Conversational intelligent securities and futures trading system based on languages and characters

Abstract

The invention relates to the technical field of security transaction assistance, and particularly discloses a language and character-based dialogue type intelligent security and futures transaction system, which comprises a voice processing module, a text analysis module, a transaction generation module, a confirmation and feedback module, a transaction execution module and a query module, according to a traditional transaction system, a user needs to manually input transaction information such as stock codes, buying and selling numbers and prices, and the process is complex and time-consuming. Through the voice and character dialogue function of the system, the input time of a user is greatly shortened through voice interaction, repeated operation is reduced through a quick input template, high-frequency transaction requirements are converted into one-key instructions, the user transaction efficiency is greatly improved, particularly, the user is helped to quickly grasp investment opportunities when high-frequency transactions or the market fluctuates severely, and the user experience is improved. The system pushes transaction states and market information through real-time messages, ensures that a user masters investment trends in real time, and provides intelligent prompting and early warning functions, such as market fluctuation risks or investment portfolio adjustment suggestions.

Description

Technical Field

[0001] The present invention belongs to the technical field of securities trading assistance, and particularly relates to a conversational intelligent securities and futures trading system based on language and text. Background Art

[0002] In existing securities (futures) trading software, the trading method is the traditional human-computer interaction method. Investors need to manually input most of the trading elements, such as the trading direction (buy or sell), security code, trading price, trading quantity, order type (limit order or market order, etc.). After the trading order is issued, it is necessary to manually switch the software interface to view the status of the entrustment, transaction details, position situation, etc. The main disadvantages and existing problems of this traditional trading method are as follows:

[0003] There are many trading elements that need to be input for the order form, and investors need to remember and calculate more content, such as stock codes, the number of shares corresponding to the purchase amount, etc., which are prone to errors.

[0004] The manual filling speed of the order form is relatively slow, especially for middle-aged and elderly users.

[0005] The feedback of the entrustment result is not timely, and users need to query it by themselves, which is easy to forget.

[0006] For users with a large number of individual stocks in their positions, users need to remember more content, which is easy to forget and make mistakes.

[0007] There is no record of the user's order input stream, and it is easy to have disputes with securities (futures) companies in case of problems. Securities (futures) companies need to invest additional resources and manpower to handle them, and it may trigger public opinion events.

[0008] Compared with the current popular program trading, the convenience and intelligence level of the traditional trading entrustment method have a huge gap. For example, program trading can easily achieve batch trading, and with the help of a computer, a large number of entrustments can be sent at the millisecond or even microsecond time level, and it can be traded before investors with manual trading. Overall, it crushes investors with manual trading in terms of trading speed and intelligence level, which is unfair to investors with manual trading.

[0009] In view of this, the inventor proposes a conversational intelligent securities and futures trading system based on language and text to solve the above problems. Summary of the Invention

[0010] The purpose of the present invention is to provide a conversational intelligent securities and futures trading system based on language and text to solve the problems raised in the above background art.

[0011] To achieve the above purpose, the present invention provides the following technical solutions:

[0012] A conversational intelligent securities and futures trading system based on language and text, comprising:

[0013] A voice processing module, configured to receive a user's voice input, convert the voice into text through a voice recognition technology based on a deep neural network, and obtain a user instruction text;

[0014] A text parsing module, configured to perform keyword extraction and semantic analysis on the user instruction text by using a bidirectional long short-term memory network, identify the user's intention, and generate a standardized trading instruction;

[0015] A trading generation module, configured to generate an order based on the trading instruction by using an automatic order generation method based on constraint optimization, and push the order to the user for confirmation to obtain a user instruction;

[0016] A confirmation and feedback module, configured to receive the user instruction and execute a confirmation or cancellation instruction for the order; when the user confirms, send a trading instruction to the trading execution module, otherwise do not send a trading instruction;

[0017] A trading execution module, configured to receive a trading instruction from the confirmation and feedback module and execute the trading instruction, and feedback the trading result to the user in real time;

[0018] A query module, configured to receive and process a query instruction by using an inverted index method, retrieve account, position or market information, and return a query result.

[0019] Preferably, the expression of the deep neural network is:

[0020]

[0021] P(W∣X): The probability of the text sequence W given the voice input X;

[0022] T: The number of time steps of the voice input;

[0023] wt: The predicted word or character at time step t;

[0024] X: Acoustic features, such as Mel frequency cepstral coefficients, MFCC.

[0025] Preferably, the formula of the bidirectional long short-term memory network is:

[0026] ht = BiLSTM(x t ,h t-1 )

[0027] xt: The word vector representation of the input text;

[0028] ht: The hidden state of the bidirectional LSTM;

[0029] αt: Attention weight, representing the importance of each time step;

[0030] ct: Semantic context vector.

[0031] Preferably, the formula of the automatic order generation method based on constraint optimization is:

[0032] min x ||x - u|| 2

[0033] s.t. x ∈ F

[0034] where x: The parameter set of the generated order, including price and quantity;

[0035] u: The initial parameters input by the user;

[0036] F: The system constraint set, including the minimum trading unit and price range.

[0037] Preferably, the formula of the inverted index method is:

[0038] Rq = ∪ t∈Q Index(t)

[0039] where Rq: The query result set;

[0040] Q: The keyword set of the query instruction;

[0041] Index(t): The inverted index list of the keyword t.

[0042] Preferably, the system further includes a chat record storage module for using hash indexing to record all interaction data between the user and the system, including voice and text instructions, and supporting fast retrieval and backtracking.

[0043] Preferably, the formula of the hash indexing is:

[0044] h(k) = (a · k + b) mod p

[0045] where h(k): The hash value of the record key k;

[0046] a, b, p: The parameters of the hash function, where p is a prime number;

[0047] k: The query conditions input by the user, including date and keyword.

[0048] Preferably, the system further includes an intelligent warning module for implementing intelligent prompts for the user's historical operations and system status according to the anomaly detection method, and providing a quick input template to simplify the operation;

[0049] The user permission management module is used to verify user identities, allocate access permissions, and protect transaction security;

[0050] The background management and logging module is used to record the system operation status, transaction logs, and provide system maintenance interfaces.

[0051] Preferably, the formula of the anomaly detection method is:

[0052]

[0053] Where S: anomaly score;

[0054] x i: market parameters, including price and trading volume;

[0055] μ, σ: mean and standard deviation of historical data;

[0056] N: number of samples of historical data.

[0057] Compared with the prior art, the beneficial effects of the present invention are:

[0058] (1) Through the voice and text dialogue functions of the present system: Voice interaction significantly shortens the user's input time; The quick input template reduces repetitive operations, converts high-frequency trading requirements into one-key instructions, and greatly improves the user's trading efficiency. Especially in high-frequency trading or when the market fluctuates violently, it helps users quickly seize investment opportunities.

[0059] (2) The system of the present invention pushes trading status and market information in real time through messages, ensuring that users can master investment dynamics in real time, and providing intelligent prompts and warning functions, such as market volatility risks or investment portfolio adjustment suggestions, reducing the learning cost and usage threshold, enabling investors, especially middle-aged and elderly users, to quickly adapt and use efficiently; Programmatic trading has a speed advantage, and it is difficult for traditional manual traders to compete with it. Through this system: It realizes batch instruction input and high-concurrency execution similar to programmatic trading, narrowing the efficiency gap between the two. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is one of the block diagrams of a dialogue-based intelligent securities and futures trading system based on language and text of the present invention;

[0061] Figure 2 is the second block diagram of a dialogue-based intelligent securities and futures trading system based on language and text of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0062] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0063] Embodiment 1:

[0064] Please refer to Figure 1 and Figure 2 As shown in, a conversational intelligent securities and futures trading system based on language and text includes:

[0065] A voice processing module for receiving a user's voice input, converting the voice into text through a voice recognition technology based on a deep neural network, and obtaining a user instruction text;

[0066] A text parsing module for extracting keywords and performing semantic analysis on the user instruction text by using a bidirectional long short-term memory network, identifying the user's intention, and generating a standardized trading instruction;

[0067] A trading generation module for generating an order form based on the trading instruction by using an automatic order form generation method based on constraint optimization, and pushing the order form to the user for confirmation to obtain a user instruction;

[0068] A confirmation and feedback module for receiving the user instruction and executing a confirmation or cancellation instruction for the order form; when the user confirms, sending a trading instruction to the trading execution module, otherwise not sending a trading instruction;

[0069] A trading execution module for receiving the trading instruction from the confirmation and feedback module and executing the trading instruction, and real-time feedback the trading result to the user;

[0070] A query module for receiving and processing a query instruction by using an inverted index method, retrieving account, position or market information, and returning a query result.

[0071] Specifically, the expression of the deep neural network is:

[0072]

[0073] P(W∣X): The probability of the text sequence W given the voice input X;

[0074] T: The number of time steps of the voice input;

[0075] wt: The predicted word or character at time step t;

[0076] X: Acoustic features, such as Mel frequency cepstral coefficients, MFCC;

[0077] Acoustic model: Extract speech features X, including MFCC or spectrogram;

[0078] Language model: Predict the most likely wt according to the context;

[0079] Decoder: Combine the outputs of the acoustic model and the language model to generate the final text;

[0080] The accuracy of speech-to-text is greatly improved, especially in noisy environments; reduce the misoperation of trading instructions caused by speech recognition errors.

[0081] Specifically, the formula of the bidirectional long short-term memory network is:

[0082] ht = BiLSTM(x t ,h t-1 )

[0083] xt: Word vector representation of the input text;

[0084] ht: Hidden state of the bidirectional LSTM;

[0085] αt: Attention weight, indicating the importance of each time step;

[0086] ct: Semantic context vector;

[0087] Word vector: Generated by a pre-trained model, and the pre-trained model includes Word2Vec, BERT;

[0088] Attention weight: Perform weighted analysis on the text according to the keywords in the instruction;

[0089] BiLSTM hidden state: Capture bidirectional context information in the text;

[0090] Effect: Improve the ability to understand fuzzy or colloquial instructions;

[0091] Significance: Improve the accuracy of text parsing and avoid misinterpretation of trading instructions.

[0092] Specifically, the formula of the automatic order generation method based on constraint optimization is:

[0093] min x ||x - u|| 2

[0094] s.t. x ∈ F

[0095] Where x: The parameter set of the generated order, including price and quantity;

[0096] u: Initial parameters input by the user;

[0097] F: The set of system constraints, including the minimum trading unit and price range;

[0098] Price range: The price fluctuation range restricted by market rules;

[0099] Quantity unit: The minimum trading unit of securities or futures;

[0100] User input: The trading instructions of the user, including the stock code and the purchase quantity;

[0101] Quickly generate an order form that complies with market rules, reduce the input workload of users, and avoid transactions that fail due to non-compliant parameters.

[0102] Specifically, the formula of the inverted index method is:

[0103] Rq = ∪ t∈Q Index(t)

[0104] where Rq: The set of query results;

[0105] Q: The set of keywords of the query instruction;

[0106] Index(t): The inverted index list of keyword t;

[0107] Keywords: The core words in the user's query instruction, including the stock code and account name;

[0108] Inverted index: Map keywords to corresponding data records;

[0109] Improve the query response speed, immediately return the information required by the user, optimize the user experience, and enhance the system response efficiency.

[0110] Specifically, the system further includes a chat record storage module for using hash indexing to record all interaction data between the user and the system, including voice and text instructions, and supporting quick retrieval and backtracking.

[0111] Specifically, the formula of the hash index is:

[0112] h(k) = (a·k + b) mod p

[0113] where h(k): The hash value of record key k;

[0114] a, b, p: The parameters of the hash function, where p is a prime number;

[0115] k: The query condition input by the user, including the date and keyword;

[0116] Key-value pair: Store the date and content of the user interaction record;

[0117] Hash function parameters: Ensure the minimization of hash collisions;

[0118] Implement fast retrieval of users' historical transaction records, reduce the search latency of the storage system, and improve the retrieval efficiency.

[0119] Specifically, the system further includes an intelligent warning module, which is used to generate intelligent prompts for users' historical operations and system status according to the anomaly detection method, and provide a quick input template to simplify the operation;

[0120] User permission management module, which is used to verify user identities, allocate access permissions, and protect transaction security;

[0121] Background management and log module, which is used to record the system operation status, transaction logs, and provide a system maintenance interface.

[0122] Specifically, the formula of the anomaly detection method is:

[0123]

[0124] Where S: Anomaly score;

[0125] x i: Market parameters, including price and trading volume;

[0126] μ, σ: Mean and standard deviation of historical data;

[0127] N: Number of samples of historical data;

[0128] Market parameters: Include stock price, volatility, trading volume;

[0129] Historical data window: Time range for calculating the mean and standard deviation;

[0130] Threshold setting: Trigger a warning when S exceeds the threshold;

[0131] The above formula is used to timely detect abnormal market fluctuations, remind users to take actions, help users quickly respond to market risks, and avoid losses.

[0132] As can be seen from the above, the interactive interface of the innovative entrusted transaction is intuitive and clear, without the need for investors to manually input, reducing the probability of input errors and ensuring the accuracy of transactions.

[0133] The trading status is pushed in a timely manner, and investors no longer need to switch interfaces to check entrustments and transactions. Greatly improve the trading efficiency.

[0134] The interaction between investors and the system is saved in the form of chat records. The instruction flow of investors is clear, and it is easy to investigate in case of disputes, and the responsibility can be quickly determined.

[0135] Text instructions can be used as a backup for voice instructions. In case the voice recognition effect is poor in a noisy environment, text instructions can be directly used. The "quick input" template set in advance can also be used.

[0136] The business entrance is unified. Investors no longer need to spend effort learning and getting familiar with various operation interfaces with different functions, which greatly enhances the expandability of the entire APP's functions. Various query functions such as conditional orders, batch orders, and positions can be easily implemented, and reminder functions can also be added.

[0137] Intelligent tools can also be used in manual trading, narrowing the huge gap between manual trading and programmed trading in terms of efficiency and fairness, improving the investment environment, and contributing to the stable and healthy development of the capital market.

[0138] In the future, it can be further expanded to non-trading operations to save labor.

[0139] Traditional trading systems require users to manually enter trading information such as stock codes, trading quantities, prices, etc. The process is complex and time-consuming. Through the voice and text dialogue functions of this system: Voice interaction significantly shortens the user's input time; the quick input template reduces repetitive operations, converts high-frequency trading requirements into one-key instructions, and greatly improves the user's trading efficiency. Especially in high-frequency trading or when the market fluctuates violently, it helps users quickly seize investment opportunities.

[0140] Voice recognition and intelligent parsing: Automatically understand the user's natural language instructions and reduce the memory burden.

[0141] Instruction confirmation mechanism: Ensure that all trading instructions are confirmed by the user before execution.

[0142] Fuzzy parsing and error correction: Intelligently identify possible input errors of the user and prompt for correction.

[0143] Reduce the trading error rate, safeguard the user's asset security, and reduce economic losses caused by operational mistakes.

[0144] Users interact with the system in natural language without having to learn complex operation processes.

[0145] The system pushes trading status and market information through real-time messages to ensure that users can grasp investment dynamics in real time, and provides intelligent tips and warning functions such as market volatility risks or investment portfolio adjustment suggestions, reducing the learning cost and usage threshold, enabling investors, especially middle-aged and elderly users, to quickly adapt and use it efficiently.

[0146] Programmed trading has a speed advantage, and it is difficult for traditional manual traders to compete with it. Through this system: Batch instruction input and high-concurrency execution similar to programmed trading are achieved, narrowing the efficiency gap between the two.

[0147] Example 2:

[0148] Application of Batch Voice Command Transactions

[0149] Scenario Description

[0150] The user hopes to issue multiple transaction instructions at once via voice:

[0151] Instruction 1: Buy 1 lot of "Kweichow Moutai" stock at market price.

[0152] Instruction 2: Buy 1,000 shares of "Industrial and Commercial Bank of China" at a limit price of 10 yuan per share.

[0153] Operation Steps and Parameters

[0154] The user inputs a voice command:

[0155] Voice Content: "Help me buy 1 lot of Kweichow Moutai at the latest price, and at the same time buy 1,000 shares of Industrial and Commercial Bank of China with a limit price of 10 yuan."

[0156] Voice Length: 5 seconds, file size 1.2MB

[0157] Background Noise Level: 50dB (quiet environment)

[0158] Voice Processing Module:

[0159] Voice Recognition Result:

[0160] Converted Text: "Help me buy 1 lot of Kweichow Moutai at the latest price, and at the same time buy 1,000 shares of Industrial and Commercial Bank of China with a limit price of 10 yuan."

[0161] Conversion Time: 1 second

[0162] Accuracy Rate: 99% (clear voice, no errors)

[0163] Text Parsing Module:

[0164] Parsing Result: Two transaction instructions:

[0165] Instruction 1: Stock Code 600519, buy, 1 lot (market price).

[0166] Instruction 2: Stock Code 601398, buy, 1,000 shares, limit price 10 yuan per share.

[0167] Parsing Time: 0.6 seconds

[0168] Transaction Generation Module:

[0169] Generation of Entrustment Form:

[0170] Instruction 1: Kweichow Moutai, buy, market price, quantity 100 shares.

[0171] Instruction 2: Industrial and Commercial Bank of China, buy, limit price 10 yuan / share, quantity 1000 shares.

[0172] Order generation time: 0.5 seconds

[0173] Confirmation and feedback module:

[0174] Feedback content:

[0175] "Confirm to buy 1 lot of Kweichow Moutai at the latest price, and the current latest price is 1519 yuan / share? At the same time, confirm to buy 1000 shares of Industrial and Commercial Bank of China at a limit price of 10 yuan per share?"

[0176] User confirmation time: 3 seconds, by replying "confirm".

[0177] Transaction execution module:

[0178] Execute instruction:

[0179] Instruction 1: Traded at the market price of 1519 yuan / share, amount 151900 yuan.

[0180] Instruction 2: Not traded at the limit price (current market price 10.1 yuan).

[0181] Execution time: 2 seconds

[0182] As can be seen from the above, the user completes the input and parsing of two complex instructions within 5 seconds, while the traditional operation takes at least 20 seconds.

[0183] Intelligent improvement: Support batch voice instructions to simplify multiple manual operations of users.

[0184] Feedback timeliness: Clearly prompt users for instructions that are not traded at the limit price to avoid omission of operations.

[0185] Example 3:

[0186] Combined condition query and intelligent prompt

[0187] Scenario description

[0188] The user hopes to query the positions in the account with a loss margin exceeding 5% and a current market value higher than 100,000 yuan, and obtain adjustment suggestions.

[0189] Operation steps and parameters

[0190] The user enters a query instruction:

[0191] Text content: "Query positions with a loss exceeding 5% and a market value higher than 100,000 yuan."

[0192] Text parsing module:

[0193] Parsing result:

[0194] Query condition 1: Loss margin > 5%.

[0195] Query condition 2: Current market value > 100,000 yuan.

[0196] Parsing time: 0.3 seconds

[0197] Query module:

[0198] Database retrieval parameters:

[0199] Query fields: Stock code, name, position quantity, cost price, current price, profit and loss margin, market value.

[0200] Query conditions: Loss > 5%, market value > 100,000 yuan.

[0201] Database results:

[0202]

[0203] Retrieval time: 1 second

[0204] Intelligent prompt module:

[0205] Prompt content:

[0206] "The current loss margin of PetroChina you hold is -6.25%, and the market value is 112,500 yuan; the current loss margin of Hengrui Medicine is -6%, and the market value is 112,800 yuan. It is recommended that you reduce your positions to avoid potential risks."

[0207] Provide quick commands: "Do you need to sell loss-making stocks? You can choose 'Sell 5,000 shares of PetroChina' or 'Sell 300 shares of Hengrui Medicine'."

[0208] User operations and feedback:

[0209] User reply: "Sell 300 shares of Hengrui Medicine."

[0210] The system generates a commission order and completes the market transaction: The current price is 94 yuan / share, and the amount is 28,200 yuan.

[0211] As can be seen from the above, the combined condition query and intelligent prompt reduce the user's screening cost, and the query time only takes 1.3 seconds.

[0212] Intelligent level: Combine the profit and loss margin and market value to prompt investment suggestions to help users optimize decisions.

[0213] Risk aversion: Timely reminder and completion of position reduction operations to avoid further losses.

[0214] As can be seen from the above, Example 2 demonstrates the high efficiency and intelligent feedback ability of the system in batch voice command processing and transactions.

[0215] Example 3 emphasizes the practical value of combined conditional queries and personalized prompts in investment decisions. These applications effectively improve the trading efficiency and investment security of users, and at the same time reflect the innovative capabilities of the system in aspects such as speech recognition, intelligent analysis, and real-time feedback.

[0216] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A conversational intelligent securities and futures trading system based on language and text, characterized in that: include: The voice processing module is used to receive the user's voice input and convert the voice into text through the speech recognition technology based on the deep neural network to obtain the user's command text; A text parsing module, for extracting keywords and performing semantic analysis on the user instruction text based on a bidirectional long short-term memory network, identifying user intentions, and generating standardized transaction instructions; A transaction generation module, used to generate a commission order from the transaction instruction based on an automatic commission order generation method based on constraint optimization, and push the commission order to the user for confirmation to obtain the user instruction; The confirmation and feedback module is used to receive the user's instructions and execute the confirmation or cancellation instructions of the order; when the user confirms, the transaction instruction is sent to the transaction execution module, otherwise the transaction instruction is not sent; The transaction execution module is used to receive the transaction instructions from the confirmation and feedback module and execute the transaction instructions, and provide real-time feedback of the transaction results to the user; The query module is used to receive and process query instructions using the inverted index method, retrieve account, position or market information, and return query results.

2. A conversational intelligent securities and futures trading system based on language and text according to claim 1, characterized in that: The expression of the deep neural network is: P(W|X): the probability of the text sequence W given the speech input X; T: the number of time steps of speech input; wt: the predicted word or character at time step t; X: Acoustic features, including Mel frequency cepstral coefficients, MFCC.

3. A conversational intelligent securities and futures trading system based on language and text according to claim 1, characterized in that: The formula of the bidirectional long short-term memory network is: ht=BiLSTM(x t ,h t-1 ) xt: word vector representation of the input text; ht: hidden state of bidirectional LSTM; αt: attention weight, indicating the importance of each time step; ct: semantic context vector.

4. A conversational intelligent securities and futures trading system based on language and text according to claim 1, characterized in that: The formula of the automatic order generation method based on constraint optimization is: min x ||coins|| 2 stx∈F Where x: the parameter set of the generated order, including price and quantity; u: initial parameter entered by the user; F: A set of system constraints, including the minimum trading unit and price range.

5. The interactive intelligent securities and futures trading system based on language and text according to claim 1 is characterized in that: The formula of the inverted index method is: Rq=∪ t∈Q Index(t) Where Rq: query result set; Q: A set of keywords for the query command.

6. A conversational intelligent securities and futures trading system based on language and text according to claim 1, characterized in that: The system also includes a chat record storage module, which uses a hash index to record all interaction data between the user and the system, including voice and text instructions, and supports fast retrieval and backtracking.

7. A conversational intelligent securities and futures trading system based on language and text according to claim 6, characterized in that: The formula of the hash index is: h(k)=(a·k+b)modp Where h(k): hash value of record key k; a, b, p: parameters of the hash function, satisfying that p is a prime number; k: Query conditions entered by the user, including date and keywords.

8. The interactive intelligent securities and futures trading system based on language and text according to claim 1 is characterized in that: The system also includes an intelligent early warning module for realizing intelligent prompts generated by user historical operations and system status according to the abnormality detection method, and providing a quick input template to simplify the operation; User rights management module, used to verify user identity, assign access rights and protect transaction security; The background management and log module is used to record the system operation status, transaction logs, and provide a system maintenance interface.

9. A language and text-based conversational intelligent securities and futures trading system according to claim 8, characterized in that: The formula of the anomaly detection method is: Where S: abnormality score; xi: market parameters, including price and volume; μ, σ: mean and standard deviation of historical data; N: The number of samples of historical data.

Images