A text information processing method and device, electronic equipment and storage medium

Abstract

The application provides a text information processing method and device, electronic equipment and a storage medium. The method comprises: obtaining to-be-processed text information; performing word segmentation processing on the to-be-processed text information to determine at least one candidate keyword vector; dynamically adjusting the candidate keyword vector to form a candidate keyword vector set; determining a keyword extraction strategy matched with the to-be-processed text information based on the candidate keyword vector set and a reward value parameter through a deep reinforcement learning network; and extracting the candidate keyword vector set based on the keyword extraction strategy to obtain at least one keyword vector as a keyword of the to-be-processed text information. Thus, the dependence on word granularity in keyword extraction is reduced, the extracted keyword of the to-be-processed text information is suitable for different use scenarios, and the user experience is improved.

Description

Technical Field

[0001] This invention relates to information processing technology, and more particularly to text information processing methods, apparatus, electronic devices, and storage media. Background Technology

[0002] Human-Computer Interaction (HCI) refers to the process of information exchange between humans and computers using a conversational language and interactive methods. With the development of HCI technology, more and more intelligent products based on HCI have emerged, such as chatbots. These intelligent products can chat and communicate with users and generate corresponding answers based on user questions. Understanding the query text is a core technology in NLP fields such as search engines and dialogue systems; a more accurate understanding of the user's request allows the system to provide better answers. A query is usually a piece of text or a sentence obtained through speech recognition, and the extraction of query keywords plays a crucial role in query understanding.

[0003] Traditional keyword extraction mainly involves preprocessing (such as Chinese word segmentation), scoring each candidate word based on statistical and sequence labeling models, and finally selecting the appropriate keywords. However, the accuracy of the keywords selected in this process is low, which affects the processing of problem texts by artificial intelligence and results in poor accuracy of the processing results. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a text information processing method, apparatus, electronic device, and storage medium, which can achieve more accurate processing of problem texts, reduce the dependence on word granularity in keyword extraction, make the extracted keywords of the text information to be processed applicable to different usage scenarios, reduce the impact of related information in the problem statement on keyword extraction, and improve the user experience.

[0005] The technical solution of this invention is implemented as follows:

[0006] This invention provides a text information processing method, including:

[0007] Obtain the text information to be processed;

[0008] The text information to be processed is segmented to determine at least one candidate keyword vector;

[0009] The candidate keyword vectors are dynamically adjusted through a word vector processing network to form a set of candidate keyword vectors, and reward value parameters corresponding to different keyword vectors in the set of candidate keyword vectors.

[0010] Using a deep reinforcement learning network, a keyword extraction strategy matching the text information to be processed is determined based on the candidate keyword vector set and the reward value parameter.

[0011] Based on the keyword extraction strategy, the candidate keyword vector set is extracted to obtain at least one keyword vector as the keyword of the text information to be processed.

[0012] This invention also provides a text information processing device, comprising:

[0013] The information transmission module is used to acquire text information to be processed;

[0014] The information processing module is used to perform word segmentation on the text information to be processed and determine at least one candidate keyword vector;

[0015] The information processing module is used to dynamically adjust the candidate keyword vectors through a word vector processing network to form a candidate keyword vector set, and reward value parameters corresponding to different keyword vectors in the candidate keyword vector set.

[0016] The information processing module is used to determine a keyword extraction strategy that matches the text information to be processed based on the candidate keyword vector set and the reward value parameter through a deep reinforcement learning network.

[0017] The information processing module is used to extract from the candidate keyword vector set based on the keyword extraction strategy, and obtain at least one keyword vector as the keyword of the text information to be processed.

[0018] In the above scheme,

[0019] The information processing module is used to trigger the corresponding word segmentation library based on the question text parameter information carried by the text information to be processed;

[0020] The information processing module is used to segment the question text using the triggered word segmentation dictionary to form different word-level question texts;

[0021] The information processing module is used to perform noise reduction processing on the different word-level question texts to form a set of word-level feature vectors corresponding to the question texts, wherein the set of word-level feature vectors includes at least one candidate keyword vector.

[0022] In the above scheme,

[0023] The information processing module is used to determine the name of the word segmentation library used when performing word segmentation processing on the problem text;

[0024] The information processing module is used to determine the parameters of a word segmentation library that matches the word-level feature vector corresponding to the question text, based on the name of the word segmentation library. The parameters of the word segmentation library include:

[0025] The types of word segmentation libraries, the names of word segmentation libraries, and the versions of word segmentation libraries.

[0026] In the above scheme,

[0027] The information processing module is used to determine a dynamic noise threshold that matches the usage environment of the word vector processing network.

[0028] The information processing module is used to perform noise reduction processing on the problem text set according to the dynamic noise threshold, and trigger a dynamic word segmentation strategy that matches the dynamic noise threshold.

[0029] The information processing module is used to perform word segmentation on the question text according to a dynamic word segmentation strategy that matches the dynamic noise threshold, thereby forming a set of dynamic word-level feature vectors corresponding to the question text.

[0030] In the above scheme,

[0031] The information processing module is used to determine a fixed noise threshold corresponding to the usage environment of the word vector processing network;

[0032] The information processing module is used to perform noise reduction processing on the problem text set according to the fixed noise threshold, and trigger a fixed word segmentation strategy that matches the fixed noise threshold.

[0033] The information processing module is used to perform word segmentation on the question text according to a fixed word segmentation strategy that matches the fixed noise threshold, and to generate a fixed set of word-level feature vectors corresponding to the question text.

[0034] In the above scheme,

[0035] The information processing module is used to delete and merge the candidate keyword vectors through the word vector processing network using the Monte Carlo tree search algorithm, thereby realizing dynamic adjustment of the candidate keyword vectors;

[0036] The information processing module is used to determine the frequency information of each candidate keyword vector in the candidate keyword vector set in the standard corpus.

[0037] The information processing module is used to determine the number of times the character vector matching each candidate keyword vector appears in the standard corpus;

[0038] The information processing module is used to determine the reward value parameters corresponding to different keyword vectors in the candidate keyword vector set based on the frequency information of the candidate keyword vectors appearing in the standard corpus and the frequency information of the character vectors appearing in the standard corpus.

[0039] In the above scheme,

[0040] The information processing module is used to determine, through a deep reinforcement learning network, based on the candidate keyword vector set and reward value parameters, to extract keywords that match the text information to be processed via the policy subnetwork of the deep reinforcement learning network, or...

[0041] The information processing module is used to determine keywords that match the text information to be processed, extracted through the value subnetwork of the deep reinforcement learning network.

[0042] In the above scheme,

[0043] The information processing module is used to determine the weight parameters that match the policy subnetwork of the deep reinforcement learning network using the reward value parameters.

[0044] The information processing module is used to respond to the weight parameters and generate keywords corresponding to the text information to be processed and the selection probability of the keywords according to the policy subnetwork of the deep reinforcement learning network based on the candidate keyword vector set.

[0045] The information processing module is used to obtain at least one keyword vector as the keyword of the text information to be processed based on the selection probability of the keyword.

[0046] In the above scheme,

[0047] The information processing module is used to determine, through the reward value parameter, the state-action value function that matches the value subnetwork of the deep reinforcement learning network;

[0048] The information processing module is used to respond to the state action value function, and through the value subnetwork of the deep reinforcement learning network, generate keywords corresponding to the text information to be processed and the selection probability of the keywords based on the candidate keyword vector set and the state value of the corresponding keyword vector;

[0049] The information processing module is used to obtain at least one keyword vector as the keyword of the text information to be processed based on the selection probability of the keyword.

[0050] In the above scheme,

[0051] The information processing module is used to determine the characteristics of the usage environment of the text information to be processed;

[0052] The information processing module is used to obtain a set of training samples from the data source that match the features of the usage environment of the text information to be processed.

[0053] The information processing module is used to train the deep reinforcement learning network based on the feature set and corresponding question text labels that match the training samples, so as to determine the model parameters that are compatible with the value subnetwork and policy subnetwork in the deep reinforcement learning network.

[0054] This invention also provides an electronic device, the electronic device comprising:

[0055] Memory, used to store executable instructions;

[0056] A processor, when executing executable instructions stored in the memory, implements a preceding text information processing method.

[0057] This invention also provides a computer-readable storage medium storing executable instructions, characterized in that the executable instructions, when executed by a processor, implement a preceding text information processing method.

[0058] The embodiments of the present invention have the following beneficial effects:

[0059] This invention acquires text information to be processed; performs word segmentation on the text information to determine at least one candidate keyword vector; dynamically adjusts the candidate keyword vectors through a word vector processing network to form a set of candidate keyword vectors, and sets reward value parameters corresponding to different keyword vectors in the set; uses a deep reinforcement learning network to determine a keyword extraction strategy that matches the text information to be processed based on the set of candidate keyword vectors and the reward value parameters; and extracts at least one keyword vector from the set of candidate keyword vectors based on the keyword extraction strategy to obtain keywords for the text information to be processed. This enables more accurate processing of problem texts, reduces the dependence on word granularity in keyword extraction, makes the extracted keywords applicable to different usage scenarios, reduces the influence of related information in the problem statement on keyword extraction, and improves the user experience. Attached Figure Description

[0060] Figure 1 This is a schematic diagram illustrating a usage scenario of the text information processing method provided in an embodiment of the present invention;

[0061] Figure 2 This is a schematic diagram of the composition structure of the text information processing device provided in an embodiment of the present invention;

[0062] Figure 3 This is a schematic diagram of the keyword extraction process for the problem text in an embodiment of the present invention;

[0063] Figure 4 A schematic flowchart of an optional text information processing method provided in an embodiment of the present invention;

[0064] Figure 5 This is a schematic diagram illustrating text information processing via a mini-program in an embodiment of the present invention;

[0065] Figure 6 A schematic flowchart of an optional text information processing method provided in an embodiment of the present invention;

[0066] Figure 7 This is a schematic diagram illustrating the text information processing effect in an embodiment of the present invention;

[0067] Figure 8 This is a schematic diagram illustrating the text information processing effect in an embodiment of the present invention;

[0068] Figure 9 This is a schematic diagram illustrating an optional process for text information processing in an embodiment of the present invention;

[0069] Figure 10 This is a schematic diagram of an optional data structure for the text information processing method in an embodiment of the present invention;

[0070] Figure 11 This is a schematic diagram of the Monte Carlo tree search algorithm processing in an embodiment of the present invention. Detailed Implementation

[0071] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings. The described embodiments should not be regarded as limitations on this invention. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0072] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0073] Before providing a further detailed description of the embodiments of the present invention, the nouns and terms involved in the embodiments of the present invention will be explained, and the nouns and terms involved in the embodiments of the present invention shall be interpreted as follows.

[0074] 1) In response to, used to indicate the conditions or states on which the operation performed depends. When the conditions or states on which it depends are met, one or more operations performed may be performed in real time or with a set delay. Unless otherwise specified, there is no restriction on the order in which the multiple operations are performed.

[0075] 2) Word segmentation: Also known as word cutting, it functions to divide the text information of a complete sentence into multiple words. For example: XX is a Chinese singer. The result after word segmentation is: XX, China, singer.

[0076] 3) Word segmentation library: also known as word segmentation library, refers to a specific word segmentation method. Different word segmentation libraries have their own corresponding word dictionaries, and can perform word segmentation processing on the corresponding text information according to their respective word dictionaries.

[0077] 4) Token: Before any actual processing is performed on the input text, it needs to be divided into language units such as words, punctuation marks, numbers, or alphanumeric characters. These units are called tokens.

[0078] 5) Softmax: The normalized exponential function, a generalization of the logistic function. It can "compress" a K-dimensional vector containing arbitrary real numbers into another K-dimensional real vector, such that each element is in the range [0, 1], and the sum of all elements is 1.

[0079] 6) Model Training: Multi-class classification learning on the image dataset. This model can be built using deep learning frameworks such as TensorFlow and Torch, employing multiple layers of neural networks like CNNs to form a multi-class classification model. The model input is a three-channel or original-channel matrix generated from images read using tools like OpenCV. The model output is the multi-class probability, ultimately outputting the webpage category through algorithms such as softmax. During training, the model approximates the correct trend using objective functions such as cross-entropy.

[0080] 7) Neural Network (NN): Artificial Neural Network (ANN), also known as neural network or neural network-like network, is a mathematical or computational model in the fields of machine learning and cognitive science that imitates the structure and function of biological neural networks (the central nervous system of animals, especially the brain) and is used to estimate or approximate functions.

[0081] 8) Encoder-decoder architecture: A commonly used network architecture in machine translation technology. It consists of two parts: an encoder and a decoder. The encoder converts the input text into a series of context vectors that can express the features of the input text. The decoder receives the output of the encoder as its own input and outputs the corresponding text sequence in another language.

[0082] 9) A mini program is a type of software developed using a front-end-oriented language (such as JavaScript) and implemented within a Hyper Text Markup Language (HTML) page. It is downloaded by a client (such as a browser or any client with an embedded browser engine) via a network (such as the Internet) and interpreted and executed within the client's browser environment, saving the step of installation on the client side. For example, a mini program can be activated via voice command on a terminal, enabling download and execution of various services such as airline ticket purchase, task processing and creation, and data display on a social network client.

[0083] 10) Query text: A request statement entered by the user in the intelligent assistant, usually containing only one intention or expectation of the user. For example: "Play Andy Lau's 'Ice Rain'"; "Tell me the story of the Foolish Old Man Who Moved Mountains"; "I want to watch the movie Infernal Affairs", etc.

[0084] Figure 1 This is a schematic diagram illustrating a usage scenario of the text information processing method provided in this embodiment of the invention. (See attached diagram.) Figure 1The terminals (including terminals 10-1 and 10-2) are equipped with corresponding clients capable of performing different functions. The clients of the terminals (including terminals 10-1 and 10-2) obtain different articles from the corresponding servers 200 through the network 300 for browsing or obtain the mini-programs or public accounts stored on the server. When the terminal runs the instant messaging client process, it can use the provided search function to search for different content such as Moments, mini-programs, articles, public accounts, novels, music, emoticons, etc., based on keywords, or search for any type of resource on the Internet. The terminal connects to server 200 via network 300, which can be a wide area network (WAN), a local area network (LAN), or a combination of both. Data transmission is achieved using a wireless link. The types of articles retrieved by the terminals (including terminals 10-1 and 10-2) from the corresponding servers 200 via network 300 are not the same. For example, terminals 10-1 and 10-2 can retrieve either a mini-program or a public account matching search command A from the corresponding server 200 via network 300, or they can retrieve only articles matching search command A from the corresponding server 200 via network 300 for browsing. During this process, the user can input the corresponding speech recognition statement as a search command through the configured speech recognition software client. The chat client can also receive the corresponding speech recognition results and display the received speech recognition result (question text) as the search command to the user, or perform a task matching the question text.

[0085] In some embodiments of the present invention, the different types of mini-programs stored in the server 200 can be written in software code environments of different programming languages, and the code objects can be different types of code entities. For example, in C language software code, a code object can be a function. In JAVA language software code, a code object can be a class, and in iOS OC language, it can be a piece of object code. In C++ language software code, a code object can be a class or a function to execute question text from different terminals. This application does not distinguish the source of the search instructions. The mini-program in the instant messaging client process can trigger the search engine. A mini-program is a program developed based on a front-end language (such as JavaScript), implementing services in Hyper Text Markup Language (HTML) pages. It is downloaded by the client (such as a browser or any client with an embedded browser core) via a network (such as the Internet) and interpreted and executed in the client's browser environment, saving the step of installation on the client. For example, by waking up a mini-program on the terminal via voice commands, users can download and run mini-programs in social network clients to provide various services such as ticket purchase, task processing and creation, and data display.

[0086] Server 200 sends corresponding search results to terminals (terminal 10-1 and / or terminal 10-2) via network 300 based on the keywords identified by the terminal. As an example, terminal 10-1 can be used to acquire text information to be processed; perform word segmentation on the text information to be processed to determine at least one candidate keyword vector; dynamically adjust the candidate keyword vectors through a word vector processing network to form a set of candidate keyword vectors, and reward value parameters corresponding to different keyword vectors in the set; determine a keyword extraction strategy matching the text information to be processed based on the set of candidate keyword vectors and the reward value parameters through a deep reinforcement learning network; and extract from the set of candidate keyword vectors based on the keyword extraction strategy to obtain at least one keyword vector as a keyword of the text information to be processed.

[0087] The text information processing method provided in this application is based on artificial intelligence (AI). AI is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, AI is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to have perception, reasoning, and decision-making functions.

[0088] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0089] In the embodiments of this application, the main artificial intelligence software technologies involved include the aforementioned speech processing technologies and machine learning. For example, it may involve Automatic Speech Recognition (ASR) technology in speech technology, including speech signal preprocessing, speech signal frequency analyzing, speech signal feature extraction, speech signal feature matching / recognition, and speech training.

[0090] For example, this could involve machine learning (ML), a multidisciplinary field encompassing probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning typically includes techniques such as deep learning, which includes artificial neural networks, such as convolutional neural networks (CNNs), recurrent neural networks (RNNs), and deep neural networks (DNNs).

[0091] It is understood that the text information processing method and voice processing provided in this application can be applied to intelligent devices. Intelligent devices can be any device with voice command recognition function, such as intelligent terminals, smart home devices (such as smart speakers, smart washing machines, etc.), smart wearable devices (such as smartwatches), in-vehicle intelligent central control systems (which wake up small programs in the terminal to perform different tasks through voice commands), or AI intelligent medical devices (which are woken up and triggered by voice commands), etc.

[0092] The structure of the text information processing device according to an embodiment of the present invention will be described in detail below. The text information processing device can be implemented in various forms, such as a dedicated terminal with text information processing function, or a server equipped with text information processing function, for example, the preceding... Figure 1 Server 200 in the middle. Figure 2 This is a schematic diagram of the composition structure of the text information processing device provided in the embodiments of the present invention. It can be understood that... Figure 2 Only an exemplary structure of the text information processing device is shown, not the entire structure; it can be implemented as needed. Figure 2 The structure shown may be part or all of the structure.

[0093] The text information processing apparatus provided in this embodiment of the invention includes: 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 via a bus system 205. It can be understood that the bus system 205 is used to realize the connection and communication between these components. In addition to a data bus, the bus system 205 also includes a power bus, a control bus, and a status signal bus. However, for clarity, in... Figure 2 The general labeled all buses as Bus System 205.

[0094] The user interface 203 may include a monitor, keyboard, mouse, trackball, click wheel, buttons, touchpad, or touch screen.

[0095] It is understood that memory 202 can be volatile memory or non-volatile memory, or both. In this embodiment of the invention, memory 202 is capable of storing data to support the operation of the terminal (e.g., 10-1). Examples of this data include any computer programs used to operate on the terminal (e.g., 10-1), such as operating systems and applications. The operating system includes various system programs, such as the framework layer, core library layer, driver layer, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications.

[0096] In some embodiments, the text information processing device provided in this invention can be implemented using a combination of hardware and software. For example, the text information processing device provided in this invention can be a processor in the form of a hardware decoding processor, which is programmed to execute the text information processing method provided in this invention. For instance, the processor in the form of a hardware decoding processor can employ one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0097] As an example of the text information processing device provided in this embodiment of the invention, which is implemented using a combination of hardware and software, the text information processing device provided in this embodiment of the invention can be directly embodied as a combination of software modules executed by processor 201. The software modules can be located in a storage medium, which is located in memory 202. Processor 201 reads the executable instructions included in the software modules in memory 202 and combines them with necessary hardware (e.g., including processor 201 and other components connected to bus 205) to complete the text information processing method provided in this embodiment of the invention.

[0098] As an example, processor 201 can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., wherein the general-purpose processor can be a microprocessor or any conventional processor, etc.

[0099] As an example of the hardware implementation of the text information processing device provided in the embodiments of the present invention, the device provided in the embodiments of the present invention can be directly executed by a processor 201 in the form of a hardware decoding processor. For example, it can be executed by one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components to implement the text information processing method provided in the embodiments of the present invention.

[0100] In this embodiment of the invention, the memory 202 is used to store various types of data to support the operation of the text information processing device. Examples of such data include: any executable instructions for operation on the text information processing device, such as executable instructions that can be included in a program implementing the text information processing method of this embodiment of the invention.

[0101] In other embodiments, the text information processing apparatus provided in this invention can be implemented in software. Figure 2A text information processing device stored in memory 202 is shown. This device can be software in the form of programs and plugins, and includes a series of modules. As an example of a program stored in memory 202, it may include the text information processing device itself. The text information processing device includes the following software modules: an information transmission module 2081 and an information processing module 2082. When the software modules in the text information processing device are read into RAM and executed by processor 201, the text information processing method provided in this embodiment of the invention will be implemented. The functions of each software module in the text information processing device in this embodiment of the invention are described below.

[0102] in,

[0103] The information transmission module 2081 is used to acquire text information to be processed.

[0104] The information processing module 2082 is used to perform word segmentation on the text information to be processed and determine at least one candidate keyword vector.

[0105] The information processing module 2082 is used to dynamically adjust the candidate keyword vectors through a word vector processing network to form a candidate keyword vector set, and reward value parameters corresponding to different keyword vectors in the candidate keyword vector set.

[0106] The information processing module 2082 is used to determine a keyword extraction strategy that matches the text information to be processed based on the candidate keyword vector set and the reward value parameter through a deep reinforcement learning network.

[0107] The information processing module 2082 is used to extract from the candidate keyword vector set based on the keyword extraction strategy, and obtain at least one keyword vector as the keyword of the text information to be processed.

[0108] according to Figure 2 The electronic device shown, in one aspect of this application, also provides a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform various embodiments and combinations of embodiments provided in the various optional implementations of the above-described text information processing method.

[0109] Combination Figure 2 The illustrated text information processing apparatus illustrates the text information processing apparatus method provided in the embodiments of the present invention. Before introducing the text information processing apparatus method provided in the embodiments of the present invention, the process of extracting keywords in related technologies is first introduced. Figure 3This is a schematic diagram of the keyword extraction process for the problem text in this embodiment of the invention. Specifically, Chinese word segmentation and statistical models can be used for processing. The following is an example of the original problem text "What should middle-aged and elderly people do when they encounter arteriosclerosis". The traditional processing process includes: forming candidate words through word segmentation: middle, elderly, encounter, artery, sclerosis, what to do. Then (1) the word frequency (TF) and inverse document frequency (IDF) are used as the weights of the candidate words; (2) topic models such as LDA are used to select topics, topic words and their probability values; (3) the weights of the candidate words are calculated based on the graph structure, with the candidate words as nodes of the graph, and the weights of the candidate words are calculated using methods such as PageRank; the probability of each candidate word being labeled as a keyword is used through the sequence labeling model, and the top k are selected as keywords based on the weights or probabilities of the candidate words. For example, the keywords are "elderly, artery, sclerosis". The drawback of this process is that it depends on the results of word segmentation in two ways: (1) it depends too much on the results of word segmentation. Word segmentation often brings some fine-grained errors. For example, here "middle" and "elderly" are not segmented into "middle-aged and elderly". (2) Based on statistical and sequence labeling models, at the word vector level, the combination of words is not considered, such as “artery” and “hardening” should be combined into a keyword “arteriosclerosis”.

[0110] To overcome the above-mentioned shortcomings, see Figure 4 , Figure 4 This is an optional flowchart illustrating a text information processing method provided in an embodiment of the present invention. It can be understood that... Figure 4 The steps shown can be performed by various electronic devices running text information processing devices, such as dedicated terminals with search command processing capabilities, mobile phones, or communication devices running search function applets. The following addresses... Figure 4 The steps shown are explained.

[0111] Step 401: The text information processing device acquires the text information to be processed in the text processing environment.

[0112] The sources of the text information to be processed can vary depending on the text processing environment. For example, in a smart medical terminal, the text information to be processed could be medical record information used to assist in diagnosis, or it could be vital sign information and question information entered by the patient. In a media asset usage environment on a mobile terminal, the text information to be processed could be news text information presented to the user by a smart speaker or an in-vehicle intelligent system.

[0113] Step 402: The text information processing device performs word segmentation on the text information to be processed and determines at least one candidate keyword vector.

[0114] In some embodiments of the present invention, the text information to be processed is segmented to determine at least one candidate keyword vector, which can be achieved in the following ways:

[0115] Based on the question text parameter information carried in the text information to be processed, a corresponding word segmentation library is triggered; the question text is segmented using the triggered word segmentation library's word dictionary to form different word-level question texts; noise reduction is performed on the different word-level question texts to form a set of word-level feature vectors corresponding to the question text, wherein the set of word-level feature vectors includes at least one candidate keyword vector. Specifically, the name of the word segmentation library used for segmenting the question text is determined; based on the name of the word segmentation library, the parameters of the word segmentation library matching the word-level feature vectors corresponding to the question text are determined, wherein the parameters of the word segmentation library include: the type of word segmentation library, the name of the word segmentation library, and the version of the word segmentation library. (Refer to...) Figure 5 , Figure 5 This is a schematic diagram illustrating text information processing via a mini-program in an embodiment of the present invention. In conjunction with the description of the preceding embodiments, different terminal devices (e.g., those in the preceding embodiments) Figure 1 The terminals 10-1 and / or 10-2 shown can provide a search bar for entering keywords to be searched on their respective search interfaces (e.g., web pages, information search apps, and WeChat search mini-programs), and a search button for performing data searches on those keywords. Users input question text into the search bar via voice commands. When the terminal device detects a click on the search button, it triggers the server to initiate a corresponding word segmentation instruction. This word segmentation instruction carries the keywords from the search bar, and the server receives the instruction. Alternatively, the terminal device can display popular search keywords on the search interface. When a click on a popular search keyword is detected, the terminal device sends the word segmentation instruction to the server, carrying the popular search keyword, and the server receives the instruction. It should be noted that this embodiment of the invention does not limit the triggering method of the word segmentation instruction.

[0116] Among them, when the same text information is processed using different word segmentation libraries, the formed word-level feature vectors are not exactly the same. Therefore, according to the name of the word segmentation library, the parameters of the word segmentation library that match the word-level feature vector corresponding to the search instruction text are determined, so as to determine the parameters of the word segmentation library used for word segmentation of the search instruction text. For example, when the search instruction text is "The Story of Time, mp3 of Singer A", after being processed using word segmentation library A, a set of word-level feature vectors A (The Story of Time; mp3 of Singer A) corresponding to the search instruction text is formed; after being processed using word segmentation library B, a set of word-level feature vectors B (The Story of Time; Singer A; mp3) corresponding to the search instruction text is formed; after being processed using word segmentation library A1, a set of word-level feature vectors A1 (Time; Story; Singer A; mp3) corresponding to the search instruction text is formed.

[0117] In some embodiments of the present invention, a dynamic noise threshold matching the usage environment of the word vector processing network can be determined; the problem text set is denoised according to the dynamic noise threshold, and a dynamic word segmentation strategy matching the dynamic noise threshold is triggered; according to the dynamic word segmentation strategy matching the dynamic noise threshold, the problem text is segmented to form a set of dynamic word-level feature vectors corresponding to the problem text. Among them, due to the different usage environments of electronic devices, the dynamic noise thresholds matching the usage environments are also different. For example, in the usage environment where a small program on a mobile terminal retrieves media resource information, the dynamic noise threshold matching the usage environment needs to be less than the dynamic noise thresholds in the environments of smart speakers and in-vehicle intelligent systems.

[0118] In some embodiments of the present invention, a fixed noise threshold corresponding to the usage environment of the word vector processing network can be determined; the problem text set is denoised according to the fixed noise threshold, and a fixed word segmentation strategy matching the fixed noise threshold is triggered; according to the fixed word segmentation strategy matching the fixed noise threshold, the problem text is segmented to form a set of fixed word-level feature vectors corresponding to the problem text. Among them, when the word vector processing network is固化 in a corresponding hardware mechanism, such as the intelligent triage system in a hospital, and the usage environment is for processing and keyword extraction of spoken instructions related to medical information, since the noise is relatively single, by setting the fixed noise threshold corresponding to the fixed word vector processing network, the training speed of the word vector processing network can be effectively improved, reducing the waiting time of users and making the word vector processing network more adaptable to the intelligent triage usage environment in the hospital.

[0119] It should be noted that the word "固化" in the original text seems to be an incorrect or non-standard term. It might be a misspelling. I've translated it as "固化" as it is in the original, but it's recommended to check and correct it if possible.Step 403: The text information processing device dynamically adjusts the candidate keyword vectors through a word vector processing network to form a set of candidate keyword vectors and reward value parameters corresponding to different keyword vectors in the set of candidate keyword vectors.

[0120] In some embodiments of the present invention, candidate keyword vectors are dynamically adjusted through a word vector processing network to form a set of candidate keyword vectors, and reward value parameters corresponding to different keyword vectors in the set of candidate keyword vectors are respectively implemented in the following ways:

[0121] The word vector processing network utilizes a Monte Carlo tree search algorithm to delete and merge candidate keyword vectors, dynamically adjusting their quantity. It determines the frequency of each candidate keyword vector in the standard corpus and the frequency of the matching character vector in the standard corpus. Based on these frequency information, it determines the reward value parameters corresponding to different keyword vectors in the set. Specifically, it can also determine the model parameters of the word vector processing network adapted to the text processing environment, and then use a matching Monte Carlo tree search algorithm to delete or merge candidate keyword vectors, dynamically adjusting their quantity. For example, the query = "What to do when middle-aged and elderly people encounter arteriosclerosis" is used. Here, the operation involves a pair of candidate words, such as "middle-aged" - "elderly people", "encounter" - "artery", "sclerosis" - "what to do", etc. After all candidate word pairs have been processed, they are grouped into a set, denoted by <>, such as <"middle-aged and elderly people", "encounter", "artery", "hardening", "what to do">. Specifically, the operations that can be performed on each candidate word pair include the following categories as shown in Table 1:

[0122] Table 1

[0123] NO_ACTION: No operation. "arteriosclerosis" MERGE_LEFT: Left merge "arteriosclerosis" MERGE_RIGHT: Right merge "Arteriosclerotic arteries" DELETE_LEFT: Left delete "hardening" DELETE_RIGHT: Right-click to delete "artery" DELETE_ALL: Delete all "”

[0124] Furthermore, the reward calculation here can begin with a relatively large corpus of articles. From this corpus, the probability of each term appearing in articles containing all the characters can be calculated. For example, if the term "artery" contains the characters "movement" and "pulse," and there are m articles in the entire corpus containing both "movement" and "pulse," with frequencies of m1 and m2 respectively, and the number of articles containing "artery" among these m articles is n, with a frequency of n1, then the reward parameter for "artery" can be calculated using Formula 1:

[0125]

[0126] Step 404: The text information processing device uses a deep reinforcement learning network to determine a keyword extraction strategy that matches the text information to be processed, based on the candidate keyword vector set and the reward value parameter.

[0127] Specifically, a deep reinforcement learning network can be used to determine, based on the candidate keyword vector set and reward value parameters, whether to extract keywords matching the text information to be processed through the policy subnetwork of the deep reinforcement learning network or to extract keywords matching the text information to be processed through the value subnetwork of the deep reinforcement learning network.

[0128] Step 405: The text information processing device extracts from the candidate keyword vector set based on the keyword extraction strategy to obtain at least one keyword vector as a keyword of the text information to be processed.

[0129] Continue to refer to Figure 6 , Figure 6 This is an optional flowchart illustrating a text information processing method provided in an embodiment of the present invention. It can be understood that... Figure 6 The steps shown can be performed by various terminals running text information processing devices, such as dedicated terminals with search command processing capabilities or electronic devices running smart search mini-programs. The following section addresses... Figure 6 The steps shown are explained.

[0130] Step 601: Determine the weight parameters that match the policy subnetwork of the deep reinforcement learning network using the reward value parameters.

[0131] Step 602: When the candidate word vector state is fixed, in response to the weight parameters, the policy subnetwork of the deep reinforcement learning network generates keywords corresponding to the text information to be processed and the selection probability of the keywords based on the candidate keyword vector set.

[0132] Step 603: Based on the selection probability of the keywords, obtain at least one keyword vector as the keyword of the text information to be processed.

[0133] Step 604: Determine the state-action value function that matches the value subnetwork of the deep reinforcement learning network using the reward value parameter.

[0134] Step 605: When the candidate word vector state is not fixed, in response to the state action value function, through the value subnetwork of the deep reinforcement learning network, based on the candidate keyword vector set and the state value of the corresponding keyword vector, generate keywords corresponding to the text information to be processed and the selection probability of the keywords.

[0135] Step 606: Based on the selection probability of the keywords, obtain at least one keyword vector as the keyword of the text information to be processed.

[0136] Thus, the keywords of the text information to be processed can be obtained.

[0137] Furthermore, the characteristics of the usage environment of the text information to be processed can be determined; a set of training samples matching the characteristics of the usage environment of the text information to be processed can be obtained from the data source; the deep reinforcement learning network can be trained according to the feature set matching the training samples and the corresponding question text labels, so as to determine the model parameters that are adapted to the value subnetwork and policy subnetwork in the deep reinforcement learning network.

[0138] refer to Figure 7 and Figure 8 , Figure 7 This is a schematic diagram illustrating the text information processing effect in an embodiment of the present invention. Figure 8 This is a schematic diagram illustrating the text information processing effect in an embodiment of the present invention. The user interface includes a first-person perspective view of the search function mini-program within the instant messaging software process. The user interface also includes a display control component. This component controls the display of search results matching the question text input by the user via voice. For example, if a user inputs the question text "What songs, movies, and TV series does celebrity A have?" via voice command in the WeChat process, the text information processing method provided in this application can generate keywords "celebrity A," "songs," "movies," and "TV series" for the text information to be processed. The provided search results are search results A1 related to the keywords "celebrity A," "songs," "movies," and "TV series"; or refer to... Figure 8 Among them, such as Figure 8As shown, the short video playback interface can be displayed in the corresponding short video APP or triggered through a WeChat mini program (the deep reinforcement learning network can be trained and packaged into the corresponding APP or saved in the WeChat mini program as a plugin). The short video can respond to the question text entered by the user through voice command through the corresponding application and recommend the results to the user. Effective subsequent recommendation of related videos can effectively improve the user experience. In the short video process, the user enters the question text "Celebrity B's works this year" through voice command. The text information processing method provided in this application determines that the keywords of the question text are "Celebrity B" and "works this year". The search results provided are works of "Celebrity B" marked as "Works this year B1".

[0139] Figure 9 This is a schematic diagram of an optional text information processing process in an embodiment of the present invention. When the word vector processing network is embedded in a corresponding hardware device, such as an intelligent diagnosis and treatment system in a hospital, and the usage environment involves processing and extracting keywords from spoken instructions related to medical information, the user inputs a voice command as the question text. The intelligent diagnosis and treatment system extracts keywords from the question text using the text information processing method provided in this application and obtains corresponding response information based on the keywords. The text information processing method of the intelligent diagnosis and treatment system is illustrated using the question text "What should middle-aged and elderly people do if they encounter arteriosclerosis?" as an example. Specifically, it includes the following steps:

[0140] Step 901: Perform Chinese word segmentation on the problem text to obtain a candidate word set S.

[0141] in, Figure 10 This diagram illustrates an optional data structure for the text information processing method in this embodiment of the invention. Specifically, it first uses a Monte Carlo Tree Search (MCTS) algorithm to sample and generate possible keyword combinations. MCTS is a general term for a class of tree search algorithms that can solve problems with huge exploration spaces; for example, Go algorithms are based on MCTS. MCTS is a tree-based search algorithm that balances exploration and utilization and remains relatively effective even with a large search space. After dynamically adjusting the candidate keyword vectors using the MCTS algorithm, the reward value corresponding to different keyword vectors is calculated. Then, a Deep Reinforcement Learning (Deep RL) network is used to learn the policy network. The Deep RL network includes a policy sub-network and a value sub-network to perform the extraction process, determining keywords that match the text information to be processed through different keyword extraction strategies.

[0142] Step 902: Determine the newly added variable F after word segmentation, store the operation status of all word pairs in the set, and determine the initial operation status as 0.

[0143] in, Figure 11 This is a schematic diagram of the Monte Carlo tree search algorithm in an embodiment of the present invention. For the text information to be processed, "What should middle-aged and elderly people do when they encounter arteriosclerosis", the initial candidate word set <"middle", "elderly", "encounter", "artery", "sclerosis", "what to do"> obtained by word segmentation is serialized in a simulated operation, and the corresponding reward value is calculated. The processing result of the Monte Carlo tree search algorithm will eventually generate a tree structure. Each node in the tree represents a combination possibility, and each node has a corresponding value. Here, the value is calculated according to the back propagation process of the reward value of each node in the sequence.

[0144] Step 903: Check the state of all word vectors in variable F and determine whether the operation state is all 1. If the operation state of all word vectors is 1, execute step 904; otherwise, execute step 908.

[0145] Step 904: Select any word vector with an operation state of 0 from variable F as the starting node of the Monte Carlo tree search algorithm.

[0146] Step 905: Update the operation state of the selected word vector to 1.

[0147] Step 906: Perform keyword extraction (action) according to the keyword extraction strategy. When the action is not NO_ACTION, a new set S' is obtained, and S is updated to S'. When the action is NO_ACTION, step 904 is executed iteratively.

[0148] Since the keyword extraction process from the entire text is a sequential task, the deep reinforcement learning network selects an appropriate keyword extraction strategy based on the simulation results of the Monte Carlo tree search algorithm. The settings for the deep reinforcement learning network are shown in Table 2.

[0149] Table 2

[0150]

[0151] Step 907: Reset variable F, update all word vectors, and adjust the operation state to 0.

[0152] Specifically, by using a deep reinforcement learning network, based on the candidate keyword vector set and reward value parameters, when determining the policy subnetwork of the deep reinforcement learning network to extract keywords that match the text information to be processed, the probability p(action|state) of the action can be maximized given the state. At the same time, the reward value is used as a weight, and the action with the highest probability is selected according to the policy during the prediction stage to obtain the keyword.

[0153] Furthermore, when it is determined that the keywords matching the text information to be processed are extracted through the value subnetwork of the deep reinforcement learning network, the state-action value function Q-function of the reward value can be fitted according to the state. In the prediction stage, the action with the largest value is selected according to the value of Q-function to obtain the keywords.

[0154] Step 908: Obtain the word vectors in the set as keywords for the target text.

[0155] The keywords obtained by the text information processing method provided in this application are shown in Table 3. Compared with the processing results of related technologies, the text information processing method of this application can extract keywords without relying on the granularity of words. The extracted keywords are applicable to the corresponding use scenarios, which facilitates the use of keywords by the neural network model.

[0156] Table 3

[0157]

[0158]

[0159] The present invention has the following beneficial technical effects:

[0160] This invention acquires text information to be processed; performs word segmentation on the text information to determine at least one candidate keyword vector; dynamically adjusts the candidate keyword vectors through a word vector processing network to form a set of candidate keyword vectors, and sets reward value parameters corresponding to different keyword vectors in the set; uses a deep reinforcement learning network to determine a keyword extraction strategy that matches the text information to be processed based on the set of candidate keyword vectors and the reward value parameters; and extracts at least one keyword vector from the set of candidate keyword vectors based on the keyword extraction strategy to obtain keywords for the text information to be processed. This enables more accurate processing of problem texts, reduces the dependence on word granularity in keyword extraction, makes the extracted keywords applicable to different usage scenarios, reduces the influence of related information in the problem statement on keyword extraction, and improves the user experience.

[0161] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A text information processing method, characterized in that, The method includes: Obtain the text information to be processed in the text processing environment; The text information to be processed is segmented into words to determine at least one candidate keyword vector; By using a word vector processing network and employing a Monte Carlo tree search algorithm, the candidate keyword vectors are deleted and merged, thereby dynamically adjusting the candidate keyword vectors to form a set of candidate keyword vectors. Determine the frequency of each candidate keyword vector in the candidate keyword vector set in the standard corpus; Determine the frequency of occurrence of the character vector matching each candidate keyword vector in the standard corpus; Based on the frequency of occurrence of the candidate keyword vectors in the standard corpus and the frequency of occurrence of the character vectors in the standard corpus, the reward value parameters corresponding to different keyword vectors in the candidate keyword vector set are determined respectively. Using a deep reinforcement learning network, a keyword extraction strategy matching the text information to be processed is determined based on the candidate keyword vector set and the reward value parameter. The deep reinforcement learning network includes a policy subnetwork and a value subnetwork. The keyword extraction strategy is to extract keywords by maximizing the action probability given a candidate word pair state through the policy subnetwork, or to extract keywords by fitting a state-action value function of the reward value according to the state through the value subnetwork. Based on the keyword extraction strategy, the candidate keyword vector set is extracted to obtain at least one keyword vector as the keyword of the text information to be processed.

2. The method according to claim 1, characterized in that, The step of segmenting the text information to be processed to determine at least one candidate keyword vector includes: Based on the question text parameter information carried by the text information to be processed, the corresponding word segmentation library is triggered. The question text parameter information is used to characterize the type of text processing environment corresponding to the question text. Different question text parameter information corresponds to different word segmentation libraries. The question text is segmented using the triggered word segmentation dictionary to form different word-level question texts; The different word-level question texts are subjected to noise reduction processing to form a set of word-level feature vectors corresponding to the question texts, wherein the set of word-level feature vectors includes at least one candidate keyword vector.

3. The method according to claim 2, characterized in that, The method further includes: Determine the name of the word segmentation library used when performing word segmentation on the problem text; Based on the name of the word segmentation library, determine the parameters of the word segmentation library that match the word-level feature vector corresponding to the question text, wherein the parameters of the word segmentation library include: The types of word segmentation libraries, the names of word segmentation libraries, and the versions of word segmentation libraries.

4. The method according to claim 2, characterized in that, The denoising process performed on the different word-level question texts to form a set of word-level feature vectors corresponding to the question texts includes: Determine a dynamic noise threshold that matches the usage environment of the word vector processing network; The problem text set is denoised according to the dynamic noise threshold, and a dynamic word segmentation strategy matching the dynamic noise threshold is triggered. The dynamic word segmentation strategy refers to a word segmentation strategy that performs word segmentation on the different word-level problem texts after denoising according to the dynamic noise threshold. Based on a dynamic word segmentation strategy that matches the dynamic noise threshold, the different word-level question texts after denoising are segmented to form a set of dynamic word-level feature vectors corresponding to the question texts.

5. The method according to claim 2, characterized in that, The denoising process performed on the different word-level question texts to form a set of word-level feature vectors corresponding to the question texts includes: Determine a fixed noise threshold corresponding to the usage environment of the word vector processing network; The problem text set is denoised according to the fixed noise threshold, and a fixed word segmentation strategy matching the fixed noise threshold is triggered. The fixed word segmentation strategy refers to the word segmentation strategy that performs word segmentation on the different word-level problem texts after denoising according to the fixed noise threshold when the word vector processing network is fixed on a specific hardware device. Based on a fixed word segmentation strategy that matches the fixed noise threshold, the different word-level question texts after denoising are segmented to form a fixed set of word-level feature vectors corresponding to the question texts.

6. The method according to claim 1, characterized in that, The method further includes: Based on the text processing environment of the text information to be processed, determine the model parameters of the word vector processing network that are adapted to the text processing environment; Based on the model parameters of the word vector processing network adapted to the text processing environment, the candidate keyword vectors are deleted or merged using a matching Monte Carlo tree search algorithm, and the number of candidate keyword vectors is dynamically adjusted.

7. The method according to claim 1, characterized in that, The step of determining a keyword extraction strategy that matches the text information to be processed based on the candidate keyword vector set and reward value parameters using a deep reinforcement learning network includes: Using a deep reinforcement learning network, based on the candidate keyword vector set and reward value parameters, a policy subnetwork is determined. Based on the state values ​​of the candidate keyword vectors in the candidate keyword vector set, the probability of each operation in the deletion and merging operations is maximized. Using the reward value parameters as weights, the operation with the highest probability is selected during the prediction phase to extract keywords that match the text information to be processed. Keywords that match the text information to be processed are extracted through the value subnetwork of the deep reinforcement learning network.

8. The method according to claim 7, characterized in that, The step of extracting at least one keyword vector from the candidate keyword vector set based on the keyword extraction strategy to obtain at least one keyword vector as a keyword for the text information to be processed includes: The weight parameters that match the policy subnetwork of the deep reinforcement learning network are determined using the reward value parameters. In response to the weight parameters, the policy subnetwork of the deep reinforcement learning network generates keywords corresponding to the text information to be processed and the probability of the keywords being selected, based on the candidate keyword vector set. Based on the selection probability of the keywords, at least one keyword vector is obtained as the keyword of the text information to be processed.

9. The method according to claim 7, characterized in that, The step of extracting at least one keyword vector from the candidate keyword vector set based on the keyword extraction strategy to obtain at least one keyword vector as a keyword for the text information to be processed includes: The reward value parameter is used to determine the state-action value function that matches the value subnetwork of the deep reinforcement learning network. The state-action value function is used to evaluate the expected reward value that can be obtained by performing each operation in the deletion and merging of the keyword vector based on the state value of the corresponding keyword vector in the candidate keyword vector set. In response to the state action value function, the value subnetwork of the deep reinforcement learning network generates keywords corresponding to the text information to be processed and the probability of selection of the keywords based on the candidate keyword vector set and the state values ​​of the corresponding keyword vectors. Based on the selection probability of the keywords, at least one keyword vector is obtained as the keyword of the text information to be processed.

10. The method according to claim 1, characterized in that, The method further includes: Determine the characteristics of the usage environment of the text information to be processed; In the data source, obtain a set of training samples that match the characteristics of the usage environment of the text information to be processed; The deep reinforcement learning network is trained based on the feature set that matches the training samples and the corresponding question text labels, so as to determine the model parameters that are suitable for the value subnetwork and policy subnetwork in the deep reinforcement learning network.

11. A text information processing device, characterized in that, The device includes: The information transmission module is used to acquire the text information to be processed in the text processing environment; The information processing module is used to perform word segmentation on the text information to be processed and determine at least one candidate keyword vector; The information processing module is used to dynamically adjust the candidate keyword vectors by deleting and merging them using a Monte Carlo tree search algorithm through a word vector processing network; determine the frequency of each candidate keyword vector in the standard corpus; determine the frequency of the character vector matching each candidate keyword vector in the standard corpus; and determine the reward value parameters corresponding to different keyword vectors in the candidate keyword vector set based on the frequency of the candidate keyword vectors and the frequency of the character vectors in the standard corpus. The information processing module is used to determine a keyword extraction strategy that matches the text information to be processed based on the candidate keyword vector set and the reward value parameter through a deep reinforcement learning network. The deep reinforcement learning network includes a policy subnetwork and a value subnetwork. The keyword extraction strategy is to extract keywords by maximizing the action probability given a candidate word pair state through the policy subnetwork, or to extract keywords by fitting a state-action value function of the reward value according to the state through the value subnetwork. The information processing module is used to extract from the candidate keyword vector set based on the keyword extraction strategy, and obtain at least one keyword vector as the keyword of the text information to be processed.

12. The apparatus according to claim 11, characterized in that, The information processing module is used to trigger the corresponding word segmentation library according to the question text parameter information carried by the text information to be processed. The question text parameter information is used to characterize the type of text processing environment corresponding to the question text. Different question text parameter information corresponds to different word segmentation libraries. The information processing module is used to segment the question text using the triggered word segmentation dictionary to form different word-level question texts; The information processing module is used to perform noise reduction processing on the different word-level question texts to form a set of word-level feature vectors corresponding to the question texts, wherein the set of word-level feature vectors includes at least one candidate keyword vector.

13. An electronic device, characterized in that, The electronic device includes: Memory, used to store executable instructions; A processor, when executing executable instructions stored in the memory, implements the text information processing method according to any one of claims 1 to 10.

14. A computer-readable storage medium storing executable instructions, characterized in that, When the executable instructions are executed by the processor, they implement the text information processing method according to any one of claims 1 to 10.

15. A computer program product comprising computer instructions, characterized in that, When the computer instructions are executed by the processor, they implement the text information processing method according to any one of claims 1 to 10.

Images