Generating narrative query responses from search-based autosuggestion queries using a generative language model.

Abstract

This disclosure describes a query gateway system that provides an efficient and flexible framework for providing context-preserving autosuggest queries from an autosuggest query system (e.g., a search engine query experience) to a generative language model system (e.g., an AI chat experience). For example, the query gateway system establishes a framework for leveraging the functionality and services of the autosuggest query system and automatically provides context-preserving queries to the generative language model system using a separate user interface that does not interrupt user navigation or require manual, redundant user input. Furthermore, the query gateway system incorporates additional enhancements, including an AI chat eligibility model and a query reformulation model, to improve the computational efficiency and accuracy of the AI ​​chat system.

Description

Background Art

[0001] Background

[0001] The ability to search websites across the vast expanse of the Internet has been significantly enhanced by search engine services. More recently, advanced chat services that utilize artificial intelligence (AI) known as generative language models (GLMs), including large language models (LLMs), have emerged. These GLMs utilize machine learning models to generate narrative-based responses based on user queries. These services typically operate independently, but certain service providers have begun incorporating links between these services. However, a framework for integrating context retention between services still does not exist. Furthermore, while a particular function can enhance one service, mechanisms for leveraging these advancements to improve other services are not yet in place.

[0002]

[0002] The following detailed description presents specific and detailed embodiments along with the respective drawings. Further, each of the figures listed below corresponds to one or more implementations considered in the present disclosure.

Brief Description of the Drawings

[0003] Brief Description of the Drawings [Figure 1]

[0003] An overview example of implementing a query gateway system that facilitates context-preserving auto-suggest queries from an auto-suggest query system to a generative language model system is shown. [Figure 2]

[0004] The system environment in which the query gateway system is implemented is shown. [Figure 3]

[0005] An example flowchart of utilizing an auto-suggest query using both an auto-suggest query system and a generative language model system is shown. [Figure 4A]

[0006] This example demonstrates a graphical user interface that automatically provides context-preserving autosuggest queries from the user interface of the autosuggest query system to the user interface of the generative language model system. [Figure 4B]

[0006] An example of a graphical user interface is shown that automatically provides context-preserving autosuggest queries from the user interface of the autosuggest query system to the user interface of the generative language model system. [Figure 5]

[0007] An exemplary block diagram is shown for providing a reformulated autosuggest query system to a generative language model system, based on an autosuggest query system. [Figure 6]

[0008] This example flowchart shows how to generate a reformulated autosuggest query from an autosuggest query. [Figure 7A]

[0009] This example demonstrates a graphical user interface that generates narrative-based responses using a generative language model from multiple autosuggestion queries. [Figure 7B]

[0009] An exemplary graphical user interface is shown that generates a narrative-based response using a generative language model from multiple autosuggest queries. [Figure 8]

[0010] This document illustrates an exemplary sequence of operations in a computer implementation method for generating narrative query responses using a generative language model. [Figure 9]

[0011] This shows the components included in an exemplary computer system. [Modes for carrying out the invention]

[0004] Detailed explanation

[0012] This disclosure describes a query gateway system that provides an efficient and flexible framework for delivering context-preserving autosuggest queries from an autosuggest query system to a generative language model system. Specifically, the query gateway system establishes a framework for leveraging the functions and services of the autosuggest query system, including relevant autosuggest queries, to enhance the queries provided to the generative language model system. Furthermore, the query gateway system incorporates additional enhancements, such as an AI chat eligibility model and a query reformulation model, to improve the computational efficiency and accuracy of the AI ​​chat system.

[0005]

[0013] In some implementations, the query gateway system combines the auto-suggestion functionality of a web-based search experience with an AI chat-based experience. For example, the query gateway system selectively generates and provides additional graphical elements within the auto-suggestion query interface (e.g., a web-based search experience) to initiate the transition to the generative language model system interface (e.g., an AI chat-based experience). Furthermore, when transitioning between the search and chat experiences, the query gateway system facilitates the preservation of the context of the auto-suggestion queries, often providing additional context to the queries provided to the AI ​​chat system. In addition, the query gateway system provides a flexible framework that minimizes navigation interruptions during transitions between search systems.

[0006]

[0014] More specifically, in some implementations, the query gateway system utilizes a generative language model to generate narrative query responses, such as AI chat responses. For example, when a text input (e.g., a text prefix) corresponding to a search query (e.g., a web search) is detected, the query gateway system identifies one or more autosuggest queries. Furthermore, in some cases, the query gateway system generates a generative language model (GLM) eligibility cache for the autosuggest queries. Then, using the generative language model eligibility cache, the query gateway system determines that the autosuggest query is eligible for a generative language model (e.g., an AI chat model), and based on this determination, provides generative language model elements for display next to the autosuggest query within the first user interface. Upon detecting the selection of a generative language model element, the query gateway system generates a reformulated autosuggest query from the original autosuggest query. Furthermore, the query gateway system provides the reformulated autosuggest queries to the generating language model and displays them in a second user interface separate from the first user interface.

[0007]

[0015] As described in this document, the query gateway system offers several significant technical advantages over existing systems in terms of computational efficiency, accuracy, and flexibility. Furthermore, the query gateway system enables several practical applications that address problems by providing a framework for retaining the context of one or more autosuggest queries from an autosuggest query system within the query, and this context is often reformulated and provided to a generative language model system, resulting in several advantages as further described below.

[0008]

[0016] For example, the query gateway system adds an AI chat element (i.e., a GLM element) to the autosuggest query pane of the search user interface for each eligible autosuggest query. Then, when the AI ​​chat element shown next to the first autosuggest query is selected, the query gateway system opens a separate AI chat user interface (i.e., a GLM interface) containing a first narrative-based response from GLM based on this first autosuggest query. When the system detects the selection of another AI chat element shown next to a second autosuggest query (since autosuggest queries are saved within the search user interface even when an AI chat element is selected), the query gateway system opens another separate AI chat user interface containing a second narrative-based response from GLM based on the second autosuggest query.

[0009]

[0017] By including a generative language model element (e.g., an AI chat button) in the user interface that triggers a transition to the GLM system, the query gateway system provides an improved user interface that significantly reduces the number of navigation steps currently required to switch from the autosuggest query system to the GLM system. In more detail, the existing system provides autosuggest queries in response to the user entering a text prefix, but these autosuggest queries only allow the user to perform a corresponding web search using the selected autosuggest query. If the user wants to switch to an AI chat-based experience, this user must navigate to the GLM user interface and manually re-enter the autosuggest query. In some cases, after a conventional web search has been performed, the existing system will offer the option to transfer the autosuggest query to the GLM system for further processing if selected. In contrast, the query gateway system allows the user to automatically transition from autosuggest queries to the GLM interface without requiring the user to manually navigate to the GLM interface or re-enter the autosuggest query.

[0010]

[0018] This advantage becomes even greater when multiple auto-suggest queries are selected to be provided to the GLM system. For example, in existing systems, if a user wants to receive responses from multiple auto-suggest queries from a text input prefix, this user may need to re-enter the text input prefix for each selected auto-suggest query, and then copy each selected auto-suggest query from the auto-suggest system to the GLM system. Each time, the user has to manually switch between each system interface, often reloading the system's user interface and re-entering the text input prefix and / or auto-suggest queries. In contrast, the query gateway system allows the user to enter the text input prefix once and select AI chat elements corresponding to multiple auto-suggest queries provided without having to leave or reload the auto-suggest query system interface. Furthermore, the query gateway system allows the user to visit different instances of the GLM system corresponding to each of the selected AI chat elements, and the corresponding auto-suggest queries are automatically loaded into the GLM system.

[0011]

[0019] By utilizing generative language model elements alongside autosuggested queries, the query gateway system also improves overall accuracy by preserving the context of the autosuggested queries. For example, the query gateway system automatically provides autosuggested queries to the GLM system without requiring the user to manually select, copy, and / or re-enter the autosuggested queries from the autosuggested query search system to the GLM system. In this way, the query gateway system ensures that the context is accurately preserved during the transition.

[0012]

[0020] Furthermore, the navigation advantages provided by generative language model elements become even more significant when computing devices have relatively small display devices. For example, devices with relatively small display devices can make navigating between interfaces such as browser tabs very difficult. Add to this users who need to manually copy and paste or re-enter autosuggest queries, and navigating between multiple interfaces becomes even more difficult. Instead, a query gateway system provides generative language model elements (e.g., AI chat elements) that users can easily select for one or more autosuggest queries, and this query gateway system automatically transfers and loads the corresponding autosuggest queries into the GLM system in a new or separate interface.

[0013]

[0021] Furthermore, in many cases, query gateway systems achieve even greater accuracy by utilizing reformulated autosuggest queries. For example, in these implementations, the query gateway system uses autosuggest queries to generate a new, reformulated autosuggest query that is better suited to the input parameters of the GLM system. In this way, this reformulated autosuggest query allows the GLM system to produce a better and more accurate response than an autosuggest query that is not sufficiently suited to the GLM system.

[0014]

[0022] Furthermore, by utilizing the eligibility cache or reformulated query cache of the generative language model, the query gateway system reduces the number of computational operations that need to be performed. For example, in some implementations, when an autosuggestion query is selected, the query gateway system determines whether this autosuggestion query is associated with a reformulated autosuggestion query. If a reformulated autosuggestion query exists in the eligibility cache of the generative language model, the query gateway system does not need to reprocess the autosuggestion query to generate a new reformulated autosuggestion query. If it is necessary to generate a new reformulated autosuggestion query, the query gateway system uses a lightweight language model to determine the reformulated autosuggestion query in real time. In addition, the query gateway system stores the newly generated reformulated autosuggestion query in the eligibility cache of the generative language model to prevent the regeneration of the reformulated autosuggestion query in the future.

[0015]

[0023] As a further example, in some implementations, the query gateway system allows for the simultaneous selection of multiple autosuggest queries. In these implementations, the query gateway system reformulates multiple autosuggest queries into a single robust AI chat query and provides it to the GLM system. In this way, the query gateway system provides a much more flexible approach than the current system, which requires the user to manually formulate the AI ​​chat query. Furthermore, the query gateway system allows the user to receive narrative-based responses from the GLM system with relatively fewer navigation steps. In addition, by generating an autosuggest query reformulated from multiple autosuggest queries, the narrative-based responses generated by the GLM system become more accurate.

[0016]

[0024] This disclosure uses several terms to describe the features and advantages of one or more implementations. For example, the term "autosuggest query search system" refers to a search engine system or another type of search system that provides a user with contextually relevant proposed search queries, called "autosuggest queries". For example, an autosuggest query search system utilizes a text input field to detect text input corresponding to a search query. "Text input" is detected in the form of typed characters, phrases, and / or prefixes when the text input is received. In response, the autosuggest query search system provides a plurality of autosuggest queries as input to the search query.

[0017]

[0025] The term "generative language model system" (GLM system) refers to an advanced computing system that uses natural language processing and machine learning to generate human-like text that is coherent and contextually relevant, referred to in this document as "narrative-based query response" or "narrative query response". An example of such a model is a large language model (LLM), which has been trained on a vast dataset and can generate fluent, coherent, topic-specific text. GLMs have applications in natural language understanding, content generation, text summarization, dialogue systems, language translation, and creative writing assistance. In some cases, a GLM system is referred to as an AI chat system. In some cases, a GLM is referred to as an AI chat model, an AI chat service, or simply an AI chat.

[0018]

[0026] In this document, a "Generative Language Model element" (GLM element) refers to a graphical user interface element connected to an auto-suggest query. In some cases, a GLM element is referred to as an AI chat element or an AI chat button. In various implementations, one or more GLM elements are displayed in the user interface of an auto-suggest query search system, such as an auto-suggest pain that indicates one or more auto-suggest queries for text input. When a GLM element modified for a given auto-suggest query is selected, as described below, the query gateway system may trigger a transition to the GLM system based on the given auto-suggest query.

[0019]

[0027] Additional details regarding an exemplary implementation of a query gateway system (i.e., an "auto-suggest query to GLM gateway system") are considered in relation to each of the attached figures. For example, FIG. 1 shows a schematic example implementing a query gateway system that delivers a context-preserving auto-suggest query from an auto-suggest query system to a generative language model system according to one or more implementations.

[0020]

[0028] As shown in FIG. 1, a series of operations 100 shows an example of a query gateway system that provides a framework for delivering a context-preserving auto-suggest query from an auto-suggest query system to a generative language model system. In various implementations, the query gateway system executes this series of operations 100. In some implementations, an auto-suggest query search system and / or a generative language model (GLM) system associated with the query gateway system executes one or more of each operation, or a part of each operation.

[0021]

[0029] Furthermore, this series of operations 100 includes an operation 102 that identifies an autosuggest query 114 based on a text input 110 for query searching. For example, in response to a user providing a text input 110, the query suggestion service determines an autosuggest query 114 from a set of autosuggest queries 112. In various implementations, this query suggestion service is part of an autosuggest query search system.

[0022]

[0030] Figure 1 also shows a first user interface 116 related to operation 102. This first user interface 116 includes a text input field for receiving text input 110 and an autosuggest pane for displaying autosuggest queries 114. In various implementations, the autosuggest query search system is displayed, such as as part of a search engine website or product search user interface. Additional details regarding generating autosuggest queries from text input are shown in relation to Figures 3 and 5.

[0023]

[0031] In Figure 1, this sequence of operations 100 includes an operation 104 that generates and displays AI chat elements for eligible autosuggest queries. For example, when an autosuggest query 114 is determined in text input 110, the query gateway system provides one or more of the autosuggest queries 114 to an AI chat eligibility model (i.e., a GLM eligibility model) having a generative language model cache 120 to determine whether a given autosuggest query is eligible to be provided to the GLM system. For eligible autosuggest queries, the query gateway system generates and provides generative language model elements 122 (GLM elements), such as AI chat elements, which are displayed with the autosuggest query in the first user interface 116. Additional details regarding determining GLM eligibility are shown in relation to Figure 5.

[0024]

[0032] As shown in the figure, this series of operations 100 includes an operation 106 that generates a reformulated autosuggestion query when it detects a selection of an AI chat element in the autosuggestion query. For example, in response to detecting a selection of a generative language model element 122 that appears with the corresponding autosuggestion query, the query gateway system sends the autosuggestion query 124 to the GLM system for processing.

[0025]

[0033] In some implementations, before providing the autosuggest query 124 to the GLM system, the query gateway system determines a reformulated autosuggest query 126 in the autosuggest query 124. For example, the query gateway system uses a query reformulation model to identify and / or generate a reformulated autosuggest query 126 in the autosuggest query 124. Generally, the reformulated autosuggest query is more comprehensive and detailed (e.g., more robust and detailed), leading to a better response from the GLM system. Additional details regarding determining GLM eligibility are shown in relation to Figures 3, 5, and 6.

[0026]

[0034] As shown in the figure, this series of operations 100 includes an operation 108 that provides the reformulated autosuggest query 126 to the AI ​​chat system in a separate user interface. For example, once the reformulated autosuggest query 126 is determined, the query gateway system delivers it to the GLM system, which then guides or causes the GLM system to begin processing the reformulated autosuggest query 126 as a narrative-based query using the generative language model 130.

[0027]

[0035] Furthermore, as described above, in various embodiments, the query gateway system generates a new user interface, such as a second user interface 132, to correspond to the GLM system. For example, if the first user interface 116 is displayed in the first window or browser tab, the query gateway system creates or generates a new window or browser tab to display the AI ​​chat system. This ensures that the first user interface 116 and the user's search experience remain uninterrupted when the AI ​​chat system is triggered to provide a narrative query response 134 for the corresponding autosuggest query. Additional details regarding triggering the GLM system are shown below in relation to Figures 3, 4A-4B, 5, and 7A-7B.

[0028]

[0036] While an overview of the query gateway system has been provided, additional details regarding the components and elements of the query gateway system (i.e., the "AutoSuggest Query-to-GLM Gateway System") are provided. For example, Figure 2 shows an exemplary system environment in which the query gateway system is implemented in one or more implementation forms. Figure 2 shows an exemplary deployment and configuration of the query gateway system, but other deployments and configurations are also possible.

[0029]

[0037] As shown in the figure, Figure 2 includes a computing environment 200 of a computing system having a server device 202 and a client device 230, which are connected by a network 240. Further details regarding the other computing devices are shown below in relation to Figure 9. Furthermore, Figure 9 also shows additional details regarding the network, such as the network 240 shown in the figure.

[0030]

[0038] As shown in the figure, the server device 202 includes a content management system 204 that manages the digital content hosted and / or accessed by the server device 202. For example, this content management system 204 facilitates users to search for digital content from websites, databases, or data storage across the internet.

[0031]

[0039] As shown in the diagram, the content management system 204 on the server device 202 includes a query gateway system 206. In some implementations, the query gateway system 206 is located outside the content management system 204. In various implementations, the various parts of the query gateway system 206 are arranged across various components.

[0032]

[0040] As described above, the query gateway system 206 implements a framework that acts as a context-preserving gateway between the auto-suggest query search system and the GLM system (e.g., an AI chat system). In numerous implementations, the query gateway system 206 implements one or more functions, such as AI chat eligibility and reformulated auto-suggest queries, to improve efficiency and accuracy when facilitating transitions between search systems. Furthermore, the query gateway system 206 utilizes a separate user interface to streamline navigation and eliminate confusion when transitioning between search systems. In addition, the content management system 204 includes the auto-suggest query system 226 and the GLM system 228, which may be located outside the content management system 204 and / or on a different computing device than the server device 202 (for example, the GLM system 228 may be located on a different cloud computing system).

[0033]

[0041] As shown in the figure, the query gateway system 206 includes various components and elements implemented in hardware and / or software. For example, the query gateway system 206 includes an autosuggest query manager 210 that communicates with the autosuggest query system 226 to generate, identify, refine, and determine autosuggest queries for text input corresponding to user search queries. Furthermore, the query gateway system 206 includes a query eligibility manager 212 that identifies, generates, and / or utilizes GLM-eligible queries 220 to determine when an autosuggest query is eligible to work successfully in the GLM system 228. Furthermore, the query gateway system 206 includes a query reformulation manager 214 for determining a reformulated query 222 (e.g., a reformulated autosuggest query) for the autosuggest query to be provided to the GLM system 228. Furthermore, the query gateway system 206 includes a GLM manager 216 that communicates with the GLM system 228 to facilitate the generation of narrative queries (e.g., AI chat responses) that utilize the GLM model.

[0034]

[0042] Furthermore, the query gateway system 206 includes a storage manager 218. In various implementations, the storage manager 218 stores data corresponding to the query gateway system 206. As shown in the figure, the storage manager 218 includes GLM-eligible queries 220, reformulated queries 222, and machine learning models 224. The machine learning models 224 may include some or all of the models corresponding to the GLM eligibility model (e.g., a classifier model), the query reformulation model (sequence-to-sequence model or lightweight language model), and / or the autosuggest query system 226 (e.g., an autosuggest query generation model) and / or the GLM system 228 (e.g., LLM).

[0035]

[0043] Furthermore, the computing environment 200 includes a client device 230 having a client application 232. In various implementations, this client device 230 is associated with a user who wants to input text and / or receive more robust narrative-based query responses from the AI ​​chat system. In many implementations, the user interacts with a server device 202 (e.g., a content management system 204 and / or a query gateway system 206) to access content and / or services. The computing environment 200 can include any number of client devices.

[0036]

[0044] As shown in the diagram, the client device 230 includes a client application 232. For example, the client application 232 is a web browser application, a mobile application, or another type of application that accesses and receives digital content by accessing internet-based content. In some implementations, the client device 230 includes a plug-in associated with a query gateway system 206 that communicates with the client application 232 and performs corresponding operations. In some implementations, a portion of the query gateway system 206 is integrated with the client application 232 to perform corresponding operations.

[0037]

[0045] Now that the foundation of the query gateway system 206 is in place, additional details regarding the various functions of the query gateway system 206 will be described below. As mentioned above, Figure 3 shows additional details of the query gateway system 206. Specifically, Figure 3 shows illustrative diagrams of how to utilize autosuggest queries using both the autosuggest query system and the generative language model system in several implementation forms.

[0038]

[0046] As shown in the figure, Figure 3 includes a flow 300 that provides an initial overview of providing query results to the user from either an auto-suggest query search system or a GLM query system (i.e., a GLM system). For example, this flow 300 includes an operation 302 that detects a prefix from a text input. For example, the user is provided with a first user interface that includes a text input field for entering characters of text into a search query. In this document, “prefix” refers to a partial text input provided by the user before any suggestions or completions are displayed.

[0039]

[0047] In response to detecting a prefix, the query suggestion service 304 (for example, part of the AutoSuggest Query System) searches for a set of AutoSuggest queries 112 to determine a relevant or matching AutoSuggest query. For example, flow 300 includes an action 306 that displays the relevant AutoSuggest query in the AutoSuggest pane. For example, once an AutoSuggest query is determined, the first user interface updates to display the AutoSuggest pane below the text input field containing the relevant AutoSuggest query.

[0040]

[0048] At this point, flow 300 may branch into two paths. In the first path corresponding to the search result 310, flow 300 includes an action 312 in which an autosuggest query is selected. For example, the user selects one of the autosuggest queries. Accordingly, the autosuggest query system updates the first user interface to display the search result corresponding to this selection, as shown in action 314.

[0041]

[0049] In the second path corresponding to the AI ​​chat narrative result 320, this flow includes the action 322 of selecting an AI chat element adjacent to the autosuggest query. As previously stated, the query gateway system 206 generates and provides AI chat elements (i.e., GLM elements) for one or more of the autosuggest queries in the first user interface. The query gateway system 206 then detects when the user selects one of the AI ​​chat elements.

[0042]

[0050] As shown in the figure, flow 300 includes a query reformulation model 324. In various implementations, when one of the AI ​​chat elements is selected, the query reformulation model 324 generates a reformulated autosuggestion query based on the autosuggestion query corresponding to the selected AI chat element. The query gateway system 206 then provides the reformulated autosuggestion query to the AI ​​chat system (i.e., the GLM system).

[0043]

[0051] Furthermore, as shown in the figure, flow 300 includes operation 326 to open a second user interface of the AI ​​chat system using a reformulated autosuggest query. For example, in relation to providing a reformulated autosuggest query to the AI ​​chat system, query gateway system 206 causes the AI ​​chat system to execute the reformulated autosuggest query using an AI chat model (e.g., GLM) and generate a narrative query response. Thus, when the user moves to the second user interface, the narrative query response is either waiting for the user or in the process of being generated.

[0044]

[0052] Figures 4A-4B show a graphical example of this process. Specifically, Figures 4A-4B show an exemplary graphical user interface that automatically provides context-preserving autosuggest queries from the user interface of the autosuggest query system to the user interface of the generative language model system. As shown in the figures, Figures 4A-4B include a client device 400 equipped with a display device (e.g., a laptop, desktop, smartphone, or tablet). The client device 400 can run an operating system (OS) that implements various systems, programs, applications, and services. For example, the client device 400 implements a client application 401 such as a web browser.

[0045]

[0053] In Figure 4A, the client application 401 includes a first user interface 402, shown as a browser tab. This first user interface corresponds to a search query page and includes an input field 403, from which the user can provide text input or another type of input (e.g., audio or image) for the search query. As shown in the figure, this input field 403 includes a text input 404 (e.g., a prefix). Also as shown in the figure, the first user interface 402 includes an autosuggestion pane 405 that displays an autosuggestion query 406 corresponding to the text input 404.

[0046]

[0054] Furthermore, the first user interface 402 displays an AI chat element 408 next to each of the auto-suggest queries 406. In various implementations, one or more of the auto-suggest queries 406 do not include an AI chat element, as described below. In some implementations, the first user interface 402 includes a popup, additional text, or other guidance indicating the purpose and function of the AI ​​chat element 408. For example, the first user interface 402 includes a tutorial explaining that the AI ​​chat element 408 will open a new, separate user interface where the corresponding auto-suggest query will be used as the initial query in the AI ​​chat system.

[0047]

[0055] As shown in the figure, Figure 4A involves using a pointer (e.g., mouse or finger) to select one of the auto-suggestion queries 406. For example, this selected auto-suggestion query 410 is highlighted with a different graphic in its AI chat element once selected.

[0048]

[0056] Upon detecting the selection of an auto-suggest query 410, the query gateway system 206 triggers a transition to the AI ​​chat system. As previously mentioned, in some implementations, the query gateway system 206 reformulates the auto-suggest query before providing it to the AI ​​chat system. In one or more implementations, the query gateway system 206 directly provides the selected auto-suggest query 410 to the AI ​​chat system, processing and / or providing the reformulation command.

[0049]

[0057] Moving on to Figure 4B, this figure shows a client device 400 having a client application 401 that then displays the second user interface 422. Note that the first user interface 402 is still present, but temporarily in the background. The second user interface 422 corresponds to the AI ​​chat system (i.e., the GLM system), providing the user with a narrative-based experience and resulting in more comprehensive query responses.

[0050]

[0058] The second user interface 422 includes a reformulated autosuggest query 424. Specifically, the selected AI chat element corresponded to the autosuggest query "Vietnam itinerary". Using this as a prompt, the query gateway system 206 generates a more robust and comprehensive query formulated into a single sentence, which is used as the input query to the AI ​​chat model. In some implementations, the query gateway system 206 provides the corresponding autosuggest query to the AI ​​chat model with or without additional text input.

[0051]

[0059] Accordingly, the AI ​​chat model generates a narrative query response 426 and provides it to the user. This narrative query response 426 serves as the starting point for the conversation with the user and provides more information about a given topic in the second user interface 422. Furthermore, as shown in the figure, the AI ​​chat system provides additional prompts to move the conversation forward.

[0052]

[0060] As described above, when the selected auto-suggest query 410 is triggered and the second user interface 422 is generated, the client application 401 retains the first user interface 402. After this selected auto-suggest query 410 is selected, the user can easily navigate to the second user interface 422, which includes an automatically generated version of the narrative query response 426. Furthermore, the user can select another AI chat element within the auto-suggest query 406, which allows the query gateway system 206 to generate a third user interface containing another instance of the AI ​​chat system, accurately and automatically applying the context of the newly selected AI chat element.

[0053]

[0061] Furthermore, for each selected AI chat element, the query gateway system 206 may trigger a new instance of the AI ​​chat system, along with a context-preserving query in the new user interface. In these implementations, the query gateway system 206 provides the user with quick access to each target autosuggest query with minimal user effort.

[0054]

[0062] Furthermore, when the reformulated autosuggest query 424 is displayed, the user can quickly navigate back to the first user interface 402, which holds the previously provided text input 404 and autosuggest query 406. This allows the user to then proceed to the search results or select another AI chat element to interact with the AI ​​chat system. In practice, the query gateway system 206 facilitates simple, efficient, accurate, and rapid navigation between the autosuggest query system and instances of the GLM system with minimal effort required from the user, reducing the potential for user errors that often occur when navigating under existing systems.

[0055]

[0063] Figure 5 shows an exemplary block diagram for providing reformulated autosuggest queries, based on an autosuggest query system, to generative language model systems in several implementation forms. Specifically, Figure 5 shows a framework 500 used by a query gateway system 206 to facilitate a smooth gateway between the autosuggest query system and a GLM system (e.g., an AI chat system). For example, this framework 500 includes an offline framework 502 and an online framework 504.

[0056]

[0064] In various implementations, the offline framework 502 enables the query gateway system 206 and / or the autosuggest query system to perform functions and operations on a periodic schedule or non-real-time, such as when a threshold is met. For example, the offline framework 502 includes a query log 506. In various implementations, this query log 506 may contain past search queries performed by the user and may include metadata associated with the searches. Furthermore, the offline framework 502 shows an autosuggest data generation pipeline 508 that generates suggestions 510 (e.g., autosuggest queries) from past text inputs in the query log 506.

[0057]

[0065] In various implementations, proposal 510 is stored in an autosuggest query cache or another type of data storage medium. Furthermore, proposal 510 is stored in various data structures, such as prefix-based tries, reverse indexes, or other data structures that pair text prefixes and corresponding metadata with autosuggest queries.

[0058]

[0066] Furthermore, the offline framework 502 includes a classifier model 512. In various implementations, this classifier model 512 utilizes a machine learning algorithm trained to assign input queries to specific categories or classes. In this case, the classifier model 512 is trained to classify whether an autosuggest query is suitable for the AI ​​chat model. More specifically, not all autosuggest queries are suitable for the AI ​​chat system. For example, autosuggest queries such as "Bing" or "Microsoft Outlook web email login" may not be suitable to be provided to the AI ​​chat system when the search user wants quick access to the requested resource and does not want to find further information about it.

[0059]

[0067] In various implementations, the query gateway system 206 can train a classifier model 512 to classify suggestions 510 using the text inputs in the query log 506. In some cases, each suggestion 510 is assigned a binary value indicating whether it is suitable for the AI ​​chat system. In some implementations, the classifier model 512 determines the probability that an autosuggest query (e.g., a suggestion) is suitable for the AI ​​chat system. As mentioned earlier, the query gateway system 206 can periodically update the suggestion classification when the classifier model 512 is updated. Again, as shown in the figure, the classifier model 512 stores the classified suggestions in the chat eligibility cache 522 (i.e., the eligibility cache for the generative language model), which will be discussed further below.

[0060]

[0068] Similarly, the query gateway system 206 can train a sequence-to-sequence model 514 (Seq2Seq model) to generate autosuggest queries (e.g., autosuggest queries) reformulated from proposal 510. In various implementations, this sequence-to-sequence model is a deep learning machine learning language model that maps input sequences to output sequences through a coding and decoding process. The sequence-to-sequence model 514 can represent various types of language models, such as large-scale language models (LLMs).

[0061]

[0069] The sequence-to-sequence model 514 receives an autosuggest query and generates a reformulated autosuggest query. In many implementations, the reformulated autosuggest query is more detailed, robust, and longer than the corresponding autosuggest query. For example, the autosuggest query may be a string of 3 to 5 words, a string of nouns, and / or an incomplete sentence. In contrast, the reformulated autosuggest query includes two to three complete sentences, a query framework, a scenario-based statement, a longer-formulated question, and / or a detailed query request that provides context for this query. In practice, the sequence-to-sequence model 514 generates a reformulated autosuggest query that results in a more accurate and complete narrative query response from the AI ​​chat system.

[0062]

[0070] As shown in the figure, the sequence-to-sequence model 514 stores the reformulated autosuggest queries in the reformulated query cache 534, which will be discussed further below. Furthermore, the query gateway system 206 periodically updates the sequence-to-sequence model 514 based on the query log 506 and proposal 510 to generate new reformulated autosuggest queries and / or update existing reformulated autosuggest queries.

[0063]

[0071] In various cases, the classifier model 512 and the sequence-to-sequence model 514 are combined into a single machine learning model. For example, the query gateway system 206 jointly trains on the same input data to generate a single multitask model that receives various prompts and provides corresponding various outputs. For example, this multitask model may be an LLM or another type of generative language model.

[0064]

[0072] In various implementations, the online framework 504 enables the query gateway system 206 to provide users with real-time functionality related to the auto-suggest query system and the GLM system. For example, the online framework 504 includes an initial trigger 515 that detects prefixes from the user's text input. For instance, when a user visits a website that provides search functionality, they enter text into a text input field in the first user interface.

[0065]

[0073] Accordingly, the autosuggest query system displays suggestions in an autosuggest pane located within the first user interface, as shown in operation 524. Specifically, the autosuggest query system utilizes a query suggestion service 516 within the online framework 504, which determines autosuggest queries from prefixes in real time. For example, this query suggestion service 516 accesses suggestions 510 (e.g., a set of autosuggest queries) in the suggestion cache and, in response to detecting one or more prefixes, determines the relevant autosuggest queries in real time. In various implementations, the query suggestion service 516 includes a prefix-based try (e.g., prefix tree) service, a non-prefix matching service, a language model-based next word service, or another service for determining autosuggest queries from a set of autosuggest queries for a given prefix.

[0066]

[0074] Furthermore, the online framework 504 includes an AI chat eligibility model 520 having a chat eligibility cache 522. In various implementations, when the query suggestion service 516 determines an autosuggest query by prefix, the query gateway system 206 determines whether this autosuggest query is suitable for the AI ​​chat system.

[0067]

[0075] More specifically, in some implementations, the query gateway system 206 uses the AI ​​chat eligibility model 520 to determine whether a given autosuggest query is located in the chat eligibility cache 522. If a match is found in the chat eligibility cache 522, the query gateway system 206 generates and provides an AI chat element to be displayed with the given autosuggest query in the autosuggest pane (e.g., operation 524). In some implementations, the query gateway system 206 determines the match based on a threshold number or threshold amount of matching words. In various implementations, the AI ​​chat eligibility model 520 is a machine learning model that determines the match based on the proximity of the autosuggest query to eligible autosuggest queries in a vector space. If no match is found in a given autosuggest query, the query gateway system 206 does not provide an AI chat element to be displayed with the given autosuggest query in the autosuggest pane of the first user interface.

[0068]

[0076] In various implementations, the chat eligibility cache 522 stores only eligible autosuggest queries. In alternative implementations, the chat eligibility cache 522 stores only ineligible autosuggest queries that the query gateway system 206 recognizes as unsuitable for the AI ​​chat system. In some implementations, the chat eligibility cache 522 stores both eligible and ineligible autosuggest queries, along with positive or negative eligibility indicators.

[0069]

[0077] In some implementations, the online framework 504 excludes or omits the AI ​​chat eligibility model 520. In these implementations, the query gateway system 206 generates AI chat elements for any autosuggest query provided within the autosuggest pane.

[0070]

[0078] As shown in the figure, the online framework 504 includes another trigger 526 that detects a selected AI chat element (for example, a clicked AI chat suggestion). For example, when a user selects an AI chat element, the query gateway system 206 identifies the corresponding autosuggest query. Often, the corresponding autosuggest query refers to an autosuggest query that appears next to or alongside the selected AI chat element. The query gateway system 206 then sends the corresponding autosuggest query to the AI ​​chat system via the query reformulation model 530.

[0071]

[0079] As shown in the figure, the query reformulation model 530 includes a lightweight language model 532 and a reformulated query cache 534. As previously mentioned, the query gateway system 206 uses the query reformulation model 530 to generate a reformulated autosuggest query from an input autosuggest query. For example, the query gateway system 206 provides a corresponding autosuggest query to the query reformulation model 530, which determines the reformulated autosuggest query.

[0072]

[0080] In some implementations, the query reformulation module 530 determines whether the corresponding autosuggest query exists in the reformulated query cache 534. If so, the query reformulation model 530 identifies the reformulated autosuggest query associated with the corresponding autosuggest query from the reformulated query cache 534. Otherwise, the query reformulation module 530 uses the lightweight language model 532 to determine the reformulated autosuggest query for the corresponding autosuggest query in real time (e.g., on the fly). Additional details regarding the determination of the reformulated autosuggest query are shown below in relation to Figure 6.

[0073]

[0081] In various implementations, the lightweight language model 532 is a smaller, simpler, and faster version of the sequence-to-sequence model 514. For example, the sequence-to-sequence model 514 is an LLM, while the lightweight language model 532 is an LSTM or another type of language machine learning model. In numerous implementations, the lightweight language model 532 is a classifier-type model that operates within specified low-latency parameters to ensure real-time processing (e.g., producing reformulated autosuggest queries within milliseconds).

[0074]

[0082] In some implementations, the online framework 504 excludes or omits the query reformulation model 530. For example, instead of determining a reformulated autosuggest query, the query reformulation model 530 directly provides the corresponding autosuggest query to the AI ​​chat system. In some cases, if a reformulated autosuggest query exists in the reformulated query cache 534, the query reformulation model 530 provides this to the AI ​​chat system; otherwise, it provides the corresponding autosuggest query to the AI ​​chat system.

[0075]

[0083] Furthermore, as shown in operation 540, the online framework 504 includes invoking the AI ​​chat system with the reformulated autosuggest query. For example, the query gateway system 206 provides the AI ​​chat system with the reformulated autosuggest query, along with an instruction to the AI ​​chat system to execute the reformulated autosuggest query.

[0076]

[0084] As mentioned above, in various implementations, the AI ​​chat system opens a new, separate user interface. For example, an instance of the AI ​​chat system appears in a new browser tab or window. In some cases, the AI ​​chat system starts within another application. Furthermore, in one or more implementations, the AI ​​chat system opens as a background service (for example, as a new tab without browser focus) and becomes visible when selected by the user. In various implementations, the AI ​​chat system comes to the forefront or becomes visible when an AI chat element is selected, triggering a transition to the AI ​​chat system.

[0077]

[0085] As mentioned above, Figure 6 provides additional details regarding the determination of the reformulated autosuggest query. Specifically, Figure 6 shows exemplary flowcharts for generating a reformulated autosuggest query from an autosuggest query in several implementation forms. As shown in the figure, Figure 6 includes a series of operations 600 along with the query reformulation model 530 included in Figure 5. Specifically, this series of operations 600 corresponds to the components of the query reformulation model 530.

[0078]

[0086] As shown in the figure, a series of operations 600 includes an operation 602 that receives an autosuggest query. For example, based on detecting that the user has selected an AI chat element, the query gateway system 206 provides a corresponding autosuggest query to the query reformulation model 530.

[0079]

[0087] Furthermore, the sequence of operations 600 includes operation 604, which determines whether the autosuggest query is in the reformulated query cache 534. For example, the query gateway system 206 queries the reformulated query cache 534 to determine whether the autosuggest query matches an entry for the autosuggest query in the reformulated query cache 534. If a match is found (for example, a cache hit), the query gateway system 206 identifies one or more reformulated autosuggest queries that map to the matching autosuggest query. In practice, as shown in operation 606, the query gateway system 206 identifies the reformulated autosuggest queries.

[0080]

[0088] In some implementations, the query reformulation model 530 is a machine learning model that determines matches based on the proximity of a provided autosuggestion query to an autosuggestion query in a reformulated query cache 534 in a vector space. For example, the query reformulation model 530 generates feature vectors in the provided autosuggestion query and determines whether those feature vectors map within a threshold distance to feature vectors of known autosuggestion queries in a trained vector space. In some implementations, the query gateway system 206 determines matches based on a threshold number or threshold amount of matching or overlapping words.

[0081]

[0089] If no match is found in the reformulated query cache 534 (for example, a cache miss), the query gateway system 206 uses the lightweight language model 532 to perform operation 608, which determines the reformulated autosuggest query on the fly. As previously mentioned, the lightweight language model 532 generates a newly reformulated autosuggest query for the provided autosuggest query within a low latency time threshold (for example, a few milliseconds). Furthermore, as shown in operation 610, the query gateway system 206 adds the newly determined reformulated autosuggest query to the reformulated query cache 534.

[0082]

[0090] In some implementations, the query gateway system 206 also provides the autosuggest query to the aforementioned classifier model to generate a reformulated autosuggest query offline and provides it to the reformulated query cache 534. In this way, the reformulated query cache 534 contains the latest version of the reformulated autosuggest query that maps to the autosuggest query.

[0083]

[0091] In some implementations, the lightweight language model (or classifier model) also utilizes additional context, such as previous searches, adjacent autosuggest queries, user information, location information, or other information, to determine a reformulated autosuggest query. For example, the lightweight language model 532 performs a forward search using the autosuggest query, accesses the top three results (e.g., title, header information, page information, and / or metadata), and uses this content to generate a reformulated, better autosuggest query for the autosuggest query.

[0084]

[0092] As mentioned above, by using the reformulated query cache 534 to identify reformulated autosuggest queries, the query gateway system 206 does not need to reprocess the same queries, reducing unnecessary processing and improving efficiency. Furthermore, by utilizing online lightweight language models and offline classifier models, the query gateway system 206 leverages resource efficiency to produce highly accurate results without sacrificing latency, allowing the user to wait while the query gateway system 206 processes requests.

[0085]

[0093] Figures 7A and 7B show exemplary graphical user interfaces for generating narrative-based responses using a generative language model from multiple autosuggest queries, in several implementation forms. For ease of explanation, Figures 7A and 7B include the aforementioned client device 400, which has a client application 401 and a first user interface 402. For example, the first user interface 402 includes the aforementioned text input 404, autosuggest pane 405, autosuggest query 406, and AI chat element 408.

[0086]

[0094] Furthermore, the client application 401 includes an autosuggest query selection element 702 and a selection confirmation element 710. For example, in the illustrated embodiment, the query gateway system 206 allows the user to select from several autosuggest queries generated in the text input 404. As shown in the figure, many of the autosuggest queries 406 are selected and indicated by check marks.

[0087]

[0095] Upon detecting the selection of the selection confirmation element 710, the query gateway system 206 can trigger the AI ​​chat system, as described above. For example, the query gateway system 206 provides each of the autosuggestion queries corresponding to the selected autosuggestion query selection element to a query reformulation model, which generates a reformulated autosuggestion query. In these implementations, the query reformulation model may process each of the provided autosuggestion queries and generate a reformulated autosuggestion query that is detailed, comprehensive, and / or summarized.

[0088]

[0096] For example, Figure 7B shows the first user interface 402 of a client application 401 on a client device 400. In particular, the first user interface 402 of the AI ​​chat system displays a reformulated autosuggest query 724 (e.g., a single query) that is much longer and more detailed than individually selected autosuggest queries. Furthermore, the reformulated autosuggest query 724 is formulated into a comprehensive, context-preserving query that causes the AI ​​chat system to provide a highly accurate and user-friendly narrative query response 726.

[0089]

[0097] Figure 7A shows the auto-suggest query selection element 702 in the auto-suggest pane of the first user interface 402, but additional implementations may include additional or different elements. For example, the auto-suggest pane 405 may include an element for opening search results in a new tab, or various AI chat elements for triggering various AI chat services. Another example is an auto-suggest query that incorporates elements that perform multiple functions, such as opening the search results of an auto-suggest query in a new tab and simultaneously triggering an AI chat system in another new tab.

[0090]

[0098] Furthermore, this disclosure has described implementations of the query gateway system 206 that provide a seamless gateway between a query search service and an AI chat system. While the auto-suggest query system has been described using the example of a search engine, the query gateway system 206 can also function with other types of auto-suggest query systems. These systems may include content-based sites that provide auto-suggest queries when a user searches for articles or other content, or e-commerce sites that suggest products and services. Moreover, the query gateway system 206 can operate wherever auto-suggest queries are provided to the user, such as in a search text field, browser bar, operating system search box, or multifunction search box.

[0091]

[0099] Next, moving to Figure 8, this figure shows an exemplary flowchart of a series of operations for using the query gateway system 206 in one or more implementation forms. Specifically, Figure 8 shows an exemplary series of operations of a computer implementation method for generating narrative query responses using a generative language model in one or more implementation forms.

[0092]

[0100] Figure 8 shows the operation in one or more implementations, but alternative implementations may omit, add, rearrange, and / or modify any of the illustrated operations. Furthermore, the operation in Figure 8 can be performed as part of a method, such as a computer implementation method. Alternatively, a persistent computer-readable medium, when executed by a processing system having a processor, may contain instructions that cause a computing device to perform the operation in Figure 8. In further implementations, a system can perform the operation in Figure 8 (for example, a processing system having a processor can execute the instructions).

[0093]

[0101] As shown in the figure, this series of operations 800 includes an operation 810 that generates an eligibility cache for autosuggest queries. For example, operation 810 includes generating an eligibility cache for the generating language models of autosuggest queries. In various implementations, operation 810 includes generating an eligibility cache for the generating language models of autosuggest queries by classifying the autosuggest queries by the generating language models.

[0094]

[0102] In some implementations, operation 810 includes generating a generative language model eligibility cache for autosuggest queries by classifying the autosuggest queries by the generative language model; determining that an autosuggest query is eligible for the generative language model by identifying the autosuggest query in the generative language model eligibility cache; and providing a generative language model element for display next to the autosuggest query in a first user interface based on the determination that the autosuggest query is eligible for the generative language model.

[0095]

[0103] Furthermore, as shown in the figure, the sequence of operations 800 includes an operation 820 that determines an autosuggestion query for text input. For example, in an exemplary implementation, this operation 820 includes determining an autosuggestion query from a set of autosuggestion queries in response to receiving text input corresponding to a search query. In one or more implementations, operation 820 includes determining an autosuggestion query from a set of autosuggestion queries in response to receiving text input corresponding to a search query.

[0096]

[0104] In various implementations, operation 820 includes receiving text input corresponding to a search query from a client device. In various implementations, operation 820 includes determining an autosuggest query from a set of autosuggest queries stored in an autosuggest query database based on the text input. In one or more implementations, operation 820 includes generating an eligibility cache for the autosuggest query generation language model based on query logs and classifier models of past text inputs.

[0097]

[0105] Furthermore, as shown in the figure, this series of operations 800 includes an operation 830 that provides a GLM element next to the autosuggest query based on the determination that the autosuggest query is eligible for a Generative Language Model (GLM). For example, in an exemplary implementation, this operation 830 includes providing a Generative Language Model element for display next to the autosuggest query in the first user interface based on the determination that the autosuggest query is eligible for a Generative Language Model using a Generative Language Model eligibility cache.

[0098]

[0106] In one or more implementations, operation 830 includes providing a generative language model element for display next to an autosuggest query in a first user interface. In some implementations, operation 810 includes providing a generative language model element for display next to an autosuggest query in a first user interface based on determining that the autosuggest query is eligible using a generative language model eligibility cache. In various implementations, operation 830 includes determining that an autosuggest query is eligible for the generative language model by identifying the autosuggest query in the generative language model eligibility cache.

[0099]

[0107] In some implementations, operation 830 includes determining that an additional autosuggest query is not eligible for a generative language model. Based on the determination that the additional autosuggest query is not eligible for a generative language model, operation 830 also includes deciding not to provide any generative language model elements for display next to the additional autosuggest query in the first user interface. In various implementations, operation 830 includes displaying the generative language model elements along with the autosuggest query in an autosuggest user interface pane, which includes a second autosuggest query displayed with a second generative language model element, and a third autosuggest query displayed without any generative language model elements.

[0100]

[0108] Furthermore, as shown in the figure, this series of operations 800 includes an operation 840 that generates a reformulated autosuggestion query from an autosuggestion query. For example, in an exemplary implementation, this operation 840 includes generating a reformulated autosuggestion query from an autosuggestion query in response to detecting a selection of a generated language model element.

[0101]

[0109] In one or more implementations, operation 840 includes generating a reformulated autosuggest query from an autosuggest query using a reformulated model, including a lightweight language model and a reformulated query cache, in response to detecting a selection of a generating language model element in a first user interface. In some implementations, operation 840 includes generating a reformulated autosuggest query from an autosuggest query using a reformulated model in response to detecting a selection of a generating language model element in a first user interface, and providing this reformulated autosuggest query to the generating language model as an autosuggest query.

[0102]

[0110] In various implementations, operation 840 includes generating a reformulated autosuggestion query from an autosuggestion query by determining that the autosuggestion query is in the reformulated query cache and identifying the reformulated autosuggestion query in the reformulated query cache associated with the autosuggestion query. In some implementations, operation 840 includes generating a reformulated autosuggestion query for the reformulated query cache from previous autosuggestion queries using a sequence-to-sequence machine learning model. According to some implementations, a reformulated autosuggestion query is generated from an autosuggestion query by determining that the autosuggestion query is not present in the reformulated query cache, determining the reformulated autosuggestion query on the fly using a lightweight language model, and adding the reformulated autosuggestion query to the reformulated query cache with the autosuggestion query.

[0103]

[0111] In some implementations, operation 840 includes generating a reformulated autosuggest query from the autosuggest query by determining that the autosuggest query is in a reformulated query cache and identifying the reformulated autosuggest query in the reformulated query cache associated with the autosuggest query. In various implementations, the reformulated autosuggest query is more detailed than the autosuggest query.

[0104]

[0112] Furthermore, as shown in the figure, this series of operations 800 includes an operation 850 that provides the reformulated autosuggestion query to the GLM. For example, in an exemplary implementation, this operation 850 includes providing the reformulated autosuggestion query to the Generating Language Model and displaying it in a second user interface separate from the first user interface.

[0105]

[0113] In one or more implementations, operation 850 includes providing the generative language model with an autosuggest query in response to detecting a selection of generative language model elements, and displaying it in a second user interface separate from the first user interface. In some implementations, operation 850 includes providing the generative language model with a reformulated autosuggest query, along with the narrative query result of the reformulated autosuggest query, which is generated by the generative language model, and displaying it in a second user interface separate from the first user interface.

[0106]

[0114] In some implementations, by providing the generative language model with a reformulated autosuggest query, the generative language model will automatically generate a narrative query response to the reformulated autosuggest query. In various implementations, operation 850 includes opening a new browser tab in the browser on the user's client device that displays a second user interface for the generative language model, which includes the reformulated autosuggest query and the narrative query response.

[0107]

[0115] In various implementations, operation 850 includes detecting the selection of additional generative language model elements that appear next to additional autosuggestion queries in the text input, and opening an additional new browser tab in the browser on the user's client device that shows additional instances of the generative language model, including the additional autosuggestion queries and additional narrative query responses in response to these additional autosuggestion queries.

[0108]

[0116] In some implementations, this set of operations 800 includes additional operations. For example, this set of operations 800 includes operations to detect selections of multiple generative language model elements corresponding to multiple autosuggestion queries provided in response to text input; operations to detect additional selections of combined generative language model elements; operations to generate a reformulated autosuggestion query from the multiple autosuggestion queries; and operations to provide the reformulated autosuggestion query to the generative language model.

[0109]

[0117] Figure 9 shows certain components that may be included in computer system 900. Computer system 900 may be used to implement various computing devices, components, and systems described herein (for example, by executing computer implementation instructions). Herein, “computing device” refers to an electronic component that performs a set of operations based on a set of programmed instructions. Computing devices include groups such as electronic components, client devices, and server devices.

[0110]

[0118] In various implementations, computer system 900 represents one or more of the aforementioned client devices, server devices, or other computing devices. For example, computer system 900 may refer to various types of network devices capable of accessing data on a network, a cloud computing system, or another system. For example, a client device may refer to a mobile device such as a mobile phone, smartphone, personal digital assistant (PDA), tablet, laptop, or wearable computing device (e.g., a headset or smartwatch). A client device may also refer to a non-mobile device such as a desktop computer, a server node (e.g., from another cloud computing system), or another non-portable device.

[0111]

[0119] The computer system 900 comprises a processing system including a processor 901. The processor 901 may be a general-purpose single-chip or multi-chip microprocessor (e.g., an Advanced Reduced Instruction Set Computer (RISC) machine (ARM)), a dedicated microprocessor (e.g., a Digital Signal Processor (DSP)), a microcontroller, a programmable gate array, etc. The processor 901 may be called a central processing unit (CPU) and may execute computer implementation instructions. The processor 901 shown in the figure is simply a single processor within the computer system 900 in Figure 9, but in an alternative configuration, a combination of processors (e.g., ARM and DSP) may also be used.

[0112]

[0120] The computer system 900 also includes a memory 903 that electronically communicates with the processor 901. The memory 903 may be any electronic component capable of storing electronic information. For example, the memory 903 may be embodied as random access memory (RAM), read-only memory (ROM), magnetic disk storage medium, optical storage medium, flash memory device in RAM, onboard memory included with the processor, erasable programmable read-only memory (EPROM), electro-erasable programmable read-only memory (EEPROM), registers, and combinations thereof.

[0113]

[0121] Instruction 905 and data 907 may be stored in memory 903. Instruction 905 may be executable by processor 901 to implement some or all of the functions disclosed herein. Executing instruction 905 may include using data 907 stored in memory 903. Any of the various examples of modules and components described herein may be partially or completely implemented as instruction 905, stored in memory 903 and executed by processor 901. Any of the various examples of data described herein may be present in data 907, stored in memory 903 and used by processor 901 during the execution of instruction 905.

[0114]

[0122] The computer system 900 may also include one or more communication interfaces 909 for communicating with other electronic devices. These one or more communication interfaces 909 may be based on wired communication technology, wireless communication technology, or both. Some examples of these one or more communication interfaces 909 include Universal Serial Bus (USB), Ethernet adapters, wireless adapters operating by the IEEE 902.11 wireless communication protocol, Bluetooth® wireless communication adapters, and infrared (IR) communication ports.

[0115]

[0123] The computer system 900 may also include one or more input devices 911 and one or more output devices 913. Some examples of the one or more input devices 911 include a keyboard, mouse, microphone, remote control, buttons, joystick, trackball, touchpad, and light pen. Some examples of the one or more output devices 913 include a speaker and a printer. A particular type of output device typically included in the computer system 900 is a display device 915. The display device 915 used in the implementations disclosed herein may utilize any suitable image projection technology, such as a liquid crystal display (LCD), light-emitting diode (LED), gas plasma, or electroluminescence. A display control device 917 may also be provided to convert data 907 stored in memory 903 into text, graphics, and / or (optionally) moving images to be displayed on the display device 915.

[0116]

[0124] The various components of the computer system 900 may be connected to one or more buses, including power buses, control signal buses, status signal buses, and data buses. For clarity, various buses are shown in Figure 9 as the bus system 919.

[0117]

[0125] This disclosure describes a query gateway system within a network framework. In this disclosure, “Network” means one or more data links that enable the transfer of electronic data between computer systems, modules, and other electronic devices. Networks may include public networks such as the Internet, as well as private networks. When information is transferred or provided over a network or another communication connection (hardwired, wireless, or both), a computer correctly perceives the connection as a medium of transmission. The medium of transmission may include networks and / or data links that carry the requested program code in the form of computer-executable instructions or data structures accessible by a general-purpose or private computer.

[0118]

[0126] Furthermore, the networks described herein may represent a network or combination of networks (such as the Internet, a corporate intranet, a virtual private network (VPN), a local area network (LAN), a wireless local area network (WLAN), a cellular network, a wide area network (WAN), a metropolitan area network (MAN), or a combination of two or more such networks) to which one or more computing devices may access the various systems described herein. In practice, the networks described herein may include one or more networks that use one or more communication platforms or technologies for transmitting data. For example, a network may include the Internet or other data links that enable the transfer of electronic data between each client device and components of a cloud computing system (e.g., server devices and / or virtual machines on them).

[0119]

[0127] Furthermore, upon reaching various computer system components, program code in the form of computer-executable instructions or data structures can be automatically transferred from the transmission medium to a persistent, computer-readable storage medium (device) (and vice versa). For example, computer-executable instructions or data structures received via a network or data link can be buffered in random-access memory (RAM) within a network interface module (NIC), and then ultimately transferred to the RAM of the computer system and / or to a less volatile computer storage medium (device) within the computer system. Therefore, it should be understood that components of a computer system that also utilize a transmission medium (or more primarily a transmission medium) can be equipped with a persistent, computer-readable storage medium (device).

[0120]

[0128] Computer-executable instructions include instructions and data that, when executed by a processor, cause a general-purpose computer, a dedicated computer, or a dedicated processing unit to perform a particular function or group of functions. In some implementations, computer-executable instructions and / or computer implementation instructions are executed by a general-purpose computer to transform that general-purpose computer into a dedicated computer implementing the elements of this disclosure. Computer-executable instructions may include, for example, binary, intermediate format instructions such as assembly language, or even source code. While subject matter has been described using terminology specific to structural features and / or methodological behavior, it should be understood that the subject matter as defined in the appended claims is not necessarily limited to the features or behaviors described above. Rather, the described features and behaviors are disclosed as exemplary forms of implementing the claims.

[0121]

[0129] Those skilled in the art will understand that this disclosure may be implemented in network computing environments having many types of computer system configurations, including personal computers, desktop computers, laptop computers, message processors, handheld devices, multiprocessor systems, microprocessor-based or programmable consumer electronics, network PCs, minicomputers, mainframe computers, mobile phones, PDAs, tablets, pagers, routers, switches, and the like. This disclosure may also be implemented in distributed system environments where local and remote computer systems linked over a network (by hardwired data links, wireless data links, or a combination of hardwired and wireless data links) both perform tasks. In a distributed system environment, program modules may reside in both local and remote memory storage.

[0122]

[0130] Each technique described herein may be implemented in hardware, software, firmware, or any combination thereof, unless specifically described as being implemented in a particular manner. Any function described as a module, component, etc., may also be implemented together in an integrated logic device, or separately as individual but interoperable logic devices. When each technique is implemented in software, it may be at least partially implemented by a persistent, processor-readable storage medium containing instructions that, when executed by at least one processor, perform one or more of the methods described herein (including computer implementation methods). Each instruction may be organized into routines, programs, objects, components, data structures, etc., which may perform specific tasks and / or implement specific data types, and may be combined or distributed as required in various implementations.

[0123]

[0131] A computer-readable medium can be any available medium that can be accessed by a general-purpose or dedicated computer system. A computer-readable medium that stores computer-executable instructions is a persistent, computer-readable storage medium (device). A computer-readable medium that carries computer-executable instructions is a transmission medium. Thus, as an example, an implementation of the present disclosure may comprise at least two distinctly different types of computer-readable mediums, namely a persistent, computer-readable storage medium (device) and a transmission medium.

[0124]

[0132] In this specification, persistent, computer-readable storage media (devices) may include RAM, ROM, EEPROM, CD-ROM, solid-state drives (SSDs) (for example, based on RAM), flash memory, phase-change memory (PCM), other types of memory, other optical disk storage devices, magnetic disk storage devices or other magnetic storage devices, or any other media that can be used to store desired program code means in the form of computer-executable instructions or data structures and that can be accessed by a general-purpose computer or a dedicated computer.

[0125]

[0133] Without departing from the scope of the claims, each step and / or operation of the method described herein may be interchanged with one another. That is, unless a specific order of steps or operations is required for the proper operation of the described method, the specific order and / or use of steps and / or operations may be modified without departing from the scope of the claims.

[0126]

[0134] The term "determining" encompasses a wide range of actions, and therefore "determining" can include calculating, arithmetic, processing, deriving, investigating, looking up (e.g., looking up in a table, data repository, or other data structure), and confirming. Similarly, "determining" can include receiving (e.g., receiving information), accessing (e.g., accessing data in memory), and resolving, selecting, choosing, and establishing.

[0127]

[0135] The terms “comprising,” “including,” and “having” are intended to be inclusive and mean that additional elements other than those enumerated may exist. Furthermore, it should be understood that any reference to “one implementation” or “multiple implementations” in this disclosure is not intended to exclude the existence of additional implementations that also incorporate the enumerated features. For example, any element or feature described in this specification with respect to one implementation may be combined with any element or feature of any other implementation described herein, provided that they are compatible.

[0128]

[0136] This disclosure may be implemented in other specific forms without departing from its spirit or characteristics. The implementations described are illustrative and not limiting. The scope of this disclosure is given in the appended claims rather than in the foregoing description. Any modifications that fall within the spirit and scope of the equivalence of the claims shall be included within that scope.

Claims

1. A computer implementation method for generating narrative query responses using a generative language model, To generate an eligibility cache (120) for the autosuggest query generation language model, In response to receiving a text input (110) corresponding to a search query, the system determines an autosuggest query (124) from a set of autosuggest queries (112), Based on the determination that the autosuggest query (124) is eligible for the generative language model (130) using the eligibility cache (120) of the generative language model, a generative language model element (122) is provided to be displayed next to the autosuggest query (124) in the first user interface (116). In response to detecting the selection of the aforementioned generated language model element (122), a reformulated autosuggest query (126) is generated from the autosuggest query (124), The reformulated autosuggest query (124) is provided to the generative language model (130) and displayed in a second user interface (132) that is different from the first user interface (116). Computer implementation methods, including those mentioned above.

2. The computer implementation method according to claim 1, further comprising generating an eligibility cache for the generating language model of an autosuggest query based on a query log and classifier model of a previous text input.

3. The computer implementation method according to claim 1, further comprising determining that the autosuggest query is eligible for the generative language model by identifying the autosuggest query in the eligibility cache of the generative language model.

4. The additional autosuggestion query is determined to be ineligible for the aforementioned generating language model, Based on the fact that the additional autosuggestion query is not eligible for the generative language model, it is decided not to provide any generative language model elements for display next to the additional autosuggestion query within the first user interface. The computer implementation method according to claim 1, further comprising:

5. The computer implementation method according to claim 1, further comprising displaying a generated language model element together with the autosuggest query in an autosuggest user interface pane, wherein the autosuggest user interface pane includes a second autosuggest query displayed together with a second generated language model element, and a third autosuggest query displayed without any generated language model element.

6. Determining that the aforementioned autosuggest query is within the reformulated query cache, and By identifying the reformulated autosuggest query in the reformulated query cache associated with the autosuggest query, The computer implementation method according to claim 1, further comprising generating the reformulated autosuggest query from the autosuggest query.

7. The computer implementation method according to claim 6, further comprising using a sequence-to-sequence machine learning model to generate reformulated autosuggest queries for the reformulated query cache from existing autosuggest queries.

8. Determine that the autosuggest query does not exist in the reformulated query cache. Using a lightweight language model, the reformulated autosuggestion query is determined on the fly, and By adding the reformulated autosuggest query to the reformulated query cache in the autosuggest query, The computer implementation method according to claim 1, further comprising generating the reformulated autosuggest query from the autosuggest query.

9. The computer implementation method according to claim 1, wherein the reformulated autosuggest query is more detailed than the autosuggest query.

10. The computer implementation method according to claim 1, wherein the reformulated autosuggest query is provided to the generative language model, causing the generative language model to automatically generate a narrative query response to the reformulated autosuggest query.

11. The computer implementation method according to claim 10, further comprising opening a new browser tab in a browser on a user's client device that displays the second user interface of the generated language model, wherein the second user interface includes the reformulated autosuggest query and the narrative query response.

12. To detect the selection of additional generative language model elements that appear next to the additional autosuggestion queries in the aforementioned text input, Opening an additional new browser tab in the browser on the user's client device that shows an additional instance of the generating language model, including the additional autosuggest queries and additional narrative query responses in response to the additional autosuggest queries. The computer implementation method according to claim 11, further comprising:

13. A computer implementation method for generating narrative query responses using a generative language model, The client device (230) receives text input (110) corresponding to the search query, Based on the text input (110), an autosuggest query (124) is determined from a set of autosuggest queries (112) stored in the autosuggest query cache, To provide a generative language model element (122) for display next to the autosuggest query (124) in the first user interface (116), In response to detecting the selection of the generated language model element (122) in the first user interface (116), a reformatting model (530) having a lightweight language model (532) and a reformatting query cache (534) is used to generate a reformatted autosuggestion query (126) from the autosuggestion query (124), The narrative query result of the reformulated autosuggest query (124) generated by the generative language model (130), along with the reformulated autosuggest query (126), is provided to the generative language model (130) and displayed in a second user interface (132) separate from the first user interface (116). Computer implementation methods, including those mentioned above.

14. By classifying the autosuggest queries using the aforementioned generative language model, a qualification cache for the generative language model of the autosuggest queries is generated. By identifying the autosuggest query in the eligibility cache of the generative language model, it is determined that the autosuggest query is eligible for the generative language model. Based on the determination that the autosuggest query is eligible for the generative language model, the generative language model element is provided for display next to the autosuggest query within the first user interface. The computer implementation method according to claim 13, further comprising:

15. To detect the selection of multiple generative language model elements corresponding to multiple autosuggestion queries provided in response to the aforementioned text input, Detecting the selection of additional elements in a combined generative language model, The process involves generating the reformulated autosuggestion query from the aforementioned multiple autosuggestion queries, To provide the reformulated autosuggestion query to the generative language model. The computer implementation method according to claim 13, further comprising:

16. Determining that the autosuggest query is within the reformulated query cache, By identifying the reformulated autosuggest query in the reformulated query cache associated with the autosuggest query, The computer implementation method according to claim 13, further comprising generating the reformulated autosuggest query from the autosuggest query.

17. The computer implementation method according to claim 16, further comprising using a sequence-to-sequence machine learning model to generate reformulated autosuggest queries for the reformulated query cache from existing autosuggest queries.

18. Determine that the autosuggest query does not exist in the reformulated query cache. Using the lightweight language model, the reformulated autosuggestion query is determined on the fly, and By adding the reformulated autosuggest query to the reformulated query cache in the autosuggest query, The computer implementation method according to claim 13, further comprising generating the reformulated autosuggest query from the autosuggest query.

19. A system for generating narrative query responses using a generative language model, Processor (901), Computer memory (903), when executed by the processor (901), By classifying the autosuggest query (114) using the generative language model (130), a qualifying cache (120) for the generative language model of the autosuggest query (114) is generated. In response to receiving a text input (110) corresponding to a search query, the system determines an autosuggest query (124) from a set of autosuggest queries (112). Based on the determination that the autosuggest query (124) is eligible for the generative language model (130) using the generative language model eligibility cache (120), a generative language model element (122) is provided to be displayed next to the autosuggest query (124) in the first user interface (116), and In response to detecting the selection of the aforementioned generative language model element (122), the autosuggest query (124) is provided to the generative language model (130) and displayed in a second user interface (132) separate from the first user interface (116). Computer memory (903) including an instruction (905) that causes the system to perform an operation including and A system that includes these features.

20. When executed by the aforementioned processor, In response to detecting the selection of the generated language model element in the first user interface, the reformulated model is used to generate a reformulated autosuggest query from the autosuggest query, and The reformulated autosuggestion query is provided to the generative language model as the autosuggestion query. The system according to claim 19, further comprising an instruction causing the system to perform an operation including the above.

Images