KR102840639B1 - Electronic apparatus and method for determining search intent - Google Patents

Abstract

The method of the present disclosure may include the steps of: obtaining search result information related to content corresponding to the search query based on the search query; determining at least one of a type of content or a type of search result based on the search result information; generating a prompt based on at least one of the search query, the search result information, the type of content, or the type of search result; and inputting the generated prompt into a language model to determine a search intent of the search query.

Description

ELECTRONIC APPARATUS AND METHOD FOR DETERMINING SEARCH INTENT

The technical idea of the present disclosure relates to a method and an electronic device for determining search intent.

Today, information retrieval technologies are exploring various methods to effectively provide relevant information based on user search queries. In particular, it can be difficult to determine a user's search intent based solely on search queries consisting of words or short sentences. Search intent refers to the purpose or need behind a user's input of a specific search term, and various data, including user log data, can be analyzed to determine this intent.
Prior Document No. 10-2316474, “Search Result Processing Server, Search Result Processing Method, and Terminal”

The technical problem to be solved by the present invention is to generate a prompt based on at least one of search result information, type of content, or type of search result, and input the prompt into a language model to determine search intent.

The technical problem to be solved by the present invention is to determine at least one of the type of content or the type of search result based on search result information.

The technical problem that the present invention seeks to solve is to determine search intent through a language model even in a situation where there is no learning data that classifies search intent into labels.

The technical problem that the present invention seeks to solve is to update a language model using learning data classified by search intent as a label.

The technical problem to be solved by the present invention is to identify at least some of the contents among the plurality of contents displayed on a search results page in order to generate a prompt.

In one embodiment, the method may include: obtaining search result information related to content corresponding to the search query based on the search query; determining at least one of a type of content or a type of search result based on the search result information; generating a prompt based on at least one of the search query, the search result information, the type of content, or the type of search result; and inputting the generated prompt into a language model to determine a search intent of the search query.

In one embodiment, the search result information may include at least one of the name of the website that posted the content, the title of the content, a summary of the content, or the Uniform Resource Locator (URL) address of the website.

In one embodiment, the type of content may be at least one of a type posted on an e-commerce webpage displayed on a search results page, a type of news-related content, a type of content posted on a community webpage, a type of content posted on a blog, a type of video content, a type of image content, or a type of file posted on a webpage.

In one embodiment, the type of search result is determined based on search result information, and may include at least one of a first type displaying the title and summary information of the content, a second type displaying in the form of an advertisement, a third type displaying in the form of news, a fourth type displaying in the form of information provision, a fifth type displaying video content or image content, a sixth type displaying a map, a seventh type displaying application installation, an eighth type displaying a URL link for website shortcuts, or a ninth type displaying a community, as a type of content displayed on a search result page.

In one embodiment The step of determining at least one of the type of content or the type of search result is: It may include a step of determining the type of content based on at least one of the protocol, domain, path, query parameter, file name, or fragment included in the URL address.

In one embodiment, the step of determining at least one of the type of content or the type of search result may include at least one of: determining the type of content as a news-related content type based on a determination that the domain corresponds to a news domain or that the path includes a news-related keyword; determining the type of content as a content type posted on a blog based on a determination that the domain corresponds to a blog domain or that the path includes a blog-related keyword; determining the type of content as a video content type based on a determination that the domain corresponds to a platform domain for uploading videos; or determining the type of content as a file type posted on a webpage based on a determination that the URL address includes a file name.

In one embodiment, the step of determining at least one of a type of content or a type of search result may include a step of determining a type of content or a type of search result corresponding to each of a plurality of contents corresponding to the search query.

In one embodiment, the method may further include the step of excluding content whose search result type is the second type or the fifth type from among the plurality of contents; and the step of generating a prompt based on at least one of the content type or the search result type of each of the remaining contents excluding the content whose search result type is the second type or the fifth type.

In one embodiment, the step of generating a prompt may include the step of generating a prompt based on at least one of a name of a website corresponding to each of a plurality of contents, a title of the contents, a summary of the contents, a URL address of the website, a type of the contents, or a type of search result.

In one embodiment, the step of generating a prompt may include the step of generating the prompt to include a query that determines a search intent based on at least one of a search query, search result information, a type of content, or a type of search result, and a query that asks for an answer to explain the basis for determining the search intent.

In one embodiment, the method may further include the steps of generating training data using the generated prompt as input data and labeling search intents corresponding to the search query; and updating the language model based on the training data.

In one embodiment, the search intent may include at least one of a first intent to seek an answer to a query, a second intent to find a particular website, a third intent to seek information about a particular product, or a fourth intent to complete a particular task.

In another embodiment, a computer-readable recording medium having recorded thereon a program for determining a search intent, wherein the program may obtain search result information related to content corresponding to the search query based on a search query, determine at least one of a type of the content or a type of the search result based on the search result information, generate a prompt based on at least one of the search query, the search result information, the type of the content, or the type of the search result, and input the generated prompt into a language model to determine the search intent of the search query.

In another embodiment, an electronic device includes a memory; and one or more processors, wherein the one or more processors are configured to obtain search result information related to content corresponding to the search query based on a search query, determine at least one of a type of the content or a type of the search result based on the search result information, generate a prompt based on at least one of the search query, the search result information, the type of the content, or the type of the search result, and input the generated prompt into a language model to determine a search intent of the search query.

According to an embodiment of the present disclosure, a prompt may be generated based on at least one of search result information, type of content, or type of search result, and the prompt may be input into a language model to determine search intent.

According to an embodiment of the present disclosure, at least one of the type of content or the type of search result can be determined based on search result information.

According to an embodiment of the present disclosure, search intent can be determined through a language model even in a situation where there is no learning data that classifies search intent into labels.

According to an embodiment of the present disclosure, a language model can be updated using learning data classified as labels with search intent.

According to an embodiment of the present disclosure, at least some of the content may be identified to generate a prompt among a plurality of contents displayed on a search results page.

FIG. 1 is a diagram illustrating a system for determining search intent according to one embodiment of the present disclosure.
FIG. 2 is a diagram illustrating a method for determining search intent using search result information according to one embodiment of the present disclosure.
FIG. 3 is a diagram illustrating content of which the type of search result is the second type or the fifth type according to one embodiment of the present disclosure.
FIG. 4 is a diagram illustrating content used to determine a prompt according to one embodiment of the present disclosure.
FIG. 5 is a diagram illustrating the types of content and the types of search results according to one embodiment of the present disclosure.
FIG. 6 is a flowchart illustrating a method by which an electronic device determines a search intent according to one embodiment of the present disclosure.
FIG. 7 is a drawing illustrating an electronic device according to various embodiments of the present disclosure.

Hereinafter, preferred embodiments of the present invention will be described with reference to the attached drawings. It should be noted that identical components in the drawings are indicated by identical reference numerals and symbols, wherever possible, even if they appear in different drawings. In the following description of the present invention, detailed descriptions of known functions or configurations will be omitted if they are deemed to unnecessarily obscure the gist of the present invention.

Additionally, when a part is said to "include" a component, this does not mean that it excludes other components, but rather that it may include other components, unless otherwise specifically stated.

The terminology used herein is for the purpose of describing embodiments and is not intended to limit the present invention. In this specification, the singular also includes the plural, unless specifically stated otherwise. In this specification, terms such as "include," "provide," "provide," or "have" do not exclude the presence or addition of one or more other components other than the mentioned components.

In this specification, terms such as “or,” “at least one,” and the like can refer to one of the words listed together, or a combination of two or more. For example, “or B” or “and at least one of B” can include only one of A or B, or can include both A and B.

In this specification, descriptions using the word “for example” or the like should not be construed as limiting the embodiments of the invention in terms of the effects of variations such as tolerances, measurement errors, limitations of measurement accuracy, and other commonly known factors, as well as the information presented, such as cited characteristics, variables, or values, may not be exactly the same.

In this specification, terms such as "first" and "second" may be used to describe various components, but the components should not be limited by these terms. Furthermore, these terms should not be construed to limit the order of each component, but rather may be used to distinguish one component from another. For example, a "first component" may be referred to as a "second component," and similarly, a "second component" may also be referred to as a "first component."

Each block of the processing flow diagrams attached to this specification and combinations of the flow diagrams can be performed by computer program instructions. These computer program instructions can be installed in a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing equipment, so that the instructions, when executed by the processor of the computer or other programmable data processing equipment, create a means for performing the functions described in the flow diagram block(s).

These computer program instructions may be stored in a computer-available or computer-readable memory that can direct a computer or other programmable data processing device to implement a function in a particular manner, and the instructions stored in the computer-available or computer-readable memory may also produce an article of manufacture that includes instruction means for performing the function described in the flowchart block(s).

Since the computer program instructions may be installed on a computer or other programmable data processing device, a series of operational steps may be performed on the computer or other programmable data processing device to create a computer-executable process, and the instructions that cause the computer or other programmable data processing device to perform the steps for performing the functions described in the flowchart block(s) may also provide steps for performing the functions described in the flowchart block(s).

Additionally, each block may represent a module, segment, or portion of code that contains one or more executable instructions for performing a specific logical function(s). Furthermore, in some alternative implementation examples, the functions described in the blocks may occur out of order. For example, two blocks depicted in succession may actually be executed substantially concurrently, or the blocks may sometimes be executed in reverse order, depending on their respective functions.

The “electronic device” or “terminal” referred to herein may be implemented as a computer or portable terminal capable of connecting to a server or other terminal via a network. Here, the computer includes, for example, a notebook, desktop, or laptop equipped with a web browser, and the portable terminal may include, for example, a wireless communication device that ensures portability and mobility, such as a communication-based terminal such as IMT (International Mobile Telecommunication), CDMA (Code Division Multiple Access), W-CDMA (W-Code Division Multiple Access), LTE (Long Term Evolution), and all types of handheld-based wireless communication devices such as smartphones and tablet PCs. In addition, the “electronic device” or “terminal” referred to herein may also include a processor, a memory that stores and executes program data, permanent storage such as a disk drive, a communication port that communicates with an external device, a user interface device such as a touch panel, a key, a button, etc.

In the present disclosure, methods implemented as software modules or algorithms may be stored on a computer-readable recording medium as computer-readable codes or program instructions executable on a processor. Here, the computer-readable recording medium may include a magnetic storage medium (e.g., read-only memory (ROM), random-access memory (RAM), floppy disks, hard disks, etc.) and an optical reading medium (e.g., CD-ROM, DVD: Digital Versatile Disc)). The computer-readable recording medium may be distributed and executed on network-connected computer systems.

An artificial intelligence model (or model) can be implemented as a neural network (or artificial neural network) and can operate based on statistical learning algorithms that mimic biological neurons in machine learning and cognitive science. A neural network can refer to a general model in which artificial neurons (nodes) that form a network through the combination of synapses learn and change the binding strength of the synapses to have problem-solving capabilities. A neural network can be composed of multiple neural network layers, and for example, a neural network can include an input layer, a hidden layer, and an output layer. Each of the multiple neural network layers can include at least one node and at least one weight, and can perform neural network operations through operations between the operation results of the previous (precious) layer and the weights. At least one weight of the multiple neural network layers can be optimized based on the learning results of the artificial intelligence model. For example, at least one weight may be updated during the learning process to reduce or minimize the loss or cost value obtained from the AI model. A neural network can infer a desired outcome from any input.

The learning methods of artificial intelligence models can be categorized into supervised learning, where input and output data are provided as training data, and the correct answer (output data) corresponding to the problem (input data) is determined, unsupervised learning, where only input data is provided without output data, and the correct answer (output data) corresponding to the problem (input data) is not determined, and reinforcement learning, where a reward is given whenever an action is taken in the current state, and learning progresses in the direction of maximizing this reward. Alternatively, they can be categorized according to the architecture, which is the structure of the learning model.

In an embodiment of the present disclosure, the artificial intelligence model is a CNN (Convolution Neural Network) such as GoogleNet, AlexNet, VGG Network, etc., R-CNN (Region with Convolution Neural Network), RPN (Region Proposal Network), RNN (Recurrent Neural Network), S-DNN (Stacking-based deep Neural Network), S-SDNN (State-Space Dynamic Neural Network), Deconvolution Network, DBN (Deep Belief Network), RBM (Restrcted Boltzman Machine), Fully Convolutional Network, LSTM (Long Short-Term Memory) Network, Classification Network, Generative Modeling, eXplainable AI, Continual AI, Representation Learning, AI for Material Design, BERT for natural language processing, SP-BERT, MRC/QA, Text Analysis, Dialog System, GPT-3, GPT-4, Visual Analytics for vision processing, Visual Understanding, Video Synthesis, ResNet for data intelligence, Anomaly Detection, Prediction, Time-Series Forecasting, Optimization, At least one of various artificial intelligence structures and algorithms, such as Recommendation, Data Creation, etc., may be used, and the above-described examples are merely listing examples of artificial intelligence structures and algorithms used according to embodiments of the present disclosure, and do not limit the artificial intelligence structures and algorithms used according to embodiments of the present disclosure.

Unless otherwise defined, all terms used herein may be used in their common sense by those of ordinary skill in the art to which the present invention pertains. Furthermore, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

Hereinafter, various embodiments of the present invention are described with reference to the attached drawings.

FIG. 1 is a diagram illustrating a system for determining search intent according to one embodiment of the present disclosure.

In one embodiment, the server (110) may be a device that determines search intent. The server (110) may be a device that provides users with the determined search intent. The server (110) may determine a search intent corresponding to a search query entered by a user through a terminal (120) and transmit the search intent to the terminal (120).

Search intent refers to the fundamental purpose or reason a user enters a specific query into a search engine. Depending on the type of information or goal the user seeks through search, search intent can be categorized into information exploration (searching to obtain information), transaction (searching to purchase or reserve), and navigation (searching to navigate to a specific site or page). The above classifications are merely examples and the present disclosure is not limited thereto. Accurately understanding search intent plays a key role in providing users with the information they desire more accurately and quickly.

In one embodiment, the external server (130) may be a search engine. A search engine is a software system that searches the web for relevant information in response to a search query (e.g., keyword) entered by a user. As one function of the search engine, the search engine may search a database storing various search queries (or keywords) to obtain related keywords that begin with or include the entered keyword. For example, when "iPhone price" is entered, the search engine may automatically generate related keywords such as "iPhone price" or "iPhone force shutdown," which are auto-complete search terms that begin with or include the entered keyword.

The server (110), terminal (120), and external server (130) can communicate with each other through a network (150). The network (150) refers to a connection structure that enables information exchange between each node, such as devices, terminals, and servers. Examples of such a network (150) include, but are not limited to, a 3GPP (3rd Generation Partnership Project) network, an LTE (Long Term Evolution) network, a 5G network, a WIMAX (World Interoperability for Microwave Access) network, the Internet, a LAN (Local Area Network), a Wireless LAN (Wireless Local Area Network), a WAN (Wide Area Network), a PAN (Personal Area Network), a Wi-Fi network, a Bluetooth network, a satellite broadcasting network, an analog broadcasting network, and a DMB (Digital Multimedia Broadcasting) network.

FIG. 2 is a diagram illustrating a method for determining search intent using search result information according to one embodiment of the present disclosure.

In one embodiment, the server (110) may obtain search result information (220) related to content corresponding to the search query based on the search query (210).

A search query (210) refers to text or other input formats entered by a user to find desired information through a search engine. Users enter various queries, such as keywords, sentences, and questions, to satisfy their specific information needs. These queries serve as crucial inputs for the search engine to provide appropriate search results. A search query may be a single word, or it may be a sentence or phrase composed of multiple words. The search engine analyzes the entered query to determine the context, meaning, and user intent, and then filters and sorts the search results based on this information before providing them to the user.

Search result information (220) is a collection of information provided by a search engine based on a search query (210), and is provided in various forms to enable users to access desired information. The search engine analyzes the entered search query and provides search results such as web pages, images, videos, and map information that are highly relevant to the search query. In this process, the search results are sorted and filtered by considering keywords matching the search query, page quality, and search intent. In addition, the search result information is composed of a summarized preview, title, and URL (Uniform Resource Locator) address, thereby helping users quickly identify the information they want.

Search result information (220) may include at least one of the name of the website that posted the content, the title of the content, a summary of the content, or the URL address of the website. The website name is text referring to the website. The title of the content may be the title of content (e.g., a post, a video, etc.) posted on the website. The summary of the content may be a summary of the content posted on the website. For example, if the content is a post, the summary may include the summarized text of the post. The URL address of the website may be a link for accessing the website.

For example, in response to a search query (210), multiple contents may be displayed as search results on a search results page. For example, contents A (230) and contents B (240) may be displayed as search results on a search results page. The search result information of contents A (230) may include at least one of the name of the website on which contents A is posted (231), the title of contents A (232), the summary information of contents A (233), or the URL address of the website (234). As another example, the search result information of contents B (240) may include at least one of the name of the website on which contents B is posted (241), the title of contents B (242), the summary information of contents B (243), or the URL address of the website (244). The contents A and B disclosed in FIG. 2 are merely examples, and more contents may be displayed as search results on a search results page, or fewer contents may be displayed as search results on a search results page.

In one embodiment, the server (110) may determine at least one of the content type and the search result type based on the search result information (220). The content type may be a classification system that considers the characteristics of the content. The content type may be predetermined and stored on the server (110) or in an external database.

The content type may be at least one of the following: e-commerce webpage type, news-related content type, community webpage type, blog type, video content type, image content type, or file posted on a webpage type. If it does not fall under the above types, it may fall under other types. If the content type is e-commerce webpage type, the search intent may be the intention to obtain information about a specific product. If the content type is blog type, the search intent may be the intention to obtain information about a specific product, the intention to obtain an answer to a question, etc. If the content type is video content type or image content type, the search intent may be the intention to obtain information about a specific product, the intention to obtain an answer to a question, etc. The file posted on a webpage type means that a file (e.g., a PDF file, a Word file, etc.) is displayed on the search results page, and the file may be downloaded when the URL address of the content is selected. When a file is posted on a webpage, the intent may be to seek an answer to a query, to complete a specific task, etc. As described above, while a certain level of search intent can be determined based on the content type, the exact search intent can be difficult to determine. Therefore, the server (110) can select at least some content from among the multiple contents included in the search results page and comprehensively consider the content item types of the selected contents to determine the search intent.

In one embodiment, the server (110) can determine the content type based on search result information. The server (110) can determine the content type based on at least one of the protocol, domain, path, query parameter, file name, or fragment included in the URL address, which is one of the search result information. The protocol is located at the beginning of the URL and specifies the communication method for accessing the resource. Commonly used protocols include http and https, which allow web browsers to securely communicate with the server. The domain follows the protocol in the URL and indicates the name of the server where the resource is located. The domain is used as a human-readable name to replace the server's IP address; for example, example.com can be a domain. The path follows the domain and indicates the directory path that specifies the location of a specific resource within the server. For example, a path such as /products/item1 navigates to a specific page or file within the domain. The query parameter begins with a question mark (?) after the path and consists of key-value pairs that transmit additional information to the server. For example, ?id=123&category=books sends the id and category parameters to the server. The filename is the part of the path that refers to a specific file, usually including the file extension. For example, index.html refers to an HTML file called index. A fragment begins with # at the end of the URL and acts as an index to a specific location within a web page. For example, #section1 would jump directly to section1 within the page.

For example, the server (110) can determine the content type (250) of content A (230) or determine the content type (260) of content B (240).

In one embodiment, the server (110) may determine the type of content as being posted on an e-commerce webpage based on a determination that the domain corresponds to an e-commerce domain, the path includes e-commerce-related keywords, or the URL address corresponds to a URL address displayed in an advertising area (e.g., a sponsored link displayed on a search results page). For example, the server (110) may analyze the URL address (234) of content A (230) and determine the type of content as being posted on an e-commerce webpage if the domain or path includes keywords such as “shop,” “store,” or “products,” the query parameter includes “item,” or “price,” or if the URL is included in an advertising area.

In one embodiment, the server (110) may determine the type of content as news-related content based on whether the domain corresponds to a news domain or whether the path contains news-related keywords. For example, the server (110) may analyze the URL address (234) of content A (230) and determine the type of content as news-related content if the domain belongs to a pre-collected list of news websites or if the domain and/or path name contains keywords such as "news" or "article."

(Example: https://www.newspim.com/news/view/202409079 )

In one embodiment, the server (110) may determine the type of content as a blog-published content type based on whether the domain corresponds to a blog domain or whether the path includes blog-related keywords. For example, the server (110) may analyze the URL address (244) of content B (240) and determine the type of content as a blog-published content type if the domain belongs to a pre-collected blog domain list or if the domain or path name includes keywords such as "blog" or "post."

In one embodiment, the server (110) may determine the type of content as a video content type based on a determination that the domain corresponds to a platform domain for uploading videos. For example, the server (110) may analyze the URL address (244) of content B (240) and determine the content type as a video content type if the domain belongs to a pre-collected list of video domains (e.g., YouTube).

In one embodiment, the server (110) may determine the type of content as a type of file posted on a webpage based on a determination that a file name is included in the URL address. For example, the server (110) may analyze the URL address (244) of content B (240) and determine the type of content as a type of file posted on a webpage if the file name included in the URL address has an extension such as “.pdf”, “.doc”, etc.

The type of search result is determined based on the search result information and may be the type of content displayed on the search result page. To make it easy for users to understand the search results at a glance, the format in which the content is displayed may vary greatly depending on the purpose. The type of search result may include at least one of the following: a first type that displays the title and summary information of the content, a second type that displays in the form of an advertisement, a third type that displays in the form of news, a fourth type that displays in the form of information provision (e.g., encyclopedia, Wikipedia, etc.), a fifth type that displays video content or image content, a sixth type that displays a map, a seventh type that displays application installation, an eighth type that displays URL links for website shortcuts, or a ninth type that displays a community (e.g., an online community). If it does not fall into any of the above types, it may be classified as a "other" type.

In one embodiment, the server (110) may determine the type of search result based on the search result information (220). For example, the server (110) may determine the type (251) of the search result of content A based on the search result information of content A (230). For another example, the server (110) may determine the type (261) of the search result of content B based on the search result information of content B (240).

For example, if both the title and summary information of content A (230) are displayed on the search results page, the server (110) may determine the type of search results as the first type, which is a general search results display type. The first type may be the most general search result format.

In one embodiment, the server (110) may infer the search intent based on the type of search result. For example, if the type of search result is the fourth type, the server (110) may determine the search intent as an intent to seek an answer to a query or to complete a specific task.

In one embodiment, the server (110) may generate a prompt (270) based on at least one of a search query, search result information, a type of content, or a type of search result. The server (110) may determine a type of content or a type of search result corresponding to each of a plurality of contents corresponding to the search query. The server (110) may determine at least a portion of content from among the plurality of contents to be used for generating the prompt (270) based on the type of search result. The server (110) may generate the prompt (270) based on at least one of a name of a website, a title of the content, summary information of the content, a URL address of the website, a type of content, or a type of search result corresponding to each of the determined at least a portion of the content (e.g., the plurality of contents). For example, the server (110) may generate a prompt (270) based on at least one of the name of the website on which content A is posted (231), the title of content A (232), the summary information of content A (233), the URL address of the website (234), the type of content A (250), the type of search result of content A (251), the name of the website on which content B is posted (241), the title of content B (242), the summary information of content B (243), the URL address of the website (244), the type of content B (260), or the type of search result of content B (261).

In one embodiment, the server (110) may exclude content not to be used in generating the prompt (270) based on the type of search result. For example, the server (110) may determine at least some content from among the plurality of contents displayed on the search results page to be used in generating the prompt (270). This is because, among the types of search results, there may be content that is not important for understanding the search intent and may actually reduce the prediction accuracy.

In one embodiment, the server (110) may exclude content whose search result type is the second or fifth type from among multiple contents. Content displayed as advertisements may not be a critical factor in determining search intent, as search engines display advertisements on search results pages regardless of the user's search intent. Furthermore, content that only displays videos or images may be difficult or even impede the ability to determine search intent, as it lacks a title or summary.

In one embodiment, the server (110) may generate a prompt (270) based on at least one of the types of the content or the types of the search results of each of the remaining contents, excluding the contents whose search result types are the second type or the fifth type. For example, if the types of the search results of contents A (230) and B (240) are the first type, the type of the search result of content C (not shown) is the eighth type, the type of the search result of content D (not shown) is the ninth type, and the type of the search result of content E (not shown) is the second type, the server (110) may generate a prompt (270) for determining the search intent based on contents A, B, C, and D, excluding content E.

In one embodiment, the server (110) may generate a prompt (270) including a query that determines a search intent based on at least one of a search query, search result information, a type of content, or a type of search result, and a query that requests an answer explaining the basis for determining the search intent. For example, the prompt (270) may be, "Analyze the user's search intent based on the search result information according to the search query, and accurately determine whether the search query is for browsing information, moving to a specific website or page, researching a product or service, or making an actual purchase or transaction, and explain the reason." As another example, the prompt (270) may be, "In response to the search query 'long padding,' the title of content A displayed on the search results page is 'long padding,' the URL address is https://www.dddd.com, the summary information includes the price of the product, the type of search result is the first type, and the type of content is the type posted on an e-commerce webpage. In this case, analyze the user's search intent."

A search intent may include at least one of a first intent to seek an answer to a query, a second intent to find a specific website, a third intent to seek information about a specific product, or a fourth intent to complete a specific task.

The primary intent may be a user's intention to find information on a specific topic. For example, if the content is Wikipedia, a blog, a news article, or an academic resource; a tutorial or educational content starting with "How to"; a question answering service like Quora or Stack Overflow; content containing definitions and explanations of a specific concept or topic; or visual content supplementing explanations with videos and images, the server (110) may input a prompt (270) into the language model (280) to obtain the primary intent.

A secondary intent may be a user's intention to directly access a specific website or page. For example, if the search result type for content is a link to the official page of a brand or site, the exact URL address of the website the user is looking for is located at the top of the search results page, a specific section within the website (e.g., a login page, a customer support page) is displayed on the search results page, or a URL address containing the brand or site name is displayed, the server (110) may input a prompt (270) into the language model (280) to acquire the secondary intent.

A third intent may be a user's intention to research, compare, and find information to make a purchase decision about a product or service. For example, if the content includes product comparisons on review sites or blogs, user reviews, "Top 10" lists, "Recommended Products," prominent PPC advertisements or sponsored links, or content describing detailed product specifications or features, the server (110) can input the prompt (270) into the language model (280) to acquire the third intent.

The fourth intent may be the user's intention to actually purchase a product or perform a specific action. If the content is a product page (e.g., a product detail page) on an e-commerce website (e.g., Amazon, eBay, etc.), a direct link to a purchase page where the user can add to their cart or checkout immediately, content regarding a discount code or promotional link, content explaining payment and shipping methods, an advertising link encouraging product purchases, or a shopping advertisement, the server (110) may input the prompt (270) into the language model (280) to acquire the fourth intent.

A language model (280) is an AI model designed to learn the patterns and structure of language based on text data, thereby generating appropriate responses to given text input or predicting the next word. This model primarily learns from a vast amount of documents to understand word relationships, grammar, and context, and uses this to perform various natural language processing tasks such as text generation, translation, summarization, and question-answering. Examples of the language model (280) include GPT, Gemma, Claude, and LlaMa.

In one embodiment, the language model (280) is a pre-trained model that can determine search intents even in situations where there is no training data labeled with search intents. Since the pre-trained language model (280) is not supervised, it is not limited to a pre-determined search intent classification system and can identify and output various search intents. In other words, the language model (280) is not limited to the first through fourth search intents and can output various other search intents.

In another embodiment, the server (110) may generate training data using the generated prompt as input data and the search intent corresponding to the search query as a label. For example, the labels may be determined as the first to fourth intents. In another example, the labels may be determined as further refinements of the first to fourth intents. For example, the first intent may be further refined into the intent to find basic facts or an overview of a specific topic or person, the intent to find detailed professional or academic information, the intent to find how-tos, tutorials, and step-by-step guides, the intent to find multimedia content such as images, videos, and music, the intent to find social media content, forums, and community-related information, the intent to find information on recent issues or events, and the intent to find information related to personal interests such as health, finance, and hobbies. The server (110) may update the language model (280) based on the training data. Through this, the language model (280) may more accurately identify the search intent.

In one embodiment, the server (110) may obtain a search intent (281) and/or a basis (282) from the language model (280). The basis (282) may refer to the basis for determining the search intent (281). For example, the basis (282) for determining the search intent (281) as the first intent may also be output. Using the basis (282), a user can verify whether the output of the language model (280) is correct.

FIG. 3 is a diagram illustrating content of which the type of search result is the second type or the fifth type according to one embodiment of the present disclosure.

On the search results page, content A (310) may be a second type displayed in the form of an advertisement. For example, if the server (110) generates a prompt (270) based on the search result information of content A (310), the type of content, and the type of search result, and inputs the prompt (270) into the language model (280), the server (110) may obtain a fourth intent, which is the user's intent to purchase a product or perform a specific action.

Content B (320) may be a fifth type that displays video content or image content. For example, if the server (110) generates a prompt (270) based on the search result information of Content B (320), the type of content, and the type of search result, and inputs the prompt (270) into the language model (280), since Content B (320) displays a product sales image, the server (110) can obtain the fourth intent, which is the user's intent to purchase a product or perform a specific action.

FIG. 4 is a diagram illustrating content used to determine a prompt according to one embodiment of the present disclosure.

The type of search result of content C (410) may be the first type. The search result information of content C (410) may include the website name (411), URL address (412), title (413), and summary information (414). Since content C (410) includes both the title (413) and summary information (414), it may be determined to be the first type.

The search result type of content D (430) may be the fourth type, which is displayed in the form of information provision. Content D (430) may be the fourth type because it displays a description of a specific term (e.g., long padding).

The search result type of content E (450) may be a third type displayed in the form of news.

The search result type for Content F (460) may be the ninth type, indicating a community, because the title and summary indicate a question posted by a specific community.

In one embodiment, the server (110) may determine the prompt (270) based on the search result information, content type, and search result type of each of Content C (410), Content D (430), Content E (450), and Content F (460). Since multiple contents on a single search results page may have different search result types or different content types, it may be difficult to accurately determine the search intent based on a single content alone. Therefore, it may be more accurate to determine the search intent by comprehensively considering multiple contents. For example, Content C (410) includes an e-commerce webpage link, and Content F (460) appears to have an intent to seek information, but ultimately seeks information for purchasing padding. Therefore, when these are combined, the search intent may be the fourth intent.

FIG. 5 is a diagram illustrating the types of content and the types of search results according to one embodiment of the present disclosure.

On the search results page, content may be displayed as shown in the left screen (500). The information on the left screen (500) may be identical to the information on the right screen (501). The information on the right screen (501) may include not only search result information, but also the type of content and search result type determined based on the search result information.

For example, based on search result information of content G (510), the content type (511) and the search result type (512) can be determined. The basic type can be the first type.

For another example, based on the search result information of content H (530), the content type (531) and the type of search result (532) can be determined.

FIG. 6 is a flowchart illustrating a method by which an electronic device determines a search intent according to one embodiment of the present disclosure.

In one embodiment, the electronic device may obtain search result information related to content corresponding to the search query based on the search query (610).

In one embodiment, the electronic device may determine at least one of a type of content or a type of search result based on the search result information (620).

In one embodiment, the electronic device may generate a prompt based on at least one of a search query, search result information, type of content, or type of search result (630).

In one embodiment, the electronic device can input the generated prompt into a language model to determine the search intent of the search query (640).

FIG. 7 is a drawing illustrating an electronic device according to various embodiments of the present disclosure.

An electronic device (800) according to one embodiment may be a server or a user terminal (e.g., a mobile device, a desktop, a laptop, a personal computer, etc.). Referring to FIG. 7, an electronic device (800) according to one embodiment may include a user interface (810), a processor (830), a display (850), and a memory (870). The user interface (810), the processor (830), the display (850), and the memory (870) may be connected to each other via a communication bus (605).

A user interface (810) encompasses everything that enables interaction between humans and machines. It may enable a user to operate and control systems, software, applications, websites, etc. For example, a user interface may include a graphical user interface, a text-based interface, a voice user interface, a natural user interface (e.g., gestures, touch, etc.), etc.

The display (850) can display information generated by the processor (830).

The memory (870) can store generated information. In addition, the memory (870) can store various pieces of information generated during the processing of the processor (830) described above. In addition, the memory (870) can store various types of data and programs. The memory (870) can include volatile memory or non-volatile memory. The memory (870) can store various types of data by using a large-capacity storage medium such as a hard disk.

In addition, the processor (830) can perform at least one method or an algorithm corresponding to at least one method described above through FIGS. 1 to 6. The processor (830) may be a data processing device implemented as hardware having a circuit having a physical structure for executing desired operations. For example, the desired operations may include code or instructions included in a program. The processor may be configured as, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or an NPU (Neural Network Processing Unit). For example, an electronic device implemented as hardware may include a microprocessor, a central processing unit, a processor core, a multi-core processor, a multiprocessor, an ASIC (Application-Specific Integrated Circuit), or an FPGA (Field Programmable Gate Array).

The processor (830) can execute a program and control an electronic device. The program code executed by the processor (830) can be stored in memory.

Meanwhile, the embodiments disclosed in this specification may be implemented in the form of a recording medium that stores computer-executable instructions. The instructions may be stored in the form of program code, and when executed by a processor, may generate program modules to perform the operations of the disclosed embodiments. The recording medium may be implemented as a computer-readable recording medium. The computer-readable recording medium may include any type of recording medium that stores instructions that can be deciphered by a computer. Examples thereof include ROM, RAM, magnetic tape, magnetic disk, flash memory, and optical data storage devices.

The above-described embodiments are specific examples for implementing the present disclosure. The present disclosure will encompass not only the above-described embodiments, but also embodiments that can be simply designed or easily modified. Furthermore, the present disclosure will encompass techniques that can be easily modified and implemented using the above-described embodiments. Therefore, the scope of the present disclosure should not be limited to the above-described embodiments, but should be defined not only by the claims set forth below, but also by equivalents of the claims of the present disclosure.

Claims (14)

In a method for determining search intent using an electronic device,
A step of obtaining search result information related to content corresponding to a search query based on a search query;
A step of determining the type of the content and the type of the search result based on the search result information;
A step of generating a prompt based on the search query, the search result information, the type of the content, and the type of the search result; and
A step of inputting the generated prompt into a language model to determine the search intent of the search query;
Including,
The type of the search result is determined based on the search result information, and includes a first type that displays the title and summary information of the content, a second type that displays in the form of an advertisement, a third type that displays in the form of news, a fourth type that displays in the form of information provision, a fifth type that displays video content or image content, a sixth type that displays a map, a seventh type that displays application installation, an eighth type that displays a URL link for a shortcut to a website, and a ninth type that displays a community.
The step of determining the type of the content and the type of the search result includes the step of determining the type of the content and the type of the search result corresponding to each of the plurality of contents corresponding to the search query,
A step of excluding contents of the second type and the fifth type, which are types of search results that lower the prediction accuracy among the plurality of contents; and a step of generating the prompt based on the type of the content and the type of the search result of each of the remaining contents excluding the contents of the second type and the fifth type, and
A step of generating learning data using the generated prompt as input data and the search intent corresponding to the search query as a label; a step of updating the language model based on the learning data; and a step of outputting the basis for determining the search intent with the language model to verify whether the output of the language model is correct.
method.
In the first paragraph,
The above search result information is,
A method comprising at least one of the name of the website that posted the content, the title of the content, summary information of the content, or the URL (Uniform Resource Locator) address of the website.
In the first paragraph,
The type of the above content is:
A method of displaying at least one of the following types of content posted on an e-commerce webpage, a news-related content type, a community webpage, a blog, a video content type, an image content type, or a file posted on a webpage:
delete In the second paragraph,
The step of determining at least one of the type of the above content and the type of the search result is:
A method comprising the step of determining the type of the content based on at least one of a protocol, domain, path, query parameter, file name, or fragment included in the URL address.
In paragraph 5,
The step of determining at least one of the type of the above content and the type of the search result is:
A step of determining the type of the content as a news-related content type based on a determination that the domain corresponds to a news domain or that the path includes a news-related keyword;
A step of determining the type of the content as a type of content posted on a blog based on a determination that the domain corresponds to a blog domain or that the path includes a blog-related keyword;
A step of determining the type of the content as a video content type based on a determination that the above domain corresponds to a platform domain for uploading videos; or
A step of determining the type of the content as a type in which a file is posted on a web page based on a determination that the above URL address includes a file name.
A method comprising at least one of:
delete delete In the first paragraph,
The steps to create the above prompt are:
A step of generating the prompt based on at least one of the name of the website corresponding to each of the plurality of contents, the title of the contents, the summary information of the contents, the URL address of the website, the type of the contents, or the type of the search result.
A method comprising:
In the first paragraph,
The steps to create the above prompt are:
A step of generating the prompt so as to include a query that determines a search intent based on at least one of the search query, the search result information, the type of the content, or the type of the search result, and a query that requires an answer to the basis for determining the search intent.
A method comprising:
delete In the first paragraph,
The above search intent is,
A method comprising at least one of a first intent to seek an answer to a question, a second intent to find a specific website, a third intent to seek information about a specific product, or a fourth intent to complete a specific task.
In a computer-readable recording medium having recorded thereon a program for determining search intent,
The above program is,
Obtain search result information related to content corresponding to the search query based on the search query,
Determine the type of the content and the type of the search result based on the above search result information,
Generate a prompt based on the search query, the search result information, the type of the content, and the type of the search result,
Input the generated prompt into the language model to determine the search intent of the search query,
The type of the search result is determined based on the search result information, and includes a first type that displays the title and summary information of the content, a second type that displays in the form of an advertisement, a third type that displays in the form of news, a fourth type that displays in the form of information provision, a fifth type that displays video content or image content, a sixth type that displays a map, a seventh type that displays application installation, an eighth type that displays a URL link for a shortcut to a website, and a ninth type that displays a community.
In order to determine the type of the content and the type of the search result, the type of the content and the type of the search result corresponding to each of the plurality of contents corresponding to the search query are determined,
Among the plurality of contents, the contents of the second and fifth types, which have lower prediction accuracy, are excluded, and the prompt is generated based on the type of the content and the type of the search result of each of the remaining contents excluding the contents of the second and fifth types, and
The above generated prompt is used as input data, learning data is generated with the search intent corresponding to the search query as a label, the language model is updated based on the learning data, and the basis for determining the search intent by the language model is output to verify whether the output of the language model is correct.
Computer-readable recording medium.
Electronic devices
memory; and
One or more processors;
Including,
One or more of the above processors,
Obtain search result information related to content corresponding to the search query based on the search query,
Determine the type of the content and the type of the search result based on the above search result information,
Generate a prompt based on the search query, the search result information, the type of the content, and the type of the search result,
Input the generated prompt into the language model to determine the search intent of the search query,
The type of the search result is determined based on the search result information, and includes a first type that displays the title and summary information of the content, a second type that displays in the form of an advertisement, a third type that displays in the form of news, a fourth type that displays in the form of information provision, a fifth type that displays video content or image content, a sixth type that displays a map, a seventh type that displays application installation, an eighth type that displays a URL link for a shortcut to a website, and a ninth type that displays a community.
In order to determine the type of the content and the type of the search result, the type of the content and the type of the search result corresponding to each of the plurality of contents corresponding to the search query are determined,
Among the plurality of contents, the contents of the second and fifth types, which have lower prediction accuracy, are excluded, and the prompt is generated based on the type of the content and the type of the search result of each of the remaining contents excluding the contents of the second and fifth types, and
The above generated prompt is used as input data, learning data is generated with the search intent corresponding to the search query as a label, the language model is updated based on the learning data, and the basis for determining the search intent by the language model is output to verify whether the output of the language model is correct.
Electronic devices.