Artificial intelligence-based flight assessment feedback recommendation and finding mapping

Abstract

To provide a method of performing an assessment of an airline company, a comment recommendation and a finding prediction mapping.SOLUTION: A method includes steps of: providing prompts of an assessment classification by a first large language model (LIM) in response to a first user partial entry into an assessment classification entry field; receiving a user entry of selection of one of the prompts or an alternative classification; providing prompts of a task description in which the first LIM is proposed in response to a partial entry of the task description; receiving a user input of selection of one of the prompts or an alternative task description; providing prompts including a sub-set of words of a comment to be proposed by the first LLM in response to the partial entry of the comment; receiving a user input of selection of one of the prompts or an alternative comment; providing predicted findings on the basis of the assessment classification, the task description, and the comment by a second LIM; and receiving a user input of selection of one of the predicted findings or the alternative finding.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001]

[0001] The present disclosure relates generally to artificial intelligence systems, and more particularly to large-scale language models for recommending and mapping user input to the outcome of airline operational evaluations. [Background technology]

[0002]

[0002] Airline Operational Safety Support (AOSS) is an operator's effectiveness initiative that identifies and evaluates an operator's readiness to continue operating aircraft safely. Under this initiative, an evaluation is conducted in collaboration with individual airlines against their own processes, procedures, and programs. This evaluates an airline's ability to continuously operate and maintain a fleet with a safe level of effectiveness and airworthiness, and identifies individual operator needs to further improve fleet performance.

[0003]

[0003] As part of this process, the safety team provides comments for each item based on the assessment checklist descriptions, ratings, etc. These comments are lengthy descriptions based on the safety team pilots' understanding and observations: operator efficiency, airline processes, procedures, and safety risks. The comments are further used to identify types of findings that may occur in the near future if corrective action is not taken. The finding classification helps AOSS identify the root causes and mitigation measures for safety issues. Summary of the Invention

[0004]

[0004] One exemplary embodiment provides a computer-implemented method for airline ratings, comment recommendations, and finding predictive mapping. The method includes, in response to a first user partial entry into a rating classification entry field, providing, by a first large-scale language model, a first number of prompts of suggested rating classifications based on several predefined rating classifications. A selection of one of the first number of prompts, or user entry of an alternative rating classification, is received. In response to a second user partial entry into a task description entry field, the first large-scale language model provides a second number of prompts of suggested complete task descriptions based on historical information related to the rating classifications. A selection of one of the second number of prompts, or user entry of an alternative complete task description, is received. In response to a third user partial entry into a comment entry field, the first large-scale language model provides a third prompt including an initial subset of words for the suggested comment based on historical information related to the task description. A selection of one of the third prompts, or user entry of an alternative comment, is received. A second large-scale language model provides several predicted findings based on the rating classification, task description, and comments, and a user input of a selection of one of the predicted findings or an alternative finding is received.

[0005] Another exemplary embodiment provides a system for airline ratings, comment recommendations, and finding predictive mapping, comprising: a storage device storing program instructions; and one or more processors operatively connected to the storage device, the one or more processors configured to execute the program instructions. When executed by one or more processors, the program instructions cause the system to: in response to a first user partial entry in the rating classification entry field, provide, via a first large-scale language model, a first number of prompts of a suggested rating classification based on several predefined rating classifications, and receive a user entry of a selection of one of the first number of prompts or an alternative rating classification; in response to a second user partial entry in the task description entry field, provide, via the first large-scale language model, a second number of prompts of a suggested complete task description based on past information related to the rating classification, and receive a user entry of a selection of one of the second number of prompts or an alternative complete task description; in response to a third user partial entry in the comment entry field, provide, via the first large-scale language model, a third prompt including an initial subset of words of a suggested comment based on past information related to the task description; and in response to a third user partial entry in the comment entry field, provide, via the second large-scale language model, several predicted findings based on the rating classification, the task description, and the comments, and receive a user entry of a selection of one of the predicted findings or an alternative finding.

[0006] Another exemplary embodiment provides a computer program product for airline ratings, comment recommendations, and finding predictive mapping. The computer program product includes a computer-readable storage medium having program instructions embodied thereon. The program instructions perform the steps of: in response to a first user partial entry in the rating classification entry field, providing, by a first large-scale language model, a first number of prompts of a suggested rating classification based on several predefined rating classifications, and receiving a user entry of a selection of one of the first number of prompts or an alternative rating classification; in response to a second user partial entry in the task description entry field, providing, by the first large-scale language model, a second number of prompts of a suggested complete task description based on past information related to the rating classification, and receiving a user input of a selection of one of the second number of prompts or an alternative complete task description; in response to a third user partial entry in the comment entry field, providing, by the first large-scale language model, a third prompt including an initial subset of words of a suggested comment based on past information related to the task description, and receiving a user input of a selection of one of the third prompts or an alternative comment; and in response to a second user partial entry in the comment entry field, providing, by the second large-scale language model, several predicted findings based on the rating classification, task description, and comments, and receiving a user input of a selection of one of the predicted findings or an alternative finding.

[0007]

[0007] These features and functions may be realized individually in various embodiments of the present disclosure or may be combined in further embodiments, further details of which can be understood by reference to the following description and drawings.

[0008] The novel features believed characteristic of the illustrative embodiments are set forth in the appended claims. However, the illustrative embodiments, as well as their preferred modes of use, further objects and features thereof, will best be understood by reading the following detailed description of illustrative embodiments of the present disclosure when read in conjunction with the accompanying drawings. [Brief explanation of the drawings]

[0009] [Figure 1]

[0009] FIG. 1 illustrates a block diagram of an airline rating system, according to an exemplary embodiment. [Figure 2]

[0010] 1 is an illustration of a diagram illustrating artificial intelligence architecture evaluation comment recommendations in accordance with an illustrative embodiment; [Figure 3]

[0011] 1 depicts a diagram illustrating a comment recommendation engine in accordance with an illustrative embodiment; [Figure 4A]

[0012] 10 is an illustration of a user interface with a blank comment field in accordance with an illustrative embodiment; [Figure 4B]

[0013] 1 illustrates a diagram of a user interface with a partial comment entry and a suggested comment, according to an example embodiment. [Figure 4C]

[0014] 1 is an illustration of a user interface with suggested findings in response to entered comments in accordance with an illustrative embodiment; [Figure 5]

[0015] 1 depicts a flowchart illustrating a process for airline ratings, comment recommendations, and finding predictive mapping in accordance with an illustrative embodiment. [Figure 6]

[0016] 1 is an illustration of a block diagram of a data processing system in accordance with an illustrative embodiment; DETAILED DESCRIPTION OF THE INVENTION

[0010]

[0017] The illustrative embodiments recognize and take into account that information collected by safety team members may involve manually writing down one or more comments into an Excel spreadsheet, averaging approximately 50-60 words for each piece of feedback, taking approximately 2-4 minutes for each observation, and introducing potential for human error, such as spelling mistakes, incorrect punctuation, or missed feedback.

[0011]

[0018] The illustrative embodiments also recognize and take into account that if one or more comments are manually entered for every observation, the AOSS team will begin reading the observations, those comments, and writing the type of finding (hazard) for each observation. This may involve mapping manual, non-standardized findings. Mapping approximately 400 observations will take a significant amount of time.

[0012]

[0019] Exemplary embodiments provide an artificial intelligence system that uses a trained GPT algorithm to predict and prompt one or more suggested comment phrases in response to a user's initial word entry into an interface. The algorithm is trained based on historical data observations to learn patterns from the historical data and find relationships between words / tokens present in aviation safety data. A second GPT algorithm is trained to map one or more comments to findings / potential hazards based on history.

[0013]

[0020] 1 is a block diagram of an airline rating system 100, according to an example embodiment. The airline rating system 100 includes a web-based front-end application 102 having a user interface 104.

[0014]

[0021] The user interface 104 includes several entry fields into which a user may enter key data. These include a rating category entry field 106, a task description entry field 108, a comment entry field 110, and a finding entry field 112. In response to a user typing a few words into the rating category entry field 106, the task description entry field 108, or the comment entry field 110, the web-based front-end application 102 makes respective application programming interface (API) calls 114 to a Python / JAVA back-end 122 (see FIG. 3). The API calls 114 include a checklist data population API call 116, a comment prediction API call 118, and a finding prediction API call 120.

[0015]

[0022] An API call 114 via a Python / JAVA backend 122 triggers a first large-scale language model 124 to generate suggested entries to display on the user interface 104 to assist the user in completing the entry (FIG. 4B). These suggestions are based on predefined rating classifications 130, past task description data 132, and past comment data 134 stored in a database 128. The first language model 124 is trained on the database 128.

[0016]

[0023] In response to entries in the comment entry field 110, the second large-scale language model 126 generates several suggested findings by mapping the comments to historical risk classification data 136. These suggested findings are displayed in the user interface 104 for selection and entry into the finding entry field 112 (see FIG. 4C).

[0017]

[0024] The airline rating system 100 may be implemented in software, hardware, firmware, or a combination thereof. When software is used, the operations performed by the airline rating system 100 may be implemented as program code configured to run on hardware, such as a processor unit. When firmware is used, the operations performed by the airline rating system 100 may be implemented as program code and data stored in permanent memory to run on a processor unit. When hardware is employed, the hardware may include circuitry that operates to perform the operations in the airline rating system 100.

[0018]

[0025] In example embodiments, the hardware may take the form of at least one of a circuit system, an integrated circuit, an application-specific integrated circuit (ASIC), a programmable logic device, or any other suitable type of hardware configured to perform certain operations. When a programmable logic device is used, the device may be configured to perform certain operations. The device may be permanently configured or may be later reconfigured to perform certain operations. Programmable logic devices include, for example, programmable logic arrays, programmable array logic, field programmable logic arrays, field programmable gate arrays, and other suitable hardware devices. Furthermore, these processes may be implemented in organic components integrated with inorganic components or may be composed entirely of non-human organic components. For example, the processes may be implemented as circuits in organic semiconductors.

[0019]

[0026] Computer system 150 is a physical hardware system that includes one or more data processing systems. When multiple data processing systems are present in computer system 150, the data processing systems communicate with each other using a communication medium. The communication medium may be a network. The data processing systems may be selected from at least one of a computer, a server computer, a mobile device such as a tablet computer, or any other suitable data processing system.

[0020]

[0027] As shown, computer system 150 includes several processor units 152 capable of executing program code 154 that implements processes in exemplary embodiments. As used herein, a processor unit of several processor units 152 is a hardware device, consisting of hardware circuitry (e.g., circuitry on an integrated circuit that responds to and processes instructions and program code to operate a computer). When several processor units 152 execute program instructions 154 for processing, several processor units 152 are one or more processor units that may be on the same computer or different computers. In other words, processing can be distributed among processor units on the same computer or different computers of a computer system. Furthermore, several processor units 152 can be the same type of processor unit or different types of processor units. For example, several processor units can be selected from at least one of a single-core processor, a dual-core processor, a multi-processor core, a general-purpose central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), or other types of processor units.

[0021]

[0028] The illustrative embodiments provide an automated, AI-based tool that recommends suggested comments and types of findings (hazards) for safety team members to select as they begin writing comments for individual checklist items. These recommendations significantly save time, reduce the incidence of human errors such as spelling and punctuation errors, and prevent missed feedback points.

[0022]

[0029] The solution uses a generative pre-trained transformer (GPT) algorithm to recommend descriptive comments and map findings for individual observations. GPT is a powerful natural language processing (NLP) algorithm that learns patterns in text data and uses them to write sentences, summarize text content, ask questions, answer questions, and classify data. Exemplary embodiments use two pre-trained GPT algorithms with over 2 billion parameters. The first algorithm is fine-tuned using up to 1,500 AOSS historical data observations (each observation is a small paragraph). This first algorithm learns patterns from historical data and finds relationships between words / tokens present in AOSS aviation safety data. These patterns assist the algorithm in recommending future comments as safety team members begin typing their observations. The second GPT pre-trained algorithm is fine-tuned using historical data with additional features, "comments," to predict findings / hazards associated with the observations.

[0023]

[0030] The AOSS team has approximately 400 to 500 predefined checklists / tasks that are given to the airline during the assessment. For each task, the evaluator observes the airline's performance and provides comments for each performed task. After the assessment is completed, the AOSS team begins mapping (classifying) findings / hazards for each individual performed task based on the comments provided for that task. A comment can be a single sentence or a small paragraph. Manually writing comments can be time-consuming and can lead to errors, such as overlooking a comment point or misspellings. Therefore, exemplary embodiments automate comment creation by suggesting comment phrases to the evaluator during the assessment and finding / hazard mapping.

[0024]

[0031] 2 illustrates a diagram illustrating an artificial intelligence architecture rating comment recommendation according to an exemplary embodiment. Architecture 200 is an exemplary implementation of airline rating system 100 of FIG.

[0025]

[0032] Trained GPT model 1 202 is a GPT-based algorithm trained on historical comment data entered in the past. The algorithm maps relationships between the area of ​​the evaluation, the task description, and the initial comment input from the evaluation to previously captured comments. When the evaluator begins writing a few words (e.g., 3-4) in the comment field 210, trained GPT model 1 202 takes these initial inputs, finds the best past comments that are contextually similar to the given input, and suggests them to the evaluator (see FIG. 4B). Comment 214 is an example suggested comment that trained GPT model 1 202 might find in response to the user's partial word entry. The user can select the best suggested comment or enter an alternative comment.

[0026]

[0033] Once a comment is selected and entered, trained GPT2 204 takes the input (area of ​​rating 206, task description 208, and AI-recommended feedback / comments 210) and classifies the comment as a risk / finding (e.g., a finding 212 associated with an element contained within the comment). Trained GPT2 204 is another GPT-based algorithm trained on past comments and risks.

[0027]

[0034] Trained GPT Model 1 202 uses a transformer architecture for training on text data, which requires three inputs from past data: a checklist / assessment 206, a task description 208, and feedback / comments 210. The algorithm attempts to find relationships between words present in a sentence or paragraph, finding the importance of words based on their position in the sentence, which assists the algorithm in finding the contextual meaning of the sentence or paragraph.

[0028]

[0035] Trained GPT model 2 204 also uses a transformer architecture. This model is trained on past data, such as checklists / assessments 206, task descriptions 208, feedback / comments 210, and hazards / findings (e.g., 212). It attempts to map sentence / paragraph context to hazard / finding types. Once it finds a relationship between comments and hazard types, it applies that relationship to new data to map hazards / findings.

[0029]

[0036] Figure 3 shows a diagram illustrating a comment recommendation engine according to an example embodiment. Figure 3 shows a high-level architectural diagram of a web application, where a front-end application communicates with a back-end tier via multiple API calls.

[0030]

[0037] The exemplary embodiments provide a web-based responsive application that can be hosted within a cloud computing system. The front-end interface 310 may employ the Angular 15+ framework and interacts with a Python / Java-based back-end 316 via several APIs that communicate with two GPT machine models 312 and 314.

[0031]

[0038] The HTTP API call 302 to load filter information brings all checklist information from a database 308 via a backend 316. The database 308 may include an Azure Structured Query Language (SQL) database. This information is displayed in a tabular format in a frontend interface 310, where each row represents one assessment item (see FIG. 4A). This information is editable by the user to write comments and map findings / hazards.

[0032]

[0039] The comment field for each rating item includes an editable text box (see Figure 4B). When a user begins writing a comment with a few words, a comment prediction API call 304 is made, feeding all the information for that rating item into a comment recommendation ML model 312 (equivalent to model 202 in Figure 2) and returning a list of suggested comments to the front-end interface 310.

[0033]

[0040] Once a comment is selected from the suggested list of comments, or an alternative comment is manually entered by the user, a finding prediction API call 306 occurs, capturing all information including the selected / entered comment, which feeds this information into a finding prediction ML model 314 (equivalent to model 204 in FIG. 2) to retrieve the assigned findings for that rating item.

[0034]

[0041] 4A-4C illustrate a front-end user interface according to one example embodiment. In this example, front-end user interface 400 includes tabular row items that separate the classification of entries into different columns, including findings 402, ratings 404, checklists 406, airplane models 408, and comments 410.

[0035]

[0042] In one embodiment shown in Figure 4A, comment entry field 412 is initially empty. Each row has its own comment entry field that functions as an intelligent search box.

[0036]

[0043] Once a user begins writing a partial entry 414 (e.g., two or three words) in the comment entry field 412, the comment recommendation ML 312 evaluates all other parameters, such as the rating 404, checklist 406, and airplane model 408, and returns a list of the most suitable suggested comments 416, 418 for that item. These suggested comments 416, 418 are displayed as selectable options below the comment entry field 412, as shown in FIG. 4B. The user can then select the appropriate comment by simply clicking the correct option. Alternatively, if none of the suggested comments adequately matches what the user wants to communicate, the user can manually enter an original comment. The original comments can be used to further train the comment recommendation ML model 312.

[0037]

[0044] Once the user selects an appropriate comment (e.g., 416) and saves it, another API call is made to retrieve the mapped, likely findings from the finding prediction ML model 314. The predicted findings 420 are displayed in the finding field 402 for that rating item, as shown in FIG. 4C.

[0038]

[0045] 5 shows a flowchart illustrating a process for airline rating, comment recommendation, and finding predictive mapping, according to an example embodiment. Method 500 can be implemented within airline rating system 100 of FIG.

[0039]

[0046] In response to a first user partial entry into the rating category entry field, the first large-scale language model provides a first number of prompts of suggested rating categories based on several predefined rating categories (step 502). The predefined rating categories include flight operations, simulator operations, training department, safety department, crew dispatch department, and crew scheduling department.

[0040]

[0047] The first large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses the rating classification, task description, and comments to find relationships between words present in a sentence or paragraph, identify the importance of words based on their position in the sentence, and find the contextual meaning of the sentence or paragraph. The GPT algorithm is integrated into a web-based front-end application with an interface that provides a table format with several intelligent search fields that trigger front-end API calls in response to the entry of several words.

[0041]

[0048] The system receives a user entry of a selection of one of a first number of prompts or an alternative rating category (step 504).

[0042]

[0049] In response to a second user partial entry into the task description entry field, the first large-scale language model provides a second number of prompts of a suggested complete task description based on past information related to the rating classification (step 506).

[0043]

[0050] The system then receives user input of a selection of one of a second number of prompts, or alternatively, a complete task description (step 508).

[0044]

[0051] In response to a third user partial entry in the feedback entry field, the first large-scale language model provides a third prompt including an initial subset of words (3-4 words) of suggested comments based on past information related to the task description (operation 510).

[0045]

[0052] The system receives user input of a selection of one of the third prompts or an alternative comment (step 512). The user entry of the alternative rating classification, alternative complete task description, or alternative comment is used to retrain and update the first large-scale language model.

[0046]

[0053] A second large-scale language model provides several predefined findings based on the rating classification, task description, and comments (step 514). The second large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses the rating classification, task description, and comments from the first large-scale language model or manually entered comments from a user to map sentence and paragraph contexts to types of findings.

[0047]

[0054] The system receives user input of one of the predefined findings or an alternative finding (step 516). The user input of the alternative finding is used to retrain and update the second large-scale language model. Process 500 then ends.

[0048]

[0055] Referring now to Figure 6, a block diagram of a data processing system is shown in accordance with an illustrative embodiment. Data processing system 600 may be used to implement computer system 150 in Figure 1. In this illustrative example, data processing system 600 includes a communications framework 602 that provides communications between a processor unit 604, a memory 606, a persistent storage unit 608, a communications unit 610, an input / output (I / O) unit 612, and a display 614. In this example, communications framework 602 takes the form of a bus system.

[0049]

[0056] Processor unit 604 functions to execute instructions for software that may be loaded into memory 606. Processor unit 604 may be several processors, a multi-processor core, or some other type of processor, depending on the particular implementation. In one embodiment, processor unit 604 includes one or more conventional general-purpose central processing units (CPUs). In an alternative embodiment, processor unit 604 includes one or more graphical processing units (GPUs).

[0050]

[0057] Memory 606 and persistent storage 608 are examples of storage device(s) 616. A storage device is any hardware capable of temporarily and / or persistently storing at least one of, for example, but not limited to, data, information such as program code in a functional form, or other suitable information. Storage device 616 may also be referred to as a computer-readable storage device, in these illustrative examples. Memory 606, in these examples, may be, for example, a random access memory or any other suitable volatile or non-volatile storage device. Persistent storage 608 may take various forms, depending on the particular implementation.

[0051]

[0058] For example, persistent storage 608 may contain one or more components or devices. For example, persistent storage 608 may be a hard drive, a flash memory, a rewritable optical disk, a rewritable magnetic tape, or some combination thereof. The medium used by persistent storage 608 may also be removable. For example, a removable hard drive may be used for persistent storage 608. In these illustrative examples, communications unit 610 provides for communication with other data processing systems or devices. In these illustrative examples, communications unit 610 is a network interface card.

[0052]

[0059] Input / output unit 612 allows for the input and output of data to and from other devices that may be connected to data processing system 600. For example, input / output unit 612 may provide a connection for user input through at least one of a keyboard, a mouse, or some other suitable input device. Further, input / output unit 612 may send output to a printer. Display 614 provides a mechanism for displaying information to a user.

[0053]

[0060] Instructions for at least one of an operating system, applications, or programs may be located in storage devices 616. Storage devices 616 are in communication with processor unit 604 via communications framework 602. The processes of the different embodiments may be executed by processor unit 604 using computer-implemented instructions that may be located in a memory, such as memory 606.

[0054]

[0061] These instructions are referred to as program code, computer usable program code, or computer readable program code, which may be read and executed by a processor in processor unit 604. The program code in different embodiments may be embodied on different physical or computer readable storage media, such as memory 606 or persistent storage 608.

[0055]

[0062] Program code 618 is located in a functional form on selectively removable computer readable media 620 and may be loaded onto or transferred to data processing system 600 for execution by processor unit 604. In these illustrative examples, program code 618 and computer readable media 620 form computer program product 622. In one example, computer readable media 620 may be computer readable storage medium 624 or computer readable signal medium 626.

[0056]

[0063] In these illustrative examples, computer readable storage medium 624 is a physical or tangible storage device used to store program code 618, rather than a medium that propagates or transmits program code 618. As used herein, computer readable storage medium 624 should not be construed as being, per se, a transitory signal such as an electric wave or other freely propagating electromagnetic wave, or an electromagnetic wave propagating through a waveguide or other transmission medium (such as an optical pulse passing through a fiber optic cable), nor should electrical signals transmitted over wires be construed as being, per se, a transitory signal such as an electric wave or other freely propagating electromagnetic wave, or an electromagnetic wave propagating through a waveguide or other transmission medium (such as an optical pulse passing through a fiber optic cable), or an electrical signal transmitted over a wire.

[0057]

[0064] Alternatively, program code 618 may be transferred to data processing system 600 using computer readable signal media 626. Computer readable signal media 626 may be, for example, a propagated data signal embodying program code 618. For example, computer readable signal media 626 may be at least one of an electromagnetic signal, an optical signal, or any other suitable type of signal. These signals may be transmitted over at least one communications link, such as wireless communications links, fiber optic cable, coaxial cable, an electrical wire, or any other suitable type of communications link.

[0058]

[0065] The different components illustrated for data processing system 600 are not meant to provide architectural limitations to the manner in which different embodiments may be implemented. The different illustrative embodiments may be implemented in a data processing system including components in addition to or instead of those illustrated for data processing system 600. Other components illustrated in FIG. 6 may differ from the illustrated example. Different embodiments may be implemented using any hardware device or system capable of executing program code 618.

[0059]

[0066] As used herein, the phrase "at least one of" used in conjunction with enumerated items means that various combinations of one or more of the enumerated items may be used, and that only one of each enumerated item may be required. In other words, "at least one of" means that any combination of items and any number of items from the list may be used, and not all of the enumerated items may be required. An item may be a specific object, article, or category.

[0060]

[0067] For example, without limitation, "at least one of item A, item B, and item C" may include item A, item A and item B, or item B. This example may also include item A, item B, and item C, or item B and item C. Of course, any combination of these items may be present. In some illustrative examples, "at least one of" may be, by way of example and not limitation, "two items A, one item B, and ten items C," "four items B, and seven items C," or other suitable combinations.

[0061]

[0068] As used herein, the term "a number of," when used in reference to an item, means one or more items. For example, "several different types of networks" refers to one or more different types of networks. In one exemplary embodiment, a "set" when referring to an item means one or more items. For example, a set of elements is one or more elements.

[0062]

[0069] The description of various exemplary embodiments has been presented for purposes of illustration and description and is not intended to be exhaustive or to limit the embodiments to the forms disclosed. Various exemplary examples describe components that perform operations or steps. In certain exemplary embodiments, the components may be configured to perform the described operations or steps. For example, the components may have a configuration or design that provides the components with the ability to perform the operations or steps described in the examples as being performed by the components. Furthermore, to the extent that the terms "includes / including," "has," "contains," and variations thereof are used herein, such terms, like the term "comprises," are intended to be inclusive, as open-ended terms that do not exclude any additional or other elements.

[0063]

[0070] Numerous modifications and variations will be apparent to those skilled in the art. Furthermore, different exemplary embodiments may offer different features as compared to other preferred embodiments. The selected embodiment(s) have been chosen and described in order to best explain the principles and practical applications of the embodiments, and to enable others skilled in the art to understand the disclosure of the various embodiments, including various modifications suitable for the particular use contemplated.

Claims

1. 1. A computer-implemented method for airline rating, comment recommendation, and finding predictive mapping, comprising: Using several processors, providing (502) a first number of prompts of suggested rating categories based on several predefined rating categories (130) by a first large-scale language model (124) in response to a first user partial entry into the rating category entry field (106); receiving (504) a user entry of a selection of one of the first number of prompts or an alternative rating category; providing (506) a second number of prompts of a suggested complete task description based on past information (132) related to the rating classification by the first large-scale language model in response to a second user partial entry into the task description entry field (108); receiving (508) a user input of a selection of one of the second number of prompts or, alternatively, a complete task description; providing (510) a third prompt including an initial subset of suggested comment words based on past information (134) related to the task description by the first large-scale language model in response to a third user partial entry into a comment entry field (110); receiving (512) a user input of a selection of one of the third prompts or an alternative comment; providing (514) a number of predicted findings based on the rating classification, the task description, and the comments by a second large-scale language model (126); and receiving (516) a user input of a selection of one of the predicted findings or an alternative finding.

2. The predefined evaluation categories are: Flight operations, Simulator operation, Training Department, Safety Department, Crew Dispatch Department, and The method of claim 1 including a crew scheduling department.

3. 2. The method of claim 1, wherein the first large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses rating classifications, task descriptions, and comments to find relationships between words present in a sentence or paragraph, identify word importance based on their position in the sentence, and find contextual meaning of the sentence or paragraph.

4. 4. The method of claim 3, wherein the GPT algorithm is integrated into a web-based front-end application (102) having an interface (104), the interface providing a table format with several intelligent search fields that triggers a front-end application programming interface call in response to the entry of several words.

5. 10. The method of claim 1, wherein the second large-scale language model includes a generative pre-trained transformer (GPT) algorithm that maps sentence and paragraph context to finding types using rating classifications, task descriptions, and comments from the first large-scale language model or manually entered comments from the user.

6. The method of claim 1 , wherein user entries of alternative rating classifications, alternative complete task descriptions, or alternative comments are used to retrain and update the first large-scale language model.

7. The method of claim 1 , wherein user input of alternative findings is used to retrain and update the second large-scale language model.

8. The method of claim 1 , wherein the initial subset of suggested comment words comprises three to four words.

9. 1. A system for airline rating, comment recommendation, and finding predictive mapping, comprising: a storage device (616) storing program instructions; and one or more processors (604) operatively connected to the storage device, the one or more processors configured to execute the program instructions, which, when executed by the one or more processors, provide the system with: providing (502) a first number of prompts of suggested rating categories based on several predefined rating categories (130) by a first large-scale language model (124) in response to a first user partial entry into the rating category entry field (106); receiving (504) a user entry of a selection of one of the first number of prompts or an alternative rating category; providing (506) a second number of prompts of a suggested complete task description based on past information (132) related to the rating classification by the first large-scale language model in response to a second user partial entry into the task description entry field (108); receiving (508) a user input of a selection of one of the second number of prompts or, alternatively, a complete task description; providing (510) a third prompt including an initial subset of suggested comment words based on past information (134) related to the task description by the first large-scale language model in response to a third user partial entry into a comment entry field (110); receiving (512) a user input of a selection of one of the third prompts or an alternative comment; providing (514) a number of predicted findings based on the rating classification, the task description, and the comments by a second large-scale language model (126); and receiving (516) a user input of a selection of one of the predicted findings or an alternative finding.

10. The predefined evaluation categories are: Flight operations, Simulator operation, Training Department, Safety Department, Crew Dispatch Department, and The system of claim 9 including a crew scheduling department.

11. 10. The system of claim 9, wherein the first large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses rating classifications, task descriptions, and comments to find relationships between words present in a sentence or paragraph, identify word importance based on position within a sentence, and find contextual meaning of a sentence or paragraph.

12. 12. The system of claim 11, wherein the GPT algorithm is integrated into a web-based front-end application (102) having an interface (104), the interface providing a tabular form with several intelligent search fields that triggers a front-end application programming interface call upon entry of several words.

13. 10. The system of claim 9, wherein the second large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses rating classifications, task descriptions, and comments from the first large-scale language model or manually entered comments from the user to map sentence and paragraph context to finding types.

14. 10. The system of claim 9, wherein user entry of alternative rating classifications, alternative complete task descriptions, or alternative comments is used to retrain and update the first large-scale language model.

15. The system of claim 9 , wherein user input of alternative findings is used to retrain and update the second large-scale language model.

16. The system of claim 9 , wherein the initial subset of suggested comment words comprises three to four words.

17. 1. A computer program product for airline rating, comment recommendation, and finding predictive mapping, comprising: a computer-readable storage medium (624) having program instructions embodied therein, the program instructions comprising: providing (502) a first number of prompts of suggested rating categories based on several predefined rating categories (130) by a first large-scale language model (124) in response to a first user partial entry into the rating category entry field (106); receiving (504) a user entry of a selection of one of the first number of prompts or an alternative rating category; providing (506) a second number of prompts of a suggested complete task description based on past information (132) related to the rating classification by the first large-scale language model in response to a second user partial entry into the task description entry field (108); receiving (508) a user input of a selection of one of the second number of prompts or, alternatively, a complete task description; providing (510) a third prompt including an initial subset of suggested comment words based on past information (134) related to the task description by the first large-scale language model in response to a third user partial entry into a comment entry field (110); receiving (512) a user input of a selection of one of the third prompts or an alternative comment; providing (514) a number of predicted findings based on the rating classification, the task description, and the comments by a second large-scale language model (126); and receiving (516) a user input of a selection of one of the predicted findings or an alternative finding.

18. The predefined evaluation categories are: Flight operations, Simulator operation, Training Department, Safety Department, Crew Dispatch Department, and 20. The computer program product of claim 17, comprising a crew scheduling department.

19. 18. The computer program product of claim 17, wherein the first large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses rating classifications, task descriptions, and comments to find relationships between words present in a sentence or paragraph, identify word importance based on position within a sentence, and find contextual meaning of a sentence or paragraph.

20. 18. The computer program product of claim 17, wherein the second large-scale language model includes a generative pre-trained transformer (GPT) algorithm that uses rating classifications, task descriptions, and comments from the first large-scale language model or manually entered comments from the user to map sentence and paragraph context to finding types.

Images