Troubleshooting method and device and electronic equipment
By acquiring the description text of home appliance malfunctions, matching it with troubleshooting text, and using a generative language model for fault diagnosis, the problem of slow home appliance fault diagnosis and waste of human resources has been solved, achieving efficient automated fault diagnosis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-09
- Publication Date
- 2026-04-10
AI Technical Summary
Current technologies for troubleshooting home appliance malfunctions are slow and waste a lot of human resources. In particular, for malfunctions that are easy to repair, after-sales personnel need to make judgments by consulting repair manuals or relying on experience, which leads to low efficiency.
By acquiring the fault description text of the target device, identifying the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, generating prompt information, and using a generative language model to conduct fault investigation, personalized fault investigation guidance is provided.
It improves the speed of troubleshooting, reduces the time spent manually reviewing documents, realizes automated troubleshooting, saves human resources, and reduces the experience and knowledge limitations of human judgment.
Smart Images
Figure CN121836658A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of computers, and more particularly, to a troubleshooting method and device and electronic equipment. BACKGROUND
[0002] Currently, when a household appliance malfunctions, the user generally reports the problem to apply for maintenance through a 400 after-sales telephone. For possible fault judgment, especially for easily repairable faults, the after-sales personnel usually need to judge by querying the relevant maintenance manual or according to their own experience, or even need to go on-site to troubleshoot, therefore, there is a problem of slow troubleshooting speed and waste of a large amount of human resources in the related art. SUMMARY
[0003] In view of the above problems, the present application provides a troubleshooting method, device and electronic equipment, which can improve the troubleshooting speed while effectively saving human resources.
[0004] In a first aspect, an embodiment of the present application provides a troubleshooting method, which comprises: obtaining a fault description text of a target device; determining a target troubleshooting text matched with the fault description text from a plurality of troubleshooting texts, each troubleshooting text corresponding to a fault type; generating prompt information for indicating fault-related information corresponding to the fault description text from the target troubleshooting text based on the target troubleshooting text and the fault description text; and performing troubleshooting based on the prompt information by using a generative language model to obtain a troubleshooting result of the target device.
[0005] In a second aspect, an embodiment of the present application provides a troubleshooting device, which comprises: a text obtaining module configured to obtain a fault description text of a target device; a text determining module configured to determine a target troubleshooting text matched with the fault description text from a plurality of troubleshooting texts, each troubleshooting text corresponding to a fault type; a prompt information generating module configured to generate prompt information for indicating fault-related information corresponding to the fault description text from the target troubleshooting text based on the target troubleshooting text and the fault description text; and a troubleshooting module configured to perform troubleshooting based on the prompt information by using a generative language model to obtain a troubleshooting result of the target device.
[0006] In a third aspect, an embodiment of the present application provides an electronic device, comprising one or more processors; a memory, one or more application programs, wherein the one or more application programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs are configured to execute the above method.
[0007] Fourthly, embodiments of this application provide a computer-readable storage medium storing program code, which can be invoked by a processor to execute the above-described method.
[0008] The fault diagnosis method, apparatus, and electronic device provided in this application embodiment acquire a fault description text of a target device; determine a target troubleshooting text that matches the fault description text from a plurality of troubleshooting texts, with each troubleshooting text corresponding to a fault type; generate prompt information based on the target troubleshooting text and the fault description text to indicate that fault-related information corresponding to the fault description text can be found from the target troubleshooting text; and perform fault diagnosis using a generative language model based on the prompt information to obtain the fault diagnosis result of the target device. Using the above method, fault diagnosis can be performed by combining the acquired fault description text with existing troubleshooting texts using an electronic device, thereby effectively improving fault diagnosis efficiency and greatly saving human resources. Attached Figure Description
[0009] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0010] Figure 1 A flowchart illustrating a troubleshooting method provided in an embodiment of this application is shown;
[0011] Figure 2 It shows Figure 1 A flowchart illustrating step S120;
[0012] Figure 3 This paper illustrates another flowchart of a troubleshooting method provided in an embodiment of the present application;
[0013] Figure 4 A timing diagram of a troubleshooting method provided in an embodiment of this application is shown;
[0014] Figure 5 This illustration shows another flowchart of a troubleshooting method provided in an embodiment of this application;
[0015] Figure 6 This paper shows a module block diagram of a fault diagnosis device provided in an embodiment of the present application;
[0016] Figure 7 A block diagram of an electronic device for performing a troubleshooting method according to an embodiment of this application is shown.
[0017] Figure 8 An embodiment of this application shows a storage unit for storing or carrying program code that implements the troubleshooting method according to an embodiment of this application. Detailed Implementation
[0018] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.
[0019] With the development of science and technology, various devices have been widely used, and consequently, equipment failures have also occurred. In related technologies, the common approach is to call the device's after-sales service hotline to report the fault details so that after-sales personnel can determine the cause of the fault and provide a repair method based on the reported details, technical documents, or their own experience. Alternatively, after-sales personnel can directly visit the site to troubleshoot the problem. Clearly, these troubleshooting methods suffer from low efficiency.
[0020] To address the aforementioned problems, the inventors, through long-term research, discovered and proposed the troubleshooting method, apparatus, and electronic device provided in this application. After obtaining the description text of the target device, by identifying the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, the time spent manually reviewing a large number of documents can be effectively reduced, improving the speed of troubleshooting. Furthermore, by generating prompt information based on the target troubleshooting text and the fault description text to indicate how to find fault-related information corresponding to the fault description text from the target troubleshooting text, personalized troubleshooting guidance can be provided to utilize generative language models for troubleshooting. The automated troubleshooting process throughout the entire troubleshooting process can complete the troubleshooting more efficiently than related technologies, while avoiding the possibility of errors caused by human judgment being easily limited by experience and knowledge.
[0021] Please see Figure 1 , Figure 1 This is a flowchart illustrating a troubleshooting method provided in one embodiment of this application. The method can be applied to electronic devices, which may be target devices, terminal devices associated with the target devices, or servers.
[0022] In some embodiments, the server can be a standalone physical server, a server cluster or distributed system consisting of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms.
[0023] Terminal devices can be smartphones, tablets, laptops, desktop computers, intelligent voice interaction devices, in-vehicle terminals, etc., but are not limited to these.
[0024] The target equipment can be household appliances (such as refrigerators, air conditioners, televisions, or robot vacuum cleaners), medical equipment (such as MRI equipment, ultrasound equipment, etc.), industrial equipment, or transportation equipment, etc., without specific limitations here.
[0025] The troubleshooting method may specifically include the following steps S110 to S140.
[0026] Step S110: Obtain the fault description text of the target device.
[0027] The fault description text of the target device refers to the specific description of the device fault provided by the user or collected by the target device when the fault occurs. This fault description text may include the fault symptoms exhibited by the target device, and may also include one or more of the following: the time of the fault occurrence, the frequency of the fault occurrence, the actions taken by the target device before the fault occurred, the operating environment at the time of the fault, and accompanying error messages.
[0028] The malfunction symptom of the target device indicates the specific problem occurring on the target device, such as "computer cannot start," "phone screen does not light up," or "washing machine does not drain." The time of malfunction occurrence indicates the time when the malfunction first appeared or the most recent occurrence. The frequency of malfunction occurrence indicates whether the malfunction is a one-off occurrence or a frequent occurrence. The actions performed at the time of malfunction indicate what actions the user performed before the malfunction occurred, such as "computer cannot start after installing new software" or "washing machine suddenly stops working while using it." The accompanying error message indicates the error code or warning message displayed when the device malfunctions, such as "blue screen error code 0x0000007B is displayed on the computer screen." The operating environment at the time of malfunction refers to the environmental information of the target device at the time of the malfunction, such as temperature and humidity.
[0029] The aforementioned methods for obtaining the description text of the target device can be: receiving fault description text for the target device uploaded by the terminal device; or obtaining fault description text generated by the fault detection device of the target device when it detects an anomaly in the target device. In another method, when the target device has a built-in diagnostic program, fault description information generated when the target device detects an anomaly during startup or periodic operation is received.
[0030] For example, taking a mobile phone as the target device, a user could input a fault description text on the terminal device as: "My phone was fully charged last night, but this morning I found it wouldn't turn on. The screen is completely dark, there's no vibration or sound feedback, and a forced restart has no effect." Taking a washing machine as another example, a user could input a fault description text on the terminal device as: "During the washing process, the washing machine cannot complete the draining operation. The drain hose is not blocked, and the washing machine door closes normally."
[0031] Step S120: Determine the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, with each troubleshooting text corresponding to a fault type.
[0032] Among them, multiple troubleshooting texts refer to documents corresponding to each type of fault, which are organized according to the equipment type of the target equipment, including common fault phenomena, fault descriptions, fault causes, troubleshooting procedures, and instructions on whether the fault can be resolved by the user. The document format can be Word, PDF, PPT, or plain text files such as TXT.
[0033] For example, taking the washing machine's inability to drain as an example, the troubleshooting text for this fault specifically includes: Fault phenomenon: During the washing process, the washing machine cannot complete the draining operation, resulting in clothes being soaked in water; Fault description: During the washing process, the washing machine cannot drain normally after the wash cycle is completed, resulting in excessively high water levels and affecting the spin-drying effect. Fault causes: 1. Clogged drain pipe, 2. Drain pump failure, 3. Control circuit board problem, 4. Door lock failure; Troubleshooting process: 1. Check the drain pipe: Check if the drain pipe is bent or blocked, and clear any blockages; 2. Check if the door lock is closed; 3. If the above checks are ineffective, professional repair is required; Troubleshooting method: Users can check the drain pipe and door lock themselves, but problems involving the circuit board require professional handling.
[0034] In one possible implementation, the method for determining the target troubleshooting text that matches the fault description text among multiple troubleshooting texts can be to calculate the relevance between the multiple troubleshooting texts and the fault description text, and determine the K troubleshooting texts with the highest relevance to the fault description text among the multiple troubleshooting texts as the target troubleshooting texts that match the fault description text.
[0035] In the above implementation method, the relevance calculation of multiple troubleshooting texts to the fault description text can be performed as follows: extract keywords from the fault description text and multiple troubleshooting texts respectively to obtain the keywords of the fault description text and the keywords of each troubleshooting text; calculate the overlap between the keywords of the fault description text and the keywords of each troubleshooting text; determine the relevance between the keywords of the fault description text and the keywords of each troubleshooting text based on the overlap between the keywords of the fault description text and the keywords of each troubleshooting text, wherein the overlap is positively correlated with the relevance.
[0036] The method for calculating the overlap between the keywords in the fault description text and the keywords in the troubleshooting text is as follows: determine the number of first keywords in the union of the keywords in the fault description text and the keywords in the troubleshooting text, and the number of second keywords in the intersection of the keywords in the fault description text and the keywords in the troubleshooting text. The ratio between the number of second keywords and the number of first keywords is used to calculate the overlap between the keywords in the fault description text and the keywords in the troubleshooting text.
[0037] One method for calculating the relevance between multiple troubleshooting texts and the fault description text is as follows: Extract the vector of the fault description text, calculate the similarity between the fault description text vector and the text vectors of each troubleshooting text, and determine the relevance between the fault description text and each troubleshooting text based on the similarity. Here, similarity and relevance are positively correlated. It is worth noting that the vectors of the fault description text and the text vectors of each troubleshooting text have the same length and are both preset lengths, and the vector extraction methods for the fault description text and the text vectors of each troubleshooting text are the same.
[0038] For example, the text vector of the troubleshooting text can be extracted by encoding the troubleshooting text using an encoding network (e.g., a bidirectional encoder based on a transformer or an encoding network based on an open artificial intelligence model) to obtain a text vector of a preset length. Alternatively, the troubleshooting text can be segmented to obtain a vocabulary list; each word in the vocabulary list can be embedded to obtain a word vector; and the word vectors of each word in the vocabulary list can be aggregated to obtain a text vector of a preset length.
[0039] One approach is to use word vector models, such as the Word2Vec word vector model or the GloVe model (Global Vectors for Word Representation), to embed each word in the vocabulary list and obtain the word vector of each word. When aggregating the word vectors of each word in the vocabulary list, simple averaging, weighted averaging, or other aggregation methods can be used to merge the word vectors of all words into a single vector.
[0040] Another method for calculating the relevance between multiple troubleshooting texts and the fault description text is as follows: The fault description text is segmented to obtain a vocabulary list; for each troubleshooting text, based on the word frequency and inverse document frequency (IVF) of a word in the vocabulary list corresponding to the index of the troubleshooting text, a relevance value is obtained between that word and the index of the troubleshooting text, where the index of the troubleshooting text includes the word frequency and IVF of multiple words in the troubleshooting text; the relevance values between each word in the vocabulary list of the fault description text and the index of the troubleshooting text are summed to obtain a first relevance score between the fault description text and the index of the troubleshooting text.
[0041] The method for obtaining the relevance value between a word and the index of the troubleshooting text based on the word frequency and inverse document frequency of the words in the vocabulary list of the fault description text corresponding to the index of the troubleshooting text can be as follows: multiply the word frequency and inverse document frequency of the words in the vocabulary list of the fault description text corresponding to the index of the troubleshooting text to obtain the relevance value between the word and the index of the troubleshooting text.
[0042] The above-described method for calculating relevance is merely illustrative. Multiple relevance calculation methods can be combined. For example, the relevances calculated by the two methods can be weighted and summed to obtain the final relevance calculation result. The target troubleshooting text that matches the fault description text can then be selected based on the final relevance calculation result. Alternatively, one or more fault matching texts with the highest relevance can be selected using each calculation method, and the selected troubleshooting text can be used as the target troubleshooting text.
[0043] In another possible implementation, a text matching prediction model can be used to predict the target troubleshooting text that matches the fault description text among multiple troubleshooting texts. This text matching prediction model is trained based on training samples, which include sample fault description texts, multiple sample troubleshooting texts, and matching tags between each sample troubleshooting text and the sample fault description text. During training, specific features of each sample fault description text and each sample troubleshooting text can be extracted. Based on the extracted features, a predicted matching result between the sample fault description text and each sample troubleshooting text is obtained. A loss is calculated based on the predicted matching result and the matching tags between the sample fault description text and each sample troubleshooting text to obtain the model loss. The model parameters of the matching prediction model are then adjusted based on the model loss.
[0044] Step S130: Generate a prompt message based on the target troubleshooting text and the fault description text to indicate that fault-related information corresponding to the fault description text can be found from the target troubleshooting text.
[0045] Prompt information, as a form of information-enhanced data, aims to make generative language models understand what tasks need to be performed and what content needs to be output.
[0046] Wherein, if the troubleshooting text includes fault phenomena, fault description, fault cause, troubleshooting process, and troubleshooting method, the fault-related information retrieved from the target troubleshooting text corresponding to the fault description text may include one or more of the fault cause, troubleshooting process, and troubleshooting method. In order to facilitate the subsequent generative language model to output more comprehensive fault investigation results based on the prompt information, in one possible embodiment of this application, the fault-related information corresponding to the fault description text includes the fault cause that caused the fault described in the fault description text and the troubleshooting method for the fault.
[0047] Accordingly, based on the target troubleshooting text and the fault description text, a prompt message is generated to indicate at least one of the fault cause that led to the fault described in the fault description text and the troubleshooting method for the fault, which can be found from the target troubleshooting text; for example, the prompt message can be generated based on the target troubleshooting text and the fault description text to indicate the fault cause that led to the fault described in the fault description text and the troubleshooting method for the fault, which can be found from the target troubleshooting text.
[0048] For example, if the fault description text is "During the washing process, the washing machine cannot complete the drainage operation. The drain pipe is checked and there is no blockage, and the washing machine door is closed normally", after finding the target troubleshooting text corresponding to this fault description text, a prompt message can be generated that includes the target troubleshooting text and the known fault description text "During the washing process, the washing machine cannot complete the drainage operation. The drain pipe is checked and there is no blockage, and the washing machine door is closed normally". Please provide the fault cause that causes the washing machine to fail to drain and the troubleshooting method for this fault in conjunction with the target troubleshooting text.
[0049] Step S140: Use a generative language model to troubleshoot based on the prompt information to obtain the troubleshooting results of the target device.
[0050] Generative language models can include, but are not limited to, GPT (Generative Pre-Trained Transformer), Bloom (a Transformer language model with only a decoder), Llama (Large Language Model Meta AI, an open and efficient large-scale basic language model), and GLM (Generative Language Model Pretraining with Autoregressive Bloom Infilling).
[0051] By employing the method described in this application, after obtaining the description text of the target device, and by identifying the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, the time spent manually reviewing a large number of documents can be effectively reduced, thus improving the speed of fault diagnosis. Furthermore, by generating prompts based on the target troubleshooting text and the fault description text to indicate how to find fault-related information corresponding to the fault description text from the target troubleshooting text, personalized fault diagnosis guidance can be provided to facilitate fault diagnosis using a generative language model. The entire fault diagnosis process can be automated, achieving more efficient fault diagnosis compared to existing technologies, while significantly saving human resources. In addition, it can avoid the possibility of errors caused by human judgment being limited by experience and knowledge.
[0052] Please see Figure 2 As shown, each of the troubleshooting texts corresponds to an index and a text vector of a preset length;
[0053] The above step S120 includes:
[0054] Step S121: Calculate the first correlation score between the fault description text and each of the troubleshooting process texts based on the indexes of each of the troubleshooting process texts and the fault description text.
[0055] Among them, the index of the troubleshooting process text may include the word frequency and inverse document frequency of multiple words in the troubleshooting process text.
[0056] The index of the troubleshooting process text can be pre-constructed, and its construction process may include: segmenting the troubleshooting process text to obtain a vocabulary list of the troubleshooting process text; obtaining the word frequency of each word in the troubleshooting process text according to the vocabulary list of the troubleshooting process text; determining the inverse document frequency of the target word according to the number of texts in the text set composed of the multiple troubleshooting process texts that contain the target word, where the target word is any word in the vocabulary list of the troubleshooting process text; constructing the index of the troubleshooting process text according to the word frequency and inverse document frequency of each word in the troubleshooting process text.
[0057] Among them, when constructing the index of the troubleshooting process text according to the word frequency and inverse document frequency of each word in the troubleshooting process text, words that meet specified conditions can be screened according to the word frequency and inverse document frequency of each word, and the index is constructed based on the selected words. Among them, the index of the troubleshooting process text can be a data structure in the form of a table or dictionary, including the selected words and their corresponding word frequencies and inverse document frequencies. The above-mentioned specified conditions can be at least one of the word frequency being greater than a preset word frequency and the inverse document frequency being greater than a preset frequency threshold.
[0058] The above step S121 can be: performing a segmentation process on the fault description text to obtain a vocabulary list of the fault description text; for each of the troubleshooting process texts, obtaining the relevance value between the word in the vocabulary list of the fault description text and the index of the troubleshooting process text according to the word frequency and inverse document frequency corresponding to the word in the index of the troubleshooting process text, where the index of the troubleshooting process text includes the word frequency and inverse document frequency of multiple words in the troubleshooting process text; accumulating the relevance values between each word in the vocabulary list of the fault description text and the index of the troubleshooting process text to obtain the first correlation score between the fault description text and the index of the troubleshooting process text.
[0059] It is worth mentioning that after the above-mentioned segmentation process on the fault description text and the troubleshooting process text, stop words in the fault description text can also be removed, such as removing "de", "shi", "zai", etc. in the text, so that each word in the finally obtained vocabulary list has practical significance.
[0060] In addition, the index of the pre-extracted troubleshooting text can be stored in the electronic device or the memory or database associated with the electronic device, so that after the fault description text is segmented, the relevance of each word in the vocabulary list of the fault description text and the index of each pre-stored troubleshooting text can be calculated.
[0061] Step S122: Calculate a second correlation score between the fault description text and each of the troubleshooting texts based on the text vector of each troubleshooting text and the fault description text.
[0062] The text vector of the troubleshooting text can be pre-extracted. For example, the troubleshooting text can be encoded using an encoding network to obtain a text vector of a preset length; or the troubleshooting text can be segmented to obtain a vocabulary list; each word in the vocabulary list can be embedded to obtain a word vector; and the word vectors of each word in the vocabulary list can be aggregated to obtain a text vector of a preset length. For details on extracting the text vector of the troubleshooting text, please refer to the specific description of step S120 in the foregoing embodiments, which will not be repeated here.
[0063] Specifically, step S122 above may involve: vectorizing the fault description text to obtain a fault description vector of a preset length; calculating the similarity between the fault description vector and the text vector of each troubleshooting text to obtain the similarity between the fault description text and each troubleshooting text; and determining a second relevance score between the fault description text and each troubleshooting text based on the similarity between the fault description text and each troubleshooting text.
[0064] It is worth mentioning that the text vector of the troubleshooting text is obtained by vectorizing the troubleshooting text in the same way as the vectorization of the fault description text, and the extracted vector lengths are the same. That is, the text vector of the troubleshooting text and the vector of the fault description text have the same length, which is a preset length.
[0065] In addition, the text vectors of the pre-extracted troubleshooting texts can be stored in electronic devices or their associated storage or databases, such as FAISS (Facebook AI Similarity Search) or ElasticSearch, so that after the vectors of the fault description texts are extracted, the similarity between the vectors of the fault description texts and the vectors of the pre-stored troubleshooting texts can be calculated.
[0066] When calculating the similarity between the fault description vector and the text vector of each troubleshooting text, cosine similarity can be used.
[0067] Step S123: The M troubleshooting texts with the highest first relevance scores are identified as target troubleshooting texts that match the fault description text.
[0068] Where M is a positive integer less than or equal to K.
[0069] For example, the value of M can be any value between 1 and 5.
[0070] Step S124: The K troubleshooting texts with the highest second relevance scores are identified as target troubleshooting texts that match the fault description text.
[0071] Where N is a positive integer less than or equal to K, and the union of the K troubleshooting texts and the M troubleshooting texts includes the K troubleshooting texts.
[0072] For example, N can be any value between 1 and 5. It should be understood that among the M troubleshooting texts selected based on the first relevance score and the N troubleshooting texts selected based on the second relevance score, there may be some troubleshooting texts that are the same. The response K is a value less than or equal to M+N.
[0073] By employing steps S121-S124 above, two different relevance calculation methods can be combined to select target troubleshooting texts. The first relevance score, determined by a vocabulary list of fault description texts obtained from the index and word segmentation of each troubleshooting text, focuses on lexical-level matching. The second relevance score, determined by the text vectors of each troubleshooting text and the vector of the fault description text obtained through vectorization, focuses on semantic-level matching. By combining these two methods, it is ensured that the obtained target troubleshooting texts include both texts with high keyword matching degrees with the fault description texts and texts with high semantic matching degrees. This improves the accuracy and reliability of subsequent fault investigation results obtained from the target troubleshooting texts.
[0074] Please see Figure 3 As shown, in one possible implementation, after performing step S140, the method further includes:
[0075] Step S150: Send the troubleshooting results to the user client associated with the target device.
[0076] Among them, the electronic device may be associated with user clients of different devices. While obtaining the fault description text of the target device, the device identifier of the target device or the user client identifier associated with the target device may also be obtained. If the device identifier of the target device is obtained, the user client identifier associated with the device identifier of the target device is searched from the database based on the device identifier of the target device, and the fault investigation result is sent to the user client indicated by the user client identifier associated with the target device.
[0077] By sending the troubleshooting results to the user client associated with the target device, when the troubleshooting method included in the troubleshooting results indicates that the user should troubleshoot the problem himself, the user can troubleshoot the problem according to the troubleshooting method for that problem, so that the target device can quickly return to normal operation.
[0078] It is worth mentioning that if users are unable to complete the troubleshooting process according to the troubleshooting method, they can also send a manual troubleshooting request to the electronic device. After receiving the manual troubleshooting request, the electronic device can also send a fault diagnosis message to the troubleshooting client according to the manual troubleshooting request so that an engineer can come to the site to troubleshoot.
[0079] In one possible implementation, the troubleshooting text includes a fault phenomenon, a fault description, a fault cause, a troubleshooting process, and a troubleshooting method. The step of using a generative language model to perform fault investigation based on the prompt information to obtain the fault investigation result of the target device includes: the fault cause leading to the fault described in the fault description text, and the troubleshooting method for that fault. The method further includes:
[0080] Step S160: If the troubleshooting method indication included in the troubleshooting result is not to be troubleshooted by the user, determine the address information of the target device based on the device identifier of the target device, determine the troubleshooting client based on the address information of the target device, and send the troubleshooting instruction information including the address information and the troubleshooting result to the troubleshooting client.
[0081] The device identifier of the target device can be obtained when acquiring the fault description text of the target device, or it can be uploaded by the target client; no specific limitation is made here.
[0082] It's worth noting that different troubleshooting clients correspond to different engineers, and each client can have a specific troubleshooting address range. By combining the target device's address information with the respective address ranges of multiple clients, the appropriate client can be determined, allowing for the assignment of the nearest engineer. Furthermore, the client can be selected based on the target device's address information, the corresponding address ranges of multiple clients, and the volume of troubleshooting, enabling faster and more efficient on-site engineer dispatch for troubleshooting.
[0083] By employing step S160 above, when a generative language model determines a fault that the user cannot resolve independently, the address information of the target device is determined based on the device identifier. A troubleshooting client is then identified based on the target device's address information, and a fault diagnosis instruction message including the address information and the fault diagnosis result is sent to the troubleshooting client. This sends a dispatch order to the client associated with the engineer performing the troubleshooting (the troubleshooting client), instructing the engineer to perform on-site repairs. Clearly, this automated dispatching method reduces the need for manual intervention, speeds up fault handling, and allows engineers to receive repair tasks and proceed to the site more quickly.
[0084] This application provides a fault troubleshooting method that can utilize electronic equipment, such as a server. Figure 4 As shown, this server can be associated with user clients and troubleshooting clients. Users can upload fault description text of the target device to the server through the user client. Please refer to... Figure 5 As shown, the server can perform the following steps:
[0085] I. Textualization of Fault Content: This involves organizing common fault phenomena, fault descriptions, fault causes, troubleshooting procedures, and instructions on whether users can resolve the fault themselves into a document corresponding to each type of fault (i.e., troubleshooting text). The document format can be Word, PDF, PPT, or a plain text file such as TXT.
[0086] 2. Constructing document indexes / extracting text vectors: Obtain the index of each troubleshooting text using the index construction method described in step S121 of the aforementioned embodiments, and extract the text vector of each troubleshooting text using the vector extraction method described in step S122 of the aforementioned embodiments.
[0087] III. Data Storage: The index of each troubleshooting text is stored in the memory of the electronic device, and the text vector of each troubleshooting text is stored in the vector database.
[0088] IV. Fault Description Text Processing: Receives fault description text of the target device uploaded by the user client. Performs word segmentation on the fault description text to obtain a vocabulary list, and extracts vectors from the fault description text to obtain its vector.
[0089] V. Document Matching: Select M target troubleshooting texts using the methods described in steps S121 and S123 of the aforementioned embodiments, and select N target troubleshooting texts using the methods described in steps S122 and S124 of the aforementioned embodiments, so as to finally obtain K target troubleshooting texts.
[0090] VI. Generate Prompt Messages: Generate prompt messages based on K target troubleshooting texts and fault description texts. For example, the prompt message might be: "The user's refrigerator has malfunctioned, and the fault description text is 'xxx'. Please use the following knowledge to determine the cause of the malfunction and whether the user can resolve it independently. If so, provide specific troubleshooting steps. If you cannot determine the cause, please answer 'I cannot determine'. Knowledge: Troubleshooting text 1, Troubleshooting text 2, ..., Troubleshooting text K."
[0091] 7. Troubleshooting: Utilize a generative language model to output troubleshooting results based on prompts.
[0092] 8. Troubleshooting: The troubleshooting results are sent to the user client associated with the target device. If the troubleshooting method included in the results indicates that the user should troubleshoot independently, the user can use the appropriate method to resolve the fault and quickly restore the target device to normal operation. If the troubleshooting method in the results indicates that the user should not troubleshoot independently, the address information of the target device is determined based on its device identifier. A troubleshooting client is then identified based on the address information, and a troubleshooting instruction message including the address information and the troubleshooting results is sent to the troubleshooting client. This facilitates the assignment of an engineer to perform on-site troubleshooting for the target device.
[0093] The above-described troubleshooting method first transcribes existing fault information into troubleshooting text. Then, it constructs an index for the troubleshooting text and extracts and stores its text vectors for later direct retrieval. For example, upon receiving a fault description text for a target device, the text is vectorized and segmented. A first relevance score is determined based on the index and segmentation of each troubleshooting text to select the target troubleshooting text, focusing on lexical matching. A second relevance score is determined based on the text vectors of each troubleshooting text and the vectorized fault description text to select the target troubleshooting text, focusing on semantic matching. By combining these two methods, the obtained target troubleshooting text is ensured to include both text with high keyword matching and semantic matching. This allows for more accurate and reliable generation of prompts using the obtained target troubleshooting text and fault description text, and the use of a generative language model to obtain troubleshooting results based on these prompts. Furthermore, the automated fault diagnosis process described above can complete fault diagnosis more efficiently than related technologies, while avoiding the possibility of errors caused by human judgment being limited by experience and knowledge.
[0094] Please see Figure 6 , Figure 6 A module block diagram of a fault diagnosis device provided in an embodiment of this application is shown. The following will focus on... Figure 6 The process is described in detail below. The fault diagnosis device 200 is applied to the aforementioned electronic equipment. The fault diagnosis device 200 includes: a text acquisition module 210, a text determination module 220, a prompt message generation module 230, and a fault diagnosis module 240, wherein:
[0095] The text acquisition module 210 is used to acquire the fault description text of the target device; the text determination module 220 is used to determine the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, with each troubleshooting text corresponding to a fault type; the prompt information generation module 230 is used to generate prompt information based on the target troubleshooting text and the fault description text to indicate that fault-related information corresponding to the fault description text can be found from the target troubleshooting text; the fault investigation module 240 is used to perform fault investigation based on the prompt information using a generative language model to obtain the fault investigation result of the target device.
[0096] Furthermore, the troubleshooting text includes a fault phenomenon, a fault description, a fault cause, a troubleshooting process, and a troubleshooting method. The fault-related information corresponding to the fault description text includes the fault cause leading to the fault described in the fault description text and the troubleshooting method for that fault. The prompt information generation module 230 is also used to generate prompt information based on the target troubleshooting text and the fault description text, indicating that the fault cause leading to the fault described in the fault description text and the troubleshooting method for that fault should be found from the target troubleshooting text. The fault troubleshooting result of the target device obtained by using a generative language model based on the prompt information includes: the fault cause leading to the fault described in the fault description text and the troubleshooting method for that fault.
[0097] Furthermore, the fault diagnosis device also includes a second information sending module, used to determine the address information of the target device based on the device identifier of the target device when the fault diagnosis result indicates that the fault diagnosis method is not to be diagnosed by the user, determine the fault diagnosis client based on the address information of the target device, and send fault diagnosis instruction information including the address information and the fault diagnosis result to the fault diagnosis client.
[0098] Furthermore, the text determination module includes a correlation calculation submodule and a text determination submodule. The correlation determination submodule is used to calculate the correlation score between the fault description text and each of the troubleshooting texts. The text determination submodule is used to determine the K troubleshooting texts with the highest correlation scores as target troubleshooting texts that match the fault description text, where K is a positive integer.
[0099] Furthermore, each of the troubleshooting texts corresponds to an index and a text vector of a preset length; the relevance determination submodule is also used to calculate a first relevance score between the fault description text and each of the troubleshooting texts based on the index of each of the troubleshooting texts and the fault description text; and to calculate a second relevance score between the fault description text and each of the troubleshooting texts based on the text vector of each of the troubleshooting texts and the fault description text; the text determination submodule is also used to determine the M troubleshooting texts with the highest first relevance score as target troubleshooting texts that match the fault description text, where M is a positive integer less than or equal to K; and to determine the K troubleshooting texts with the highest second relevance score as target troubleshooting texts that match the fault description text, where N is a positive integer less than or equal to K, and the union set formed by the K troubleshooting texts and the M troubleshooting texts includes the K troubleshooting texts.
[0100] Furthermore, the relevance determination submodule is also used to perform word segmentation on the fault description text to obtain a vocabulary list of the fault description text; for each troubleshooting text, based on the word frequency and inverse document frequency of the words in the vocabulary list of the fault description text corresponding to the index of the troubleshooting text, a relevance value between the word and the index of the troubleshooting text is obtained, wherein the index of the troubleshooting text includes the word frequency and inverse document frequency of multiple words in the troubleshooting text; the relevance values between each word in the vocabulary list of the fault description text and the index of the troubleshooting text are accumulated to obtain a first relevance score between the fault description text and the index of the troubleshooting text.
[0101] Furthermore, the fault diagnosis device also includes: a word segmentation module, a word frequency acquisition module, an inverse document frequency (IVF) determination module, and an index construction module. The word segmentation module is used to segment the fault diagnosis text into words to obtain a word list for the fault diagnosis text; the word frequency acquisition module is used to obtain the word frequency of each word in the fault diagnosis text based on the word list; the IVF determination module is used to determine the IVF of the target word based on the number of texts containing the target word in the text set composed of the multiple fault diagnosis texts, where the target word is any word in the word list of the fault diagnosis text; and the index construction module is used to construct an index for the fault diagnosis text based on the word frequency and IVF of each word in the fault diagnosis text.
[0102] Furthermore, the relevance determination submodule is also used to vectorize the fault description text to obtain a fault description vector of a preset length; calculate the similarity between the fault description vector and the text vector of each troubleshooting text to obtain the similarity between the fault description text and each troubleshooting text; and determine a second relevance score between the fault description text and each troubleshooting text based on the similarity between the fault description text and each troubleshooting text.
[0103] Furthermore, the fault diagnosis device also includes a vector extraction module, used to encode the troubleshooting text using an encoding network to obtain a text vector of a preset length; or to perform word segmentation on the troubleshooting text to obtain a vocabulary list of the troubleshooting text; to embed each word in the vocabulary list of the troubleshooting text to obtain a word vector of each word; and to aggregate the word vectors of each word in the vocabulary list of the troubleshooting text to obtain a text vector of a preset length.
[0104] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described device and module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0105] In the several embodiments provided in this application, the coupling between modules can be electrical, mechanical, or other forms of coupling.
[0106] Furthermore, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.
[0107] Please see Figure 7 , Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of this application is shown. The electronic device 300 may include the following components: a memory 310, one or more processors 320, and one or more application programs, wherein the one or more application programs are stored in the memory 310 and are used to cause the electronic device 400 to execute a method for generating a food purchase list applied to the electronic device when invoked by one or more processors 320.
[0108] The processor 320 may include one or more processing cores. The processor 320 connects to various parts within the electronic device 400 using various interfaces and lines, and performs various functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory, and by calling data stored in memory. Optionally, the processor may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. It is understood that the aforementioned modem may also not be integrated into the processor 320, but may be implemented separately through a communication chip.
[0109] The memory 310 may include random access memory (RAM) or read-only memory (ROM). The memory 310 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 310 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function, instructions for implementing the various method embodiments described above, etc. The data storage area may store data created during the use of the electronic device.
[0110] Please see Figure 8 This diagram illustrates a structural block diagram of a computer-readable storage medium according to an embodiment of this application. The computer-readable medium 400 stores program code that can be invoked by a processor to execute the methods described in the above method embodiments.
[0111] The computer-readable storage medium 400 may be an electronic memory such as flash memory, EEPROM (Electrically Erasable Programmable Read-Only Memory), EPROM, hard disk, or ROM. Optionally, the computer-readable storage medium 400 includes a non-volatile computer-readable storage medium. The computer-readable storage medium 400 has storage space for program code 410 that performs any of the method steps described above. This program code can be read from or written to one or more computer program products. The program code 410 may be compressed, for example, in a suitable form.
[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.
Claims
1. A troubleshooting method, characterized in that, Applied to electronic devices, the method includes: Obtain the fault description text of the target device; Identify the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, with each troubleshooting text corresponding to a fault type; Based on the target troubleshooting text and the fault description text, a prompt message is generated to indicate that fault-related information corresponding to the fault description text can be found from the target troubleshooting text; The fault diagnosis results of the target device are obtained by using a generative language model based on the prompt information.
2. The method according to claim 1, characterized in that, The troubleshooting text includes the fault phenomenon, fault description, fault cause, troubleshooting process, and troubleshooting method. The fault-related information corresponding to the fault description text includes the fault cause that led to the fault described in the fault description text and the troubleshooting method for the fault. The step of generating prompt information based on the target troubleshooting text and the fault description text to indicate that fault-related information corresponding to the fault description text can be found from the target troubleshooting text includes: Based on the target troubleshooting text and the fault description text, generate prompt information to indicate the cause of the fault described in the fault description text and the troubleshooting method for the fault, which can be found from the target troubleshooting text. The troubleshooting results of the target device obtained by using a generative language model based on the prompt information include: the cause of the fault described in the fault description text and the troubleshooting method for the fault.
3. The method according to claim 2, characterized in that, After using a generative language model to troubleshoot based on the prompt information and obtaining the troubleshooting results, including the cause of the fault described in the fault description text and the troubleshooting method for the fault, the method further includes: The troubleshooting results are sent to the user client associated with the target device.
4. The method according to claim 2, characterized in that, After using a generative language model to troubleshoot based on the prompt information and obtaining the troubleshooting results, including the cause of the fault described in the fault description text and the troubleshooting method for the fault, the method further includes: If the troubleshooting method instruction included in the troubleshooting results is not to be troubleshooted by the user, the address information of the target device is determined based on the device identifier of the target device, the troubleshooting client is determined based on the address information of the target device, and a troubleshooting instruction message including the address information and the troubleshooting results is sent to the troubleshooting client.
5. The method according to claim 1, characterized in that, The step of determining the target troubleshooting text that matches the fault description text among multiple troubleshooting texts includes: Calculate the correlation score between the fault description text and each of the troubleshooting texts; The K troubleshooting texts with the highest relevance scores are identified as target troubleshooting texts that match the fault description text, where K is a positive integer.
6. The method according to claim 5, characterized in that, Each of the troubleshooting texts corresponds to an index and a text vector of a preset length; The calculation of the correlation score between the fault description text and each of the troubleshooting texts includes: A first relevance score is calculated between the fault description text and each of the troubleshooting texts based on the index of each troubleshooting text and the fault description text. A second correlation score is calculated between the fault description text and each of the troubleshooting texts based on the text vector of each troubleshooting text and the fault description text. The step of identifying the K troubleshooting texts with the highest relevance scores as target troubleshooting texts that match the fault description text includes: The M troubleshooting texts with the highest first relevance scores are identified as target troubleshooting texts that match the fault description text, where M is a positive integer less than or equal to K; The K troubleshooting texts with the highest second relevance scores are identified as target troubleshooting texts that match the fault description text. N is a positive integer less than or equal to K, and the union set formed by the K troubleshooting texts and the M troubleshooting texts includes the K troubleshooting texts.
7. The method according to claim 6, characterized in that, The calculation of a first relevance score between the fault description text and each of the troubleshooting texts, based on the index of each troubleshooting text and the fault description text, includes: The fault description text is segmented into words to obtain a vocabulary list of the fault description text; For each troubleshooting text, the relevance value between the word and the index of the troubleshooting text is obtained based on the word frequency and inverse document frequency of the words in the vocabulary list of the fault description text corresponding to the index of the troubleshooting text. The index of the troubleshooting text includes the word frequency and inverse document frequency of multiple words in the troubleshooting text. The relevance values between each word in the vocabulary list of the fault description text and the index of the troubleshooting text are summed to obtain the first relevance score between the fault description text and the index of the troubleshooting text.
8. The method according to claim 6, characterized in that, The indexes of the various troubleshooting texts are constructed in the following manner: The troubleshooting text is segmented into words to obtain a vocabulary list of the troubleshooting text; The word frequency of each word in the troubleshooting text is obtained from the vocabulary list of the troubleshooting text. The inverse document frequency of the target word is determined based on the number of texts containing the target word in the text set consisting of the multiple troubleshooting texts. The target word is any word in the word list of the troubleshooting texts. An index of the troubleshooting text is constructed based on the word frequency and inverse document frequency of each word in the troubleshooting text.
9. The method according to claim 6, characterized in that, The calculation of a second relevance score between the fault description text and each of the troubleshooting texts, based on the text vectors of each troubleshooting text and the fault description text, includes: The fault description text is vectorized to obtain a fault description vector of a preset length; The similarity between the fault description vector and the text vector of each troubleshooting text is calculated to obtain the similarity between the fault description text and each troubleshooting text. A second relevance score is determined based on the similarity between the fault description text and each of the troubleshooting texts.
10. The method according to claim 6, characterized in that, The text vectors corresponding to each of the multiple troubleshooting texts are obtained in the following way: The troubleshooting text is encoded using an encoding network to obtain a text vector of a preset length; or The troubleshooting text is segmented into words to obtain a vocabulary list of the troubleshooting text. The words in the vocabulary list of the troubleshooting text are embedded to obtain the word vectors of each word. The word vectors of each word in the vocabulary list of the troubleshooting text are aggregated to obtain a text vector of a preset length.
11. A fault diagnosis device, characterized in that, Applied to electronic devices, the device includes: The text acquisition module is used to acquire the fault description text of the target device; The text determination module is used to determine the target troubleshooting text that matches the fault description text among multiple troubleshooting texts, and each troubleshooting text corresponds to a fault type. The prompt information generation module is used to generate prompt information based on the target troubleshooting text and the fault description text, indicating that fault-related information corresponding to the fault description text can be found from the target troubleshooting text; The fault diagnosis module is used to perform fault diagnosis based on the prompt information using a generative language model, and obtain the fault diagnosis results of the target device.
12. An electronic device, characterized in that, include: One or more processors; Memory; One or more applications, wherein the one or more applications are stored in the memory and configured to be executed by the one or more processors, the one or more applications being configured to perform the method as described in any one of claims 1-11.
13. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1-11.