A vehicle-mounted question-answering method, device, equipment and storage medium based on vehicle networking

Abstract

Embodiments of the present invention provide an in-vehicle question-and-answer method, apparatus, system, device and medium based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the node to which they belong. This method enables the data to be located and retrieved more quickly and accurately after storage. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the node to which it belongs to the vehicle, so that the user can adjust the driving route and plan the itinerary in time through the obtained answers, thereby improving the convenience and efficiency of travel.

Description

Technical Field

[0001] The present invention relates to the field of vehicle technology, and in particular to an in-vehicle question-and-answer method, device, equipment and storage medium based on an Internet of Vehicles. Background Art

[0002] C-V2X (Cellular Vehicle-to-Everything) is a key technology that supports wireless communication between vehicles and other vehicles (V2V), infrastructure (V2I), pedestrians (V2P) and networks (V2N), aiming to improve road safety and traffic efficiency. With the rapid development of intelligent driving technology, in-vehicle voice assistants have become an important way for drivers to interact with the system. However, in existing technologies, it is difficult for in-vehicle question-and-answer systems to accurately retrieve and integrate multi-source information such as external traffic, navigation and vehicle status, which affects the system's intelligent decision-making capabilities. Summary of the invention

[0003] In view of the above problems, embodiments of the present invention are proposed to provide an in-vehicle question-and-answer method, apparatus, system, device and storage medium based on the Internet of Vehicles that overcome the above problems or at least partially solve the above problems.

[0004] In order to solve the above problems, an embodiment of the present invention discloses a vehicle-mounted question-answering method based on the Internet of Vehicles, which is applied to a server. The method includes:

[0005] Distribute and store the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data;

[0006] Adding the embedding vector corresponding to the newly added traffic data to the index of the corresponding node;

[0007] Constructing a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes;

[0008] When a query request from a vehicle is received, a target answer is searched for in the question-and-answer database according to the query request, and the target answer is sent from the node to which it belongs to the vehicle.

[0009] Optionally, the index includes a temporary index and a long-term index, and adding the embedding vector corresponding to the newly added traffic data to the index of the corresponding node includes:

[0010] Adding the embedding vector corresponding to the newly added traffic data to the temporary index of the corresponding node;

[0011] Also includes:

[0012] After a preset time interval, the embedding vector corresponding to the newly added data stored in the temporary index is migrated to the long-term index;

[0013] The step of constructing a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes includes:

[0014] The question-and-answer database is constructed according to the embedding vector corresponding to the newly added traffic data in the temporary index and the embedding vector corresponding to the newly added traffic data in the long-term index.

[0015] Optionally, the method further comprises:

[0016] Recording the storage time of the embedded vector stored in the long-term index;

[0017] When the storage time of the embedded vectors stored in the long-term index exceeds a preset time threshold, the embedded vectors exceeding the preset time threshold are deleted from the long-term index.

[0018] Optionally, constructing a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes includes:

[0019] Identify the embedding vectors corresponding to the newly added traffic data to obtain keywords and identification entities;

[0020] Generate candidate answers based on the keywords and the identified entities;

[0021] According to the correspondence between the keywords and the candidate answers, the candidate answers are stored in the question and answer database.

[0022] Optionally, adding the embedding vector corresponding to the newly added traffic data to the index of the corresponding node includes:

[0023] Determine the similarity between the embedding vector corresponding to the newly added traffic data and the embedding vectors respectively stored in the plurality of nodes;

[0024] If the similarity between the embedding vector corresponding to the newly added traffic data and the embedded vector stored in the target node is greater than a preset similarity threshold, the embedding vector corresponding to the newly added traffic data is added to the index to which the target node belongs.

[0025] Optionally, searching the question-and-answer database for a target answer according to the query request includes:

[0026] Parsing the query request to obtain question features;

[0027] Querying the question-answer database for candidate answers associated with the question feature, and determining the similarity between the question feature and the associated candidate answers;

[0028] According to the similarity between the question feature and the associated candidate answers, a target answer is determined from the associated candidate answers.

[0029] Optionally, it also includes:

[0030] When the acquired new traffic data contains preset keywords, an alarm message is output.

[0031] The present invention also discloses a vehicle-mounted question-answering method based on the Internet of Vehicles, which is applied to a vehicle, and the method comprises:

[0032] Collecting new traffic data around the vehicle, and uploading the new traffic data to the Internet of Vehicles;

[0033] Get the query request entered by the user;

[0034] Sending the query request to the server so that the server searches for the target answer in the question-and-answer database; the question-and-answer database is obtained by the server distributing and storing the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generating an embedding vector corresponding to the newly added traffic data, which is constructed according to the embedding vector corresponding to the newly added traffic data;

[0035] The target answer sent by the server is received.

[0036] The present invention also discloses a vehicle-mounted question-answering system based on the Internet of Vehicles, the system comprising a server and a vehicle;

[0037] The vehicle is used to collect new traffic data around the vehicle, upload the new traffic data to the Internet of Vehicles, obtain a query request input by a user, send the query request to the server, and receive a target answer sent by the server;

[0038] The server is used to distribute and store the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data; add the embedding vector corresponding to the newly added traffic data to the index of the node to which it belongs; build a question and answer database based on the embedding vector corresponding to the newly added traffic data in the index of the multiple nodes; when receiving a query request from a vehicle, query the question and answer database for a target answer based on the query request, and send the target answer from the node to which it belongs to the vehicle.

[0039] The present invention also discloses a vehicle-mounted question-answering device based on the Internet of Vehicles, which is applied to a server, and the device includes:

[0040] A storage module, used to store the newly added traffic data obtained from the Internet of Vehicles in a distributed manner in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data;

[0041] An adding module, used to add the embedding vector corresponding to the newly added traffic data to the index of the corresponding node;

[0042] A construction module, configured to construct a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes;

[0043] The query module is used to query the target answer in the question-and-answer database according to the query request when receiving a query request from the vehicle, and send the target answer from the corresponding node to the vehicle.

[0044] Optionally, the index includes a temporary index and a long-term index, and the adding module includes:

[0045] A first adding submodule, used to add the embedding vector corresponding to the newly added traffic data to the temporary index of the corresponding node;

[0046] Also includes:

[0047] A migration submodule, configured to migrate the embedding vector corresponding to the newly added data stored in the temporary index to the long-term index after a preset time interval;

[0048] The building blocks include:

[0049] The first construction submodule is used to construct the question-answer database according to the embedding vector corresponding to the newly added traffic data in the temporary index and the embedding vector corresponding to the newly added traffic data in the long-term index.

[0050] Optionally, it also includes:

[0051] A recording submodule, used to record the storage time of the embedded vector stored in the long-term index;

[0052] A deletion submodule is used to delete the embedded vectors stored in the long-term index when the storage time of the embedded vectors exceeds the preset time threshold from the long-term index.

[0053] Optionally, the building block includes:

[0054] An identification submodule, used to identify the embedding vector corresponding to the newly added traffic data to obtain keywords and identification entities;

[0055] A generating submodule, for generating candidate answers according to the keywords and the identified entities;

[0056] The second construction submodule is used to store the candidate answers in the question and answer database according to the corresponding relationship between the keywords and the candidate answers.

[0057] Optionally, the adding module includes:

[0058] A similarity determination module, used to determine the similarity between the embedding vector corresponding to the newly added traffic data and the embedding vectors respectively stored in the plurality of nodes;

[0059] The second adding submodule is used to add the embedding vector corresponding to the newly added traffic data to the index to which the target node belongs if the similarity between the embedding vector corresponding to the newly added traffic data and the embedded vector stored in the target node is greater than a preset similarity threshold.

[0060] Optionally, the query module includes:

[0061] A parsing submodule, used to parse the query request to obtain question features;

[0062] A query submodule, configured to query the question-answer database for candidate answers associated with the question feature, and determine the similarity between the question feature and the associated candidate answers;

[0063] The determination submodule is used to determine a target answer from the associated candidate answers according to the similarity between the question feature and the associated candidate answers.

[0064] Optionally, it also includes:

[0065] The output module is used to output warning information when the acquired new traffic data contains preset keywords.

[0066] The present invention also discloses a vehicle-mounted question-answering device based on the Internet of Vehicles, which is applied to a vehicle, and the device comprises:

[0067] A collection module, used for collecting new traffic data around the vehicle and uploading the new traffic data to the Internet of Vehicles;

[0068] An acquisition module is used to obtain a query request input by a user;

[0069] A sending module, used for sending the query request to the server, so that the server searches for the target answer in the question-and-answer database; the question-and-answer database is obtained by the server from the Internet of Vehicles by distributing and storing the newly added traffic data obtained in multiple nodes, and generating an embedding vector corresponding to the newly added traffic data, which is constructed according to the embedding vector corresponding to the newly added traffic data;

[0070] A receiving module is used to receive the target answer sent by the server.

[0071] The present invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to execute the steps of the above-mentioned in-vehicle question-and-answer method based on the Internet of Vehicles.

[0072] The present invention also discloses a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the in-vehicle question-and-answer method based on the Internet of Vehicles are implemented as described above.

[0073] The embodiments of the present invention include the following advantages:

[0074] The present invention discloses an in-vehicle question-and-answer method based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the corresponding node. This method enables the data to be located and retrieved more quickly and accurately after being stored. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the corresponding node to the vehicle, so that the user can adjust the driving route and plan the itinerary in time through the obtained answer, thereby improving the convenience and efficiency of travel. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] Figure 1 is a flowchart of the steps of an in-vehicle question-answering method based on the Internet of Vehicles provided by an embodiment of the present invention;

[0076] Figure 2 It is a flowchart of an in-vehicle question-answering method based on Internet of Vehicles provided by an embodiment of the present invention;

[0077] Figure 3 is a flowchart of another in-vehicle question-answering method based on the Internet of Vehicles provided by an embodiment of the present invention;

[0078] Figure 4 is a structural block diagram of an in-vehicle question-answering system based on the Internet of Vehicles provided by an embodiment of the present invention;

[0079] Figure 5 is a structural block diagram of a vehicle-mounted question-and-answer device based on the Internet of Vehicles provided by an embodiment of the present invention;

[0080] Figure 6 It is a structural block diagram of another vehicle-mounted question-and-answer device based on the Internet of Vehicles provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0081] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and understandable, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0082] One of the core concepts of an embodiment of the present invention is that the present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the node to which it belongs. This method enables the data to be located and retrieved more quickly and accurately after storage; the present invention forms an ordered structure in the index, and when specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency; when the present invention receives a query request from a vehicle, it can use the index structure constructed based on the embedded vector in the question and answer database to quickly query the target answer therein and send it from the node to which it belongs to the vehicle, so that the user can adjust the driving route and plan the itinerary in time through the obtained answers, thereby improving the convenience and efficiency of travel.

[0083] Reference Figure 1 , shows a flowchart of a vehicle-based question-and-answer method provided by an embodiment of the present invention, which is applied to a server. The method may include the following steps:

[0084] In step 101, the newly added traffic data obtained from the Internet of Vehicles is distributed and stored in multiple nodes, and an embedding vector corresponding to the newly added traffic data is generated.

[0085] In an embodiment of the present invention, firstly, the server can receive new traffic data uploaded from various vehicle-mounted devices (such as vehicle-mounted sensors, driving recorders, etc.) and road infrastructure (such as intelligent traffic lights, roadside monitoring equipment, etc.) through a connection interface with the Internet of Vehicles. The new traffic data has various data types and can include real-time road condition information (such as vehicle speed, traffic volume, congestion, etc. of each road section), traffic accident information (location of occurrence, severity, etc.), road construction conditions, the impact of weather on traffic, and other related content.

[0086] Then the server can use a distributed storage system (such as common distributed file systems such as Hadoop Distributed File System, HDFS, or distributed databases such as Cassandra, etc.) to store the acquired new traffic data in multiple nodes according to pre-set rules. For example, the data can be divided and stored according to the source area (different cities, different road sections, etc.), data type (road conditions, accidents, etc.) or timestamps, so as to improve the reliability and scalability of data storage and facilitate subsequent efficient retrieval and processing. In a distributed system, each computing node is responsible for processing the traffic information of a part of the vehicles and generating embedded vectors for these vehicles. The generated vectors will be distributed and stored on multiple nodes. Each node maintains its own local index. When it is necessary to query the traffic conditions of a certain area, the system will query the indexes on all relevant nodes and summarize the results. This architecture not only improves the scalability of the system, but also avoids single-point performance bottlenecks, ensuring that the system can run stably under high concurrency.

[0087] Furthermore, natural language processing (NLP) or machine learning related technologies can be used to generate embedded vectors corresponding to the newly added traffic data. For example, for traffic data described in text form (such as "Due to heavy rain, the road was flooded in section XX, and vehicles were moving slowly"), the word embedding model can be used to map the words in the text to a low-dimensional vector space to obtain the vector representation corresponding to the entire text data. For numerical data (such as traffic flow values, vehicle speed values, etc.), they can be combined into a suitable vector form through standardization and normalization, or combined with some feature engineering methods to integrate them into the vector generated based on the text content, and finally obtain an embedded vector that can represent the semantics and features of the newly added traffic data.

[0088] Step 102: Add the embedding vector corresponding to the newly added traffic data to the index of the corresponding node.

[0089] In an embodiment of the present invention, each node maintains its own index structure, which can help quickly locate the location of data and information related to it. The index can be constructed based on different data structures, such as a hash index, which can quickly find the corresponding data through a specific key value, for example, establish a mapping relationship between a key-value pair and an embedded vector, and ensure that the corresponding key-value pair can be quickly found and matched each time a new embedded vector is generated.

[0090] Step 103: construct a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes.

[0091] In an embodiment of the present invention, the index of each node can be traversed to collect the embedded vectors corresponding to the newly added traffic data and the metadata associated therewith (such as data source, generation time, road section and other auxiliary information). These metadata are helpful in providing more comprehensive and accurate answers in subsequent queries and performing operations such as result screening.

[0092] A relational database, such as MySQL, Oracle, etc., can be used to set a suitable table structure (for relational databases) or collection and document structure (for non-relational databases) in the database, and store the embedded vectors and corresponding metadata according to logical relationships to facilitate subsequent efficient query operations. For example, a "question and answer data" table can be created, in which the column fields include embedded vector fields, traffic data description fields, source fields, related road section fields, etc., and each row corresponds to a complete data record related to traffic questions and answers.

[0093] Step 104, when a query request from a vehicle is received, a target answer is searched in a question-and-answer database according to the query request, and the target answer is sent from the node to which it belongs to the vehicle.

[0094] In an embodiment of the present invention, the server receives a query request sent from the vehicle through a network communication interface. The vehicle's query request is usually initiated by a vehicle-mounted terminal device (such as a vehicle-mounted navigation system, an intelligent vehicle computer, etc.). The request content may be about the road condition information of the current section of the road (such as "Is the section of XX kilometers ahead congested"), nearby traffic accidents (such as "Is there a traffic accident within 5 kilometers of me"), optional alternative routes (such as "Due to road construction ahead, a detour route is recommended") and other issues related to transportation.

[0095] The server can parse the received query request and extract the key information and intent. For example, it can use natural language understanding technology to analyze the keywords in the request statement, such as the road section name, distance range, road condition keywords "congestion" and "construction" in the above example), semantic roles (such as whether the subject of the inquiry is road conditions, accidents or routes, etc.) and request restrictions (such as distance range restrictions, time restrictions, etc.), so as to accurately convert them into a form that can be searched in the question and answer database.

[0096] The server can search the constructed question-and-answer database based on the parsed query request content. If the query is based on text keywords, the server can use the established index in the database to find the embedded vector corresponding to the traffic data record containing the relevant keywords. For queries involving conditions such as numerical ranges or geographical location ranges (such as distance ranges, road section area ranges, etc.), the server can use database query statements, such as the WHERE clause in SQL statements, to perform conditional screening and other operations to locate answers that meet the requirements. Then, based on evaluation methods such as the similarity of embedded vectors, such as calculating the cosine similarity between the query request vector and the embedded vectors of each traffic data in the database, the server can further screen out the traffic data answer corresponding to the most matching target answer.

[0097] When the traffic data record corresponding to the target answer is found, the complete target answer information can be extracted from the node that stores the record, and the answer data can be sent back to the vehicle-mounted terminal device through network communication protocols, such as the common TCP / IP protocol. After the vehicle receives the answer, it can display it on the on-board screen or inform the driver and passengers in the car through voice broadcasts, etc., to assist them in making reasonable traffic travel decisions.

[0098] The present invention discloses an in-vehicle question-and-answer method based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the corresponding node. This method enables the data to be located and retrieved more quickly and accurately after being stored. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the corresponding node to the vehicle, so that the user can adjust the driving route and plan the itinerary in time through the obtained answer, thereby improving the convenience and efficiency of travel.

[0099] In one embodiment of the present invention, the index includes a temporary index and a long-term index, and the embedding vector corresponding to the newly added traffic data is added to the index of the node to which it belongs, including: adding the embedding vector corresponding to the newly added traffic data to the temporary index of the node to which it belongs; also including: after a preset time interval, migrating the embedding vector corresponding to the newly added data stored in the temporary index to the long-term index; constructing a question and answer database based on the embedding vectors corresponding to the newly added traffic data in the indexes of multiple nodes, including: constructing a question and answer database based on the embedding vectors corresponding to the newly added traffic data in the temporary index and the embedding vectors corresponding to the newly added traffic data in the long-term index.

[0100] In the embodiment of the present invention, the temporary index is mainly used to quickly process the embedded vectors corresponding to the newly accessed traffic data in a short period of time. Its structure is relatively more flexible and the update operation is more frequent. It aims to capture and record the latest data information in a timely manner so that these new data can be quickly used to respond in subsequent instant query scenarios. The temporary index can adopt a form similar to a memory index, such as a memory-based hash table structure, etc., which can achieve very fast data insertion and preliminary retrieval operations, which is particularly suitable for coping with the continuous influx of real-time new traffic data in the Internet of Vehicles.

[0101] For example, the Internet of Vehicles may receive new road condition information reported by different vehicle sensors every second. For example, if a vehicle detects that the speed of the vehicle on the road section it is on has suddenly dropped, the traffic data containing this information will be generated into an embedded vector and immediately added to the temporary index. This allows other vehicles to quickly find relevant information in the temporary index and respond to it at almost the same time, thereby improving real-time response efficiency.

[0102] The preset time interval needs to be set based on a variety of factors. On the one hand, the timeliness and stability of the data must be considered. For example, for road condition information, real-time changes within a few minutes may be placed in a temporary index first. However, if there is no update after a period of time (such as 5 minutes), it can be considered relatively stable and suitable for migration to a long-term index for subsequent longer-term query references. On the other hand, the system's storage resources and query performance balance must be considered. If the temporary index accumulates too much data for a long time, it will affect its advantage of fast retrieval and will also take up a lot of memory resources. Therefore, some relatively stable data needs to be transferred out regularly.

[0103] In one example, suppose a car is driving, and the onboard sensors continuously collect information such as the vehicle's speed, location, and surrounding environment, and generate embedded vectors in real time. These vectors will first be inserted into a temporary index based on a hash table. This temporary index is used to quickly retrieve the most recently generated vectors to ensure that the traffic information system can respond to the current traffic conditions in a timely manner. At regular intervals (for example, every 5 minutes), the system will transfer the vectors in the temporary index in batches to the long-term index, which is used to store historical traffic data for long-term trend analysis and pattern recognition.

[0104] When building a question-and-answer database, it is necessary to consider the data in both temporary and long-term indexes, because they provide valuable traffic information from different perspectives. The embedding vectors and related metadata (such as data source, generation time, etc.) corresponding to the latest and most real-time traffic data are obtained from the temporary index. This information is critical for answering questions related to the current traffic conditions. The embedding vectors and metadata corresponding to relatively stable data with certain historical reference value are extracted from the long-term index, such as the time periods when a certain section of road often experienced congestion in the past period of time, the traffic patterns of certain roads under specific weather conditions, and other information, which can be used to more comprehensively answer various traffic-related questions.

[0105] The embedded vectors and related metadata collected from the temporary index and long-term index are stored and organized according to the design architecture of the question-and-answer database. For example, in the "Question-and-Answer Data" table of the relational database, in addition to the originally designed embedded vector field, traffic data description field, source field, related road section field and other regular fields, an additional field can be added to identify whether the data comes from the temporary index or the long-term index, so that data from different sources can be flexibly filtered and utilized according to actual needs during subsequent queries.

[0106] In one embodiment of the present invention, the method further includes: recording the storage time of the embedded vector stored in the long-term index; when the storage time of the embedded vector stored in the long-term index exceeds a preset time threshold, deleting the embedded vector exceeding the preset time threshold from the long-term index.

[0107] In the embodiment of the present invention, traffic data is time-sensitive. As time goes by, the reference value of some earlier data for current and subsequent traffic questions and answers will gradually decrease. It may even occupy too many storage resources due to long-term accumulation, affecting system performance. The storage time of the embedded vector in the long-term index is recorded to accurately track the length of time the data exists, so as to determine whether it needs to be cleaned up based on the set time threshold in the future, ensuring that the data retained in the long-term index is relatively valuable and meets the system resource management requirements.

[0108] The determination of the preset time threshold needs to be comprehensively considered in combination with the actual application scenarios of traffic data and the system resource status. For some urban roads or highway sections where road conditions change frequently, a shorter time threshold, such as a few hours or a day, may be more appropriate, because old road condition data will soon lose its reference value, and new data is constantly generated and can better reflect the current actual situation. For some relatively stable areas where traffic patterns do not change much, such as remote rural roads, traffic flow and road conditions are relatively stable for a long time, and the time threshold can be appropriately extended, such as setting it to one week or even one month. Retaining historical data for a relatively long time can still provide a certain reference for occasional traffic travel planning and other issues. At the same time, factors such as server storage capacity and database performance should also be considered to avoid excessive backlogs of expired data, resulting in tight storage resources and reduced query efficiency.

[0109] The system will periodically start the checking mechanism, traverse the storage time fields corresponding to all embedded vector records in the long-term index, calculate the time difference between the current time and the storage time, and determine whether it exceeds the preset time threshold. For example, the current time is "2024-11-2415:00:00" (timestamp is 1700830800), and the preset time threshold is 3 hours (converted to seconds is 10800 seconds). If the storage timestamp of an embedded vector record is 1700812200, by calculating that the time difference between the two is greater than the preset time threshold, it is determined that the traffic data corresponding to the embedded vector has expired.

[0110] For embedded vectors that are determined to have exceeded a time threshold and need to be deleted, a deletion operation is performed according to the deletion operation mechanism of the storage system used by the long-term index to accurately remove expired embedded vectors and their associated traffic data from the long-term index. The present invention can effectively manage the life cycle of traffic data embedded vectors in the long-term index, while ensuring that the question and answer database can continue to provide valuable traffic information, optimize the system storage resource utilization and maintain a good performance state, and better serve the vehicle-mounted question and answer system based on the Internet of Vehicles.

[0111] In one embodiment of the present invention, when multiple vehicles report traffic data at the same time, multiple nodes can be used to process the traffic data separately. For example, on the main roads of a city, a large number of vehicles report traffic information at the same time. This information will be divided into multiple small batches (micro-batches), and each batch is processed in parallel by different computing nodes (such as GPUs or servers). Each node is responsible for generating embedding vectors for a part of the vehicles and passing the results to the downstream index module through the stream processing engine. This method can significantly improve the generation speed of the embedded vectors and ensure that the traffic information system can process large amounts of data in real time.

[0112] In one embodiment of the present invention, a question-and-answer database is constructed based on the embedding vectors corresponding to the newly added traffic data in the indexes of multiple nodes, including: identifying the embedding vectors corresponding to the newly added traffic data to obtain keywords and identified entities; generating candidate answers based on the keywords and identified entities; and storing the candidate answers in the question-and-answer database based on the correspondence between the keywords and the candidate answers.

[0113] In an embodiment of the present invention, in the traffic data scenario of the Internet of Vehicles, the newly added traffic data often contains rich and complex information. The corresponding embedded vectors are identified to obtain keywords and identify entities, so as to accurately generate candidate answers that match traffic-related questions in the future, and it is also convenient to more efficiently organize and retrieve this information in the question and answer database. Keywords are usually words that can reflect the key characteristics of traffic conditions, such as "congestion", "accident", "construction", "speed limit", etc.; identification entities are more directed to specific traffic-related objects, such as road names (such as "XX Avenue"), places (such as "XX intersection"), vehicle types (such as "large trucks"), etc. By extracting these contents, the traffic data can be presented in a form that is easier to understand and process, which is convenient for the subsequent question and answer logic construction.

[0114] For traffic data in text form, we can use lexical analysis tools in natural language processing. For example, we can use part-of-speech tagging tools to determine the part of speech of each word, mark "congestion" as an adjective, and "XX intersection" as a noun phrase, etc., so as to screen out nouns, verbs, adjectives and other words that have a substantial description of traffic conditions as keywords. At the same time, we can use named entity recognition (NER) technology to identify entities. Common NER models (such as the model architecture of BERT+NER annotation layer based on deep learning) can accurately identify entity information such as roads, places, and vehicles from text, and extract them as identified entities.

[0115] For traffic data that is not in text form but has been converted into embedded vectors through the previous steps, the corresponding keywords and identification entities can be reversely derived by analyzing the characteristics of the embedded vectors in the vector space. For example, by training a vector-keyword mapping model with a large amount of labeled data in advance, it can be determined that the vector dimension characteristics of a specific area are often associated with a certain road section name. When encountering new traffic data with similar vector characteristics, the corresponding road section name can be mapped as the identification entity based on this. Similarly, by analyzing the association between the values ​​of each dimension of the vector and the traffic condition description words, keywords such as "congestion" that can reflect the characteristics of the traffic status can be extracted.

[0116] Based on the extracted keywords and recognized entities, combined with the information contained in the traffic data itself and some preset traffic rules and knowledge, candidate answers are generated. For example, if the keyword includes "congestion" and the recognized entity includes "XX intersection", then the candidate answer can be "XX intersection is currently congested, it is recommended to plan a detour route in advance", or "XX intersection is congested, the expected travel time will be extended, please wait patiently" and other descriptive statements related to the traffic conditions and locations, which are used to answer questions that vehicle drivers and passengers may raise about the road conditions at this intersection. For example, if the keyword is "construction" and the recognized entity includes "XX section", the candidate answer can be "XX section is under construction, half of the road is closed, please slow down" and other content that informs of the construction situation and corresponding precautions.

[0117] When the database is a relational database (such as MySQL), you can create a "question and answer information table" with "keyword" fields, "candidate answer" fields and other auxiliary fields (such as data source, generation time, etc.) set in the table. For example, a record may have a "keyword" field storing "XX intersection, congested" and a "candidate answer" field storing "XX intersection is currently congested, it is recommended to plan a detour route in advance". Through such a table structure, the correspondence between many keywords and candidate answers is recorded one by one, making it easy to quickly find matching candidate answer content based on the input keywords in subsequent queries.

[0118] In one embodiment of the present invention, the embedding vector corresponding to the newly added traffic data is added to the index of the node to which it belongs, including: determining the similarity between the embedding vector corresponding to the newly added traffic data and the embedded vectors stored in multiple nodes respectively; if the similarity between the embedded vector corresponding to the newly added traffic data and the embedded vector stored in the target node is greater than a preset similarity threshold, then adding the embedded vector corresponding to the newly added traffic data to the index to which the target node belongs.

[0119] In an embodiment of the present invention, in a distributed storage environment of traffic data in an Internet of Vehicles, by calculating the similarity between the embedding vector corresponding to the newly added traffic data and the embedded vector stored in each node, it is possible to determine the degree of association between the new data and the existing data in terms of semantics, features, etc., and then reasonably determine to which node the embedding vector corresponding to the new data is added to the index, which helps to improve the logic of data storage and the efficiency of subsequent query retrieval.

[0120] Cosine similarity is a commonly used and effective method. It measures the cosine value of the angle between two vectors in the vector space. The value range is between [-1, 1]. The closer the value is to 1, the more similar the two vectors are. and an embedding vector stored in the node The calculation formula is: in is the dot product of the vectors, and are the modulus lengths of the two vectors respectively. For example, if the cosine similarity of the two traffic data embedding vectors is calculated to be 0.8, it means that they have high similarity in feature representation.

[0121] The setting of the preset similarity threshold needs to take into account many factors. On the one hand, the diversity and correlation characteristics of the traffic data itself should be considered. If the threshold is set too low, a large amount of data with weak correlation may be added to the same node index, making the data messy during subsequent queries and difficult to obtain accurate and useful information; and if the threshold is set too high, some valuable related data may not be added to the appropriate node index, resulting in unreasonable data dispersion and affecting query efficiency. For example, for road condition data, if similar road condition information (such as the same light congestion state but different sections) is to be classified and stored as much as possible for query comparison, the similarity threshold can be set to about 0.6; but if you want to classify the data more finely and only classify the data of almost the same section and similar congestion level into one category, the threshold may need to be increased to above 0.8.

[0122] After completing the similarity calculation with the embedded vector stored in each node, the target node whose similarity is greater than the preset similarity threshold is selected. For example, the similarity between the newly added traffic data embedding vector calculated above and an embedded vector in node A is 0.75, the preset similarity threshold is 0.7, and this similarity value is the largest in the comparison with the three nodes, then node A is the target node.

[0123] Once the target node is determined, the embedded vector corresponding to the newly added traffic data is added according to the index structure and addition rules of the node, and then its storage location in the hash index is determined based on the unique hash value of the data, so that the embedded vector is completely added to the index of the target node, which facilitates the subsequent rapid retrieval and utilization of the newly added data in the node index.

[0124] It should be noted that when a car encounters traffic congestion at an intersection, the system will generate a new embedding vector to represent the current traffic conditions. This vector will be added to the index through local insertion, and the system will automatically find similar "neighbor" vectors, such as the traffic conditions of other vehicles at the same location during the same time period. In this way, the system can quickly update traffic condition information without rebuilding the entire index. This helps improve the real-time responsiveness and efficiency of the system.

[0125] In one embodiment of the present invention, a target answer is queried in a question and answer database based on a query request, including: parsing the query request to obtain question features; querying the question and answer database for candidate answers associated with the question features, and determining the similarity between the question features and the associated candidate answers; and determining the target answer among the associated candidate answers based on the similarity between the question features and the associated candidate answers.

[0126] In the embodiment of the present invention, a lexical analysis tool can be used to perform word segmentation on the query request sentence, and the sentence can be split into individual words. For example, for the query request "Is the XX intersection ahead congested now?", after word segmentation, the words "ahead", "XX intersection", "now", "congested", "?", etc. are obtained. Then, the part of speech of each word is determined by part of speech tagging, such as "XX intersection" is marked as a noun phrase (indicating a specific location), "congested" is marked as an adjective (describing traffic conditions), etc., so as to screen out the words corresponding to the part of speech that have a substantial descriptive effect on the problem as part of the problem feature.

[0127] Through syntactic analysis, the grammatical structure tree of the query request is constructed, and the grammatical relationship between words is analyzed to assist in understanding the overall semantics of the sentence. For example, it is determined that "congestion" is used to describe the state of "XX intersection" and "ahead" limits the relative position of "XX intersection". It is further clarified that the core intention of the question is to inquire about the current traffic congestion situation at the specific intersection ahead. At the same time, combined with technologies such as semantic role labeling, the semantic roles such as subject and action in the sentence are determined to more deeply grasp the semantics and intention of the query request and extract more comprehensive and accurate problem features, such as a comprehensive feature description such as "the current traffic condition (congested or not) at the specific location ahead (XX intersection)".

[0128] According to the question features obtained by analysis, matching is performed in the question and answer database to find relevant candidate answers. If the question and answer database is stored in a relational database (such as MySQL), and key information such as keywords has been stored in corresponding fields when the database is designed (such as the "keyword" field in the "question and answer information table" mentioned above), SQL query statements can be used to use operators such as LIKE keywords (for fuzzy matching) or = (for exact matching, depending on the specific requirements of the question features) to find records containing words related to the question features in the "keyword" field, thereby obtaining the corresponding candidate answers, and then calculating the similarity between the question features and the associated candidate answers through the similarity calculation formula.

[0129] Usually a similarity threshold is set. When the similarity between a candidate answer and the question feature is greater than the threshold, it is regarded as an answer that preliminarily meets the requirements. Then, from these candidate answers that meet the requirements, one or more with the highest similarity are selected as the target answer. For example, after calculation and screening, there are three candidate answers with similarities to the question features of 0.85, 0.72, and 0.68, respectively. The threshold is set to 0.7, then the first two candidate answers meet the requirements. After further comparison, the candidate answer with a similarity of 0.85 is selected as the target answer.

[0130] In an embodiment of the present invention, it also includes: outputting alarm information when the acquired new traffic data contains preset keywords.

[0131] In an embodiment of the present invention, during the information extraction stage, keywords can be identified through a keyword recognition component, which can identify safety-related keywords or phrases, such as "accident", "congestion", "emergency stop", etc., and then different thresholds are set according to the importance or urgency of the keywords. The early warning mechanism will only be triggered when the keywords reach a certain threshold.

[0132] For example, for keywords such as "bridge collapse" and "multiple car rear-end collisions with casualties" that have a significant impact on traffic safety and traffic flow and cover a wide range, a lower threshold can be set, such as 0.8 (assuming a threshold measurement scale between 0 and 1, the closer the value is to 1, the easier it is to trigger an early warning). This means that as long as such keywords are detected in new traffic data, the threshold can be easily reached, thereby quickly triggering the early warning mechanism and promptly notifying relevant personnel to take countermeasures, because these situations need to be dealt with in a race against time to avoid more serious consequences.

[0133] For example, keywords corresponding to relatively small impact situations such as "fallen trees on the roadside did not affect traffic" and "traffic signs on a certain road section are slightly blurred" can be set with a higher threshold, such as 0.9. Only when all aspects of the situation combined really indicate that the situation needs attention, such as similar problems occurring multiple times at the same location, will the warning be triggered. This prevents too much unimportant information from triggering the warning mechanism, allowing the recipient to focus on more critical traffic warning information.

[0134] The warning information can include key traffic warning content. For example, when the keyword "water accumulation" is detected, the warning message can be "There is water accumulation on the road in section XX, please slow down and drive safely", clearly indicating the affected section and corresponding recommended measures; if the keyword "rear-end collision" is detected, the warning message can be "A rear-end collision occurred at the XX intersection, please slow down and avoid in advance, and obey the instructions of the traffic police on the scene", informing the location of the accident and response tips, so that the driver and passengers can make reasonable responses in time.

[0135] In one embodiment of the present invention, the server can also identify contextual information based on the user's query request. For example, in a C-V2X system, if the user first asks "Is there a traffic jam ahead?" and gets an affirmative answer, and then the user asks "Are there any detour suggestions?" LLM will provide a detour route suggestion based on the contextual information of the traffic jam ahead.

[0136] The server can also infer the user's actual needs from the user's vague or incomplete questions. For example, in the C-V2X system, if the user urgently asks "Hurry, I need help!" LLM will infer that the user may need emergency rescue based on the vehicle status (such as vehicle failure, accident, etc.) and the user's location information, and quickly provide instructions to contact emergency services.

[0137] The server can also fuse information collected from different C-V2X devices and sensors to provide users with comprehensive information services. For example, when a user asks, "What potential dangers are there around my vehicle?" LLM will integrate information from multiple sources such as other vehicles, traffic signals, roadside sensors, etc. to tell the user whether there are pedestrians, construction areas or other obstacles nearby.

[0138] like Figure 2 , shows a flow chart of a vehicle-mounted question-and-answer method based on the Internet of Vehicles provided by an embodiment of the present invention. New traffic data, such as information of nearby vehicles, traffic signals and road condition data, can be received through a data source access module. Then, a key-value storage system in a memory is used to quickly store and retrieve key-value pair data through a data stream processing module. Irrelevant data is filtered according to predefined rules, and relevant data is aggregated to generate an embedded vector of the new traffic data. A mapping relationship between key-value pairs and embedded vectors is further established through an embedded vector management module to ensure that the corresponding key-value pairs can be quickly found and matched each time a new embedded vector is generated. Then, the embedded vector is stored in a temporary index and a long-term index of a vector database. A question-and-answer database is generated according to the embedded vectors in the vector database through a knowledge base construction module. The early warning module can identify safety-related keywords or phrases that meet preset conditions during the knowledge base construction process and give an alarm. The user's question is received through an intelligent answer module to determine whether it is a routine question. If so, the user is directly answered. If not, the keyword is retrieved and the user is answered in combination with the context. Then, it is determined whether to end, that is, whether to continue the question-and-answer session. If not, the data is updated according to the real-time traffic data obtained, so that the data in the system is kept up to date.

[0139] The present invention discloses an in-vehicle question-and-answer method based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the corresponding node. This method enables the data to be located and retrieved more quickly and accurately after being stored. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the corresponding node to the vehicle, so that the user can adjust the driving route and plan the itinerary in time through the obtained answer, thereby improving the convenience and efficiency of travel.

[0140] like Figure 3 , shows a flowchart of a vehicle-based question-and-answer method provided by an embodiment of the present invention, which is applied to a vehicle. The method may include the following steps:

[0141] Step 201, collecting new traffic data around the vehicle and uploading the new traffic data to the Internet of Vehicles;

[0142] Step 202, obtaining a query request input by a user;

[0143] Step 203, sending a query request to the server so that the server searches for a target answer in a question-and-answer database; the question-and-answer database is obtained by the server from the Internet of Vehicles, storing the newly added traffic data distributedly in multiple nodes, and generating an embedding vector corresponding to the newly added traffic data, which is constructed according to the embedding vector corresponding to the newly added traffic data;

[0144] Step 204, receiving the target answer sent by the server.

[0145] The present invention discloses an in-vehicle question-and-answer method based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the corresponding node. This method enables the data to be located and retrieved more quickly and accurately after being stored. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the corresponding node to the vehicle, so that the user can adjust the driving route and plan the itinerary in time through the obtained answer, thereby improving the convenience and efficiency of travel.

[0146] It should be noted that, for the sake of simplicity, the method embodiments are described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0147] Reference Figure 4 , shows a structural block diagram of a vehicle-mounted question-answering system 30 based on the Internet of Vehicles provided by an embodiment of the present invention, and the system may include a server 301 and a vehicle 302;

[0148] Vehicle 302 is used to collect new traffic data around the vehicle, upload the new traffic data to the Internet of Vehicles, obtain a query request input by a user, send the query request to a server, and receive a target answer sent by the server;

[0149] Server 301 is used to distribute and store the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data; add the embedding vector corresponding to the newly added traffic data to the index of the node to which it belongs; build a question and answer database based on the embedding vectors corresponding to the newly added traffic data in the indexes of multiple nodes; when receiving a query request from a vehicle, query the question and answer database for a target answer based on the query request, and send the target answer from the node to which it belongs to the vehicle.

[0150] Reference Figure 5 , shows a structural block diagram of a vehicle-mounted question-and-answer device based on the Internet of Vehicles provided by an embodiment of the present invention, which is applied to a server and may include the following modules:

[0151] The storage module 401 is used to store the newly added traffic data obtained from the Internet of Vehicles in a distributed manner in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data;

[0152] An adding module 402 is used to add the embedding vector corresponding to the newly added traffic data to the index of the corresponding node;

[0153] A construction module 403, configured to construct a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes;

[0154] The query module 404 is used to query the target answer in the question and answer database according to the query request when receiving the query request from the vehicle, and send the target answer from the corresponding node to the vehicle.

[0155] In one embodiment of the present invention, the index includes a temporary index and a long-term index, and the adding module includes:

[0156] A first adding submodule is used to add the embedding vector corresponding to the newly added traffic data to the temporary index of the corresponding node;

[0157] Also includes:

[0158] A migration submodule, used to migrate the embedding vectors corresponding to the newly added data stored in the temporary index to the long-term index after a preset time interval;

[0159] Building blocks, including:

[0160] The first construction submodule is used to construct a question-answering database according to the embedding vectors corresponding to the newly added traffic data in the temporary index and the embedding vectors corresponding to the newly added traffic data in the long-term index.

[0161] In one embodiment of the present invention, it also includes:

[0162] A recording submodule for recording the storage time of the embedded vectors stored in the long-term index;

[0163] The deletion submodule is used to delete the embedded vectors stored in the long-term index when the storage time of the embedded vectors exceeds the preset time threshold from the long-term index.

[0164] In one embodiment of the present invention, a building block comprises:

[0165] The recognition submodule is used to identify the embedding vectors corresponding to the newly added traffic data to obtain keywords and identify entities;

[0166] The generation submodule is used to generate candidate answers based on keywords and recognized entities;

[0167] The second construction submodule is used to store the candidate answers in the question-answering database according to the corresponding relationship between the keywords and the candidate answers.

[0168] In one embodiment of the present invention, adding a module includes:

[0169] A similarity determination module, used to determine the similarity between the embedding vector corresponding to the newly added traffic data and the embedding vectors respectively stored in the multiple nodes;

[0170] The second adding submodule is used to add the embedding vector corresponding to the newly added traffic data to the index to which the target node belongs if the similarity between the embedding vector corresponding to the newly added traffic data and the embedded vector stored in the target node is greater than a preset similarity threshold.

[0171] In one embodiment of the present invention, the query module includes:

[0172] The parsing submodule is used to parse the query request and obtain the question features;

[0173] A query submodule, for querying candidate answers associated with question features in the question-answering database, and determining the similarity between the question features and the associated candidate answers;

[0174] The determination submodule is used to determine the target answer from the associated candidate answers based on the similarity between the question features and the associated candidate answers.

[0175] In an embodiment of the present invention, it also includes: an output module, which is used to output warning information when the acquired new traffic data contains preset keywords.

[0176] The present invention discloses an in-vehicle question-and-answer device based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the corresponding node. This method enables the data to be located and retrieved more quickly and accurately after being stored. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the corresponding node to the vehicle, so that the user can adjust the driving route and plan the itinerary in time according to the obtained answer, thereby improving the convenience and efficiency of travel.

[0177] Reference Figure 6 , shows a structural block diagram of a vehicle-mounted question-answering device based on the Internet of Vehicles provided by an embodiment of the present invention, which is applied to a vehicle, and the device includes:

[0178] A collection module 501 is used to collect new traffic data around the vehicle and upload the new traffic data to the Internet of Vehicles;

[0179] The acquisition module 502 is used to acquire the query request input by the user;

[0180] The sending module 503 is used to send a query request to the server so that the server can query the target answer in the question-answer database; the question-answer database is obtained by the server from the Internet of Vehicles, and the new traffic data is distributed and stored in multiple nodes, and the embedding vector corresponding to the new traffic data is generated, and the embedding vector corresponding to the new traffic data is constructed;

[0181] The receiving module 504 is used to receive the target answer sent by the server.

[0182] The present invention discloses an in-vehicle question-and-answer device based on the Internet of Vehicles. The present invention generates corresponding embedded vectors for newly added traffic data and adds them to the index of the corresponding node. This method enables the data to be located and retrieved more quickly and accurately after being stored. The present invention forms an ordered structure in the index. When specific information needs to be found, the location of the relevant data can be quickly located based on the similarity between vectors and other characteristics, thereby greatly improving the query efficiency. When the present invention receives a query request from a vehicle, it can quickly query the target answer in the question-and-answer database with the help of the index structure constructed based on the embedded vector, and send it from the corresponding node to the vehicle, so that the user can adjust the driving route and plan the itinerary in time according to the obtained answer, thereby improving the convenience and efficiency of travel.

[0183] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0184] An embodiment of the present invention further provides an electronic device, including:

[0185] It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned in-vehicle question and answer method embodiment based on the Internet of Vehicles are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0186] An embodiment of the present invention also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned embodiment of the in-vehicle question and answer method based on the Internet of Vehicles are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0187] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0188] It will be appreciated by those skilled in the art that the embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the embodiments of the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0189] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0190] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0191] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0192] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0193] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "including one..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0194] The above is a detailed introduction to the in-vehicle question and answer method, device, system, equipment and storage medium based on the Internet of Vehicles provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A vehicle-mounted question-answering method based on Internet of Vehicles, characterized in that: Applied to a server, the method comprises: Distribute and store the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data; Adding the embedding vector corresponding to the newly added traffic data to the index of the corresponding node; Constructing a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes; When a query request from a vehicle is received, a target answer is searched for in the question-and-answer database according to the query request, and the target answer is sent from the node to which it belongs to the vehicle.

2. The method according to claim 1, characterized in that The index includes a temporary index and a long-term index, and the adding of the embedding vector corresponding to the newly added traffic data to the index of the corresponding node includes: Adding the embedding vector corresponding to the newly added traffic data to the temporary index of the corresponding node; Also includes: After a preset time interval, the embedding vector corresponding to the newly added data stored in the temporary index is migrated to the long-term index; The step of constructing a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes includes: The question-and-answer database is constructed according to the embedding vector corresponding to the newly added traffic data in the temporary index and the embedding vector corresponding to the newly added traffic data in the long-term index.

3. The method according to claim 2, characterized in that The method further comprises: Recording the storage time of the embedded vector stored in the long-term index; When the storage time of the embedded vectors stored in the long-term index exceeds a preset time threshold, the embedded vectors exceeding the preset time threshold are deleted from the long-term index.

4. The method according to claim 1, characterized in that: The step of constructing a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes includes: Identify the embedding vectors corresponding to the newly added traffic data to obtain keywords and identification entities; Generate candidate answers based on the keywords and the identified entities; According to the correspondence between the keywords and the candidate answers, the candidate answers are stored in the question and answer database.

5. The method according to claim 1, characterized in that: The step of adding the embedding vector corresponding to the newly added traffic data to the index of the corresponding node includes: Determine the similarity between the embedding vector corresponding to the newly added traffic data and the embedding vectors respectively stored in the plurality of nodes; If the similarity between the embedding vector corresponding to the newly added traffic data and the embedded vector stored in the target node is greater than a preset similarity threshold, the embedding vector corresponding to the newly added traffic data is added to the index to which the target node belongs.

6. The method according to claim 1, characterized in that The step of searching the question-answer database for a target answer according to the query request includes: Parsing the query request to obtain question features; Querying the question-answer database for candidate answers associated with the question feature, and determining the similarity between the question feature and the associated candidate answers; According to the similarity between the question feature and the associated candidate answers, a target answer is determined from the associated candidate answers.

7. The method according to claim 1, characterized in that Also includes: When the acquired new traffic data contains preset keywords, an alarm message is output.

8. A vehicle-mounted question-answering method based on Internet of Vehicles, characterized in that: Applied to a vehicle, characterized in that the method comprises: Collecting new traffic data around the vehicle, and uploading the new traffic data to the Internet of Vehicles; Get the query request entered by the user; Sending the query request to the server so that the server searches for the target answer in the question-and-answer database; the question-and-answer database is obtained by the server distributing and storing the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generating an embedding vector corresponding to the newly added traffic data, which is constructed according to the embedding vector corresponding to the newly added traffic data; The target answer sent by the server is received.

9. An in-vehicle question-answering system based on Internet of Vehicles, characterized in that: The system includes a server and a vehicle; The vehicle is used to collect new traffic data around the vehicle, upload the new traffic data to the Internet of Vehicles, obtain a query request input by a user, send the query request to the server, and receive a target answer sent by the server; The server is used to distribute and store the newly added traffic data obtained from the Internet of Vehicles in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data; and add the embedding vector corresponding to the newly added traffic data to the index of the node to which it belongs; A question-and-answer database is constructed based on the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes; when a query request from a vehicle is received, a target answer is queried in the question-and-answer database based on the query request, and the target answer is sent from the node to which it belongs to the vehicle.

10. A vehicle-mounted question-answering device based on Internet of Vehicles, characterized in that: Applied to a server, the device comprises: A storage module, used to store the newly added traffic data obtained from the Internet of Vehicles in a distributed manner in multiple nodes, and generate an embedding vector corresponding to the newly added traffic data; An adding module, used to add the embedding vector corresponding to the newly added traffic data to the index of the corresponding node; A construction module, configured to construct a question-answer database according to the embedding vectors corresponding to the newly added traffic data in the indexes of the multiple nodes; The query module is used to query the target answer in the question-and-answer database according to the query request when receiving a query request from the vehicle, and send the target answer from the corresponding node to the vehicle.

11. A vehicle-mounted question-answering device based on Internet of Vehicles, characterized in that: Applied to a vehicle, the device is characterized in that it comprises: A collection module, used for collecting new traffic data around the vehicle and uploading the new traffic data to the Internet of Vehicles; An acquisition module is used to obtain a query request input by a user; A sending module, used for sending the query request to the server, so that the server searches for the target answer in the question-and-answer database; the question-and-answer database is obtained by the server from the Internet of Vehicles by distributing and storing the newly added traffic data obtained in multiple nodes, and generating an embedding vector corresponding to the newly added traffic data, which is constructed according to the embedding vector corresponding to the newly added traffic data; A receiving module is used to receive the target answer sent by the server.

12. An electronic device, characterized in that: It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor is used to execute the steps of the in-vehicle question-and-answer method based on the Internet of Vehicles as described in any one of claims 1-7 or 8.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by the processor, the steps of the in-vehicle question-and-answer method based on the Internet of Vehicles as described in any one of claims 1-7 or 8 are implemented.

Images