Intelligent question and answer processing method and device, storage medium and computer equipment

Abstract

The invention discloses an intelligent question and answer processing method and device, a storage medium and computer equipment, relates to the technical field of machine learning, and mainly aims to solve the problems that an intelligent question and answer system is low in answer accuracy and cannot meet diversified language requirements, and comprises the steps of obtaining multi-source heterogeneous data of different language types, and constructing a corresponding knowledge graph; obtaining a large-scale multi-language question and answer corpus, and training a corresponding question and answer model; performing semantic alignment processing on the knowledge maps corresponding to the various language types to obtain a multi-language shared knowledge map; performing model enhancement processing on the question and answer model corresponding to each language type by adopting a multi-language shared knowledge graph to obtain an enhanced question and answer model with cross-language knowledge association capability; and receiving a user question, and determining a target enhanced question and answer model from the plurality of enhanced question and answer models based on the language type corresponding to the user question, so that the target enhanced question and answer model generates a reply result corresponding to the user question.

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to an intelligent question-answering processing method and device, a storage medium, and a computer device. Background Art

[0002] The research on intelligent question answering technology has made great progress in the application context of deep learning architecture, especially in the financial technology and smart medical scenarios.

[0003] Existing intelligent question-answering systems primarily rely on rule engines and simple machine learning models. While they can meet basic question-answering needs to a certain extent, their limitations are becoming increasingly prominent. First, due to limited semantic understanding capabilities, intelligent question-answering systems struggle to accurately parse complex natural language questions, resulting in low accuracy and reliability of answers, impacting user experience. Second, faced with the diverse language demands of globalization, existing intelligent question-answering systems perform poorly when handling multilingual questions and struggle to meet the demands of cross-language communication scenarios. The existence of the aforementioned problems severely restricts the application effect and promotion scope of intelligent question-answering systems. Therefore, there is an urgent need to enhance the system's intelligent question-answering capabilities and practicality through technological innovation and optimization. Summary of the Invention

[0004] In view of this, the present invention provides an intelligent question-answering processing method and apparatus, a storage medium, and a computer device, the main purpose of which is to solve the problems of low answer accuracy and inability to meet diversified language requirements of the intelligent question-answering system.

[0005] According to one aspect of the present invention, there is provided an intelligent question-answering processing method, comprising:

[0006] Acquire multi-source heterogeneous data of different language types respectively, and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data;

[0007] Obtain large-scale multilingual question-answering corpus, and train corresponding question-answering models according to different language types based on the knowledge graph;

[0008] Performing semantic alignment processing on the knowledge graphs corresponding to various language types so as to enable cross-language knowledge sharing among the knowledge graphs to obtain a multilingual shared knowledge graph;

[0009] Using the multilingual shared knowledge graph to perform model enhancement processing on the question-answering model corresponding to each language type, thereby obtaining an enhanced question-answering model with cross-language knowledge association capabilities;

[0010] A user question is received, and a target enhanced question-answering model is determined from a plurality of enhanced question-answering models based on a language type corresponding to the user question, so that the target enhanced question-answering model generates a reply result corresponding to the user question.

[0011] Furthermore, the knowledge graphs corresponding to various language types are constructed based on the multi-source heterogeneous data, including:

[0012] Acquire target multi-source heterogeneous data of the same language type, and preprocess the target multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data;

[0013] Using a pre-trained entity recognition model to perform entity recognition processing on the pre-processed multi-source heterogeneous data to obtain entity information; and using a pre-trained relationship extraction model to determine the semantic relationship between the entity information;

[0014] The knowledge graph corresponding to the language type is constructed based on the entity information and the semantic relationship.

[0015] Furthermore, the acquisition of large-scale multilingual question-answering corpus and training of corresponding question-answering models according to different language types in combination with the knowledge graph include:

[0016] Dividing the large-scale multilingual question-answering corpus into corpora corresponding to various language types;

[0017] Determine the resource situation of each corpus and use a transfer learning strategy to enhance the resources of the corpus with insufficient resources so that each corpus meets the resource requirements;

[0018] In combination with the knowledge graph, the multilingual pre-training model is trained using corpora corresponding to various language types to obtain the question-answering models corresponding to various language types.

[0019] Furthermore, the method combines the knowledge graph and uses corpora corresponding to various language types to perform model training on the multilingual pre-training model to obtain the question-answering model corresponding to various language types, including:

[0020] Performing data cleaning and preprocessing on the corpus corresponding to each language type to obtain a preprocessed corpus;

[0021] Combined with the knowledge graph, the corpus is processed for long text, polysemy, and complex semantic relationships to obtain processed corpus-related information corresponding to various language types;

[0022] Model training is performed on the multilingual pre-training model based on the processed corpus-related information to obtain the question-answering models corresponding to various language types.

[0023] Furthermore, the semantic alignment processing is performed on the knowledge graphs corresponding to various language types so that each of the knowledge graphs can perform cross-language knowledge sharing to obtain a multilingual shared knowledge graph, including:

[0024] Map the same entity information in different language types to the same knowledge node; and map the same semantic relationship in different language types to the same relationship type;

[0025] Cross-language knowledge sharing is performed based on the same knowledge node and the same relationship type to obtain the multilingual shared knowledge graph.

[0026] Furthermore, before performing cross-language knowledge sharing based on the same knowledge node and the same relationship type, the method further includes:

[0027] Determining an incomplete relationship chain from the knowledge graphs corresponding to various languages ​​based on the same knowledge node and the same relationship type;

[0028] The incomplete relationship chains are subjected to knowledge completion processing based on the complete relationship chains in each knowledge graph to obtain a completed knowledge graph, so as to enable cross-language knowledge sharing based on the completed knowledge graph.

[0029] Furthermore, the multilingual shared knowledge graph is used to perform model enhancement processing on the question-answering model corresponding to each language type to obtain an enhanced question-answering model with cross-language knowledge association capability, including:

[0030] Get the target relationship chain in the knowledge graph corresponding to the current language type;

[0031] Based on the semantic alignment in the multilingual shared knowledge graph, a plurality of related relationship chains corresponding to the target relationship chain are obtained from the knowledge graphs corresponding to other language types except the current language type;

[0032] Based on the target relationship chain and the related relationship chain, the question-answering model corresponding to the current language type is enhanced to obtain the enhanced question-answering model with cross-language knowledge association capability.

[0033] According to another aspect of the present invention, there is provided an intelligent question-answering processing device, comprising:

[0034] A graph construction module is used to obtain multi-source heterogeneous data of different language types and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data;

[0035] A model training module is used to obtain large-scale multilingual question-answering corpora and train corresponding question-answering models according to different language types in combination with the knowledge graph;

[0036] A knowledge fusion module is used to perform semantic alignment processing on the knowledge graphs corresponding to various language types, so as to enable cross-language knowledge sharing among the knowledge graphs and obtain a multilingual shared knowledge graph;

[0037] A model enhancement module, configured to perform model enhancement processing on the question-answering model corresponding to each language type using the multilingual shared knowledge graph to obtain an enhanced question-answering model with cross-language knowledge association capabilities;

[0038] The question-answering module is used to receive user questions and determine a target enhanced question-answering model from multiple enhanced question-answering models based on the language type corresponding to the user questions, so that the target enhanced question-answering model generates a response result corresponding to the user questions.

[0039] Furthermore, the graph construction module is also used to:

[0040] Acquire target multi-source heterogeneous data of the same language type, and preprocess the target multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data;

[0041] Using a pre-trained entity recognition model to perform entity recognition processing on the pre-processed multi-source heterogeneous data to obtain entity information; and using a pre-trained relationship extraction model to determine the semantic relationship between the entity information;

[0042] The knowledge graph corresponding to the language type is constructed based on the entity information and the semantic relationship.

[0043] Furthermore, the model training module also includes:

[0044] A corpus division unit, configured to divide the large-scale multilingual question-answer corpus into corpora corresponding to various language types;

[0045] A resource enhancement unit is used to judge the resource situation of each corpus and use a transfer learning strategy to enhance the resources of the corpus with insufficient resources so that each corpus meets the resource requirement;

[0046] The training unit is used to combine the knowledge graph and use the corpus corresponding to various language types to perform model training processing on the multilingual pre-training model to obtain the question-answering model corresponding to various language types.

[0047] Furthermore, the training unit in the model training module is also used to:

[0048] Performing data cleaning and preprocessing on the corpus corresponding to each language type to obtain a preprocessed corpus;

[0049] Combined with the knowledge graph, the corpus is processed for long text, polysemy, and complex semantic relationships to obtain processed corpus-related information corresponding to various language types;

[0050] Model training is performed on the multilingual pre-training model based on the processed corpus-related information to obtain the question-answering models corresponding to various language types.

[0051] Furthermore, the knowledge fusion module is also used to:

[0052] Map the same entity information in different language types to the same knowledge node; and map the same semantic relationship in different language types to the same relationship type;

[0053] Cross-language knowledge sharing is performed based on the same knowledge node and the same relationship type to obtain the multilingual shared knowledge graph.

[0054] Furthermore, the knowledge fusion module is also used to:

[0055] Determining an incomplete relationship chain from the knowledge graphs corresponding to various languages ​​based on the same knowledge node and the same relationship type;

[0056] The incomplete relationship chains are subjected to knowledge completion processing based on the complete relationship chains in each knowledge graph to obtain a completed knowledge graph, so as to enable cross-language knowledge sharing based on the completed knowledge graph.

[0057] Furthermore, the model enhancement module is also used to:

[0058] Get the target relationship chain in the knowledge graph corresponding to the current language type;

[0059] Based on the semantic alignment in the multilingual shared knowledge graph, a plurality of related relationship chains corresponding to the target relationship chain are obtained from the knowledge graphs corresponding to other language types except the current language type;

[0060] Based on the target relationship chain and the related relationship chain, the question-answering model corresponding to the current language type is enhanced to obtain the enhanced question-answering model with cross-language knowledge association capability.

[0061] According to another aspect of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned intelligent question and answer processing method.

[0062] According to another aspect of the present invention, a computer device is provided, comprising a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0063] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned intelligent question and answer processing method.

[0064] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0065] The present invention provides an intelligent question-answering processing method and apparatus, a storage medium, and a computer device. Compared with the prior art, the present invention separately obtains multi-source heterogeneous data of different language types and constructs knowledge graphs corresponding to various language types based on the multi-source heterogeneous data. This introduces streaming data processing technology, which can achieve real-time acquisition, efficient processing, and dynamic updating of multi-source heterogeneous data, thereby ensuring the timeliness and accuracy of the knowledge graph, providing continuously updated knowledge support for the intelligent question-answering system, and significantly improving the intelligence level and practical application effect of the system. The present invention obtains large-scale multilingual question-answering corpus and trains corresponding question-answering models according to different language types in combination with the knowledge graph; performs semantic alignment processing on the knowledge graphs corresponding to various language types so that each of the knowledge graphs can share knowledge across languages, thereby obtaining a multilingual shared knowledge graph; and uses the multilingual shared knowledge graph to perform model enhancement processing on the question-answering models corresponding to each language type, thereby obtaining an enhanced question-answering model with cross-language knowledge association capabilities. This achieves efficient support and intelligent processing of multilingual question-answering interactions, thereby improving the system's multilingual adaptability and global user coverage.

[0066] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are specifically listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the present invention. The same reference symbols are used throughout the drawings to represent the same components. In the drawings:

[0068] Figure 1 A schematic diagram of an application environment of the intelligent question-answering processing method provided by an embodiment of the present invention is shown;

[0069] Figure 2 A schematic diagram showing a flow chart of an intelligent question-answering processing method provided by an embodiment of the present invention is shown;

[0070] Figure 3 A schematic diagram showing a flow chart of another intelligent question-answering processing method provided by an embodiment of the present invention;

[0071] Figure 4 A schematic diagram illustrating a flow chart of another intelligent question-answering processing method provided by an embodiment of the present invention is shown;

[0072] Figure 5 A schematic diagram illustrating a flow chart of another intelligent question-answering processing method provided by an embodiment of the present invention is shown;

[0073] Figure 6 A schematic diagram showing the structure of an intelligent question-answering processing device provided by an embodiment of the present invention is shown;

[0074] Figure 7 A schematic structural diagram of a computer device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0075] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0076] The intelligent question-answering processing method provided by the embodiment of the present invention can be applied in Figure 1In the application environment, the server is the main executor of the intelligent question-answering processing method of the present invention. The client communicates with the server through a network. The server can obtain multi-source heterogeneous data of different language types from multiple clients, and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data; and the server can also obtain large-scale multilingual question-answering corpora generated by multiple clients, and train corresponding question-answering models according to different language types in combination with the knowledge graphs. The server performs semantic alignment processing on the knowledge graphs corresponding to various language types, so that each of the knowledge graphs can share knowledge across languages ​​and obtain a multilingual shared knowledge graph; the multilingual shared knowledge graph is used to perform model enhancement processing on the question-answering models corresponding to each language type to obtain an enhanced question-answering model with cross-language knowledge association capabilities; thereafter, the server receives a user question sent by a client, and determines a target enhanced question-answering model from multiple enhanced question-answering models based on the language type corresponding to the user question, so that the target enhanced question-answering model generates a reply result corresponding to the user question.

[0077] In the present invention, for FinTech scenarios, the intelligent question-and-answer processing method can be used to provide investors with professional investment consulting services and personalized investment advice. It can also analyze customer questions to identify potential risk points and provide corresponding risk warnings and solutions. For example, when a customer inquires about a loan, the system can combine the customer's historical data and current market conditions to assess their credit risk and provide reasonable advice. Furthermore, it can be used for intelligent outbound calling, proactively pushing financial product information to customers and reminding them to complete transactions. For smart healthcare scenarios, the intelligent question-and-answer processing method can be used to provide patients with 24-hour online consulting services. After the patient enters symptoms and questions, the system analyzes them based on a medical knowledge base and patient information, providing a preliminary diagnosis and recommendations. It can also analyze information such as the patient's medical history, test results, and imaging data, and combine it with the medical knowledge base to provide preliminary diagnostic recommendations, helping doctors improve diagnostic accuracy and efficiency. Furthermore, it can provide patients with services such as disease information and medication guidance, helping them better understand their condition and treatment options. The client can include, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented as a standalone server or a server cluster consisting of multiple servers. The present invention is described in detail below through specific examples.

[0078] See also Figure 2 As shown, Figure 2 A flowchart of the intelligent question-answering processing method provided by an embodiment of the present invention includes the following steps:

[0079] 101. Acquire multi-source heterogeneous data of different language types, and construct knowledge graphs corresponding to the various language types based on the multi-source heterogeneous data;

[0080] In this embodiment, the current execution end can collect raw data streams in real time from multi-source heterogeneous data such as the Internet, social media platforms (such as Weibo, Twitter), news websites, and professional databases through API interfaces, web crawlers, message queues, etc. The collected data forms include text, pictures, videos, etc., but mainly text data. In order to meet diverse language requirements, the multi-source heterogeneous data obtained by the current execution end includes different language types, such as Chinese, English, French, Japanese, German, etc., which are not specifically limited in the embodiment of the present invention.

[0081] In addition, the current execution end also constructs knowledge graphs corresponding to various language types based on multi-source heterogeneous data, that is, integrates and processes the knowledge content corresponding to various language types, and describes and stores them in the form of knowledge graphs.

[0082] It should be noted that for different business scenarios, multi-source heterogeneous data can be selectively processed to generate knowledge graphs relevant to the business scenario. For example, in a fintech scenario, data related to investments, financial products, and financial market conditions can be selected from multi-source heterogeneous data, although this is not specifically limited in this embodiment of the present invention.

[0083] In addition, in the smart medical scenario, data related to medical records, data related to laboratory tests, data related to medication, information related to diseases, etc. can be selected from multi-source heterogeneous data, and the embodiments of the present invention do not make specific limitations.

[0084] 102. Obtain large-scale multilingual question-answering corpus, and train corresponding question-answering models according to different language types in combination with the knowledge graph;

[0085] In this embodiment, to enable the model to answer questions in various languages, the current execution end first needs to obtain a large-scale multilingual question-answer corpus. This large-scale multilingual question-answer corpus contains question-answer pairs in multiple languages ​​(such as Chinese, English, French, Japanese, and German). This data includes publicly available multilingual question-answer datasets, historical user interaction records, and high-quality parallel corpora generated by translation tools to ensure semantic consistency between languages.

[0086] In addition, the current execution end also combines the knowledge graph constructed in step 101 to train corresponding question-answering models according to different language types, that is, each language type corresponds to a question-answering model, and the number of question-answering models is consistent with the number of language types.

[0087] 103. Performing semantic alignment processing on the knowledge graphs corresponding to various language types, so as to enable cross-language knowledge sharing among the knowledge graphs and obtain a multilingual shared knowledge graph;

[0088] In this embodiment, to enhance cross-language question-answering capabilities, the current execution end performs semantic alignment on the knowledge graphs corresponding to various language types constructed in step 101. This allows each knowledge graph to share knowledge across languages, resulting in a multilingual shared knowledge graph. This multilingual shared knowledge graph facilitates cross-language knowledge sharing and integration. Semantic alignment includes entity alignment, relationship alignment, and other processes, which are not specifically limited in this embodiment.

[0089] 104. Using the multilingual shared knowledge graph, perform model enhancement processing on the question-answering model corresponding to each language type to obtain an enhanced question-answering model with cross-language knowledge association capability;

[0090] In this embodiment, the current execution end uses a semantically aligned multilingual shared knowledge graph to perform model enhancement processing on the question-answering models corresponding to each language type, resulting in an enhanced question-answering model with cross-lingual knowledge association capabilities. This model enhancement process is primarily used for multilingual context expansion and cross-lingual reasoning. By integrating cross-lingual knowledge graphs and generating answers, the processing efficiency and accuracy of multilingual question-answering are further optimized.

[0091] 105. Receive a user question, and determine a target enhanced question-answering model from multiple enhanced question-answering models based on the language type corresponding to the user question, so that the target enhanced question-answering model generates a response result corresponding to the user question.

[0092] In this embodiment, the current execution end receives a user question sent by the client. To facilitate user operation, the current execution end can design a unified question-and-answer interface that supports user input of questions in any language and automatically identifies the language type. Next, based on the language type identification result, the current execution end selects a corresponding model for processing. Specifically, the current execution end determines a target enhanced question-and-answer model from multiple enhanced question-and-answer models, so that the determined target enhanced question-and-answer model generates a response corresponding to the user question. Furthermore, a multilingual translation function can be designed as needed to output high-quality multilingual responses.

[0093] It should be noted that the current execution end determines the target enhanced question-answering model based on the language type while also determining the knowledge graph corresponding to the current language type. For example, if the user inputs a question in Chinese, the Chinese knowledge graph is determined at the same time as the target enhanced question-answering model corresponding to Chinese. The current execution end first calls the Chinese knowledge graph in the multilingual shared knowledge graph based on the Chinese question; if the Chinese knowledge is insufficient, the English knowledge graph in the multilingual shared knowledge graph is called through semantic alignment processing; through the cross-language association of the knowledge graph, the processing efficiency and accuracy of multilingual question answering are optimized.

[0094] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to effectively integrate the knowledge content in multi-source heterogeneous data, another intelligent question answering processing method is provided, such as Figure 3 As shown, the steps of constructing knowledge graphs corresponding to various language types based on the multi-source heterogeneous data include:

[0095] 201. Acquire target multi-source heterogeneous data of the same language type, and preprocess the target multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data;

[0096] In this embodiment, the current execution end obtains target multi-source heterogeneous data of the same language type. Because the original data often contains noise, duplicate information, and inconsistent formats, the target multi-source heterogeneous data must first be preprocessed to improve data quality. Preprocessing can utilize natural language processing (NLP) techniques to perform data cleaning, word segmentation, stop word removal, and standardization. This is not specifically limited in this embodiment of the present invention. After preprocessing the target multi-source heterogeneous data, preprocessed multi-source heterogeneous data is obtained.

[0097] It should be noted that, in this embodiment, the current execution end may also perform timestamp marking on the pre-processed multi-source heterogeneous data to determine the freshness of the information.

[0098] 202. Perform entity recognition processing on the pre-processed multi-source heterogeneous data using a pre-trained entity recognition model to obtain entity information; and determine semantic relationships between the entity information using a pre-trained relationship extraction model;

[0099] In this embodiment, the current execution end uses a pre-trained entity recognition model to perform entity recognition processing on the pre-processed multi-source heterogeneous data obtained in step 201 to obtain entity information in the data. For example, in a financial technology scenario, entity recognition is performed on the data "What is the annualized rate of return of Ping An Yueyuean's financial product in one year?", and the entity information of "Ping An Yueyuean", "financial product", "one year", and "annualized rate of return" can be obtained. The embodiment of the present invention does not make specific limitations. In this embodiment, the pre-trained entity recognition model is preferably a BERT-NER model, and the embodiment of the present invention does not make specific limitations. In addition, the current execution end also needs to use a pre-trained relationship extraction model to determine the semantic relationship between each entity information, including belonging to, located in, etc., which is not specifically limited in the embodiment of the present invention.

[0100] 203. Construct the knowledge graph corresponding to the language type based on the entity information and the semantic relationship.

[0101] In an embodiment of the present invention, the current execution end uses the entity information and semantic relationship obtained in step 202 as nodes and edges in the knowledge graph, respectively, to obtain a knowledge graph corresponding to each language type.

[0102] It should be noted that to accommodate the rapidly changing information environment, the current execution end can also perform incremental updates on the constructed knowledge graph. Specifically, the current execution end compares newly identified entity information and semantic relationships with the existing knowledge graph, using entity alignment and knowledge fusion algorithms to determine whether the knowledge is new or an update to existing knowledge. For new or updated knowledge, the current execution end efficiently integrates it into the knowledge graph through an incremental graph database update mechanism, avoiding the need to rebuild the entire knowledge base and thus improving update efficiency.

[0103] In addition, in order to improve the relevance and personalization of intelligent question and answer, the current execution end can also dynamically adjust the weights of entity information and semantic relationships in the knowledge graph based on information such as the credibility of the data source, timestamp, and user feedback. Specifically, during the knowledge graph construction phase, the current execution end assigns an initial weight to each entity information and semantic relationship. In this embodiment, the weight calculation can be based on indicators such as the authority of the data source (such as news websites, social media), release time, number of citations, semantic relevance, and user feedback to form a multi-dimensional weight evaluation system. Then, the current execution end uses a weight adjustment algorithm based on reinforcement learning or online learning to dynamically update the weights of relevant entity information and semantic relationships in the knowledge graph based on indicators such as the authority of the data source (such as news websites, social media), release time, number of citations, semantic relevance, and user feedback. During the question and answer generation phase, the current execution end can perform personalized sorting and recommendation of possible answer results based on the user's current query intent and historical behavior, combined with the dynamic weights of entity information and semantic relationships in the knowledge graph. For example, in a polysemous word or multiple answer scenario, the current execution end prioritizes returning answer results with higher weights that are more in line with user preferences.

[0104] It's important to note that to prevent outdated information from disrupting the system, the current execution side also implements a weight decay mechanism. This means that over time, the weight of older knowledge decreases, while the weight of newer knowledge increases rapidly based on its source and user feedback. This mechanism ensures that the system always prioritizes the latest, most relevant, and most user-friendly knowledge.

[0105] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to enable the model to have the ability to answer questions in various language types, another intelligent question answering processing method is provided, such as Figure 4 As shown, the steps are to obtain large-scale multilingual question-answering corpus and train corresponding question-answering models according to different language types in combination with the knowledge graph, including:

[0106] 301. Divide the large-scale multilingual question-answering corpus into corpora corresponding to various language types;

[0107] In this embodiment, since the large-scale multilingual question and answer corpus contains question and answer pair data in multiple language types (Chinese, English, French, Japanese, German, etc.), in order to perform separate analysis and training on the corpus corresponding to each language type, the current execution end needs to first divide the large-scale multilingual question and answer corpus into corpora corresponding to various language types.

[0108] 302. Determine the resource status of each corpus and use a transfer learning strategy to enhance the resources of the corpus with insufficient resources so that each corpus meets the resource requirement.

[0109] In this embodiment, the current execution end determines the resource status of each corpus, that is, determines whether the data volume of the question-answer pair data in the corpus meets the requirements. For corpora that do not meet the requirements, the data volume needs to be increased. In this embodiment, the current execution end uses a transfer learning strategy to perform resource enhancement processing on corpora with insufficient resources, that is, corpora with insufficient data volume. For example, relevant knowledge of mainstream languages ​​(such as Chinese, English, etc.) is transferred to the target language (such as Spanish, Taushiro, etc.) to ensure that each of the corpora meets the resource volume requirements. This embodiment of the present invention does not make specific limitations.

[0110] 303. In combination with the knowledge graph, a corpus corresponding to each language type is used to perform model training processing on the multilingual pre-training model to obtain the question-answering model corresponding to each language type.

[0111] In this embodiment, the current execution end combines the knowledge graph and uses corpora corresponding to various language types to perform model training on the multilingual pre-trained model, specifically including:

[0112] (1) performing data cleaning and preprocessing on the corpus corresponding to each language type to obtain a preprocessed corpus;

[0113] In this embodiment, since the question-answer data in the corpus may contain meaningless symbols, extra spaces, illegal characters, etc., the current execution end can delete the meaningless symbols, extra spaces, advertising links, illegal characters, etc. through data cleaning. In addition, the current execution end can also perform word segmentation on the question-answer text. For different languages, corresponding word segmentation tools are used. For example, the NLTK library is used for word segmentation for English, and the Jieba word segmentation tool is used for Chinese, etc., which is not specifically limited in the embodiment of the present invention. In addition, the current execution end can also perform part-of-speech tagging, named entity recognition and other operations on the text after word segmentation in order to better understand the structure and meaning of the text. For example, the nouns, verbs, names, place names and other information in the text are marked to obtain a preprocessed corpus.

[0114] (2) performing long text processing, polysemy processing, and complex semantic relationship processing on the corpus in combination with the knowledge graph to obtain processed corpus-related information corresponding to various language types;

[0115] (3) Performing model training on the multilingual pre-training model based on the processed corpus-related information to obtain the question-answering model corresponding to various language types.

[0116] In an embodiment of the present invention, the current execution end combines knowledge graphs corresponding to various language types to perform long text processing, polysemy processing, and complex semantic relationship processing on the corpus, thereby obtaining information related to the processed corpus corresponding to each language type. For long text processing, a Transformer-based model structure can be used. Its self-attention mechanism can capture long-range dependencies, effectively understanding key information in long text. Furthermore, the system introduces segmentation and context splicing mechanisms to split long text into multiple segments for processing. This contextual information is integrated during the model inference phase, ensuring that the system can accurately understand the semantic logic in long questions or answers. Secondly, when processing polysemy, the system combines context-aware word sense disambiguation technology with entity information in the knowledge graph for joint judgment. During the model training phase, the system uses a multi-task learning approach to simultaneously train word sense recognition and question-answering tasks, enabling the model to automatically identify the correct meaning of words based on context. During the inference phase, the system matches polysemous words in the question with entities in the knowledge graph and performs semantic disambiguation based on the entity's contextual information, thereby improving the accuracy of understanding polysemous words. Thirdly, when processing complex semantic relationships, for problems involving multiple entities, multi-hop reasoning, or implicit relationships, the system combines the relationship path reasoning in the knowledge graph with the semantic modeling capabilities of the deep learning model to achieve the parsing of complex semantic relationships. Specifically, in the problem understanding stage, the system uses the pre-trained model to extract the key entities and intentions in the problem, and reasoning through the entity relationship chain in the knowledge graph to identify the implicit logical relationship in the problem. It should be noted that the system can also introduce graph neural networks (GNNs) to perform graph structure modeling of entities and relationships in the knowledge graph, further enhancing the understanding and reasoning capabilities of complex semantic relationships.

[0117] In an embodiment of the present invention, multilingual pre-trained models include mBERT, XLM-R, mT5, etc., which are not specifically limited in the embodiment of the present invention. These models are pre-trained on large-scale multilingual corpora and have cross-language semantic understanding capabilities. On this basis, the system further performs cross-language fine-tuning training, inputs multilingual question and answer pairs into the model, enables it to learn the semantic mapping relationship between different languages, and improves the accuracy of cross-language question and answer. This embodiment introduces Transformer-based pre-trained language models, context-aware word sense disambiguation mechanisms, knowledge graph relationship path reasoning, and graph neural networks, so that the system can effectively process long texts, polysemous words, and complex semantic relationships, thereby improving the accuracy and semantic understanding of questions and answers, and meeting users' intelligent interaction needs in diverse scenarios.

[0118] Furthermore, as a refinement and extension of the specific implementation of the above embodiment, in order to enhance the cross-language question-answering capability, another intelligent question-answering processing method is provided, which performs semantic alignment processing on the knowledge graphs corresponding to various language types to enable cross-language knowledge sharing of each knowledge graph to obtain a multilingual shared knowledge graph, including:

[0119] Map the same entity information in different language types to the same knowledge node; and map the same semantic relationship in different language types to the same relationship type;

[0120] Cross-language knowledge sharing is performed based on the same knowledge node and the same relationship type to obtain the multilingual shared knowledge graph.

[0121] In this embodiment, the current execution end maps the same entity information in different language types to the same knowledge node based on the embedded alignment method (such as TransE, TransR, etc.); for example, mapping "Apple Company" (Chinese) and "Apple Inc." (English) to the same knowledge node can effectively avoid entity recognition errors caused by language differences (such as "apple" may refer to fruit or company), and ensure the consistency and searchability of knowledge in a multilingual environment. Similarly, a similar alignment process can be performed to map the same semantic relationship in different language types to the same relationship type to ensure that the same semantic relationship in different languages ​​can be correctly mapped. Then, the current execution end performs cross-language knowledge sharing based on the above-mentioned same knowledge node and the same relationship type to obtain the multilingual shared knowledge graph.

[0122] It should be noted that, before the step of performing cross-language knowledge sharing based on the same knowledge node and the same relationship type, the method further includes:

[0123] Determining an incomplete relationship chain from the knowledge graphs corresponding to various languages ​​based on the same knowledge node and the same relationship type;

[0124] The incomplete relationship chains are subjected to knowledge completion processing based on the complete relationship chains in each knowledge graph to obtain a completed knowledge graph, so as to enable cross-language knowledge sharing based on the completed knowledge graph.

[0125] In this embodiment, the current execution end determines incomplete relationship chains from the corresponding knowledge graphs in various languages ​​based on the same knowledge node and the same relationship type. For example, in a smart healthcare scenario, the Chinese knowledge graph contains the Chinese relationship chain "adenoids - bacteria - allergy - nasal congestion - runny nose - otitis media." However, the English knowledge graph for the condition "adenoids" lacks "allergy" and "otitis media." Therefore, the relationship chain corresponding to the condition "adenoids" in the English knowledge graph is an incomplete relationship chain, and so on. This embodiment of the present invention does not specifically limit this. The current execution end needs to perform knowledge completion processing on the incomplete English relationship chain based on the Chinese relationship chain, that is, to add "allergy" and "otitis media" to the above-mentioned English relationship chain, thereby obtaining a completed knowledge graph, so that cross-language knowledge sharing can be carried out based on the completed knowledge graph. Combined with transfer learning, the structure and relationships of the English knowledge graph are mapped to the Chinese knowledge graph, improving the knowledge coverage of low-resource languages.

[0126] Furthermore, as a refinement and extension of the above embodiment, in order to improve the accuracy of multilingual question answering, another intelligent question answering processing method is provided, such as Figure 5 As shown, the step uses the multilingual shared knowledge graph to perform model enhancement processing on the question-answering model corresponding to each language type to obtain an enhanced question-answering model with cross-language knowledge association capabilities, including:

[0127] 401. Obtain the target relationship chain in the knowledge graph corresponding to the current language type;

[0128] 402. Based on the semantic alignment in the multilingual shared knowledge graph, obtain multiple related relationship chains corresponding to the target relationship chain from the knowledge graphs corresponding to other language types other than the current language type;

[0129] 403. Perform model enhancement processing on the question-answering model corresponding to the current language type based on the target relationship chain and the related relationship chain to obtain the enhanced question-answering model with cross-language knowledge association capability.

[0130] In an embodiment of the present invention, the current execution end obtains the target relationship chain in the knowledge graph corresponding to the current language type, such as the relationship chain of the cause, symptoms, and treatment methods of a disease in the Chinese knowledge graph in a smart medical scenario, which is not specifically limited by the embodiment of the present invention. The current execution end obtains multiple related relationship chains corresponding to the disease from other knowledge graphs such as the English knowledge graph and the Japanese knowledge graph based on the semantic alignment method in the multilingual shared knowledge graph. Then, the current execution end performs model enhancement processing on the Chinese question-answering model based on multiple relationship chains such as the relationship chain in the text of the renamed disease, the English relationship chain, the Japanese relationship chain, etc., to obtain an enhanced question-answering model with cross-language knowledge association capabilities.

[0131] The embodiment of the present invention provides an intelligent question-answering processing method. Compared with the existing technology, the present invention introduces streaming data processing technology by separately acquiring multi-source heterogeneous data of different language types and constructing knowledge graphs corresponding to various language types based on the multi-source heterogeneous data. It can achieve real-time acquisition, efficient processing and dynamic updating of multi-source heterogeneous data, thereby ensuring the timeliness and accuracy of the knowledge graph, providing continuously updated knowledge support for the intelligent question-answering system, and significantly improving the intelligence level and practical application effect of the system. The present invention acquires large-scale multilingual question-answering corpus and trains corresponding question-answering models according to different language types in combination with the knowledge graph; performs semantic alignment processing on the knowledge graphs corresponding to various language types so that each of the knowledge graphs can share knowledge across languages, thereby obtaining a multilingual shared knowledge graph; and uses the multilingual shared knowledge graph to perform model enhancement processing on the question-answering models corresponding to each language type, thereby obtaining an enhanced question-answering model with cross-language knowledge association capabilities. This achieves efficient support and intelligent processing of multilingual question-answering interactions, and improves the system's multilingual adaptability and global user coverage.

[0132] As the above Figure 1 The embodiment of the present invention provides an intelligent question-answering processing device, such as Figure 6 As shown, the device includes:

[0133] A graph construction module 51 is used to obtain multi-source heterogeneous data of different language types and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data;

[0134] A model training module 52 is used to obtain large-scale multilingual question-answering corpus and train corresponding question-answering models according to different language types in combination with the knowledge graph;

[0135] The knowledge fusion module 53 is used to perform semantic alignment processing on the knowledge graphs corresponding to various language types, so as to enable cross-language knowledge sharing among the knowledge graphs and obtain a multilingual shared knowledge graph;

[0136] A model enhancement module 54 is configured to perform model enhancement processing on the question-answering model corresponding to each language type using the multilingual shared knowledge graph to obtain an enhanced question-answering model with cross-language knowledge association capability;

[0137] The question-answering module 55 is used to receive user questions and determine a target enhanced question-answering model from multiple enhanced question-answering models based on the language type corresponding to the user questions, so that the target enhanced question-answering model generates a response result corresponding to the user questions.

[0138] Furthermore, the graph construction module 51 is also used to:

[0139] Acquire target multi-source heterogeneous data of the same language type, and preprocess the target multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data;

[0140] Using a pre-trained entity recognition model to perform entity recognition processing on the pre-processed multi-source heterogeneous data to obtain entity information; and using a pre-trained relationship extraction model to determine the semantic relationship between the entity information;

[0141] The knowledge graph corresponding to the language type is constructed based on the entity information and the semantic relationship.

[0142] Furthermore, the model training module 52 also includes:

[0143] A corpus division unit, configured to divide the large-scale multilingual question-answer corpus into corpora corresponding to various language types;

[0144] A resource enhancement unit is used to judge the resource situation of each corpus and use a transfer learning strategy to enhance the resources of the corpus with insufficient resources so that each corpus meets the resource requirement;

[0145] The training unit is used to combine the knowledge graph and use the corpus corresponding to various language types to perform model training processing on the multilingual pre-training model to obtain the question-answering model corresponding to various language types.

[0146] Furthermore, the training unit in the model training module 52 is also used to:

[0147] Performing data cleaning and preprocessing on the corpus corresponding to each language type to obtain a preprocessed corpus;

[0148] Combined with the knowledge graph, the corpus is processed for long text, polysemy, and complex semantic relationships to obtain processed corpus-related information corresponding to various language types;

[0149] Model training is performed on the multilingual pre-training model based on the processed corpus-related information to obtain the question-answering models corresponding to various language types.

[0150] Furthermore, the knowledge fusion module 53 is also used to:

[0151] Map the same entity information in different language types to the same knowledge node; and map the same semantic relationship in different language types to the same relationship type;

[0152] Cross-language knowledge sharing is performed based on the same knowledge node and the same relationship type to obtain the multilingual shared knowledge graph.

[0153] Furthermore, the knowledge fusion module 53 is also used to:

[0154] Determining an incomplete relationship chain from the knowledge graphs corresponding to various languages ​​based on the same knowledge node and the same relationship type;

[0155] The incomplete relationship chains are subjected to knowledge completion processing based on the complete relationship chains in each knowledge graph to obtain a completed knowledge graph, so as to enable cross-language knowledge sharing based on the completed knowledge graph.

[0156] Furthermore, the model enhancement module 54 is further configured to:

[0157] Get the target relationship chain in the knowledge graph corresponding to the current language type;

[0158] Based on the semantic alignment in the multilingual shared knowledge graph, a plurality of related relationship chains corresponding to the target relationship chain are obtained from the knowledge graphs corresponding to other language types except the current language type;

[0159] Based on the target relationship chain and the related relationship chain, the question-answering model corresponding to the current language type is enhanced to obtain the enhanced question-answering model with cross-language knowledge association capability.

[0160] The embodiment of the present invention provides an intelligent question-answering processing device. Compared with the existing technology, the present invention separately obtains multi-source heterogeneous data of different language types and constructs knowledge graphs corresponding to various language types based on the multi-source heterogeneous data. It introduces streaming data processing technology, which can achieve real-time acquisition, efficient processing and dynamic updating of multi-source heterogeneous data, thereby ensuring the timeliness and accuracy of the knowledge graph, providing continuously updated knowledge support for the intelligent question-answering system, and significantly improving the intelligence level and practical application effect of the system. The present invention obtains large-scale multilingual question-answering corpus and trains corresponding question-answering models according to different language types in combination with the knowledge graph; performs semantic alignment processing on the knowledge graphs corresponding to various language types to enable cross-language knowledge sharing of each knowledge graph to obtain a multilingual shared knowledge graph; uses the multilingual shared knowledge graph to perform model enhancement processing on the question-answering models corresponding to each language type to obtain an enhanced question-answering model with cross-language knowledge association capabilities, thereby achieving efficient support and intelligent processing of multi-language question-answering interactions, and improving the system's multilingual adaptability and global user coverage.

[0161] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer-executable instruction can execute the intelligent question-answering processing method in any of the above method embodiments.

[0162] Figure 7 A schematic structural diagram of a computer device provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computer device.

[0163] like Figure 7 As shown, the computer device may include: a processor (processor) 602 , a communication interface (Communications Interface) 604 , a memory (memory) 606 , and a communication bus 608 .

[0164] The processor 602 , the communication interface 604 , and the memory 606 communicate with each other via a communication bus 608 .

[0165] The communication interface 604 is used to communicate with other devices such as clients or other servers.

[0166] The processor 602 is used to execute the program 610, and specifically can execute the relevant steps of the above-mentioned intelligent question and answer processing method.

[0167] Specifically, the program 610 may include program codes, which include computer operation instructions.

[0168] Processor 602 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention. The one or more processors included in a computer device may be of the same type, such as one or more CPUs, or may be of different types, such as one or more CPUs and one or more ASICs.

[0169] The memory 606 is used to store the program 610. The memory 606 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0170] The program 610 may be specifically configured to enable the processor 602 to perform the following operations:

[0171] Acquire multi-source heterogeneous data of different language types respectively, and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data;

[0172] Obtain large-scale multilingual question-answering corpus, and train corresponding question-answering models according to different language types based on the knowledge graph;

[0173] Performing semantic alignment processing on the knowledge graphs corresponding to various language types so as to enable cross-language knowledge sharing among the knowledge graphs to obtain a multilingual shared knowledge graph;

[0174] Using the multilingual shared knowledge graph to perform model enhancement processing on the question-answering model corresponding to each language type, thereby obtaining an enhanced question-answering model with cross-language knowledge association capabilities;

[0175] A user question is received, and a target enhanced question-answering model is determined from a plurality of enhanced question-answering models based on a language type corresponding to the user question, so that the target enhanced question-answering model generates a reply result corresponding to the user question.

[0176] Obviously, those skilled in the art will appreciate that the various modules or steps of the present invention described above can be implemented using a general-purpose computing device, centralized on a single computing device, or distributed across a network of multiple computing devices. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. In some cases, the steps shown or described can be performed in a different order than that shown, or can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0177] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

Claims

1. An intelligent question-answering processing method, characterized in that: include: Acquire multi-source heterogeneous data of different language types respectively, and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data; Obtain large-scale multilingual question-answering corpus, and train corresponding question-answering models according to different language types based on the knowledge graph; Performing semantic alignment processing on the knowledge graphs corresponding to various language types so as to enable cross-language knowledge sharing among the knowledge graphs to obtain a multilingual shared knowledge graph; Using the multilingual shared knowledge graph to perform model enhancement processing on the question-answering model corresponding to each language type, thereby obtaining an enhanced question-answering model with cross-language knowledge association capabilities; A user question is received, and a target enhanced question-answering model is determined from a plurality of enhanced question-answering models based on a language type corresponding to the user question, so that the target enhanced question-answering model generates a reply result corresponding to the user question.

2. The method according to claim 1, characterized in that The step of constructing knowledge graphs corresponding to various language types based on the multi-source heterogeneous data includes: Acquire target multi-source heterogeneous data of the same language type, and preprocess the target multi-source heterogeneous data to obtain preprocessed multi-source heterogeneous data; Using a pre-trained entity recognition model to perform entity recognition processing on the pre-processed multi-source heterogeneous data to obtain entity information; and using a pre-trained relationship extraction model to determine the semantic relationship between the entity information; The knowledge graph corresponding to the language type is constructed based on the entity information and the semantic relationship.

3. The method according to claim 1, characterized in that The method of acquiring large-scale multilingual question-answering corpus and training corresponding question-answering models according to different language types in combination with the knowledge graph includes: Dividing the large-scale multilingual question-answering corpus into corpora corresponding to various language types; Determine the resource situation of each corpus and use a transfer learning strategy to enhance the resources of the corpus with insufficient resources so that each corpus meets the resource requirements; In combination with the knowledge graph, the multilingual pre-training model is trained using corpora corresponding to various language types to obtain the question-answering models corresponding to various language types.

4. The method according to claim 3, characterized in that The method of combining the knowledge graph and using corpora corresponding to various language types to perform model training on the multilingual pre-trained model to obtain the question-answering model corresponding to various language types includes: Performing data cleaning and preprocessing on the corpus corresponding to each language type to obtain a preprocessed corpus; Combined with the knowledge graph, the corpus is processed for long text, polysemy, and complex semantic relationships to obtain processed corpus-related information corresponding to various language types; Model training is performed on the multilingual pre-training model based on the processed corpus-related information to obtain the question-answering models corresponding to various language types.

5. The method according to claim 1, wherein The semantic alignment processing of the knowledge graphs corresponding to various language types is performed to enable cross-language knowledge sharing among the knowledge graphs to obtain a multilingual shared knowledge graph, including: Map the same entity information in different language types to the same knowledge node; and map the same semantic relationship in different language types to the same relationship type; Cross-language knowledge sharing is performed based on the same knowledge node and the same relationship type to obtain the multilingual shared knowledge graph.

6. The method according to claim 5, characterized in that Before performing cross-language knowledge sharing based on the same knowledge node and the same relationship type, the method further includes: Determining an incomplete relationship chain from the knowledge graphs corresponding to various languages ​​based on the same knowledge node and the same relationship type; The incomplete relationship chains are subjected to knowledge completion processing based on the complete relationship chains in each knowledge graph to obtain a completed knowledge graph, so as to enable cross-language knowledge sharing based on the completed knowledge graph.

7. The method according to any one of claims 1 to 6, characterized in that The method of using the multilingual shared knowledge graph to perform model enhancement processing on the question-answering model corresponding to each language type to obtain an enhanced question-answering model with cross-language knowledge association capability includes: Get the target relationship chain in the knowledge graph corresponding to the current language type; Based on the semantic alignment in the multilingual shared knowledge graph, a plurality of related relationship chains corresponding to the target relationship chain are obtained from the knowledge graphs corresponding to other language types except the current language type; Based on the target relationship chain and the related relationship chain, the question-answering model corresponding to the current language type is enhanced to obtain the enhanced question-answering model with cross-language knowledge association capability.

8. An intelligent question-answering processing device, characterized in that: include: A graph construction module is used to obtain multi-source heterogeneous data of different language types and construct knowledge graphs corresponding to various language types based on the multi-source heterogeneous data; A model training module is used to obtain large-scale multilingual question-answering corpora and train corresponding question-answering models according to different language types in combination with the knowledge graph; A knowledge fusion module is used to perform semantic alignment processing on the knowledge graphs corresponding to various language types, so as to enable cross-language knowledge sharing among the knowledge graphs and obtain a multilingual shared knowledge graph; A model enhancement module, configured to perform model enhancement processing on the question-answering model corresponding to each language type using the multilingual shared knowledge graph to obtain an enhanced question-answering model with cross-language knowledge association capabilities; The question-answering module is used to receive user questions and determine a target enhanced question-answering model from multiple enhanced question-answering models based on the language type corresponding to the user questions, so that the target enhanced question-answering model generates a response result corresponding to the user questions.

9. A storage medium, characterized in that: The storage medium stores at least one executable instruction, and the executable instruction executes an operation corresponding to the intelligent question-answering processing method according to any one of claims 1 to 7.

10. A computer device, characterized in that: comprising a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the intelligent question and answer processing method according to any one of claims 1 to 7.

Images