Problem processing method and device based on artificial intelligence, computer equipment and medium

Abstract

The invention belongs to the field of artificial intelligence and the field of financial science and technology, and relates to an artificial intelligence-based question processing method, which comprises the following steps of: querying a plurality of knowledge blocks matched with a query question input by a user from a target knowledge base; performing correlation evaluation on the knowledge blocks and the query question, and generating a correlation score between each knowledge block and the query question; performing sorting processing on the knowledge blocks based on a sequence of the correlation scores from high to low to obtain a knowledge base list; processing the knowledge base list based on the matched knowledge length, and generating target model input corresponding to a preset large model; inputting the target model input and the query problem into the large model, and receiving a query result returned by the large model; and returning the query result to the user. The invention further provides a problem processing device based on artificial intelligence, computer equipment and a storage medium. According to the method, based on the use of the correlation score and the matching knowledge length, the accuracy of the query result output by the large model is effectively improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence development technology and the field of financial technology, and in particular to artificial intelligence-based problem-solving methods, devices, computer equipment and storage media. Background Art

[0002] With the rapid development of artificial intelligence technology, intelligent question-answering systems for human-computer interactive dialogues have been widely used in the financial field, such as insurance plan communication, house purchase plan communication, and so on. In the financial field, large models based on the RAG architecture are often used in the intelligent question-answering systems of financial enterprises. As an innovative large model application, retrieval augmentation generation (RAG) technology significantly improves the knowledge acquisition ability of large models through external knowledge bases, and can incorporate new knowledge without additional training. The core of the large model based on the RAG architecture is to match user queries with knowledge blocks in the knowledge base, and then combine these related information to generate accurate answers.

[0003] However, in the prior art, when processing long documents, large models based on the RAG architecture usually use a fixed length for segmentation to adapt to the model input requirements. Although this processing method is simple, it has significant defects: on the one hand, if the segmentation length is too short, it will destroy the coherence and integrity of the document, causing key information to be omitted or split during the matching process, affecting the accuracy of the final answer. For example, the query may mislead the model due to the failure to match key paragraphs; on the other hand, if the segmentation length is too long, short knowledge segments are easily surrounded by lengthy and irrelevant information, increasing the difficulty of matching and reducing matching efficiency, further limiting the accuracy of the answers output by the large model in complex document processing. Summary of the invention

[0004] The purpose of the embodiments of the present application is to propose an artificial intelligence-based problem-solving method, apparatus, computer equipment and storage medium to solve the technical problem that when processing long documents, the existing large model based on the RAG architecture usually adopts a fixed length for block processing to adapt to the model input requirements. This processing method will result in low accuracy of the answers output by the large model.

[0005] In order to solve the above technical problems, the embodiment of the present application provides a problem solving method based on artificial intelligence, which adopts the following technical solutions:

[0006] Receive query questions input by users;

[0007] Querying a plurality of knowledge blocks matching the query question from a pre-built target knowledge base;

[0008] Performing relevance evaluation on all the knowledge blocks and the query question, and generating a relevance score between each knowledge block and the query question;

[0009] Sorting all the knowledge blocks in descending order based on the relevance scores to obtain a sorted knowledge base list;

[0010] Processing the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model;

[0011] Inputting the target model and the query question into the large model, and receiving the corresponding query result returned by the large model;

[0012] The query result is returned to the user.

[0013] Furthermore, the step of processing the knowledge base list based on the preset matching knowledge length to generate a target model input corresponding to the preset large model specifically includes:

[0014] According to the sorting order of all the knowledge blocks, the knowledge blocks in the knowledge base list are gradually added to the preset model input, and it is detected whether the total length of the current first model input is greater than the matching knowledge length;

[0015] If the total length of the first model input is greater than the matching knowledge length, calculating the difference between the total length of the first model input and the matching knowledge length;

[0016] Determining whether the difference is within a preset value range;

[0017] If the difference is within the preset numerical range, the target model input corresponding to the large model is generated based on the first model input.

[0018] Furthermore, after the step of determining whether the difference is within a preset value range, the method further includes:

[0019] If the difference is not within the preset value range, selecting a first knowledge block with the smallest relevance score from the first model input;

[0020] Eliminating the first knowledge block from the first model input to obtain a corresponding second model input;

[0021] The target model input corresponding to the large model is generated based on the second model input.

[0022] Furthermore, the step of generating the target model input corresponding to the large model based on the first model input specifically includes:

[0023] Get the preset logical order;

[0024] Obtaining all second knowledge blocks contained in the first model input;

[0025] Performing sequence adjustment processing on all the second knowledge blocks based on the logical sequence to obtain an adjusted third knowledge block;

[0026] Performing splicing processing on all the third knowledge blocks to obtain corresponding spliced ​​knowledge block data;

[0027] The spliced ​​knowledge block data is input as the target model corresponding to the large model.

[0028] Furthermore, the step of evaluating the relevance between all the knowledge blocks and the query question to generate a relevance score between each knowledge block and the query question specifically includes:

[0029] Call the preset relevance assessment model;

[0030] Based on the relevance evaluation model, a relevance evaluation is performed on the designated knowledge block and the query question to obtain a relevance evaluation result between the designated knowledge block and the query question; wherein the designated knowledge block is any one of all the knowledge blocks;

[0031] The relevance evaluation result is used as a specified relevance score between the specified knowledge block and the query question.

[0032] Furthermore, the step of returning the query result to the user specifically includes:

[0033] Optimizing the query results to obtain corresponding target query results;

[0034] Determine a target push method corresponding to the target query result;

[0035] Based on the target push mode, the target query result is returned to the user.

[0036] Furthermore, before the step of searching for a plurality of knowledge blocks matching the query question from the pre-built target knowledge base, the method further includes:

[0037] Acquire pre-collected knowledge content and build a corresponding original document library based on the knowledge content;

[0038] Receiving the formatting specification processing of the original knowledge base performed by a designated user to obtain a corresponding document library;

[0039] Calling a preset processing script to perform knowledge block extraction processing on the document library to obtain a corresponding fourth knowledge block;

[0040] Calling the preset initial knowledge base;

[0041] The fourth knowledge block is stored in the initial knowledge base to obtain the target knowledge base.

[0042] In order to solve the above technical problems, the embodiment of the present application further provides a problem processing device based on artificial intelligence, which adopts the following technical solution:

[0043] A first receiving module is used to receive a query question input by a user;

[0044] A query module, used to query a plurality of knowledge blocks matching the query question from a pre-built target knowledge base;

[0045] An evaluation module, used to evaluate the relevance between all the knowledge blocks and the query question, and generate a relevance score between each knowledge block and the query question;

[0046] A sorting module, used to sort all the knowledge blocks in descending order based on the relevance scores to obtain a sorted knowledge base list;

[0047] A generation module, used for processing the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model;

[0048] An input module, used to input the target model and the query question into the large model, and receive the corresponding query result returned by the large model;

[0049] The returning module is used to return the query result to the user.

[0050] In order to solve the above technical problems, the embodiment of the present application further provides a computer device, which adopts the following technical solution:

[0051] Receive query questions input by users;

[0052] Querying a plurality of knowledge blocks matching the query question from a pre-built target knowledge base;

[0053] Performing relevance evaluation on all the knowledge blocks and the query question, and generating a relevance score between each knowledge block and the query question;

[0054] Sorting all the knowledge blocks in descending order based on the relevance scores to obtain a sorted knowledge base list;

[0055] Processing the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model;

[0056] Inputting the target model and the query question into the large model, and receiving the corresponding query result returned by the large model;

[0057] The query result is returned to the user.

[0058] In order to solve the above technical problems, the embodiment of the present application further provides a computer-readable storage medium, which adopts the following technical solution:

[0059] Receive query questions input by users;

[0060] Querying a plurality of knowledge blocks matching the query question from a pre-built target knowledge base;

[0061] Performing relevance evaluation on all the knowledge blocks and the query question, and generating a relevance score between each knowledge block and the query question;

[0062] Sorting all the knowledge blocks in descending order based on the relevance scores to obtain a sorted knowledge base list;

[0063] Processing the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model;

[0064] Inputting the target model and the query question into the large model, and receiving the corresponding query result returned by the large model;

[0065] The query result is returned to the user.

[0066] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0067] The present application first receives a query question input by a user; then queries a plurality of knowledge blocks matching the query question from a pre-constructed target knowledge base; then performs a correlation evaluation on all the knowledge blocks and the query question, and generates a correlation score between each knowledge block and the query question; subsequently, all the knowledge blocks are sorted in descending order based on the correlation score to obtain a sorted knowledge base list; further, the knowledge base list is processed based on a preset matching knowledge length to generate a target model input corresponding to a preset large model; then, the target model input and the query question are input into the large model, and the corresponding query result returned by the large model is received; and finally, the query result is returned to the user. After searching for multiple knowledge blocks that match the query question from a pre-built target knowledge base, the present application will intelligently evaluate the relevance of all the knowledge blocks to the query question, generate a relevance score between each knowledge block and the query question, and process the knowledge base list obtained after sorting the relevance score based on the preset matching knowledge length, thereby generating a target model input corresponding to the preset large model, and subsequently inputting the target model input and the query question into the large model, receiving the corresponding query result returned by the large model, and returning the query result to the user. The present application provides a non-cutting method based on the use of relevance scores and matching knowledge lengths, which enables the large model to match the complete knowledge block that matches the query question, so that the large model can accurately output matching query results by processing the target model input and the query question, effectively improving the intelligence and accuracy of query problem processing and the accuracy of the query results obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the scheme in the present application, a brief introduction is given below to the drawings required for use in the description of the embodiments of the present application. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0069] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0070] Figure 2 A flowchart of an embodiment of the problem-solving method based on artificial intelligence according to the present application;

[0071] Figure 3 is a structural schematic diagram of an embodiment of an artificial intelligence-based problem processing device according to the present application;

[0072] Figure 4It is a structural diagram of an embodiment of a computer device according to the present application. DETAILED DESCRIPTION

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by technicians in the technical field of the present application; the terms used in the specification of the application herein are only for the purpose of describing specific embodiments and are not intended to limit the present application; the terms "including" and "having" and any variations thereof in the specification and claims of the present application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of the present application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0074] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0075] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings.

[0076] like Figure 1 As shown, the system architecture 100 may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0077] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, 103, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0078] The terminal devices 101, 102, and 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, desktop computers, and the like.

[0079] The server 105 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal devices 101 , 102 , and 103 .

[0080] It should be noted that the artificial intelligence-based problem handling method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the artificial intelligence-based problem handling device is generally set in the server / terminal device.

[0081] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0082] Continue to refer Figure 2 , showing a flow chart of an embodiment of an artificial intelligence-based problem handling method according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The artificial intelligence-based problem handling method provided in the embodiment of the present application can be applied to any scenario where query problem handling is required, and the artificial intelligence-based problem handling method can be applied to products in these scenarios, for example, query problem handling in the financial insurance field. The artificial intelligence-based problem handling method comprises the following steps:

[0083] Step S201: receiving a query question input by a user.

[0084] In this embodiment, the problem-solving method based on artificial intelligence runs on the electronic device (e.g. Figure 1The server / terminal device shown in the figure) can obtain the query question input by the user through a wired connection method or a wireless connection method. It should be pointed out that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, Wi-Fi connection, Bluetooth connection, Wi MAX connection, Zigbee connection, UWB (Ultra Wi DeBand) connection, and other wireless connection methods currently known or to be developed in the future. The present application can be applied to query problem processing scenarios in the financial field. The above-mentioned query question may be the question data that needs to be queried that is input by the user during the interaction with the large model. The query question can be question information in any format. Exemplarily, in the business scenario of insurance product consultation in the financial insurance field, the above-mentioned question information may include information on consulting insurance costs, insurance terms confirmation information, information requesting the recommendation of suitable insurance products, and so on.

[0085] Step S202: Query a plurality of knowledge blocks matching the query question from a pre-built target knowledge base.

[0086] In this embodiment, multiple knowledge blocks matching the query question can be retrieved from the target knowledge base by utilizing the index constructed in the target knowledge base.

[0087] Step S203: perform relevance evaluation on all the knowledge blocks and the query question, and generate a relevance score between each knowledge block and the query question.

[0088] In this embodiment, the above-mentioned specific implementation process of performing relevance evaluation on all the knowledge blocks and the query question to generate the relevance score between each knowledge block and the query question will be described in further details in subsequent specific embodiments of the present application and will not be elaborated in detail here.

[0089] Step S204: sorting all the knowledge blocks based on the correlation scores from high to low to obtain a sorted knowledge base list.

[0090] In this embodiment, all the knowledge blocks are sorted in descending order based on the relevance scores to obtain sorted knowledge blocks, which are then stored in a list template to obtain the above-mentioned knowledge base list.

[0091] Step S205, processing the knowledge base list based on the preset matching knowledge length to generate a target model input corresponding to the preset large model.

[0092] In this embodiment, the above-mentioned matching knowledge length is the length data corresponding to the number of knowledge blocks or the number of paragraphs. The specific value of the matching knowledge length is not limited and can be set according to the actual business processing requirements. Among them, there is no specific limitation on the selection of the above-mentioned large model. It is only necessary to ensure that the large model has retrieval enhancement generation (RAG), that is, the large model is a large language model based on the RAG architecture. Exemplarily, the above-mentioned large model can adopt a model of the GPT series, such as the GPT-4 model. The GPT series is a language model based on the Transformer structure developed by OpenAI, which performs well in the fields of text generation, dialogue system, question and answer system, etc. The GPT-4 model is the latest member of the GPT series, with more powerful generation capabilities and higher accuracy. In addition, the above-mentioned specific implementation process of processing the knowledge base list based on the preset matching knowledge length to generate the target model input corresponding to the preset large model will be further described in detail in the subsequent specific embodiments of this application, and will not be elaborated on here.

[0093] Step S206, inputting the target model and the query question into the large model, and receiving the corresponding query result returned by the large model.

[0094] In this embodiment, after the target model input and the query question are input into the large model, the large model will generate a matching answer, ie, the above-mentioned query result, according to the input target model input and the query question.

[0095] Step S207: returning the query result to the user.

[0096] In this embodiment, the specific implementation process of returning the query result to the user will be described in further detail in subsequent specific embodiments of the present application, and will not be elaborated on here.

[0097] The present application first receives a query question input by a user; then queries a plurality of knowledge blocks matching the query question from a pre-constructed target knowledge base; then performs a correlation evaluation on all the knowledge blocks and the query question, and generates a correlation score between each knowledge block and the query question; subsequently, all the knowledge blocks are sorted in descending order based on the correlation score to obtain a sorted knowledge base list; further, the knowledge base list is processed based on a preset matching knowledge length to generate a target model input corresponding to a preset large model; then, the target model input and the query question are input into the large model, and the corresponding query result returned by the large model is received; and finally, the query result is returned to the user. After searching for multiple knowledge blocks that match the query question from a pre-built target knowledge base, the present application will intelligently evaluate the relevance of all the knowledge blocks to the query question, generate a relevance score between each knowledge block and the query question, and process the knowledge base list obtained after sorting the relevance score based on the preset matching knowledge length, thereby generating a target model input corresponding to the preset large model, and subsequently inputting the target model input and the query question into the large model, receiving the corresponding query result returned by the large model, and returning the query result to the user. The present application provides a non-cutting method based on the use of relevance scores and matching knowledge lengths, which enables the large model to match the complete knowledge block that matches the query question, so that the large model can accurately output matching query results by processing the target model input and the query question, effectively improving the intelligence and accuracy of query problem processing and the accuracy of the query results obtained.

[0098] In some optional implementations, step S205 includes the following steps:

[0099] According to the sorting order of all the knowledge blocks, the knowledge blocks in the knowledge base list are gradually added to the preset model input, and it is detected whether the total length of the current first model input is greater than the matching knowledge length.

[0100] In this embodiment, the model input may be a pre-built blank document. The knowledge blocks in the knowledge base list are gradually added to the model input according to the sorting order of all the knowledge blocks, that is, the order of relevance scores from high to low, and in the process of adding knowledge blocks to the model input each time, it is detected in real time whether the total length of the current first model input is greater than the matching knowledge length. The total length of the first model input may specifically refer to the number of all knowledge blocks contained in the first model input, or may also refer to the number of all paragraphs contained in the first model input.

[0101] If the total length of the first model input is greater than the matching knowledge length, the difference between the total length of the first model input and the matching knowledge length is calculated.

[0102] In this embodiment, the difference between the total length of the first model input and the matching knowledge length is calculated by subtracting the total length of the first model input from the matching knowledge length to obtain the difference between the total length of the first model input and the matching knowledge length.

[0103] It is determined whether the difference is within a preset value range.

[0104] In this embodiment, there is no specific limitation on the value of the above-mentioned preset numerical range, and it can be set according to the actual length evaluation requirements. If the difference between the total length of the first model input and the length of the matching knowledge is within the above-mentioned preset numerical range, it indicates that the length difference between the total length of the first model input and the length of the matching knowledge is small and can be ignored. If the difference between the total length of the first model input and the length of the matching knowledge is not within the above-mentioned preset numerical range, it indicates that the length difference between the total length of the first model input and the length of the matching knowledge is large.

[0105] If the difference is within the preset numerical range, the target model input corresponding to the large model is generated based on the first model input.

[0106] In this embodiment, the specific implementation process of generating the target model input corresponding to the large model based on the first model input will be described in further detail in subsequent specific embodiments of the present application and will not be elaborated on here.

[0107] The present application gradually adds the knowledge blocks in the knowledge base list to the preset model input according to the sorting order of all the knowledge blocks, and detects whether the total length of the current first model input is greater than the matching knowledge length; if the total length of the first model input is greater than the matching knowledge length, calculates the difference between the total length of the first model input and the matching knowledge length; subsequently determines whether the difference is within a preset numerical range; if the difference is within the preset numerical range, generates the target model input corresponding to the large model based on the first model input. The present application gradually adds the knowledge blocks in the knowledge base list to the preset model input according to the sorting order of all the knowledge blocks. If it is detected that the total length of the first model input is greater than the matching knowledge length, the difference between the total length of the first model input and the matching knowledge length is further calculated. If it is determined that the difference is within the preset numerical range, the target model input corresponding to the large model is generated based on the first model input, thereby realizing fast and accurate processing of the knowledge base list based on the matching knowledge length and the preset numerical range to generate the target model input corresponding to the preset large model, improving the generation efficiency of the target model input corresponding to the large model, and ensuring the accuracy of the generated target model input.

[0108] In some optional implementations of this embodiment, after the step of determining whether the difference is within a preset value range, the method further includes:

[0109] If the difference is not within the preset numerical range, the first knowledge block with the smallest relevance score is screened out from the first model input.

[0110] In this embodiment, the first knowledge block also refers to the knowledge block that is last added to the first model input.

[0111] The first knowledge block is eliminated from the first model input to obtain a corresponding second model input.

[0112] In this embodiment, the above-mentioned elimination process refers to the process of deleting the above-mentioned first knowledge block in the first model input.

[0113] The target model input corresponding to the large model is generated based on the second model input.

[0114] In this embodiment, the implementation process of generating the target model input corresponding to the large model based on the second model input can specifically refer to the implementation process of generating the target model input corresponding to the large model based on the first model input, and no further details are given here.

[0115] If the present application detects that the difference is not within the preset numerical range, the first knowledge block with the smallest relevance score is screened out from the first model input; then the first knowledge block is eliminated from the first model input to obtain the corresponding second model input; and the target model input corresponding to the large model is subsequently generated based on the second model input. When the present application detects that the difference between the total length of the first model input and the matching knowledge length is not within the preset numerical range, it will determine that the total length of the current first model input exceeds the above matching knowledge length by a large amount, and then the current first model input will be intelligently truncated, specifically by eliminating the first knowledge block from the first model input to obtain the corresponding second model input, and then generating the target model input corresponding to the large model based on the second model input, so as to realize the rapid and accurate processing of the knowledge base list based on the matching knowledge length and the preset numerical range to generate the target model input corresponding to the preset large model, thereby improving the generation efficiency of the target model input corresponding to the large model and ensuring the accuracy of the generated target model input.

[0116] In some optional implementations, generating the target model input corresponding to the large model based on the first model input comprises the following steps:

[0117] Gets the preset logical order.

[0118] In this embodiment, there is no specific limitation on the setting of the above logical sequence, which can be set according to actual business processing requirements. Specifically, the time sequence, importance sequence, etc. can be used as the above logical sequence.

[0119] Obtain all second knowledge blocks contained in the first model input.

[0120] In this embodiment, all the second knowledge blocks mentioned above refer to the knowledge blocks that have been input into the first model input from the knowledge base list.

[0121] All the second knowledge blocks are sequentially adjusted based on the logical sequence to obtain adjusted third knowledge blocks.

[0122] In this embodiment, the sorting numbers corresponding to all the second knowledge blocks can be generated according to the logical order, and then all the second knowledge blocks can be sequentially adjusted according to the sorting numbers to obtain the adjusted third knowledge block.

[0123] All the third knowledge blocks are spliced ​​together to obtain corresponding spliced ​​knowledge block data.

[0124] In this embodiment, a splicing operation can be performed using all the third knowledge blocks to obtain corresponding spliced ​​knowledge block data. Specifically, the splicing operation can be implemented by a string concatenation operator (such as + in Python, + or template string in JavaScript, + or StringBuilder / StringBuffer in Java, etc.).

[0125] The spliced ​​knowledge block data is input as the target model corresponding to the large model.

[0126] The present application obtains a preset logical order; then obtains all the second knowledge blocks included in the first model input; then adjusts all the second knowledge blocks in sequence based on the logical order to obtain an adjusted third knowledge block; subsequently splices all the third knowledge blocks to obtain corresponding spliced ​​knowledge block data, and uses the spliced ​​knowledge block data as the target model input corresponding to the large model. After obtaining the first model input, the present application obtains all the second knowledge blocks included in the first model input, and intelligently uses the preset logical order to adjust all the second knowledge blocks in sequence to obtain the adjusted third knowledge block, and then splices all the third knowledge blocks to obtain corresponding spliced ​​knowledge block data and uses them as the final target model input corresponding to the large model, effectively ensuring the accuracy of the generated target model input.

[0127] In some optional implementations, step S203 includes the following steps:

[0128] Call the preset relevance evaluation model.

[0129] In this embodiment, the above-mentioned relevance evaluation model is a model generated after training the machine learning model based on the pre-constructed annotated sample data. The construction process of the relevance evaluation model includes: 1. Data preparation: Annotation data: Prepare a certain number of annotated query data-document pairs, where each document is marked as the relevance level with the query data (such as "relevant", "partially relevant", "irrelevant", etc.). Feature extraction: Extract features from query data and documents, and the features may include term frequency, TF-IDF, word vector, document structure features, etc. 2. Model selection: Select a machine learning model, for example, SVM or LSTM model can be selected. 3. Model training: Divide the above-mentioned query data-document into a training set and a validation set, use the training set to train the selected machine learning model, and optimize the performance of the model on the training set by adjusting the model parameters. 4. Feature optimization and model evaluation: Perform feature selection or feature engineering on the model to improve model performance. Use the validation set to evaluate the model performance and make necessary adjustments to obtain the above-mentioned relevance evaluation model.

[0130] Based on the relevance evaluation model, the relevance between the designated knowledge block and the query question is evaluated to obtain a relevance evaluation result between the designated knowledge block and the query question; wherein the designated knowledge block is any one of all the knowledge blocks.

[0131] In this embodiment, the designated knowledge block and the query question are input into the relevance evaluation model, so that the relevance evaluation model is used to evaluate the relevance of the designated knowledge block and the query question, and outputs a corresponding relevance evaluation result.

[0132] The relevance evaluation result is used as a specified relevance score between the specified knowledge block and the query question.

[0133] The present application calls a preset relevance evaluation model; then based on the relevance evaluation model, a relevance evaluation is performed on the designated knowledge block and the query question to obtain a relevance evaluation result between the designated knowledge block and the query question; and the relevance evaluation result is subsequently used as a designated relevance score between the designated knowledge block and the query question. The present application calls a preset relevance evaluation model to perform a relevance evaluation on the designated knowledge block and the query question, so that the designated relevance score between the designated knowledge block and the query question can be generated quickly and accurately, effectively ensuring the data accuracy of the generated relevance score.

[0134] In some optional implementations of this embodiment, step S207 includes the following steps:

[0135] The query result is optimized to obtain a corresponding target query result.

[0136] In this embodiment, the above optimization processing may specifically include processing such as removing redundant information and adjusting word order.

[0137] Determine a target push mode corresponding to the target query result.

[0138] In this embodiment, the above-mentioned target push mode refers to a user-friendly mode, which may specifically include any one of various forms such as text, picture, table, etc.

[0139] Based on the target push mode, the target query result is returned to the user.

[0140] In this embodiment, the target query result is returned to the user according to the communication address obtained by obtaining the user's communication address, wherein the communication address may include an email address, a mobile phone number, etc.

[0141] This application optimizes the query results to obtain the corresponding target query results; then determines the target push method corresponding to the target query results; and subsequently returns the target query results to the user based on the target push method. After receiving the corresponding query results returned by the large model, this application will also intelligently optimize the query results to obtain the target query results, ensuring the accuracy and adaptability of the generated target query results. And it will further use the target push method corresponding to the target query results to return the target query results to the user, thereby improving the push intelligence of the target query results and helping to improve the user experience.

[0142] In some optional implementations of this embodiment, before step S202, the electronic device may further perform the following steps:

[0143] Acquire pre-collected knowledge content, and build a corresponding original document library based on the knowledge content.

[0144] In this embodiment, relevant knowledge contents are collected in advance from various sources (such as Word documents, Excel tables, TXT files, etc.), and a corresponding original document library is constructed.

[0145] Receive the formatting specification processing of the original knowledge base performed by the designated user to obtain the corresponding document library.

[0146] In this embodiment, the above-mentioned formatting standard processing specifically includes title marking, text organization, table and image processing, and file name naming standard processing. Title marking: In a Word document, different title styles (such as first-level title, second-level title, etc.) are used to clearly distinguish different parts of the document. These titles will serve as themes or chapter titles of knowledge blocks. Text organization: Ensure that the text content is arranged in a logical order, and try to keep each paragraph or section around a central idea. Table and image processing: Add titles or descriptions to tables and images so that their contents can be identified and converted in subsequent processing. File name naming standards: Ensure that the name of each file contains the subject content for easy management and retrieval.

[0147] A preset processing script is called to perform knowledge block extraction processing on the document library to obtain a corresponding fourth knowledge block.

[0148] In this embodiment, the process of calling an automated script to perform knowledge block extraction processing on the document library includes: creating a Hash dictionary: when uploading a document library file, first create a Hash dictionary to record the file name and its mapping key. This mapping key will be used for subsequent knowledge block extraction and storage. Reading document content: using Python's docx and other libraries to read the content of the Word document and identify the styles therein (such as title, text, table title, figure title, etc.). Knowledge block extraction: 1) When encountering a title style, check whether a title of this level already exists in the Hash dictionary. If a title of this level exists, submit the currently accumulated knowledge block (if any), and reset the knowledge block string, that is, delete all titles in the hash dictionary with a title level less than the current level and initialize the single knowledge string; if a title of this level does not exist, record the title in the Hash dictionary. 2) When encountering a text style, if the title has been read before, add the text content to the current knowledge block string; if the title has not been read, try to use the existing title in the Hash dictionary as the start of the knowledge block. That is, all the title-level text in the current hash dictionary is added to the single knowledge string one by one, so that each piece of knowledge can be guaranteed to have a subject title, which makes it easier to index when matching. 3) When encountering a table title, the table is read and the table content is converted into text in Markdown format (or HTML format), and the extra spaces after markdown formatting are deleted. The subsequent processing refers to step 2) of the text style. This can save input length overhead and realize the structured data to text operation. 4) When encountering a figure title, the picture is stored in the database in the form of base64 encoding or file path, and the Markdown picture representation is used in the knowledge block string to reference it. Specifically, the above database is specifically an SQL database. When storing in the SQL database, the file name and the figure title are used as the primary key for storage. Then use the markdown picture representation, that is! ()[], which is displayed as a hyperlink style (blue clickable), where the content in the round brackets () is the text of the hyperlink, and the content in the square brackets [] is the link to store the hyperlink. By putting the text of the picture title in parentheses (), and putting the database primary key content in [], for example! (This is a picture) [File.docxThis is a picture.j pg], a picture with a picture title is converted into text, and then transferred to the text style processing of step 2). Since the big model can generate text, it can return! ()[] This content, so the front end can get the! ()[] in the big model reply, or it can get the content in [], that is, the database primary key, so it can get the picture from the database, and then render and display the picture on the front end.Submit knowledge blocks: After the document processing is completed, check and submit the last knowledge block (if any), because the last knowledge block may not be submitted when encountering new titles.

[0149] In this embodiment, the knowledge blocks are manually determined, and the pictures and table data in the document are processed and stored in the form of markdown, so that RAG can reply to the tables and pictures in the knowledge base.

[0150] Call the preset initial knowledge base.

[0151] In this embodiment, the initial knowledge base is a pre-built database for storing knowledge blocks.

[0152] The fourth knowledge block is stored in the initial knowledge base to obtain the target knowledge base.

[0153] In this embodiment, the target knowledge base is obtained by storing the extracted fourth knowledge block in the initial knowledge base. In addition, an index is created for each knowledge block in the target knowledge base to enable fast retrieval. The index may include information such as keywords, topics, and categories.

[0154] The present application obtains the knowledge content collected in advance and builds the corresponding original document library based on the knowledge content; then receives the formatting specification processing of the original knowledge library performed by the designated user to obtain the corresponding document library; then calls the preset processing script to extract the knowledge block of the document library to obtain the corresponding fourth knowledge block; subsequently calls the preset initial knowledge base; finally stores the fourth knowledge block in the initial knowledge base to obtain the target knowledge base. After obtaining the knowledge content collected in advance and building the corresponding original document library based on the knowledge content, the present application obtains the document library by formatting the original knowledge library based on the formatting specification processing performed by the designated user, and extracts the knowledge block of the document library based on the processing script to obtain the fourth knowledge block, and then stores the fourth knowledge block in the initial knowledge base, so as to realize the intelligent conversion of the original and disorganized knowledge content into structured, easy-to-manage and retrieve knowledge blocks and store them in the initial knowledge base to generate the final target knowledge base, which is conducive to the subsequent use of the target knowledge base to quickly and accurately retrieve the knowledge blocks related to the query questions input by the user.

[0155] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0156] It should be emphasized that in order to further ensure the privacy and security of the above query results, the above query results can also be stored in a node of a blockchain.

[0157] The blockchain referred to in this application is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0158] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (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 knowledge to obtain the best results.

[0159] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0160] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, the aforementioned storage medium can be a non-volatile storage medium such as a disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0161] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0162] Further references Figure 3 , as a response to the above Figure 2 The present application provides an embodiment of a problem-solving device based on artificial intelligence. Figure 2 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0163] like Figure 3 As shown, the artificial intelligence-based problem processing device 300 described in this embodiment includes: a first receiving module 301, a query module 302, an evaluation module 303, a sorting module 304, a generation module 305, an input module 306 and a return module 307. Among them:

[0164] The first receiving module 301 is used to receive a query question input by a user;

[0165] A query module 302 is used to query a plurality of knowledge blocks matching the query question from a pre-built target knowledge base;

[0166] An evaluation module 303 is used to evaluate the relevance between all the knowledge blocks and the query question, and generate a relevance score between each knowledge block and the query question;

[0167] A sorting module 304 is used to sort all the knowledge blocks based on the correlation scores from high to low to obtain a sorted knowledge base list;

[0168] A generation module 305 is used to process the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model;

[0169] An input module 306, used to input the target model and the query question into the large model, and receive the corresponding query result returned by the large model;

[0170] The returning module 307 is used to return the query result to the user.

[0171] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0172] In some optional implementations of this embodiment, the generating module 305 includes:

[0173] A detection submodule, used to gradually add the knowledge blocks in the knowledge base list to the preset model input according to the sorting order of all the knowledge blocks, and detect whether the total length of the current first model input is greater than the matching knowledge length;

[0174] a calculation submodule, configured to calculate a difference between the total length of the first model input and the matching knowledge length if the total length of the first model input is greater than the matching knowledge length;

[0175] A first judgment submodule, used to judge whether the difference is within a preset value range;

[0176] The first generating submodule is used to generate the target model input corresponding to the large model based on the first model input if the difference is within the preset numerical range.

[0177] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0178] In some optional implementations of this embodiment, the generating module 305 further includes:

[0179] a screening submodule, configured to screen out a first knowledge block with a minimum relevance score from the first model input if the difference is not within the preset numerical range;

[0180] A removal submodule, used for removing the first knowledge block from the first model input to obtain a corresponding second model input;

[0181] The second generating submodule is used to generate the target model input corresponding to the large model based on the second model input.

[0182] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0183] In some optional implementations of this embodiment, the first generating submodule includes:

[0184] A first acquisition unit, used to acquire a preset logical sequence;

[0185] A second acquisition unit, used to acquire all second knowledge blocks contained in the first model input;

[0186] an adjustment unit, configured to perform sequential adjustment processing on all the second knowledge blocks based on the logical sequence to obtain an adjusted third knowledge block;

[0187] A splicing unit, used for splicing all the third knowledge blocks to obtain corresponding spliced ​​knowledge block data;

[0188] A determination unit is used to input the spliced ​​knowledge block data as the target model corresponding to the large model.

[0189] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0190] In some optional implementations of this embodiment, the evaluation module 303 includes:

[0191] A calling submodule is used to call a preset correlation evaluation model;

[0192] An evaluation submodule, configured to perform a correlation evaluation between a designated knowledge block and the query question based on the correlation evaluation model, and obtain a correlation evaluation result between the designated knowledge block and the query question; wherein the designated knowledge block is any one of all the knowledge blocks;

[0193] The first determination submodule is used to use the relevance evaluation result as a designated relevance score between the designated knowledge block and the query question.

[0194] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0195] In some optional implementations of this embodiment, the returning module 307 includes:

[0196] An optimization submodule, used to optimize the query results to obtain corresponding target query results;

[0197] A second determination submodule is used to determine a target push mode corresponding to the target query result;

[0198] The return submodule is used to return the target query result to the user based on the target push mode.

[0199] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0200] In some optional implementations of this embodiment, the problem processing device based on artificial intelligence further includes:

[0201] A construction module, used to obtain pre-collected knowledge content and construct a corresponding original document library based on the knowledge content;

[0202] A second receiving module is used to receive the formatting specification processing of the original knowledge base performed by a designated user to obtain a corresponding document library;

[0203] An extraction module, used for calling a preset processing script to perform knowledge block extraction processing on the document library to obtain a corresponding fourth knowledge block;

[0204] A calling module is used to call a preset initial knowledge base;

[0205] A storage module is used to store the fourth knowledge block into the initial knowledge base to obtain the target knowledge base.

[0206] In this embodiment, the operations performed by the above modules or units correspond one-to-one to the steps of the artificial intelligence-based problem-solving method in the aforementioned implementation mode, and will not be repeated here.

[0207] To solve the above technical problems, the present application also provides a computer device. Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0208] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with components 41-43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), digital processors (DSPs), embedded devices, etc.

[0209] The computer device may be a computing device such as a desktop computer, a notebook, a PDA, a cloud server, etc. The computer device may interact with a user through a keyboard, a mouse, a remote controller, a touch pad, or a voice control device.

[0210] The memory 41 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 41 can be an internal storage unit of the computer device 4, such as a hard disk or memory of the computer device 4. In other embodiments, the memory 41 can also be an external storage device of the computer device 4, such as a plug-in hard disk equipped on the computer device 4, a smart memory card (Smart Med ia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. Of course, the memory 41 can also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions of a problem-solving method based on artificial intelligence, etc. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0211] The processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chips in some embodiments. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to run the computer-readable instructions stored in the memory 41 or process data, such as running the computer-readable instructions of the problem-solving method based on artificial intelligence.

[0212] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0213] The present application also provides another embodiment, namely, providing a computer-readable storage medium, wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the artificial intelligence-based problem-solving method as described above.

[0214] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0215] Obviously, the embodiments described above are only some embodiments of the present application, rather than all embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application is described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions recorded in the aforementioned specific implementation methods, or to perform equivalent replacement of some of the technical features therein. Any equivalent structure made using the contents of the specification and drawings of this application, directly or indirectly used in other related technical fields, is similarly within the scope of patent protection of this application.

Claims

1. A problem-solving method based on artificial intelligence, characterized in that: The steps include: Receive query questions input by users; Querying a plurality of knowledge blocks matching the query question from a pre-built target knowledge base; Performing relevance evaluation on all the knowledge blocks and the query question, and generating a relevance score between each knowledge block and the query question; Sorting all the knowledge blocks in descending order based on the relevance scores to obtain a sorted knowledge base list; Processing the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model; Inputting the target model and the query question into the large model, and receiving the corresponding query result returned by the large model; The query result is returned to the user.

2. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The step of processing the knowledge base list based on the preset matching knowledge length to generate a target model input corresponding to the preset large model specifically includes: According to the sorting order of all the knowledge blocks, the knowledge blocks in the knowledge base list are gradually added to the preset model input, and it is detected whether the total length of the current first model input is greater than the matching knowledge length; If the total length of the first model input is greater than the matching knowledge length, calculating the difference between the total length of the first model input and the matching knowledge length; Determining whether the difference is within a preset value range; If the difference is within the preset numerical range, the target model input corresponding to the large model is generated based on the first model input.

3. The problem-solving method based on artificial intelligence according to claim 2 is characterized in that: After the step of determining whether the difference is within a preset value range, the method further includes: If the difference is not within the preset value range, selecting a first knowledge block with the smallest relevance score from the first model input; Eliminating the first knowledge block from the first model input to obtain a corresponding second model input; The target model input corresponding to the large model is generated based on the second model input.

4. The problem-solving method based on artificial intelligence according to claim 2, characterized in that: The step of generating the target model input corresponding to the large model based on the first model input specifically includes: Get the preset logical order; Obtain all second knowledge blocks contained in the first model input; Performing sequence adjustment processing on all the second knowledge blocks based on the logical sequence to obtain an adjusted third knowledge block; Performing splicing processing on all the third knowledge blocks to obtain corresponding spliced ​​knowledge block data; The spliced ​​knowledge block data is input as the target model corresponding to the large model.

5. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The step of evaluating the relevance between all the knowledge blocks and the query question to generate a relevance score between each knowledge block and the query question specifically includes: Call the preset relevance assessment model; Based on the relevance evaluation model, a relevance evaluation is performed on the designated knowledge block and the query question to obtain a relevance evaluation result between the designated knowledge block and the query question; wherein the designated knowledge block is any one of all the knowledge blocks; The relevance evaluation result is used as a specified relevance score between the specified knowledge block and the query question.

6. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: The step of returning the query result to the user specifically includes: Optimizing the query results to obtain corresponding target query results; Determine a target push method corresponding to the target query result; Based on the target push mode, the target query result is returned to the user.

7. The problem-solving method based on artificial intelligence according to claim 1, characterized in that: Before the step of searching for multiple knowledge blocks matching the query question from the pre-built target knowledge base, the method further includes: Acquire pre-collected knowledge content and build a corresponding original document library based on the knowledge content; Receiving the formatting specification processing of the original knowledge base performed by a designated user to obtain a corresponding document library; Calling a preset processing script to perform knowledge block extraction processing on the document library to obtain a corresponding fourth knowledge block; Calling the preset initial knowledge base; The fourth knowledge block is stored in the initial knowledge base to obtain the target knowledge base.

8. A problem-solving device based on artificial intelligence, characterized in that: include: A first receiving module is used to receive a query question input by a user; A query module, used to query a plurality of knowledge blocks matching the query question from a pre-built target knowledge base; An evaluation module, used to evaluate the relevance between all the knowledge blocks and the query question, and generate a relevance score between each knowledge block and the query question; A sorting module, used to sort all the knowledge blocks in descending order based on the relevance scores to obtain a sorted knowledge base list; A generation module, used for processing the knowledge base list based on a preset matching knowledge length to generate a target model input corresponding to a preset large model; An input module, used to input the target model and the query question into the large model, and receive the corresponding query result returned by the large model; The returning module is used to return the query result to the user.

9. A computer device comprising a memory and a processor, wherein the memory stores computer-readable instructions, and when the processor executes the computer-readable instructions, the steps of the artificial intelligence-based problem-solving method as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-readable instructions, which, when executed by a processor, implement the steps of the artificial intelligence-based problem-solving method according to any one of claims 1 to 7.

Images