Providing operating information about a household appliance

Abstract

A method (200) for providing operating information about a household appliance (105) comprises steps of capturing search queries by a plurality of users (110) regarding operating information about household appliances (105); combining the search queries with basic metadata about the household appliances (105); assigning terms that match the search query; creating data records each on the basis of a search query, combined basic metadata and assigned terms; and creating articles (160) in human-readable form each on the basis of a data record.

Description

[0001] 202401088

[0002] 1 / 25

[0003] Provision of operating information about a household appliance

[0004] The present invention relates to the provision of operating information about a device, in particular a household appliance.

[0005] A household appliance is designed for use in a home. Modern household appliances can be complex and comprise a multitude of components that work together in a predetermined sequence. Should a malfunction occur, the appliance may no longer function properly. A user of the appliance cannot usually easily distinguish whether a malfunction requires an action they can perform, the replacement of a component, or the replacement of the entire appliance.

[0006] The appliance can automatically detect and differentiate between certain error states. An indication of a detected error can be displayed as an error code on the device. However, this generally does not yet allow for a definitive determination of the cause. An error code may be accompanied by information that makes it easier for a user to find and resolve the cause and restore the appliance to working order. Such information may be provided by the appliance manufacturer.

[0007] In cases where this information is insufficient, the user can seek advice from other sources, particularly from an online forum or information repository. However, these sources do not always provide helpful tips or comprehensible explanations. Occasionally, incorrect or incomprehensible information is provided, or dialogue with someone via a forum may require many rounds, making it a lengthy process to obtain a result. Finally, certain interventions in a household appliance are generally not recommended, and in some cases, there is a risk of personal injury or further damage to the appliance. 202401088

[0008] 2 / 25

[0009] One of the problems underlying the present invention is to provide a technique for providing improved operating information about a household appliance. The invention solves this problem by means of the subject matter of the independent claims. Dependent claims describe preferred embodiments.

[0010] According to a first aspect of the present invention, a method for providing operational information about a household appliance comprises the steps of capturing search queries from a large number of users regarding operational information about household appliances; combining the search queries with basic metadata about the household appliances; assigning terms that match the search query; creating data records based on each search query, combined basic metadata, and assigned terms; and creating articles in human-readable form based on each data record. The method is preferably computer-implemented.

[0011] The operating information may include a description of a predetermined fault pattern. Preferably, the operating information includes a suggestion or instructions that can be used to correct a possible underlying cause of the fault pattern. It may also include usage instructions that explain how the non-defective appliance can be used to solve a predetermined problem. For example, the search query might be "how do I get grass stains out of a shirt," and the operating information might include a preferred wash program and / or detergent.

[0012] The operating information can explain which mechanism of the appliance might be malfunctioning and / or which component may not be working correctly. In some cases, it may also include a warning that a fault appears to exist that cannot be rectified by the user, meaning the appliance requires professional service or disposal. 202401088

[0013] 3 / 25

[0014] Search queries can be submitted by users to an internet search engine. These queries can be provided by the search engine operator or by a third-party product or service, such as Semrush or Ahrefs. A typical search query includes a description of a problem or error related to a household appliance. For example, a search query might be "washing machine won't drain."

[0015] Optionally, an interested party can also use the described mechanism to specifically feed in new or revised information. For example, a manufacturer of household appliances could place targeted search queries for problems or basic metadata into the search engine. New information related to search queries, such as an updated user manual, can be made available to the search engine to establish a correlation between a search query and the new information.

[0016] The basic metadata can include, in particular, a brand, an error code, a category of the household appliance, and / or an error level. The brand is assigned to a manufacturer. If the brand is not explicitly stated, it can be derived, for example, from a device name. Categories can be predefined and, for example, group together appliances for laundry care, dishwashing, or cooking. Further categories can be predefined, such as cleaning appliances, personal care and bathroom appliances, pet care and grooming appliances, handicrafts appliances, lighting appliances, as well as air conditioning and heating appliances, general electrical appliances, fitness and wellness equipment, or security devices. A category can be derived from a generic term or a model name.The error code can be provided by the appliance, for example, on the appliance itself or via a message, such as to an application on a mobile device. On the appliance, the error code can be provided, for example, by means of a light, a graphic display, an error tone, or an audiovisual display. 202401088.

[0017] 4 / 25

[0018] An error level can describe the severity of the error. For example, a first error level might simply include information for the user, a second level warnings, and a third level actual errors. An example of a first-level error on a dishwasher might involve a low rinse aid level. The appliance can usually still be used, possibly with slight reductions in functionality. An example of a second-level error on the same dishwasher might include a warning that a predetermined action could not be performed, such as drying the dishes after cleaning. The appliance can still be used, but its functionality is usually no longer complete or not guaranteed to the usual standard. An example of a third-level error might prevent the appliance from performing a further function.For example, the dishwasher may refuse to continue cleaning if water has been detected in the base tray.

[0019] Terms that match the error description can be added based on knowledge of the function and operation of the household appliances. This allows similar error patterns to be grouped. Effects that typically occur together can be better accounted for. This knowledge can be derived from predetermined sources, particularly publicly accessible sources such as operating manuals, forum posts, spare parts collections, or manufacturer's instructions. Optionally, non-publicly accessible sources can also be used, such as maintenance manuals, training materials for maintenance technicians, internal findings from maintenance or customer surveys, as well as data collected during the operation of the household appliances, design or manufacturing information, or test results.

[0020] In a particularly preferred embodiment, the articles are generated using a Large Language Model (LLM) based on general information with which the LLM has been trained. An LLM is a language model characterized by its ability to generate texts in a non-specific manner. It is a computational linguistic probability model that derives statistical word and sentence sequence relationships from a large number of text documents through a computationally intensive training process.

[0021] 5 / 25

[0022] has learned. These text documents typically include those taken from the internet or an internet archive. LLMs acquire these capabilities by using extremely large datasets to learn very large sets of parameters during training. LLMs are, in a broader sense, artificial neural networks, specifically so-called transformers, and are usually trained a priori either through self-supervised learning or semi-supervised learning methods.

[0023] Due to the numerous and diverse texts used for training, the LLM encompasses a kind of world knowledge that includes operational information about household appliances. Domain-specific expertise, i.e., specific information about household appliances, their operation, and use, can be specifically acquired through fine-tuning after the initial learning process. Any known, already trained LLM can be used for the technique presented here, for example, ChatGPT or an LLM from the Phi-3 family.

[0024] It is still preferred that the terms be assigned using a Learning Management Model (LLM). The LLM can determine the terms based on the search queries or based on its initially trained world knowledge. In general, an LLM as described herein can be used for various purposes, but different LLMs can also be used.

[0025] The articles can be created in a uniform, predetermined style. This allows a human reader to better understand the content and more easily compare or relate articles to one another. For example, a first article that generically describes a problem with a household appliance and a second article that describes the same problem with reference to a specific model or manufacturer can both use the same style.

[0026] If the article is created using a LLM (Language Lifecycle Management) system, a prompt can be provided to instruct the LLM to create the article in the predetermined style. The prompt can include detailed instructions regarding the desired style, structure, and / or restrictions on the content to be generated. The style can include a writing style characterized, for example, by a 202401088

[0027] 6 / 25

[0028] Style is characterized by word choice, sentence structure, text flow, and readability. It can also encompass the degree of formality, language, dialect, and personal address (formal / friendly). Style can further include external criteria such as the organization of information, the inclusion of headings, text alignment, and so on. Where possible, appearance characteristics can also be controlled externally. For example, the article can be subdivided into HTML objects whose appearance can be controlled using CSS instructions. The CSS instructions can be in a separate file, and one CSS file can be used to display different articles.

[0029] Preferably, a generated article is checked for compliance with predetermined formal criteria. Only an article that sufficiently meets the criteria can be made available. The check can, for example, relate to syntax, formatting, length, language choice, or adherence to other specifications. This step is preferably also performed by a Learning Management Developer (LLM), who is provided with a corresponding prompt requesting a criteria check and an article to be reviewed. An article that does not meet the criteria can be discarded and regenerated, with a prompt optionally providing further details regarding any identified deficiencies. The article can be regenerated as many times as necessary until the result meets the specifications. If this is not achieved after a predetermined number of attempts, the generation process can be aborted.

[0030] Separately verifying compliance with formal criteria ensures that the articles provided are of high quality and offer significant practical value to household appliance users. Users can quickly find assistance in an article to resolve a problem with their appliance. The information provided covers the full range of measures that are reasonable for a typical user. If these measures are unsuccessful, the user can then decide whether to contact a professional for repair or consider replacing the appliance. 202401088

[0031] 7 / 25

[0032] Preferably, the article is further enhanced with additional information to improve its discoverability by a search engine. In one embodiment, keywords characterizing the content are assigned to the article based on the data set. These keywords can be used to index the article, thus improving its searchability. The keywords are preferably chosen such that an internet search engine associates a relevant search query with the article. A search program ("web crawler") of the search engine can traverse the provided articles and evaluate the keywords.

[0033] The article can be saved at an address (URL) that contains one or more keywords in a human-readable format. Such a part of a URL is also called a slug. Furthermore, keywords can be included in the article's main heading. A main heading can be defined within an HTML structural element.<hl> ...< / hl> Furthermore, the machine-readable output format of the LLM can define an article structure that is easily processed by a search engine. This can facilitate finding an article with relevant content to a given question.

[0034] The search engine can index keywords and associate them with the content of the articles. When the search engine receives a query, it can find a matching article and provide a link to it. In effect, a user can use a familiar search engine to enter a brief description of a problem with their household appliance, and the search engine can provide a link to a relevant article that explains how to proceed. This eliminates the need for the user to first access a collection of articles or to actively use a service that manages the articles.

[0035] Keywords are preferably assigned using a LLM (Language Lifecycle Management) system. Alternatively, keywords are preferably determined within the style guidelines for an article. A created article can contain human-readable information and keywords that are optimized for processing by a search engine. 202401088

[0036] 8 / 25

[0037] According to a further aspect of the present invention, a device for providing operational information about a household appliance comprises an interface for capturing search queries from a multitude of users regarding operational information about household appliances; and a processing unit. The processing unit is configured to combine search queries with basic metadata about the household appliances; to assign terms that match the error description; to create data records based on a search query, combined basic metadata, and assigned terms; and to generate articles in human-readable form based on each data record.

[0038] The interface can connect the device to a data network to which a variety of devices are connected. For example, a service that collects search queries can be connected via the data network.

[0039] The processing equipment may be configured to partially or completely execute a method described herein. For this purpose, the processing equipment may be electronic and may, for example, include a programmable microcomputer or microcontroller. The method may be in the form of a computer program product containing program code. The computer program product may also be stored on a computer-readable data carrier. Features or advantages of the method may be transferred to the equipment and vice versa.

[0040] In one embodiment, the processing unit includes or implements an LLM (Large Learning Module). In another embodiment, the LLM is connected via an interface. For example, the LLM can be offered as a service on a server or in the cloud, or it can be maintained in a local installation. The LLM is trained on the basis of a large number of texts that can be found, for example, on the internet. Such texts naturally also include those dealing with household appliances.

[0041] Non-limiting embodiments of the invention are now described in more detail with reference to the accompanying figures, in which: 202401088

[0042] 9 / 25

[0043] Figure 1 a system;

[0044] Figure 2 shows a flowchart of a process; and

[0045] Figure 3 is a flowchart

[0046] represents.

[0047] Figure 1 shows a system 100. A household appliance 105, here exemplified as a washing machine, can be used by a user 110. Should a fault occur, the user 110 can seek help in a data network 120, such as the internet, using a data device 115. The data device 115 could be, for example, a desktop computer, a laptop computer, or a personal mobile device, especially a smartphone. Information about the household appliance 105 could be stored, for example, in an information service 125, which could include a forum, a wiki, a collection of tips, or a collection of operating instructions. Typically, many different information sources 125 are connected to the data network 120.

[0048] A search engine 130 is configured to search information sources 125 for information and assign the found information to specific keywords. If the user 110 submits a search query to the search engine 130, the engine can recognize a keyword in the query and determine the associated information. The user 110 can then be provided with a link to the information at the information source 125.

[0049] To improve the quality of information that the user 110 can find in this way via their home appliance 105 in the data network 120, a device 135 is proposed. The device 135 comprises a processing unit 140, which is connected to the data network 120 via an interface 145, and an optional data storage unit 150. Furthermore, the processing unit 140 is connected to a large language model 155, also known as the Large Language Model (LLM) 155. Generally, Eie- 202401088

[0050] 10 / 25

[0051] The elements of the depicted system 100 can be communicatively connected to each other in other ways instead of via the data network 120, for example directly or via a dedicated additional data network.

[0052] The LLM 155 is trained in a known manner with a large number of texts originating from various information sources 125. As a result, the LLM reflects a kind of world knowledge from the training data, which can be used, for example, to provide meaningful answers to questions in a dialogue or to perform certain tasks, particularly those involving text processing. The LLM 155 can be implemented independently and connected to the data network 120; alternatively, an embodiment is also conceivable in which the LLM is part of the device 135. For example, the LLM 155 can be implemented by the processing unit 140, or the processing unit 140 can include an LLM 140.

[0053] It is proposed that the device 135, using the LLM 155, collects information regarding the household appliance 105 and presents it in the form of articles 160, in which the user 110 can easily absorb the information. Furthermore, it should be ensured that the search engine 130 can index the articles 160 effectively, so that a search query from the user 110 to the search engine 130 produces a response that includes a reference to a relevant article 160.

[0054] Articles 160 can be stored in the data storage 150 and delivered by the processing unit 140 when a suitable request is received via the data network 120. An Article 160 can, for example, be in the form of an HTML document. A multitude of Articles 160 can be organized according to a predetermined key, for example, relating to a model or category of household appliance 105, a fault that has occurred, a potentially defective component, a manufacturer, or a brand. An Article 160 can relate to specific appliances 105 or a group or class of appliances 105. 202401088

[0055] 11 / 25

[0056] Ideally, the articles should be divided into predefined sections. A first example section addresses common errors. Here, the entries are sorted by device category and then by the number of search engine queries (volume). This section can also be organized by internal data such as page views. A second example section addresses error codes for household appliances. These can be sorted by appliance category, then by brand, and finally alphabetically. A third example section provides the technical background of a problem.

[0057] To provide Articles 160, search queries from users 110 to the search engine 130 can first be identified that relate in some way to household appliances 105. These search queries may, in particular, include information on the operation, use, troubleshooting, or repair of a household appliance 105. Information sought by a user 110 in this way is referred to herein as operating information for household appliances 105. The search queries can be obtained directly from the search engine 130 or from a service provider that collects and analyzes search queries submitted to the search engine.

[0058] In a first processing stage, the search queries are combined with basic metadata about the household appliances 105. This basic metadata can include, in particular, a brand, an error code, an error level, and / or a category of the household appliance 105. Furthermore, the search queries can be enriched with additional relevant terms. This processing stage is preferably performed by the LLM 155 by providing a list of search queries and generating a prompt that includes a request to process the search queries as described. An output can comprise a number of records, each containing a search query along with associated basic metadata and certain additional terms. The output typically comprises a cluster of multiple search queries. For example, for the search query "washing machine not draining," 661 combinations were identified in a standard SEO analysis tool. The data- 202401088

[0059] 12 / 25

[0060] Sentences are preferably provided in machine-readable form and can, for example, be stored in data storage 140.

[0061] In a second processing stage, human-readable articles (160) can be generated from the datasets. To do this, the LLM (155) can be prompted via another command to generate an article (160) in a predetermined style based on a dataset. The styles of the generated articles (160) can be similar or identical. Furthermore, a list of keywords can be assigned to an article (160). These keywords are chosen using search engine optimization (SEO) principles so that the search engine (130) can establish an association between a search query and an article (160) with relevant content. Keywords can be stored in the article in such a way that they are not typically perceived by humans but are processable by a machine, such as a web crawler.

[0062] Optionally, a created Article 160 can be checked for compliance with predetermined formal criteria. An Article 160 that fails the check can be discarded and recreated. If the Article 160 passes the check, it can be published online. Furthermore, the search engine 130 can be notified of the article to expedite its indexing.

[0063] Figure 2 shows a flowchart of a procedure 200 for providing operational information about a household appliance 105.

[0064] In step 205, search queries can be determined that a large number of users 105 have submitted to the search engine 130 within a predetermined period regarding household appliances 105.

[0065] In step 210, the search queries can be sorted or grouped, for example by the language in which the search query was made, a category of a relevant household appliance 105, or a specification of a concrete 202401088

[0066] 13 / 25

[0067] ten household appliance 105, for example in the form of a designation, a trade name or a device code.

[0068] In step 215, a data record can be determined for a search query using LLM 155. Basic metadata about an affected household appliance 105 can be assigned to the search query; furthermore, additional terms that match the search query can be assigned. The search query typically includes a fault description, and the additional terms can expand upon this description. For example, if the search query includes a fault code displayed by household appliance 105, the additional terms can provide a brief description of the fault.

[0069] The data records are provided in a predetermined, machine-readable format, in particular in YAML format, and can be stored in data storage 150 in a single step 220. It is preferred that a large number of search queries are processed in this way before further processing takes place based on the provided data records.

[0070] In step 225, a further prompt can be defined, which includes a request to the LLM 155 to create an article 160 based on a data set. This prompt, also referred to here as the style prompt, also contains instructions regarding the style in which the article 160 should be written. Furthermore, it preferably includes a instruction to request the LLM 155 to determine meta-information for the article 160. This meta-information can, in particular, include keywords that can be used to optimize the article 160 for indexing by the search engine 130.

[0071] The specified style prompt can be provided to the LLM 155 along with a data set. Subsequently, in step 230, the human-readable portion of an article 160 can be determined, and in step 235, machine-readable metadata can be determined. A completed article 160 preferably includes both types of information. The LLM 155 can construct the content of an article 160 from the knowledge used in its training. This knowledge can include specific information such as an entry in an instruction manual, 202401088.

[0072] 14 / 25

[0073] but it can also include general knowledge, such as a causal relationship between certain events or world knowledge. World knowledge might include, for example, the general understanding that escaping water tends to flow downwards. Such knowledge can be combined with specific knowledge about a household appliance 105 to form a helpful statement in Article 160. For example, it can be determined that water that has collected in a drip tray of a household appliance 105 may have previously leaked from a component located above it. The LLM 155 can use knowledge from its own resources regarding the provided dataset to create an Article 160.

[0074] In step 240, a created article 160 can be checked for compliance with predetermined formal criteria. Such criteria can relate, for example, to formatting, readability, or length.

[0075] If an Article 160 passes the review, it can be published in step 245. Afterwards, the Article 160 can be requested and delivered via the data network 120. In one embodiment, a web offering is used, comprising a multitude of Articles 160 created in the manner described herein. Publication may include providing a request or invitation to the search engine 130 to index the Article 160.

[0076] Once search engine 130 has indexed article 160, it can provide a user 110 with a link to a matching article 160 in response to a search query. User 110 can follow the link and request article 160 from device 135 or any other device on which article 160 has been published.

[0077] It is generally preferred that the procedure 200 be executed regularly to match new search queries with existing or new articles 160. Existing articles 160 can be revised, expanded, or replaced in the process. This allows the quality of existing articles to be improved when new information becomes available or when the LLM acquires new skills. 202401088

[0078] 15 / 25

[0079] Figure 3 shows a flowchart 300 for providing operational information about a household appliance 105 in an exemplary embodiment. Specifications for the flowchart 300 can be provided by an administrator 305.

[0080] Furthermore, the flowchart includes 300 functional components that can be implemented by the processing unit 140 or the LLM 155. In the illustrated embodiment, such functional components include a data transformation engine 310, an LLM keyword enrichment engine 315, a raw data base 320, an article generation engine 325, a generated content database 330, and a website 335.

[0081] In step 350, the administrator can provide 305 collected data on device errors from household appliances 105 to the data transformation 310. The data can be normalized and stored in the database 320 (step 352).

[0082] In step 354, the administrator can provide lists of search queries or keywords that users have searched for in relation to household appliances. In step 356, metadata corresponding to the keywords can be determined. The determined data can then be provided to the enrichment process. In step 358, this enriched keywords or data records, which include search queries, additional terms, and basic metadata, can be provided to the database.

[0083] Furthermore, the administrator can provide a first prompt (system prompt) to the article creation process (325) in step 360 and a second prompt (instruction prompt) in step 362. Additionally, normalized data can be provided to article creation process (325) in step 364, and enriched keywords from database 320 can be provided in step 366. 202401088

[0084] 16 / 25

[0085] The article creation process (325) can, in step 368, create an article (160) and store it in the article database (330). From there, article (160) can be transferred to the web offering (335) in step 370.

[0086] In step 372, user 110 can request an article 160 from the website 335. To do this, user 110 can follow a link provided by the search engine 130 in response to a search query from user 110. The article 160 can then be delivered to user 110. Access analytics data can be provided to the administrator 305 in step 374.

[0087] The following are two example prompts that can be used for a technique presented herein to control LLM 155. Prompt 1 concerns the creation of a data record in step 215 of procedure 200. Prompt 2, the style prompt, concerns the creation of an article 160 based on a data record; see steps 225 to 235 of procedure 200. For practical use, it is recommended that the prompts be formulated in the language in which LLM is best controlled, usually English.

[0088] Prompt 1 :

[0089] You are writing a manual for quick troubleshooting of household appliances. You receive the contents of an error code in YAML format. You are to structure your response as a YAML file. Respond ONLY with a YAML file that has the following properties:

[0090] • category is a copy of the category field

[0091] • slug is a copy of the slug field

[0092] • brand is a copy of the field brand

[0093] • keywords is a comma-separated list of 5 to 8 SEO-friendly keywords that match the error description; try to place popular keywords first. No other content.

[0094] Your answer should be in German.

[0095] Your answer should use the informal "you".

[0096] { {File content}} 202401088

[0097] 17 / 25

[0098] Prompt 2 :

[0099] You are writing a manual for quick troubleshooting of household appliances. You receive the contents of an error code in YAML format. You are to structure your response in JSON format. Your response should have the following JSON file structure:

[0100] • The headline contains important keywords that clarify the problem and is 40 to 60 characters long. • The short title summarizes the problem in 3 to 6 words and tries to include high-quality keywords. • The brand is a reference to the brand.

[0101] • Keywords is a comma-separated list of 5 to 8 SEO-friendly keywords that match the error description; try to place popular keywords first.

[0102] • slug is a copy of the input slug

[0103] • errorCode is a copy of the input error code

[0104] • category is only a copy of the input category

[0105] • The description is a summary of the problem in 2 paragraphs.

[0106] • ranking: Copy of the number from the input ranking

[0107] • repairCostRange: is an object with the properties min and max, both of which are numbers.

[0108] • repairRecommended: is a Boolean value based on the repairCostRange, true if the estimated repairCostRange is less than 450C.

[0109] • errorLevel: assesses the severity of the problem between low, medium and high, assigning a severity level of severe if it is not possible to fix the problem without a professional mechanic.

[0110] • Troubleshooting steps should contain between 3 and 5 tips, each 1 paragraph long; each troubleshooting step should consist of a short title of 3 to 4 words and a description property called content 202401088

[0111] 18 / 25

[0112] MAKE SURE ALL FEATURES ARE INCLUDED

[0113] Use language that is easy for a customer to understand.

[0114] Write in a friendly tone.

[0115] Your answer should be SEO-friendly.

[0116] Include all keywords in your answer.

[0117] Your answer should be in German.

[0118] Your answer should use the informal "you".

[0119] I don't recommend a professional mechanic.

[0120] I do not recommend any other websites or manuals.

[0121] NO ADDITIONAL CONTENT RETURN ONLY THE VALID JSON OBJECT NO FORMATTING APPLY

[0122] { { File content}}

[0123] The output in this example (error E16 on a Bosch washing machine) is a JSON file containing structure and content data for further use, e.g., on a homepage.

[0124] Example output JSON file:

[0125] {

[0126] "Headline": "Error code E16 Bosch washing machine: Problems with the door"

[0127] "shortTitle" : "Error E16 - Door problems" ,

[0128] "brand" : "bosch" ,

[0129] Keywords: "Error code E16, Bosch washing machine, door problems, locking mechanism, door not closed, washing machine troubleshooting"

[0130] "slug" : "el6" ,

[0131] "errorCode" : "el 6" ,

[0132] "category" : "washing machines" ,

[0133] "Description": "The error message E16 displayed on your Bosch washing machine indicates possible door problems. This error can occur if the door of your washing machine is not completely closed or if there is a problem with the door mechanism. 202401088

[0134] 19 / 25

[0135] The locking mechanism is faulty. It is important to rectify this fault quickly to prevent further damage to the device and to ensure safe use.

[0136] "ranking": 2

[0137] "repairCostRange" : {

[0138] "min" : null ,

[0139] "max" : null

[0140] } ,

[0141] "repairRecommended" : true ,

[0142] "errorLevel" : "medium" ,

[0143] "troubleshootingSteps": [

[0144] {

[0145] "title": "Check the tap"

[0146] "content": "Make sure the tap is fully open and that the water pressure is sufficient. This could be a possible reason for error code E16 on your Bosch washing machine."

[0147] } ,

[0148] {

[0149] "title": "Check the door"

[0150] "content": "Make sure the door of your washing machine is securely closed. Sometimes a small object or a piece of clothing caught in it can prevent the door from closing properly. This could be the reason for the error code E16 on your Bosch washing machine."

[0151] } ,

[0152] {

[0153] "title": "Cleaning door seals"

[0154] "content": "Inspect and clean the door seal and locking mechanism. Dirt and deposits can impair proper function and thus lead to the display of error code E16."

[0155] } ,

[0156] {

[0157] "title": "Check locking mechanism"

[0158] "content": "If the door appears to be properly closed at first glance, but the error code E16 is still displayed 202401088

[0159] 20 / 25

[0160] If this happens, there could be a problem with the locking mechanism. Examine it for possible defects.

[0161] } ,

[0162] {

[0163] "title": "Restarting the washing machine"

[0164] "content": "Try unplugging your washing machine for a few minutes and then plugging it back in. This can sometimes fix minor software problems that may be causing error messages."

[0165] }

[0166] ]

[0167] }

[0168] The provided file in a machine-readable format, especially JSON, is the technical basis for using and displaying the help or SEO (Search Engine Optimized) content as article 160 on a website 335. For example, if a user 110 encounters error code E16 on a Bosch washing machine 105, they can find the relevant article 160 by entering the error code into a standard search engine 130. As a first step, a general description of the error code can be displayed, allowing the user 110 to verify that their problem has been understood. Further specific tips and instructions to help them resolve the error can be accessed within article 160, for example, by clicking or scrolling. 202401088

[0169] 21 / 25

[0170] Reference sign

[0171] 100 System

[0172] 105 household appliance

[0173] 110 users

[0174] 115 Data device

[0175] 120 Data network, Internet

[0176] 125 Information Service

[0177] 130 search engine

[0178] 135 Device

[0179] 140 processing equipment

[0180] 145 Interface

[0181] 150 data storage devices

[0182] 155 Large Language Model, LLM 160 articles

[0183] 200 procedures

[0184] 205 Recording search queries

[0185] 210 Sorting by language, category, device 215 Creating a record with LLM

[0186] 220 Store data record in machine-readable format 225 Determine style prompt

[0187] Format 230 articles for human reading

[0188] Determine 235 meta-information for search engines; check 240 articles

[0189] Publish 245 articles

[0190] 300 Flowchart

[0191] 305 Administrator

[0192] 310 Data Transformation

[0193] 315 Enrichment of keywords

[0194] 320 database

[0195] 325 article creation 202401088

[0196] 22 / 25 330 article database

[0197] 335 Web offer

[0198] 350 collected data points about device errors; 352 normalized data points

[0199] 354 Keyword Lists

[0200] 356 Keyword Metadata

[0201] 358 enriched keywords

[0202] 360 System Prompt

[0203] 362 Instruction prompt

[0204] 364 normalized data points

[0205] 366 enriched keywords

[0206] 368 generated items

[0207] 370 items

[0208] 372 analysis data

[0209] 374 Providing analysis data

Claims

202401088 23 / 25 PATENT CLAIMS 1. Method (200) for providing operational information about a household appliance (105), wherein the method (200) comprises the following steps: - Capturing search queries from a large number of users (110) regarding operating information of household appliances (105); - Combining search queries with basic metadata about the household appliances (105); - Matching terms that correspond to the search query; - Creating data records based on a search query, combined basic metadata, and associated terms; and - Creating articles (160) in human-readable form, each based on a data set.

2. Method (200) according to claim 1, wherein the articles (160) are created using an LLM (155) on the basis of general information with respect to which the LLM (155) is trained.

3. Method (200) according to claim 1 or 2, wherein the terms are assigned by means of an LLM (155).

4. Method (200) according to any of the preceding claims, wherein the articles (160) are produced in a uniform, predetermined style.

5. Method (200) according to claims 3 and 4, wherein a prompt is provided to direct the LLM (155) to create an article (160) in the predetermined style.

6. Method (200) according to one of the preceding claims, wherein a manufactured article (160) is checked for compliance with predetermined formal criteria; and only an article (160) that sufficiently meets the criteria is provided. 202401088 24 / 25 7. Method (200) according to one of the preceding claims, wherein keywords characterizing the content are assigned to the article (160) on the basis of the data set.

8. Method (200) according to claim 7, wherein the keywords are chosen to allow an internet search engine to associate a suitable search query with the article (160).

9. Method (200) according to one of claims 7 or 8, wherein the keywords are assigned by means of an LLM (155).

10. Device (135) for providing operating information about a household appliance (105), wherein the device (135) comprises the following elements: - an interface (145) for capturing search queries from a large number of users (110) regarding operational information from household appliances (105); and - a processing facility (140) which is equipped for this purpose: ° To combine search queries with basic metadata about the household appliances (105); ° to assign terms that match the search query; ° to create data records based on a search query, combined basic metadata and associated terms; and ° to create articles (160) in human-readable form based on a data record.

11. Device (110) according to claim 10, wherein the processing device (140) implements or comprises an LLM (155).

Images