Controlling the generation of multi-input search results
By generating predicted search input and confidence on the server side, combined with the initial and final input of the client device, the generation process of multi-input search results is optimized, which solves the problem of long time to generate search results and outdated results, and improves the response speed and Customer retention rate.
Patent Information
- Application Number
- CN201980029270.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2018-04-19
- Filing Date
- 2019-04-18
- Publication Date
- 2024-12-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In multi-input search query, the prior art is difficult to effectively reduce the time of search result generation, resulting in the client device being idle while waiting for the result, which may lead to reduced customer retention, and the cached pre-calculated results may be outdated and incorrect Search results.
By receiving a subset of temporary search inputs on the server side, predicted search inputs and related confidence are generated, and a search operation is initiated when the confidence exceeds the threshold, predicted search results are generated, and sent to the client device; when the confidence does not exceed the confidence level; Wait for further query at threshold; when receiving the final input set, compare it with the initial input set, send the prediction result if it matches, and initiate a new search operation to generate the final search result if it doesn't match.
Reduces the time to generate search results, improves the response speed, reduces the risk of incorrect results, and improves the utilization rate and customer retention rate of client devices.
Smart Images

Figure CN112041835B_ABST
Abstract
Description
Technical Field
[0001] The present specification relates generally to the generation of search results, and more particularly to systems and methods for controlling the generation of multi-input search results. Background Art
[0002] Various types of data searches, such as those related to travel, involve search queries containing a large number of variable inputs that must be matched to a large variable set of possible search results. Performing searches under such conditions is computationally expensive, and thus can take a significant amount of time to complete. When searches are performed as soon as input is received, the time it takes to produce results can exceed several seconds, during which time the client device making the request is idle. Furthermore, the above period of time can result in reduced customer retention.
[0003] Some attempts to solve the above problems rely on pre-calculating and caching search results for several queries. Incoming queries are compared to pre-calculated queries, and when a match is found, the cached results are returned. While this approach can provide improved response times, the cached pre-calculated results may be outdated and, therefore, may produce incorrect search results. Summary of the invention
[0004] On the one hand, a method for controlling the generation of search results based on multiple search inputs is provided, the method comprising: at a server: receiving a temporary query containing a subset of temporary search inputs; based on the subset of temporary search inputs, generating: a subset of predicted search inputs; and confidence associated with the subset of predicted search inputs; when the confidence exceeds a threshold, initiating a search operation based on an initial input set to generate predicted search results, the initial input set comprising: (i) a subset of temporary search inputs and (ii) a subset of predicted search inputs; when the confidence does not exceed the threshold, waiting for additional queries without initiating a search operation; and sending a search response based on at least one of the temporary query and the additional query.
[0005] According to another aspect, sending the search response includes, in response to generating the predicted search results, sending the predicted search results to the client device for display at the client device after receiving a final set of inputs that match the initial set of inputs at the client device.
[0006] According to another aspect, sending the search response includes, in response to generating the predicted search results, storing the predicted search results until a final query is received.
[0007] Another aspect includes: receiving a final query including a final input set; in response to determining that the final input set matches the initial input set, sending predicted search results; and in response to determining that the final input set does not match the initial input set, initiating additional search operations to generate final search results based on the final input set, and sending the final search results.
[0008] Another aspect includes: receiving a final query including a final input set that does not match the initial input set; and initiating additional search operations to generate final search results based on the final input set; wherein sending the search response also includes sending the final search results to the client device.
[0009] According to another aspect, the final query further includes a search command instructing selection of a search element at the client device.
[0010] Another aspect includes: receiving an additional query; and when the additional query is an additional temporary query that includes a subset of additional temporary search inputs, repeatedly generating predicted search inputs and confidences associated with the predicted search inputs; and when the additional query is a final query that includes a final input set, initiating an additional search operation to generate final search results based on the final input set, and sending the final search results.
[0011] According to another aspect, the subset of provisional inputs corresponds to a first subset of predefined input fields presented at the client device; and the subset of predicted inputs corresponds to a remaining subset of predefined input fields.
[0012] According to another aspect, the method further includes, in response to receiving the provisional query, prior to generating the subset of predictive search inputs, retrieving media elements and sending the media elements to the client device for storage at the client device until search results are received.
[0013] Another aspect includes, prior to retrieving and sending the media element, generating a preliminary confidence score indicating a likelihood that the client device will generate a final query; and performing the retrieval and sending only if the preliminary confidence score exceeds a preliminary threshold.
[0014] According to another aspect, generating the subset of predicted inputs and the confidences includes retrieving a history of received queries and corresponding responses and comparing the history to the subset of provisional inputs.
[0015] According to another aspect, a server is provided, which is arranged to control the generation of search results based on multiple search inputs. The server comprises a memory; a communication interface; and a processor interconnected with the memory and the communication interface and configured to perform any of the aforementioned methods.
[0016] According to another aspect, a computer program product is provided, comprising program code instructions stored on a computer readable medium for implementing the method steps according to the aforementioned method when said program is executed on a computer.
[0017] According to another aspect, a method for controlling the supply of search results based on multiple search inputs at a client device is provided. The client device method includes: at the client device, receiving a subset of temporary search inputs; generating and sending a temporary query containing the subset of temporary search inputs; in response to sending the temporary query, receiving and storing: an initial input set including (i) a subset of temporary search inputs and (ii) a subset of predicted search inputs; and predicted search results corresponding to the initial input set; receiving a final input set; and presenting a search response on a display of the client device based on whether the final input set matches the initial input set.
[0018] According to another aspect, presenting a search response includes retrieving an initial input set from storage; comparing the initial input set to a final input set; and in response to determining that the final input set matches the initial input set, retrieving predicted search results and presenting the predicted search results on a display.
[0019] Another aspect includes, in response to determining that the final set of inputs does not match the initial set of inputs, sending a final query including the final inputs, and in response to the final query, receiving and presenting on a display final search results.
[0020] Another aspect includes, in response to determining that the final set of inputs does not match the initial set of inputs, discarding the predicted search inputs.
[0021] According to another aspect, generating and sending the interim query includes monitoring a plurality of input fields; and when at least one input field is completed, generating and sending the interim query.
[0022] According to another aspect, a client device is provided, which is arranged to control the generation of search results based on multiple search inputs. The client device includes a memory; a communication interface; and a processor interconnected with the memory and the communication interface and configured to perform any of the above-mentioned client device methods.
[0023] According to another aspect, a computer program product is provided, comprising program code instructions stored on a computer readable medium for implementing the method steps according to the aforementioned client device method when said program is executed on a computer.
[0024] According to yet another aspect, a system is provided that is arranged to control the generation of search results based on multiple search inputs. The system includes a server and a client device coupled to the server via a network. The server may include a memory, a communication interface, and a processor that is interconnected with the memory and the communication interface and configured to perform any of the aforementioned server methods. The client device may include a memory, a communication interface, and a processor that is interconnected with the memory and the communication interface and configured to perform any of the aforementioned client device methods.
[0025] According to yet another aspect, a computer program product is provided, comprising program code instructions stored on one or more computer readable media to implement the method steps according to the above-described server and client device methods when said program is executed on one or more computers. BRIEF DESCRIPTION OF THE DRAWINGS
[0026] Embodiments are described with reference to the following figures, in which:
[0027] Figure 1 A system for controlling generation of multi-input search results is described;
[0028] Figure 2 A method for controlling generation of multi-input search results is described;
[0029] Figure 3A Depicted in Figure 2 The search input interface used in the method;
[0030] Figure 3B Depicts a partially completed Figure 3A Search input interface;
[0031] Figure 4 Describes the execution Figure 2 The transmission of temporary queries during the method;
[0032] Figure 5 Describes the execution Figure 2 During the method an initial search input and predicted search results are received;
[0033] Figure 6 Depicted Figure 3A The search input interface is completed;
[0034] Figure 7 Describes the execution Figure 2 The search results generated by the method;
[0035] Figure 8 Describes the execution Figure 2 The transmission of the final query during the method;
[0036] Fig. 9 A method for controlling generation of multi-input search results according to another embodiment is depicted; and
[0037] Fig.10 A method for preemptively retrieving media elements for presenting search results is described. DETAILED DESCRIPTION
[0038] Figure 1 A system 100 for controlling the generation of multi-input search results is depicted. The system 100 includes a server 104 configured to receive requests for search results from at least one client computing device 154 (also referred to herein as client device 154 or simply device 154), such as via a network 150. In response to such requests, the server 104 is configured to generate search results and provide the search results to the client device 154 via the network 150.
[0039] In this example, the search results are travel-related search results. In particular, the examples discussed below illustrate the search results as flights. That is, each search result corresponds to a specific flight from a given departure airport to a given destination airport on a specific date and at a specific price. Various other types of search results may also be considered, including other travel-related search results (e.g., ground travel, such as train and bus excursions, hotel reservations, vehicle rentals, etc.) and search results that are not related to travel.
[0040] As will be apparent to those skilled in the art, generating search results such as those described above at the server 104 typically requires matching multiple inputs provided by the client device 154 with a large number of potential results, which themselves may change frequently. Therefore, the generation of such search results may be complex enough to consume a significant period of time to complete at the server 104. Therefore, a significant period of time may elapse between the submission of a search request by the client device 154 and the receipt and presentation of the search results at the client device 154.
[0041] In the context of the present disclosure, a not short period of time indicates a period of time that is easily perceived by an operator of the client device 154. For example, a not short period of time may be any period of time greater than about 1 second. For example, a flight search may take between about 2 and 20 seconds to complete. It is understood that the above time periods are provided for illustrative purposes only; the functionality of the system 100 may also be deployed for the generation of search results that require shorter or longer time periods than the above examples.
[0042] As will be discussed in more detail below, the server 104 and the client device 154 are configured to interact to enable, under certain conditions, preemptive initiation of a search computation at the server 104. Preemptive initiation of a search computation at the server 104 can, in turn, reduce the length of time that elapses between submitting a search request and receiving and presenting search results at the client device 154, while minimizing negative impacts on the accuracy of the search results.
[0043] Before discussing the functionality of the system 100 in more detail, certain components of the server 104 and the client device 154 will be described. The server 104 includes at least one processor 108, such as a central processing unit (CPU), etc. The processor 108 is interconnected with a memory 112, which is implemented as a suitable non-transitory computer-readable medium (e.g., a suitable combination of non-volatile and volatile memory subsystems, including any one or more of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic computer storage devices, etc.). The processor 108 and the memory 112 are typically composed of one or more integrated circuits (ICs).
[0044] The processor 108 is also interconnected with a communication interface 116 that enables the server 104 to communicate with other computing devices (e.g., client devices 154) via a network 150. Thus, the communication interface 116 includes any necessary components (e.g., a network interface controller (NIC), a radio unit, etc.) for communicating via the network 150. The specific components of the communication interface 116 are selected based on the nature of the network 150, which includes any suitable combination of wired and / or wireless local area networks, and wired and / or wireless wide area networks. The server 104 may also include input and output devices connected to the processor 108, such as a keyboard, a mouse, a display, etc. (not shown).
[0045] The above-mentioned components of the server 104 can be deployed in a single chassis or in a distributed format. In some examples, the server 104 includes multiple processors, which either share the memory 112 and the communication interface 116, or each has a different associated memory and communication interface.
[0046] The memory 112 stores a plurality of computer-readable programming instructions executable by the processor 108 in the form of various applications including a search application 120. As will be appreciated by those skilled in the art, the processor 108 executes the instructions of the application 120 (and any other suitable application) in order to perform various actions defined by the instructions contained therein. In the following description, the processor 108, and more generally the server 104, is referred to as being configured to perform those actions. It should be understood that they are configured by executing (by the processor 108) the instructions of the application stored in the memory 112. As will be discussed below, the execution of the search application 120 configures the server 104 to receive search requests from the client device 154, generate search results based on the request, and return the search results to the client device 154.
[0047] The memory 112 also stores a database 124 of primary data from which the above-described search results are generated. Thus, in this example, the database 124 contains attributes defining a plurality of flights (e.g., departure and destination locations, dates, airline identifiers, fares, available seats, etc.). As will be apparent to those skilled in the art, the server 104 may also be configured to periodically update the database to reflect changes to scheduling data, availability, prices, etc. The specific format of the records in the database 124 and the management of these records are not discussed in detail in this disclosure.
[0048] The client device 154 is implemented as any of a variety of computing devices including a desktop computer, a laptop computer, a smart phone, a tablet computer, etc. In some examples, the client device 154 may also be implemented as a kiosk, for example, deployed at an airport and operated by an airline.
[0049] The client device 154 includes at least one processor 158, such as a central processing unit (CPU) or the like. The processor 158 is interconnected with a memory 162, which is implemented as a suitable non-transitory computer-readable medium (e.g., a suitable combination of non-volatile and volatile memory subsystems, including any one or more of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory, magnetic computer storage devices, etc.). The processor 158 and the memory 162 are typically composed of one or more integrated circuits (ICs).
[0050] Processor 158 is also interconnected with communication interface 166, which enables client device 154 to communicate with other computing devices (e.g., server 104) via network 150. Thus, communication interface 166 includes any necessary components (e.g., network interface controller (NIC), radio unit, etc.) for communicating via network 150. The specific components of communication interface 166 are selected based on the nature of network 150, including any suitable combination of wired and / or wireless local area networks, and wired and / or wireless wide area networks.
[0051] The client device 154 also includes at least one input component 167 configured to receive input (e.g., from an operator of the client device 154) and provide the input to the processor 158. The input component 167 includes any one or any suitable combination of a keyboard, a mouse, a touch screen, a microphone, etc. The client device 154 also includes a display 168 interconnected with the processor 158 and controllable by the processor to present various information (including input fields selectable via the input component 167 as discussed below, and search results received from the server 104) to the operator of the client device 154. In some examples, the client device 154 also includes additional output components other than the display 168, such as a speaker (not shown).
[0052] The memory 162 stores a plurality of computer-readable programming instructions executable by the processor 158 in the form of various applications including the client search application 170. As will be appreciated by those skilled in the art, the processor 158 executes the instructions of the application 170 (and any other suitable application) so as to perform various actions defined by the instructions contained therein. In the following description, the processor 158, and more generally the client device 154, is referred to as being configured to perform those actions. It should be understood that they are configured by executing the instructions of the application stored in the memory 162 (by the processor 158). As will be discussed below, the execution of the application 170 configures the client device 154 to receive input and send requests to the server 104, as well as to receive search results from the server 104 and present the search results via the display 168. In some examples, the application 170 is a dedicated application that configures the client device 154 to interact with the server 104 for the specific purpose of obtaining search results. In other examples, the application 170 is a web browser that enables the client device 154 to interact with both the server 104 and various other computing devices (not shown).
[0053] Now go to Figure 2 , the operation of the system 100 will be described in more detail. Specifically, Figure 2A method 200 of controlling the generation of search results is illustrated. The method 200 will be described in conjunction with its execution within the system 100. Thus, in the following discussion, certain blocks of the method 200 are performed by the client device 154 via the processor 158 executing the application 170, while other blocks of the method 200 are performed by the server 104 via the processor 108 executing the application 120.
[0054] At block 205, client device 154 is configured to wait for input, for example, from input component 167. Prior to block 205, it is assumed that client device 154 has retrieved and presented on display 168 a plurality of prompts for search input (e.g., from data stored in association with application 170 or by requesting a web page from server 104). Figure 3A , illustrates an example search interface generated on the display 168 prior to executing block 205 .
[0055] Figure 3A The search interface shown in includes a plurality of optional (via input component 167) input prompts, each corresponding to a search input to be provided to the server 104. In the example shown, the interface includes prompts corresponding to a set of seven search inputs. In the following discussion, prompts are generally referred to as "fields", but it should be understood that a variety of formats (e.g., text fields, drop-down menus, etc.) can be used to collect search inputs. In particular, the first search field 300 includes a pair of optional radio buttons (usually in a mutually exclusive manner so that only one button can be marked as selected at a time) to indicate whether a round-trip flight or a one-way flight is requested. The second and third search fields 304 and 308 are fields in which the departure location and the destination location are provided, respectively. The fourth and fifth fields 312 and 316 are fields in which the departure and return dates (if applicable, based on the selection associated with the field 300) are provided, respectively. The sixth and seventh fields 320 and 324 are fields in which the number of adult passengers and the number of child passengers are provided, respectively. As will be apparent to those skilled in the art, various other search fields may also be provided. For example, the search interface may include other fields for indicating whether to limit the search to any one or more of nonstop (i.e., direct) flights, cabin preferences, airline preferences, etc. The interface also includes a selectable search element 328 for submitting a search query to the server 104 containing the input provided in fields 300-324.
[0056] return Figure 2, so in this example, the client device 154 is configured to wait for receiving input associated with the interface shown in Figure 3. The client device 154 is configured to take different actions depending on whether the input received at block 205 is a temporary input or a search command. The search command is a selection of the search element 328. In other words, the search command is an explicit instruction received at the input component 167 to send a search query to the server 104.
[0057] On the other hand, temporary input is the reception of data in any one of fields 300-324. For example, client device 154 can be configured to detect temporary input by receiving input data corresponding to a particular field and then selecting a different field (e.g., placing focus on it). Thus, receiving a departure location in field 304 and then shifting focus to destination location field 308 (e.g., via activating a "tab" key on a keyboard, selecting field 308 with a mouse or touch screen) is detected as temporary input. Various other mechanisms for detecting temporary input will also be apparent to those skilled in the art.
[0058] return Figure 2 , the processing of temporary input will be discussed first. After receiving the temporary input, the client device 154 is configured to proceed to block 210. Figure 3B The illustration shows receiving an example provisional input in the form of a departure location provided in field 304, followed by moving focus to field 308, indicating completion of the search input in field 304. At block 210, client device 154 is configured to send a provisional query containing a subset of the provisional search input to server 104 via network 150. The provisional query contains not only the search input most recently received at block 205, but also any other input received in relation to fields 300-324.
[0059] Go to Figure 4 , illustrates an example temporary query 400 sent from a client device 154 to a server 104 via a network 150. The temporary query 400 includes a subset of the temporary search input 408, which includes the departure location discussed above and is shown as Figure 3A and 3B The temporary query 400 may also include a tag or other indicator 404 that distinguishes the temporary query from the final search query resulting from the selection of the search element 328. In this example, the query 400 includes only the term "temporary" as an indicator. However, in other examples, various other means of distinguishing the temporary query from the final query may be used. For example, the final query may include the indicator, while the temporary query may omit the indicator.
[0060] In this example, at block 210, the client device automatically sends the interim query 400. That is, no input is required to initiate the transmission of the interim query 400 other than receiving the interim input at block 205. The client device 154 also does not need to present any indication (e.g., via the display 168 or other output component) that the interim query 400 has been sent. That is, the interim query 400 can be sent in a manner that is not visible to an operator of the client device 154.
[0061] Prior to transmission of the above-described interim query, the server 104 is configured to await receipt of a query at block 215. As will be apparent in the following discussion, the server 104 is configured to process interim queries such as query 400 differently than a final search query resulting from selection of the search element 328 at the client device 154. Thus, in response to receiving the interim query 400 at block 215, the server 104 proceeds to block 220.
[0062] At block 220, the server 104 is configured to receive the provisional query 400 and generate a subset of predicted search inputs based on the provisional inputs 408 included in the provisional query 400. Figure 4 As will be apparent from the illustrated interim query 400, a subset of the interim inputs received from the client device 154 may be incomplete. That is, the interim inputs may not include a value for each input in the set of inputs required for the server 104 to initiate a search. For example, the set of inputs required to initiate a search at the server 104 may include search inputs corresponding to at least fields 300, 304, 308, 312, and 316 as shown in FIG. 3. The subset of predicted inputs generated at block 220 includes predicted inputs corresponding to those search inputs in the set of required search inputs for which the interim query does not include a value. That is, the interim query may include a subset of the required inputs, and the predicted inputs represent the remainder of the required inputs. The predicted inputs may also include additional inputs that may not be necessary to initiate a search (e.g., values corresponding to passenger type code fields 320 and 324).
[0063] Various mechanisms for generating predictive search inputs are contemplated. In general, server 104 generates predictive search results based not only on temporary search inputs, but also on the history of final search queries. Such records can be stored in database 124 or in another suitable repository in memory 112.
[0064] In some examples, the server 104 is configured to generate a predicted input by comparing the provisional input (either individually or in aggregate) to the historical records to identify those historical records that contain a search input that matches the provisional input. Thus, in this example where the provisional input includes a trip type of "round trip" and a departure location of "Toronto (YYZ)", the server 104 can be configured to identify all historical records with the same trip type and departure location. For each of the remaining fields not considered by the provisional input (again either individually or in aggregate), the server 104 is then configured to identify the value or values that occur most frequently in the identified historical records.
[0065] For example, the history can contain Figure 4 1 and 2. The server 104 may be configured to rank the plurality of historical records according to the most common destination locations. The ranking may reveal, for example, that a first portion (e.g., 75%) of the records have a given destination (e.g., Boston (BOS)), another portion (e.g., 10%) of the records have another destination (e.g., Montreal (YUL)), and so on. The server 104 may be configured to repeat the same process for each additional input to be predicted.
[0066] As will be apparent, for certain predicted inputs, such as departure and return dates (i.e., inputs corresponding to fields 312 and 316), server 104 is configured to select not dates that appear in the historical records, but dates (departure and / or return dates) that are based on dates in the historical records. For example, a first portion (e.g., 60%) of the historical records having a trip type and a departure location that match the trip type and departure location in query 400 may have a departure date that is twenty days in the future from when the historical search query was received at server 104. Therefore, server 104 may be configured to select a date that is twenty days in the future from when query 400 was received as the predicted departure date corresponding to query 400.
[0067] In some examples, server 104 is configured to implement a more complex mechanism for generating predicted search inputs at block 220. For example, server 104 may store a previously configured (i.e., trained on a sample of the above historical records) prediction model for generating predicted search inputs. In such an example, the generation at block 220 is performed by processing the temporary input received in the temporary query via the above model. Any of a variety of machine learning models may be employed, including, for example, a random forest model. The generation of predicted search inputs may be performed by executing a plurality of such models, each of which is configured to generate a specific predicted search input (e.g., a return date), each of which employs a different subset of the temporary inputs to generate a corresponding predicted search input.
[0068] The generation of the predictive search input at block 220 may also be based on additional data in addition to the provisional input itself. Such additional data may include data derived from the provisional input. For example, the additional data may include the continent on which the departure location is located, the day of the week on which the query 400 is received, the day of the week on which the provisional departure date is, etc. Other examples of additional data derived from the provisional query 400 include the time of day on which the query 400 is received.
[0069] The generation of the predicted search input at block 220 may also be based on other data that is not derived directly from the provisional input. For example, the client device 154 may be configured to send to the server 104, along with or as part of the provisional query 400, a cookie or other data stored in the memory 162 that contains input stored from a previous search requested by the client device 154. For example, the client device 154 may be configured to store any input entered in fields 320 and 324 for future reporting to the server 104 (although not as provisional input). Other examples of data that is not derived directly from the provisional input in the query 400 but is employed in the generation at block 220 include, for example, an operator location based on a network address of the client device 154, or based on an operator email address stored at the client device 154 and provided to the server 104 along with the query 400.
[0070] Table 1 below illustrates the results of executing block 220. In particular, Table 1 lists the provisional input received in query 400, and the predicted input generated based on the provisional input. It will be appreciated that server 104 need not store the provisional input and predicted input in the format shown below.
[0071] Table 1: Temporary input and predicted input
[0072]
[0073]
[0074] The set of search inputs including the provisional inputs and the predicted inputs is referred to herein as an initial set of search inputs.
[0075] At box 220, the server 104 is also configured to generate a confidence level corresponding to the predicted search input. In general, the confidence level provides an assessment of the accuracy of the predicted search input. Taking the example above, where the predicted search input is generated by selecting an input from a historical record that matches the temporary input, the confidence level associated with the destination Boston (BOS) location is 75%. The server 104 is configured to generate an aggregated confidence level for the entire subset of predicted search inputs, for example, based on the individual confidence levels of each predicted search input. The aggregated confidence level can be, for example, an average of each individual confidence level. For example, when a random forest model is used to generate predicted search results, the confidence level of each predicted input can be the score of the classification tree that voted for the selected predicted input. Therefore, the confidence level generated at box 220 can be an aggregation of the above scores.
[0076] Then, at block 225, the server 104 is configured to determine whether the confidence generated for the subset of predictive search inputs at block 220 exceeds a previously configured threshold (e.g., stored as a setting in the application 120). As will be discussed below, one or more subsets of the predictive search inputs that exceed the confidence threshold are employed in the generation of search results. Because the generation of search results consumes a portion of the limited computing resources of the server 104, the threshold is set to limit the contribution of computing resources to only those sets of predictive inputs that are sufficiently likely to be accurate. For example, the threshold may be set to 50% (or any equivalent expression of the same fraction). Obviously, a variety of other thresholds may be employed based on any one or more of the computing resources available to the server 104, the volume of queries processed by the server, the complexity of the search process itself, etc.
[0077] When the determination at block 225 is negative, the predicted search input is discarded and the server 104 returns to block 215 to wait for further queries without initiating a search operation. As will now be apparent, the next query may be another ad hoc query with additional input that allows a more accurate set of predicted inputs (i.e., with greater confidence) to be generated at block 220.
[0078] When the determination at block 225 is positive, the server 104 proceeds to block 230. At block 230, the server 104 is configured to initiate a search operation based on the initial input described above (i.e., a combination of the provisional search input and the predicted search input) to generate predicted search results. The search operation itself is implemented according to any suitable search mechanism, and various search mechanisms will occur to those skilled in the art. In general, the server 104 is configured to select one or more records from the database 124 that meet the criteria set by the initial input. The selected record is a predicted search result (referred to as a prediction because the operation that causes its selection is based in part on the predicted search input).
[0079] In some implementations of method 200, the generation of predicted search inputs at block 220 may result in multiple subsets of predicted search inputs, where more than one subset may satisfy the threshold evaluated at block 225. In such examples, server 104 is configured to initiate a different search operation for each subset of predicted search inputs at block 230. Thus, at block 230, server 104 generates multiple sets of predicted search results, each corresponding to a different set of initial inputs.
[0080] When multiple subsets of predictive search inputs are generated at block 220 and a threshold is met at block 225, in some examples, the server 104 is configured to select only a portion of the subsets of predictive search inputs for which to initiate execution of a search operation at block 230. For example, if four subsets of predictive search inputs generated at block 220 meet the threshold at block 225, the server 104 may be configured to select only three subsets of predictive search inputs with the highest confidence for use in block 230. In other words, the server 104 may initiate a number of search operations at block 230 equal to the number of subsets of predictive search inputs that passed evaluation at block 225, up to a configurable maximum number.
[0081] At block 235, the server 104 is configured to send the predicted search results and the corresponding initial input set to the client device 154. Figure 5 , the predicted search results are shown as being selected from the database 124 via the execution of the application 120 and sent to the client device 154 via the network 150. At block 240, the client device 154 is configured to receive and store the predicted search results and the initial input in the memory 162. For example, Figure 5The predicted search result set 500 and the initial input 504 used to generate the predicted search result 500 are illustrated. The initial input 504 may include, for example, the initial input shown above in Table 1. When more than one predicted search result set is generated at block 230, each predicted search result set is sent to the client device 154 along with the corresponding initial input set for storage in the memory 162.
[0082] After transmitting the predicted search results and the initial input at block 235, the server 104 returns to block 215 to await further queries. In addition, after receiving and storing the predicted search results and the initial input at block 240, the client device 154 is configured to return to block 205 to await further input. Note that at block 240, in response to receiving the predicted search results 500, the predicted search results 500 are not displayed on the client device 154. That is, in this example, the receipt and storage of the predicted search results are not visible to the operator of the client device 154.
[0083] As will be apparent, additional temporary input (e.g., completion of the destination location field 308) is typically received at the client device 154. Such additional temporary input causes the client device 154 to issue additional temporary queries to the server 104, which is configured to process the additional input queries as described above. In practice, the server 104 may process multiple temporary queries in parallel. For example, a subsequent temporary query may be received at block 220 before the generation of search results for a previous temporary query is completed at block 230. The server 104 is configured to continue to send predicted search results and initial input at block 235 as predicted search results become available (by completing the corresponding search operation). At the same time, the client device 154 may be configured to store only the most recently received predicted search results, but is preferably configured to store all received sets of predicted search results until a search command is received at block 205.
[0084] When the client device 154 receives a search command at block 205 (e.g., in the form of selecting the search element 328 shown in FIG. 3 via the input component 167), the client device 154 proceeds to block 245. At block 245, the client device 154 is configured to compare the final set of inputs present in the fields 300-324 when the search command is received with the initial inputs 504 stored in the memory 162. Thus, the client device 154 is configured to retrieve the initial inputs 504 and compare the initial inputs 504 with the final inputs. Figure 6 The final set of inputs received at the client device 154 via the input component 167 is illustrated. Figure 6As can be clearly seen in , the final input matches the initial input 504 (shown in Table 1). Figure 2 , so the determination at block 245 is affirmative. Therefore, the client device 154 proceeds directly to block 270 where the search results stored in memory 162 in association with the matching initial input (i.e., in this example, the predicted search results 500) are presented on the display 168.
[0085] Figure 7 The presentation of search results 700-1, 700-2, and 700-3 on the display 168 is illustrated (three search results 700 are shown, but it should be understood that any number of search results may be presented). The search results include, for example, the departure and destination locations corresponding to the departure and destination locations in the initial and final inputs, as well as departure and arrival times and dates and prices. Additional attributes may be presented in association with the search results, such as airline name, stopover locations (if applicable), etc. As will now be apparent, after the affirmative determination at block 245, the search results 700 may be displayed without further request to the server 104.
[0086] Presenting the search results may include presenting additional media elements 704, such as advertisements, images, etc. Elements 704 may also be preemptively retrieved from server 104, as discussed in more detail below.
[0087] return Figure 2 When the determination at block 245 is negative, either because the initial input 504 does not match the final input, or because no initial input is stored in memory 162, the client device 154 proceeds to block 250. The client device 154 is also configured to discard any initial input and corresponding predicted search results stored in memory 162 after a negative determination at block 245.
[0088] At block 250 , the client device 154 is configured to send a final query including the final input to the server 104 . Figure 8 The transmission of the final query 800 from the client device 154 to the server 104 is illustrated. Figure 8 As shown, after the negative determination at block 245, the predicted search results 500 and the initial input 504 have been discarded. The final query includes a final set of inputs 804 (where the departure and return dates do not match those predicted in the initial input 504 described above). The final query 800 also includes an indication 808 that distinguishes the final query 800 from a provisional query (such as the previously mentioned query 400). The indicator 808 can take a variety of forms and generally marks the query 800 as an explicit search instruction to the server 104, rather than a query that may or may not result in a search being performed.
[0089] return Figure 2 , the server 104 is configured to receive the final query at block 255 and initiate a single search operation to generate final search results corresponding to the final input. At block 260, the server 104 is configured to send the final search results to the client device 154 for storage and display. The client device 154 is in turn configured to receive and store the final search results at block 265, and then proceed to block 270 and display the stored search results (i.e., the final search results received at block 265).
[0090] Typically, a search command is not received at block 205 until after multiple interim queries have been sent via successive executions of block 210. However, in some cases, for example, if fields 300-324 shown in FIG3 are completed and search element 328 is selected in a time shorter than required for server 104 to complete generation of predicted search results for a first interim query, a search command may be received before a response to any interim query has been received at block 240. In such a case, after receiving a search command at block 205, the determination at block 245 is negative because no initial input is stored in memory 162.
[0091] Therefore, the client device 154 is configured to send the final query at block 250. However, the response to the interim query may arrive at the client device 154 before the final search results. Therefore, even after sending the final query at block 250, the client device 154 may be configured to repeat the determination at block 245 in response to receipt of any search results and associated initial input sets.
[0092] In some examples, the client device 154 is also configured to provide feedback to the server 104 in response to a negative determination at block 245. For example, the client device 154 may be configured to send a copy of the initial input that failed to match the final input along with the final query at block 250. The server 104 may then be configured to update the prediction model (e.g., the random forest described above) based on an indication that the prediction of the search input was unsuccessful. The server 104 is also configured to store the final search input in the aforementioned history record. To further enable updating of the history record, the client device 154 may be configured to send the final input to the server 104 even after a positive determination is made at block 245. In this case, the final input may be sent to the server and an indication that the sending is only for logging purposes (i.e., no search results are requested).
[0093] Go to Fig. 9, illustrates another method 900 for controlling the generation of search results. Several blocks of method 900 are performed as described above in connection with method 200; these blocks are numbered with a prefix "9" and a two-digit suffix corresponding to the two-digit suffix of the corresponding block in method 200. Thus, for example, block 905 is performed by client device 154, as described above in connection with block 205. Other blocks of method 900 represent instances in which method 900 differs from method 200—those blocks do not share a suffix with any block of method 200.
[0094] At block 905, the client device 154 is configured to wait for input, as previously discussed. Upon receiving the provisional input, the client device 154 sends the provisional query at block 910, as discussed above in connection with method 200. However, after sending the provisional query, the client device 154 returns to block 905; as will be seen below, the client device 154 does not receive the predicted search results before receiving the search command.
[0095] The server 104 is configured to receive the provisional query at block 915 and proceed to block 920, where a predicted search input and a corresponding confidence level are generated. Then, at block 925, the server 104 is configured to determine whether the confidence level meets a pre-configured threshold. As described above in conjunction with method 200, when the determination at block 925 is negative, the server 104 proceeds to block 915, and when the determination at block 925 is positive, the server 104 proceeds to block 930. At block 930, the server 104 initiates one or more search operations, as described above in conjunction with block 230.
[0096] At block 931, server 104 is configured to store the predicted search results and the corresponding initial input in memory 112, rather than sending the predicted search results and the corresponding initial input to client device 154 as in method 200. Server 104 then returns to block 915 to await further queries.
[0097] In response to receiving the search command at block 905, the client device 154 sends a final query including the final input at block 950. In other words, the client device 154 sends a final query for each search command received, rather than sending a final query only when the final input does not match the initial input as in method 200. In fact, the client device 154 in method 900 cannot evaluate whether the final input matches the initial input because the initial input is not sent to the client device 154.
[0098] When the query received at block 915 is a final query, the server 104 proceeds to block 946. At block 946, a determination is made as to whether the final input received from the client device 154 matches the initial input stored at block 931. That is, the determination at block 946 is similar to the determination performed at block 245 of method 200, except that the determination is performed by the server 104 instead of the client device 154.
[0099] When the determination at block 946 is negative (i.e., when the initial search input stored at block 931 does not match the final input), the server 104 is configured to discard the predicted search results and initiate a search operation at block 950 to generate the final search results. The final search results are stored in the memory 112 for transmission to the client device 154 at block 961. When the determination at block 946 is positive, the server 104 proceeds directly to block 961, where the stored search results are sent to the client device 154 for receipt and display by the client device 154 at block 971. As will now be apparent, the server 104 is also configured to update the history record based on the final query and the determination results at block 946, and optionally update the predictive model employed at block 925.
[0100] As previously described, presenting search results at the client device 154 may include displaying media elements such as Figure 7 In some examples, media element 704 may be preemptively retrieved from server 104. Fig.10 , illustrates a method 1000 of preemptively retrieving media elements for presenting search results as performed by a server 104 .
[0101] In particular, at block 1005, the server 104 is configured to, upon receiving a provisional query from the client device 154, generate a preliminary confidence indicating the likelihood that the client device 154 will generate a final query. The preliminary confidence may be generated using a mechanism similar to that described in conjunction with block 220. In some examples, the preliminary confidence may be generated based on the number of provisional inputs contained in the provisional query. For example, if the provisional query contains a single input, the preliminary confidence may be set to 70%; and if the provisional query contains more than one input, the preliminary confidence may be set to 100%. In other examples, the history stored by the server 104 may also include incomplete searches (e.g., when one or more provisional queries are received but no final query is present). Thus, the server 104 may determine the preliminary confidence based on any suitable combination of factors, including inputs, time of day, etc. (e.g., based on a predictive model such as the random forest described above).
[0102] Then, at block 1010, server 104 is configured to determine whether the preliminary confidence level satisfies a configurable preliminary threshold value. The preliminary threshold value is set to a level indicating a sufficient likelihood that client device 154 will submit a final query to justify the contribution of computing resources and network bandwidth required to send element 704 to client device 154.
[0103] When the determination at block 1010 is negative, the server 104 returns to block 215 or 915 without executing block 220 or 920. However, when the determination at block 1010 is positive, at block 1015, the server 104 retrieves one or more media elements (e.g., media element 704) and sends them to the client device 154 for storage prior to receiving the final query. The server 104 then proceeds to block 220 or 920.
[0104] Variations on the above are contemplated. For example, the client device 154 may be configured to delay transmission of a provisional query until a minimum number of search inputs have been received, or until one or more specific inputs (eg, a departure location and a departure date) have been received.
[0105] In another example, the server 104 may be configured to store the predicted search results for use in the event that the corresponding predicted search input does not match the final search input. For example, if the final input does not match the predicted input, and only the passenger type code input is incorrectly predicted, the server 104 may be able to reuse the search results that are generally not dependent on the passenger type code. The generation of the final search results then requires only evaluating the usability of the selected search results (that match all remaining final inputs).
[0106] In yet other examples, the thresholds applied at blocks 225 and 925 may be dynamically adjusted by the server 104. For example, the history may include an indication of which searches resulted in a booking (i.e., a ticket for a flight was purchased). When the predicted search input generated at blocks 220 or 920 corresponds to a search input associated with a high likelihood of resulting in a booking, the server 104 may be configured to lower the thresholds applied at blocks 225 or 925 in order to increase the number of predicted search result sets prepared for the client device 154.
[0107] Those skilled in the art will recognize that in some embodiments, the functionality of applications 120 and 170 may be implemented using pre-programmed hardware or firmware elements (e.g., application specific integrated circuits (ASICs), electrically erasable programmable read-only memories (EEPROMs), field programmable gate arrays (FPGAs), etc.) or other related components.
[0108] In the specification, terms such as temporary query, additional query and final query have their common and common meanings in the art as those skilled in the art will understand. In the above example, temporary query can also be referred to as partial query or incomplete query, that is, query that does not include the value of each input in the input set required for the search initiated by the server 104. Moreover, temporary query is not generated by the selection of search element 328 and thus search command (that is, receiving explicit instructions at input component 167). In this regard, the term "temporary" should be understood as non-final, tentative, preliminary, temporary, etc. On the other hand, final query can also be referred to as complete query, that is, query that includes the value of each input in the input set required for the search initiated by the server 104. Moreover, the difference between final query and temporary query is that it is generated by the selection of search element 328 and thus search command (that is, receiving explicit instructions at input component 67). In this regard, the term "final" should be understood as complete, determined, non-tentative, etc. As used in this example, additional query can refer to another query as temporary query or final query.
[0109] The scope of the claims should not be limited by the embodiments set forth in the examples above, but should be given the broadest interpretation consistent with the entire specification.
Claims
1. A method, include: receiving, at the server from the client device, a temporary query related to the travel event and comprising a subset of the temporary search input; generating, at the server, a preliminary confidence score indicating a likelihood that a subset of client devices will generate a final query based on the provisional search input; When the preliminary confidence exceeds a first threshold, retrieving and sending the intermediate search results to the client device for storage in the client device until the intermediate search results are received; Generate, at the server, a subset of predicted search inputs and confidences associated with the subset of predicted search inputs based on the subset of temporary search inputs: When the confidence exceeds a second threshold, initiating a search operation based on an initial set of inputs to generate predicted search results, the initial set of inputs comprising a subset of the provisional search inputs and a subset of the predicted search inputs; When the confidence level does not exceed a second threshold, waiting for another query without initiating a search operation; and A search response is sent from the server to the client device based on the interim search results associated with the temporary query or the additional search results associated with the additional query.
2. The method of claim 1, wherein sending a search response include: In response to the generation of the predicted search results, the predicted search results are sent to the client device for display at the client device after receiving a final set of inputs that match the initial set of inputs at the client device.
3. The method of claim 1, wherein sending a search response include: In response to the generation of the predicted search results, the predicted search results are stored until a final query is received.
4. The method according to claim 3, further comprising: include: receiving a final query including a final input set; In response to determining that the final input set matches the initial input set, sending predicted search results; as well as In response to determining that the final input set does not match the initial input set, additional search operations are initiated to generate final search results based on the final input set, and the final search results are transmitted.
5. The method of claim 3 or claim 4, wherein the final query further comprises a search command instructing selection of a search element at the client device.
6. The method according to claim 1 or claim 2, further comprising: include: receiving additional inquiries; as well as when the additional query is an additional provisional query that includes a subset of the additional provisional search input, repeatedly generating the predicted search input and the confidence associated with the predicted search input; as well as When the additional query is a final query including the final input set, another search operation is initiated to generate final search results based on the final input set, and the final search results are sent.
7. The method of any one of claims 1 to 4, wherein the subset of temporary inputs corresponds to a first subset of predefined input fields presented at the client device; and A subset of the prediction inputs corresponds to the remaining subset of the predefined input fields.
8. The method according to any one of claims 1 to 4, further comprising: include: In response to receiving the provisional query, prior to generating the subset of predictive search inputs, media elements are retrieved and sent to the client device for storage at the client device until search results are received.
9. The method of any one of claims 1 to 4, wherein generating the subset of predicted inputs and the confidence comprises retrieving a history of received queries and corresponding responses, and comparing the history of received queries to the subset of provisional inputs.
10. A method, include: receiving, at a client device, a subset of the interim search inputs; generating and sending an ad hoc query related to the travel event and containing a subset of the ad hoc search input; as well as In response to sending an ad hoc query: receiving and storing the intermediate search results based on whether a preliminary confidence level indicates a likelihood that a subset of client devices will generate a final query based on the interim search input exceeds a first threshold; receiving and storing an initial set of inputs, including a subset of the provisional search inputs and a subset of the predicted search inputs; and predicted search results corresponding to the initial input set; receiving a final set of inputs at a client device; as well as The search response is presented by the client device on a display of the client device based on whether the final input set matches the initial input set.
11. The method of claim 10, wherein presenting a search response include: retrieving an initial input set from a storage device; Compare the initial input set to the final input set; In response to determining that the final input set matches the initial input set, retrieving predicted search results and presenting the predicted search results on a display; as well as In response to determining that the final input set does not match the initial input set, a final query including the final input set is sent, and in response to the final query, final search results are received and presented on a display.
12. A server, include: Memory; Communication interface; as well as a processor interconnected with the memory and the communication interface, The memory includes instructions that, when executed by the processor, cause the processor to perform operations including: receiving, from a client device, an ad hoc query related to a travel event and comprising a subset of the ad hoc search input; generating a preliminary confidence level indicating a likelihood that a subset of client devices will generate a final query based on the provisional search input; When the preliminary confidence exceeds a first threshold, retrieving and sending the intermediate search results to the client device for storage in the client device until the intermediate search results are received; Based on the subset of temporary search inputs, a subset of predicted search inputs and confidences associated with the subset of predicted search inputs are generated: When the confidence exceeds a second threshold, initiating a search operation based on an initial set of inputs to generate predicted search results, the initial set of inputs comprising a subset of the provisional search inputs and a subset of the predicted search inputs; When the confidence level does not exceed a second threshold, waiting for another query without initiating a search operation; and A search response is sent to the client device based on the interim search results associated with the temporary query or the additional search results associated with the additional query.
13. A client device, include: Input devices; Memory; Communication interface; as well as A processor interconnected with a memory, an input device, and a communication interface, the memory including instructions that, when executed by the processor, cause the client device to perform operations including: receiving a subset of temporary search input; generating and sending an ad hoc query related to the travel event and containing a subset of the ad hoc search input; and In response to sending an ad hoc query: receiving and storing the intermediate search results based on whether a preliminary confidence level indicates a likelihood that a subset of client devices will generate a final query based on the interim search input exceeds a first threshold; Receiving and storing an initial set of inputs, including a subset of the provisional search inputs and a subset of the predicted search inputs; and predicted search results corresponding to the initial set of inputs associated with the intermediate search results; receiving a final input set; and A search response is presented on a display of the client device based on whether the final input set matches the initial input set associated with the intermediate search results.
