of the present invention. For example, a processor in the form of a hardware decoding processor may employ one or more application specific integrated circuits (ASICs, application Specific Integrated Circuit), DSPs, programmable logic devices (PLDs, programmable Logic Device), complex programmable logic devices (CPLDs, complex Programmable Logic Device), field programmable gate arrays (FPGAs, field-Programmable Gate Array), or other electronic components.

As an example of implementation of the text information processing apparatus provided by the embodiment of the present invention by combining software and hardware, the text information processing apparatus provided by the embodiment of the present invention may be directly embodied as a combination of software modules executed by the processor 201, the software modules may be located in a storage medium, the storage medium is located in the memory 202, and the processor 201 reads executable instructions included in the software modules in the memory 202, and performs the text information processing method provided by the embodiment of the present invention in combination with necessary hardware (including, for example, the processor 201 and other components connected to the bus 205).

By way of example, the processor 201 may be an integrated circuit chip having signal processing capabilities such as a general purpose processor, such as a microprocessor or any conventional processor, a digital signal processor (DSP, digital Signal Processor), or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, or the like.

As an example of implementation of the text information processing apparatus provided by the embodiment of the present invention by hardware, the apparatus provided by the embodiment of the present invention may be implemented directly by the processor 201 in the form of a hardware decoding processor, for example, by one or more application specific integrated circuits (ASIC, application Specific Integrated Circuit), DSPs, programmable logic devices (PLDs, programmable Logic Device), complex programmable logic devices (CPLDs, complex Programmable Logic Device), field programmable gate arrays (FPGAs, field-Programmable Gate Array), or other electronic components.

The memory 202 in the embodiment of the present invention is used to store various types of data to support the operation of the text information processing apparatus. Examples of such data include: any executable instructions, such as executable instructions, for operating on a text information processing apparatus, a program implementing a method of processing slave text information according to an embodiment of the present invention may be included in the executable instructions.

In other embodiments, the text information processing apparatus provided in the embodiments of the present invention may be implemented in a software manner, and fig. 2 shows the text information processing apparatus stored in the memory 202, which may be software in the form of a program, a plug-in, or the like, and includes a series of modules, and as an example of the program stored in the memory 202, may include the text information processing apparatus, and the text information processing apparatus includes the following software module information transmission module 2081 and information processing module 2082. When software modules in the text information processing apparatus are read into the RAM by the processor 201 and executed, the text information processing method provided by the embodiment of the present invention will be implemented, where functions of each software module in the text information processing apparatus include:

the information transmission module 2081 is configured to obtain text information to be processed in response to a text information processing request.

And the information processing module 2082 is used for carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed.

The information processing module 2082 is configured to determine, according to a result of the syntactic analysis of the text information to be processed, multi-hop information of different levels of the syntactic analysis result.

The information processing module 2082 is configured to determine, based on the multi-hop information, a multi-level graph neural network that matches the text information to be processed.

The information processing module 2082 is configured to process the text information to be processed through the multi-level graph neural network, so as to obtain a classification result of the text information to be processed, so as to determine an emotion state of the text information to be processed through the classification result.

According to the electronic device shown in fig. 2, in one aspect of the present application, there is also provided a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The computer instructions are read from the computer-readable storage medium by a processor of a computer device, and executed by the processor, cause the computer device to perform the methods provided in various alternative implementations of the text information processing method described above.

The text information processing method provided in the present application will be further described with reference to the usage scenario shown in fig. 1, where the terminal (including the terminal 10-1 and the terminal 10-2) obtains, through the network 300, the quotation text information of the financial resource allocation process such as funds, stocks, etc. from the corresponding server 200, and comments the generated comments may be used as the text information to be processed, for example, comments on quotations through platforms such as change money, stock transaction APP, etc. Because the text content field span is large, the used text information emotion classification technology is mainly based on a Long Short-Term Memory network (LSTM), but if the text information is longer, the method can lose a large amount of key information, so that the final emotion classification effect is poor; another common technique is to use convolutional neural network (Convolutional Neural Networks, CNN), and when CNN is used, the characteristics with different spans are extracted due to the window characteristics, so that the parallelism is better, the model is easy to train, but the relation between the front and the back of the word cannot be grasped, the position characteristics cannot be grasped, the analysis of the emotion state of the text information is influenced, and especially, the adaptability is poor in the text information processing in the financial field, and the use experience of a user is influenced.

To solve the above-mentioned drawbacks, referring to fig. 3, fig. 3 is a schematic flow chart of an alternative text information processing method according to an embodiment of the present invention, in which a target object may select different financial scenarios for use, it will be understood that the steps shown in fig. 3 may be executed by various electronic devices running the text information processing apparatus, for example, a dedicated terminal (stock machine or mobile phone) with text information processing functions, an electronic device or a financial applet, and the steps shown in fig. 3 are described below.

Step 301: the text information processing device responds to the text information processing request and acquires the text information to be processed.

In some embodiments of the present invention, taking a financial text information processing scenario as an example, acquiring financial text information to be processed may be implemented by:

analyzing the text information processing request, and determining a target object and a financial scene corresponding to the target object, which are included in the text information processing request; in a financial scene, determining historical behavior parameters of a target object and historical parameters of the financial scene; and obtaining the text information to be processed matched with the target object by carrying out data cross screening processing on the historical behavior parameters of the target object and the historical parameters of the financial scene. For example, when making a stock recommendation by stock exchange software, a user sends a text information processing request to a financial server by a client of the stock exchange software, wherein the text information processing request includes: the method comprises the steps that a target object identification parameter and a scene parameter are analyzed, after the text information processing request is obtained, the target object identification parameter and the scene parameter are continuously mapped according to the target object identification parameter, a target object and a data interface corresponding to the target object can be determined, historical behavior parameters of the target object can be called in a database of a financial server through the data interface, for example, the target object collects a certain stock or reviews a certain stock, or purchases a certain stock; the data interfaces of the scene parameters and the scene corresponding to the target object can be determined through the mapping relation of the scene parameters, and as the data types in the financial scene are more, the privacy data of the target object are related at the same time, the data interface of each scene parameter only corresponds to one fixed financial server; the method comprises the steps of utilizing a data interface of scene parameters, retrieving historical parameters of a corresponding scene from a historical information database through a financial server, wherein the historical parameters can be, for example, large-disc fluctuation data in stock trading, or stock trading volume change data of a certain plate, utilizing the historical behavior parameters of a target object to conduct data cross screening in the historical parameters of the financial scene, determining operation information of the target object in different periods in the financial scene as text information to be processed matched with the target object, for example, conducting data cross screening can obtain 'which stock is purchased by the target object along with the change of the large-disc fluctuation data', and can also obtain 'which comment is done by the target object on a certain stock along with the change of the stock trading volume of a certain plate'.

In some embodiments of the present invention, the target object is a user participating in stock exchange, the historical parameters of the financial scene may be exchange information of any stock on the market of the stock exchange, or may be individual stock data corresponding to individual stocks and large disc data corresponding to large discs, and the historical behavior parameters of the target object may be text record information of various financial products operated by the user, such as stocks, securities, futures, funds, and historical evaluation of the above financial products. It should be noted that, in the embodiments of the present invention, any credit tool that can be used as an economic benefit credential of a user may be referred to as a financial product, such as securities, bonds/derivative market products (e.g. futures, options, equity futures, etc.), and the specific use of which financial product is not particularly limited in the present application.

Specifically, corresponding stock data can be obtained as a history parameter of the financial scene based on an identification (such as a code) corresponding to the individual stock or the large disc. By the text information processing method provided by the application, when hope to obtain preference information of any stock by analyzing a text information processing request, historical behavior parameters of a target object can be utilized to carry out data cross screening in historical parameters of a financial scene to obtain market text expression of any stock and text description information of transaction amount, for example, the obtained text information to be processed matched with the target object can be 1) "XXX stock is good, recent trend is good, I have purchased stock" 2) "XXX stock of XXX is very bad, I have all cleared" I have cleared "

It will be appreciated that in the specific embodiments of the present application, user-related data such as historical behavioral parameters of a target object are involved, and when the embodiments of the present application are applied to specific products or technologies, user permissions or consents need to be obtained, and the collection, use and processing of the related data need to comply with relevant laws and regulations and standards of relevant countries and regions.

In some embodiments, the text information processing apparatus may collect the historical parameters corresponding to the stock by at least one of the following collection methods: acquiring stock data corresponding to stocks in a database; calling an application program interface (API, application Programming Interface) to acquire stock data corresponding to the stock; the historical data in the web page is crawled through the web crawler and the historical behavior parameters of the target object (whether the stock is selected as a choice stock or whether there is a transaction record).

For example, the text information processing apparatus may be a server provided by an operator and provided with a database for storing historical stock data, the database storing stock data of a plurality of stocks for a period of time (for example, one month), and large disc data for a period of time, for example, when the server needs to acquire the stock data (which may also be the large disc data), based on a code of the individual stock to be acquired, the database is queried for the individual stock data corresponding to the code. Illustratively, when the server needs to acquire real-time data of stocks, an API is called to acquire real-time data and large disc data of individual stocks to be acquired through a data interface with a stock exchange. For example, when the server needs to obtain individual stock data (which may also be large disc data), the server may crawl the stock data corresponding to the individual stock on the relevant external website through the web crawler. In the operation of financial products, comment sentences of a certain stock by a user or trade remark information of a certain stock block can be obtained as processing text information matched with a target object, and stocks (or stock types of a certain stock block) corresponding to the text information in the forward emotion state can be recommended to the user in a financial applet by analyzing the emotion state of the processing text information.

Step 302: and the text information processing device carries out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed.

In some embodiments of the present invention, the parsing processing is performed on the text information to be processed, so as to obtain a parsing result of the text information to be processed, which may be implemented in the following manner:

triggering a corresponding word segmentation library according to the identification environment of the text information to be processed; performing word segmentation processing on the text information to be processed through the triggered word dictionary of the word segmentation library, extracting Chinese character texts, and forming different word-level feature vectors; performing dependency syntax processing on the word-level feature vector to obtain at least one dependency relationship; and analyzing the word-level feature vector according to the main-predicate relation in the at least one dependency relation to obtain a syntactic analysis result in the text to be processed. Taking a recognition environment of text information to be processed as a financial scene as an example of stock trading, when a syntactic analysis result is obtained, firstly triggering a financial scene dictionary and a stock trade word segmentation dictionary, and for Chinese financial text information, using a Chinese word segmentation tool Jieba to segment Chinese text. Wherein, for the text information to be processed, "the highest market value of the stock 6001 occurs at two zero for one year", the word segmentation by the word segmentation dictionary becomes "the stock/6001/highest/market value/occurrence/at/two/zero/one year". Wherein "stock" and "6001" have noun meanings as partitionings; each word is a word or phrase, i.e. the minimum semantic unit with definite meaning; for the recognition environment of the text information to be processed, the minimum semantic units contained in the text information to be processed are different, for example, the minimum semantic units are required to be word groups for financial scenes, and when the emotion states of the user preference scenes are recorded, the minimum semantic units are required to be single characters, so that the adjustment of the minimum semantic units is required to be timely made, and the accuracy of the emotion states is ensured; for the text information to be processed of Chinese, since words serving as the minimum semantic units are often composed of different numbers of words, no natural distinguishing mark in alphabetic writing such as blank partition exists among the words, and therefore, for Chinese, it is an important step to accurately segment words to obtain reasonable word segmentation objects.

Further, the dependency syntax processing (Dependency Parsing, DP) used in the embodiments of the present application is also one of the key technologies in the NLP field, and refers to determining the dependency relationship between different words in the text, such as a master-predicate relationship and a dynamic guest relationship. Since the subject in the text to be processed is usually the main component, after obtaining a plurality of word-level feature vectors in the text, dependency syntax processing is further performed on the plurality of word-level feature vectors to obtain dependency relationships between the word-level feature vectors, where different word feature vectors may be divided into dependent words: one word modifies the other, dominant word: the manner in which the modified term is processed by the dependency syntax is not limited in the embodiments of the present application. And then, according to the main-predicate relation in the dependency relations, screening the word-level feature vector with the grammar type as the main language from the word-level feature vectors, so that the type of the word-level feature vector of the screened word-level feature vector is conveniently matched with the types of the set word-level feature vectors. Through the method, the word-level feature vectors with higher importance degree can be screened out, so that the accuracy of the follow-up knowledge selection extraction template and the guest relation are further improved.

In some embodiments of the present invention, dependency syntax processing is performed on a word-level feature vector to obtain at least one dependency relationship, and when a syntax analysis result in a text to be processed is obtained by analyzing the word-level feature vector according to a master predicate relationship in the at least one dependency relationship, processing of the dependency syntax may be implemented by calling a dependency syntax processing module, where an interface of the dependency syntax processing module requests a domain name nlp.

TABLE 1

Analyzing the word-level feature vector, when the syntactic analysis result in the text to be processed is obtained, for example, the text information to be processed is "XX company board Zhong Zhao XX collection and purchase XX stock 100 hand", the syntactic analysis results were:

syntactic analysis results of person names and stocks: [ [ "Zhao XX", "XX stock" ];

syntactic analysis results of person name and institution: [ [ "Zhao XX", "Dong Long", "XX company ];

results of syntactic analysis of stocks and numbers: [ [ "XX stock", 100 hands ] ];

syntactic analysis of person name and behavior results: [ "Zhao XX", "collection and purchase" ] ].

Step 303: and the text information processing device determines different levels of multi-hop information of the syntactic analysis result according to the syntactic analysis result of the text information to be processed.

In some embodiments of the present invention, according to the syntactic analysis result of the text information to be processed, determining the multi-hop information of different levels of the syntactic analysis result may be implemented by:

determining the syntactic structure information of different word-level feature vectors according to the syntactic analysis result of the text information to be processed; determining different levels of multi-hop information of the syntactic analysis result according to the syntactic structure information of the different word level feature vectors, wherein the multi-hop information at least comprises: first hop information and second hop information. Referring to fig. 4, fig. 4 is a schematic diagram of an optional processing procedure of multi-hop information in the embodiment of the present invention, taking text information to be processed as "It has bad money but a good battery ife" as an example, subscripts 1,2,3 in fig. 4 indicate that a current word vector needs to undergo several relation steps to reach another word vector, and in the dependency syntax in the embodiment of the present application, if two words (a, b) have a dependency relationship (the embodiment in the tree structure is that there is an edge connection directly, and a- > b) then a- > b can be defined as one-hop information of multi-hop information of different levels. Similarly, if there is a dependency relationship b- > c, but a- > c is that there is no direct dependency relationship, then a- > b- > c is defined as two-hop information in multi-hop information of different levels, so the text to be processed can include multi-hop information of different levels and can be recorded as one-hop information and two-hop information … N-hop information.

Step 304: and the text information processing device determines a multi-level graphic neural network matched with the text information to be processed based on the multi-hop information.

In some embodiments of the present invention, determining, based on the multi-hop information, a multi-level graph neural network that matches the text information to be processed may be implemented by:

determining a local map neural network matched with the text information to be processed according to multi-hop information of different levels in the multi-hop information; and determining a global graph neural network matched with the text information to be processed according to the complete sentence information in the text to be processed, wherein the multi-level graph neural network comprises the global graph neural network and at least one local graph neural network. Referring to fig. 5, fig. 5 is an optional structural schematic diagram of a multi-level graph neural network in an embodiment of the present invention, where Local graph neural network local_1gcn only uses one-hop information in syntactic analysis when performing text information processing, all the other results are set to 0, local graph neural network local_2gcn only uses two-hop information in syntactic analysis when performing text information processing, all the other results are set to 0, global graph neural network Global GCN may use the result of all hops obtained by syntactic analysis (i.e. one-hop information+two-hop information … N-hop information, and thus may also become Global multi-hop information). In the process of determining the multi-level graph neural network, the structure of the multi-level graph neural network can be adjusted because the multi-hop information of the text information to be processed is different, for example, taking the text information to be processed as "i collect XX stocks" as an example, each word vector "i", "collect", "XX stocks" has a interdependence relationship (the embodiment in the tree structure is that the word vector "i", "collect", "XX stocks" directly have edge connection and are "i" - > "collect" and "collect" > "XX stocks"), then the multi-hop information of different levels of i "- >" collect "can be defined, and meanwhile, the dependency relationship" i ">" XX "is two-hop information in the multi-hop information of different levels, so that the text to be processed can include 2 levels of multi-hop information, which can be recorded as one-hop information and two-hop information. The multi-level graph neural network required to process the text information to be processed comprises: local map neural network local_1gcn, local map neural network local_2gcn, and Global map neural network Global GCN.

In some embodiments of the present application, when the text information to be processed includes: when the first-hop information, the second-hop information and the third-hop information are used, the number of the local map neural networks is 3, and the local map neural networks are respectively as follows: the Local map neural network local_1GCN, the Local map neural network local_2GCN and the Local map neural network local_3GCN form a multi-level map neural network needed to be used for processing text information to be processed by calling the trained Global map neural network Global GCN and the 3 trained Local map neural networks local_GCN, and when the multi-hop information of the text information to be processed changes, the structure of the multi-level map neural network used by the text information processing method provided by the application also changes, so that the method can adapt to the requirements of text information processing in different scenes.

The Global graph neural network Global GCN is adopted, the meaning of a sentence can be understood from the whole angle of a complete sentence of a text to be processed, so that the sentence is not offset in understanding, the Local graph neural network local_N GCN is used for realizing that the multistage graph neural network pays attention to the attribute closest to the sentence, so that the correlation of attribute result analysis is ensured for a main body corresponding to any one-hop information in emotion analysis, further, when information between the texts to be processed is acquired through one-hop information, two-hop information and three-hop information, the information of a syntax structure is captured, the information is converted into information between hops, the information of the hops is converted into an adjacent matrix of the multistage graph neural network GCN, the data required to be input into the multistage graph neural network can be obtained through the series of conversions, and the multistage graph neural network can effectively obtain the correlation information between the elements by means of the information extraction capability of the multistage graph neural network GCN.

Step 305: the text information processing device processes the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed, so that the emotion state of the text information to be processed is determined through the classification result.

In some embodiments of the present invention, referring to fig. 6, fig. 6 is an optional flowchart of a text information processing method provided in an embodiment of the present invention, which specifically includes the following steps:

step 601: and processing the text information to be processed through the local map neural network to obtain a first processing result.

In one embodiment of the present application, the text information to be processed is respectively:

1) The operating experience of the financial transaction applet is very excellent, but its funds are always lost.

2) The fund of the financial transaction applet is very excellent, but it is always a loss.

Taking the number of Local neural networks as 2 as an example, the Local neural network local_1gcn only uses the one-hop information in the syntactic analysis when processing the text information, the rest of results are all set to 0, the Local neural network local_1gcn only uses the two-hop information in the syntactic analysis when processing the text information, the rest of results are all set to 0, the obtained first processing result can be a feature vector for representing the correlation, the Local neural network is used for focusing on the latest attribute of the text information, and the correlation of the attribute result analysis is ensured.

Wherein the graph neural network (Graph Neural Network, GNN) can ignore the input order of the nodes; in the calculation process, the representation of the node is influenced by the neighboring nodes around the node, and the connection of the graph is unchanged; the graph neural network consists of two modules: a propagation Module (Propagation Module) for transferring information between nodes in the graph and updating the state, and an Output Module (Output Module) for defining an objective function according to different tasks based on vector representations of nodes and edges of the graph. The graphic neural network has: graph convolutional neural networks (Graph Convolutional Networks, GCNs), gated graph neural networks (Gated Graph Neural Networks, GGNNs), and attention mechanism based graph attention neural networks (Graph Attention Networks, GAT). The prediction of the target stock through the graph neural network has the advantages that each node in the graph network can automatically transmit all characteristic (trend) information of the node to adjacent neighbor nodes based on the constructed graph network, and through information transmission among the neighbors for a plurality of times, each node in the graph network can contain attribute information of the node directly or indirectly connected with the node.

In some embodiments of the present invention, referring to fig. 7, fig. 7 is a schematic diagram of a neural network structure according to an embodiment of the present invention, and since nodes with direct or indirect connection in the global graph network provided in the present application mostly have similar behaviors. The propagation manner between layers of the graph neural network is referred to as formula 1:

wherein: a=a+i, I is the identity matrix, D is the degree matrix of a; h is a feature of each layer; w is a parameter matrix from an input layer to a hidden layer in the graph neural network. The constructed neural network has N nodes, each representing an associated object of the target object, whose characteristics form an N X D-dimensional matrix X, and then the relationships between the nodes form an N X N-dimensional matrix a, also referred to as an adjacency matrix. X and a are inputs to the neural network. Wherein tan is the activation function between the multi-layer networks referring to equation 2:

Z＝tanh(A tanh(A tanh(AXW ⁽⁰⁾ )W ⁽¹⁾ )W ⁽²⁾ ) Equation 2

Thereafter, wherein A is an input between the multi-layer networks, W ⁽⁰⁾ W ⁽¹⁾ And W is ⁽²⁾ For parameter matrix between different hidden layers

Mapping the learned distributed feature representation to a corresponding sample marking space by adopting two full connection layers to improve the accuracy of a final classification result, and finally, carrying out normalization operation on the vector to obtain the maximum value in the vector; and then mapping the attribute to the corresponding emotion label, namely the most probable emotion state of the attribute.

Step 602: and processing the text information to be processed through the global graph neural network to obtain a second processing result.

Wherein the second processing result can be a feature vector calculated by a Global graph neural network, the Global graph neural network Global GCN uses the result of all hops obtained by syntactic analysis to understand the meaning of the text information from the integral angle of the text, so that the meaning of the identified text information can not deviate, the identification accuracy is ensured,

step 603: and carrying out fusion processing on the first processing result and the second processing result based on an attention mechanism to obtain a fusion processing result, and carrying out normalization processing on the fusion processing result to obtain a classification result of the text information to be processed. According to the text information processing method, different attribute expression results are obtained according to the diversified GCN, when emotion analysis results of specific attributes are effectively guaranteed, emotion can be divided into positive (positive direction), negative (negative direction) and neutral 3 types, so that corresponding labels can be set for each emotion classification, vector maximum values in the fusion processing results are obtained through normalization processing of the fusion processing results, and the vector maximum values are mapped to the corresponding emotion labels, so that emotion states can be obtained. Further, the target object can adjust the labels of emotion classification according to different use requirements, for example, only the forward emotion labels are reserved, so that only the forward emotion analysis result in the text to be processed is obtained.

In some embodiments of the present application, for text information: 1) The operating experience of the financial transaction applet is very excellent, but its funds are always lost. 2) The fund of the financial transaction applet is very excellent, but it is always a loss. Different classification results can be obtained by fusing the processing results: 1) The main body is as follows: recommended fund, (emotional state negative-1), 2) subject is: operational experience, (emotional state is forward 1). Thus, analyzing the emotion of the funds recommended by the financial transaction applet may result in a negative, but positive, when the body of text information through 2) becomes the operational experience of the financial transaction applet. Through the multi-level graph neural network provided by the application, text information of different subjects can be accurately processed, so that the emotion states of different subjects can be effectively identified.

Meanwhile, in practical application, the scheme of the application can be realized through the financial APP, and meanwhile, the scheme of the application can be realized through an instant messaging software applet, so that a user can rapidly predict financial text information of different objects through a text information processing model deployed in a blockchain network when the terminal is replaced through the financial blockchain storage multi-level graph neural network.

Specifically, the target object identification, the model parameters of the text information processing model, and the target object identification may be sent to a blockchain network, so that nodes of the blockchain network populate new blocks with the target object identification, the model parameters of the text information processing model, and the target object identification, and when the new blocks are consistent, the new blocks are appended to the tail of the blockchain.

The embodiment of the invention can be realized by combining Cloud technology, wherein Cloud technology (Cloud technology) refers to a hosting technology for integrating hardware, software, network and other series resources in a wide area network or a local area network to realize calculation, storage, processing and sharing of data, and can also be understood as the general term of network technology, information technology, integration technology, management platform technology, application technology and the like applied based on a Cloud computing business model. Background services of technical network systems require a large amount of computing and storage resources, such as video websites, picture websites and more portal websites, so cloud technologies need to be supported by cloud computing.

It should be noted that cloud computing is a computing mode, which distributes computing tasks on a resource pool formed by a large number of computers, so that various application systems can acquire computing power, storage space and information service as required. The network that provides the resources is referred to as the "cloud". Resources in the cloud are infinitely expandable in the sense of users, and can be acquired at any time, used as needed, expanded at any time and paid for use as needed. As a basic capability provider of cloud computing, a cloud computing resource pool platform, referred to as a cloud platform for short, is generally called infrastructure as a service (IaaS, infrastructure as a Service), and multiple types of virtual resources are deployed in the resource pool for external clients to select for use. The cloud computing resource pool mainly comprises: computing devices (which may be virtualized machines, including operating systems), storage devices, and network devices.

As shown in fig. 1, the text information processing method provided by the embodiment of the present invention may be implemented by a corresponding cloud device, for example: the terminals (including the terminal 10-1 and the terminal 10-2) are connected to the server 200 located at the cloud through the network 300, and the network 300 may be a wide area network or a local area network, or a combination of the two. It should be noted that the server 200 may be a physical device or a virtualized device.

The beneficial technical effects are as follows:

according to the embodiment of the invention, the text information to be processed is obtained by responding to the text information processing request; carrying out syntactic analysis processing on the text information to be processed to obtain a syntactic analysis result of the text information to be processed; determining multi-hop information of different levels of the syntactic analysis result according to the syntactic analysis result of the text information to be processed; determining a multi-level graph neural network matched with the text information to be processed based on the multi-hop information; the text information to be processed is processed through the multi-level graph neural network, so that a classification result of the text information to be processed is obtained, the emotion state of the text information to be processed is determined according to the classification result, the emotion state of the text information can be analyzed through diversification of the multi-level graph neural network, complexity of emotion analysis of the text information is reduced, accuracy of emotion analysis is improved, and user experience is improved.

The foregoing description of the embodiments of the invention is not intended to limit the scope of the invention, but is intended to cover any modifications, equivalents, and alternatives falling within the spirit and principles of the invention.

Claims

1. A text information processing method, characterized in that the method comprises:

and processing the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed.

2. The method of claim 1, wherein the obtaining text information to be processed in response to the text information processing request comprises:

analyzing the text information processing request, and determining a target object and a financial scene corresponding to the target object, which are included in the text information processing request;

In the financial scene, determining historical behavior parameters of the target object and historical parameters of the financial scene;

and based on the target object, performing data cross screening processing on the historical behavior parameters of the target object and the historical parameters of the financial scene to obtain the text information to be processed, which is matched with the target object.

3. The method according to claim 1, wherein the subjecting the text information to be processed to a syntactic analysis process to obtain a syntactic analysis result of the text information to be processed includes:

triggering a corresponding word segmentation library according to the identification environment of the text information to be processed;

performing word segmentation processing on the text information to be processed through the triggered word dictionary of the word segmentation library, extracting Chinese character texts, and forming different word-level feature vectors;

performing dependency syntax processing on the word-level feature vector to obtain at least one dependency relationship;

and analyzing the word-level feature vector according to the main-predicate relation in the at least one dependency relation to obtain a syntactic analysis result in the text to be processed.

4. The method of claim 3, wherein the determining different levels of multi-hop information for the parsing result based on the parsing result of the text information to be processed comprises:

Determining the syntactic structure information of different word-level feature vectors according to the syntactic analysis result of the text information to be processed;

determining different levels of multi-hop information of the syntactic analysis result according to the syntactic structure information of the different word-level feature vectors, wherein the different levels of multi-hop information at least comprise: first hop information and second hop information.

5. The method of claim 1, wherein the determining a multi-level graph neural network that matches the text information to be processed based on the multi-hop information comprises:

determining a local map neural network matched with the text information to be processed according to multi-hop information of different levels in the multi-hop information;

and determining a global graph neural network matched with the text information to be processed according to the complete sentence information in the text to be processed, wherein the multi-level graph neural network comprises the global graph neural network and at least one local graph neural network.

6. The method according to claim 5, wherein the processing the text information to be processed through the multi-level graph neural network to obtain the classification result of the text information to be processed includes:

Processing the text information to be processed through the local map neural network to obtain a first processing result;

processing the text information to be processed through the global graph neural network to obtain a second processing result;

and carrying out fusion processing on the first processing result and the second processing result based on an attention mechanism to obtain a fusion processing result, and carrying out normalization processing on the fusion processing result to obtain a classification result of the text information to be processed.

7. A text information processing apparatus, characterized in that the apparatus comprises:

The information processing module is used for processing the text information to be processed through the multi-level graph neural network to obtain a classification result of the text information to be processed.

8. An electronic device, the electronic device comprising:

a memory for storing executable instructions;

a processor for implementing the text information processing method of any one of claims 1 to 6 when executing the executable instructions stored in the memory.

9. A computer program product comprising a computer program or instructions which, when executed by a processor, implements the text information processing method of any of claims 1 to 6.

10. A computer-readable storage medium storing executable instructions which, when executed by a processor, implement the text information processing method of any one of claims 1 to 6.