Automatic decision making for scaffold selection after partially correct answers in a conversational intelligent tutor system

By utilizing hardware processors and prediction models in the conversational intelligent tutor system, combining students' contextual knowledge and rule sets, automatic decision-making and selection of brackets is solved, and the learning experience problem in the processing of some correct answers is improved, improving the flexibility and effectiveness of bracket selection sex.

CN112199476BActive Publication Date: 2024-12-03INTERNATIONAL BUSINESS MACHINE CORPORATION
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010577460.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2019-06-23
Filing Date
2020-06-22
Publication Date
2024-12-03
Estimated Expiration
2040-06-22

AI Technical Summary

Technical Problem

The existing conversational intelligent tutor system is difficult to effectively select the appropriate stent when processing part of the correct answers, resulting in poor learning experience and lack of practical limitation assessments for different types of stents.

Method used

By leveraging hardware processors to access students’ contextual knowledge in a conversational intelligent tutor system, combining rule sets and prediction models, automatic decision-making to select appropriate bracket options, dynamically update the bracket selection framework in response to different types of learning gains and student characteristics.

Benefits of technology

Improved bracket selection for some correct answers is achieved, the learning experience is improved, the actual limitations of each bracket type is solved, and the flexibility and effectiveness of bracket functions are increased.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112199476B_ABST
    Figure CN112199476B_ABST
Patent Text Reader

Abstract

At least one hardware processor implementing a conversational intelligent tutor system obtains a partially correct natural language answer to a question posed by the conversational intelligent tutor system from a student. The at least one hardware processor implementing the conversational intelligent tutor system accesses contextual knowledge related to the student. Based on the partially correct answer and the contextual knowledge, the at least one hardware processor applies at least one of a rule set and a predictive model to select at least one scaffold option for the student. The at least one scaffold option is displayed to the student.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the electrical, electronic and computer fields, and more particularly, to a conversational intelligent tutor system (ITS). Background Art

[0002] Conversational ITS use natural language processing to chat with students about a topic, helping them improve their mastery. Natural language answers to questions posed by an ITS can be correct, partially correct, or incorrect relative to a reference answer. In one current ITS, 10% to 20% of student responses to prompt questions are scored as partially correct. Partially correct answers should provide students with some follow-up help, called scaffolding. Treating partially correct answers the same way as incorrect answers leads to a poor learning experience. Conversational ITS needs to decide which scaffold to use after each partially correct answer to help students improve their understanding. Summary of the invention

[0003] The principles of the present invention provide an automatic decision-making technique for selecting a scaffold after a partially correct answer in a conversational intelligent tutor system (ITS). In one aspect, an exemplary method includes the following steps: at least one hardware processor implementing the conversational intelligent tutor system obtains a partially correct natural language answer to a question posed by the conversational intelligent tutor system from a student; the at least one hardware processor implementing the conversational intelligent tutor system accesses contextual knowledge related to the student; based on the partially correct answer and the contextual knowledge, the at least one hardware processor applies at least one of a rule set and a predictive model to select at least one scaffold option for the student; and causes the at least one scaffold option to be displayed to the student.

[0004] On the other hand, an exemplary device includes a memory; and at least one processor coupled to the memory and operable to: implement a conversational intelligent tutor system, which obtains partially correct natural language answers to questions posed by the conversational intelligent tutor system from a student; accesses contextual knowledge related to the student; based on the partially correct answers and the contextual knowledge, applies at least one of a rule set and a predictive model to select at least one scaffolding option for the student; and causes the at least one scaffolding option to be displayed to the student.

[0005] As used herein, "facilitating" an action includes performing the action, making the action easier, assisting in performing the action, or causing the action to be performed. Thus, by way of example and not limitation, instructions executing on one processor may facilitate an action performed by instructions executing on a remote processor by sending appropriate data or commands to cause or assist in performing the action. For the avoidance of doubt, where a participant facilitates an action by means other than performing the action, the action is still performed by some entity or combination of entities.

[0006] One or more embodiments of the present invention or its elements may be implemented in the form of a computer program product including a computer-readable storage medium having a computer-usable program code for performing the indicated method steps. In addition, one or more embodiments of the present invention or its elements may be implemented in the form of a system (or device) including a memory and at least one processor coupled to the memory and operable to perform the exemplary method steps. On the other hand, on the other hand, one or more embodiments of the present invention or its elements may be implemented in the form of a device for performing one or more method steps described herein; however, the present invention is not limited thereto. The device may include (i) a hardware module, (ii) a software module stored in a computer-readable storage medium (or multiple such media) and implemented on a hardware processor, or (iii) a combination of (i) and (ii); any one of (i)-(iii) implements the specific techniques described herein.

[0007] The technology of the present invention can provide substantial beneficial technical effects. For example, one or more embodiments provide one or more of the following:

[0008] Improved scaffolding in response to partially correct answers in conversational ITS;

[0009] the ability to address the practical limitations of each different type of stent;

[0010] The ability to add improved scaffolding functionality to an existing conversational ITS by implementing the new functionality as a scaffold selection framework that integrates with the existing conversational ITS and interfaces with other elements such as knowledge databases, rule sets, and / or predictive models.

[0011] These and other features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 An example of partial inflation that may be used in one or more embodiments of the present invention is shown.

[0013] Figure 2shows examples of text links that may be used in one or more embodiments of the present invention;

[0014] Figure 3 An example of a stent diagram according to an aspect of the present invention is shown;

[0015] Figure 4 A block diagram of an exemplary system according to an aspect of the present invention is shown;

[0016] Figure 5 A flowchart illustrating an exemplary method according to one aspect of the present invention; and

[0017] Figure 6 A computer system that can be used to implement one or more aspects and / or elements of the present invention according to an embodiment of the present invention is shown, which also represents a cloud computing node. DETAILED DESCRIPTION

[0018] A conversational intelligent tutor system (ITS) can ask a question, a student can answer the question, and then there may be a conversation or interaction between the tutor and the student to resolve the problem around a specific topic (e.g., the "scaffolding" discussed above in the case of partially correct answers). In current systems, there are typically a certain number of questions and a certain number of answers, and the system attempts to match the student's answer to those assertions (answers). The system attempts to determine whether the student is close to the expected answer. In principle, the answer can be identified as correct, partially correct, or incorrect. One or more embodiments are useful in the case of partially correct answers - providing tutor feedback in the case of partially correct answers. In a typical but non-limiting example, 10-20% of the answers may be partially correct. This will depend on the audience and the subject area.

[0019] As noted, the assistance provided to the tutored student is referred to as scaffolding.One or more embodiments help students learn when there is a partially correct answer in a way that is better than simply treating the partially correct answer as an incorrect answer.

[0020] Current systems typically offer a limited set of bracket options. Figure 1 The partial pump shown includes the instructor prompting a second attempt at the question. The student is expected to expand his or her answer and provide something that more closely matches the assertion of the question. Specifically, Figure 1In a non-limiting example, at 101, the instructor asks the student “Which elements are found in lower concentrations in lunar rocks compared to Earth rocks?” At 103, the student responds “There is less iron and gold in lunar rocks.” At 105, the instructor prompts the student for a second attempt “That looks partially correct, keep going. You can also add more to your answer.” At 107, the student provides another answer “I meant iron and nickel.” At 109, the instructor compliments the student for providing a more complete answer and summarizes the model answer.

[0021] In fill-in-the-blank, the instructor asks students to fill in the missing words in the expected answer. Figure 2 In the text-linking shown, the student is shown the relevant paragraphs for the current assertion. Note the prompt at 201, and the relevant paragraphs at 203. In one approach to enabling questionins, the system continues to ask simpler or more focused questions about the topic. One or more embodiments are related to many different types of scaffolds.

[0022] Depending on the type of stent selected, stenting can be one-step or multi-step.

[0023] Currently, only some conversational ITS evaluate partially correct answers, resulting in a poor learning experience. Even those conversational ITS that do evaluate partially correct answers rely on a teaching model to determine the scaffolding to use after each partially correct answer. Advantageously, in contrast to the prior art, one or more embodiments take into account the student's general knowledge of the subject matter and the gap between the student's partially correct answer to the current question and the model answer.

[0024] In practice, one or more embodiments select one or more scaffolding options after a partially correct answer based on: (i) contextual knowledge that the ITS has collected about the student for the current question posed by the ITS and in the past (depending on the level of interaction between the student and the tutor in the past), and (ii) a set of rules or predictive models that specify scaffolds and their combinations and take into account any practical limitations of each type of scaffold. In one or more embodiments, the rules and models are iteratively improved by evaluating the learning gains generated by different types of scaffolds (and their combinations) in the student population. That is, a dynamic rule set or dynamic predictive model is employed in one or more embodiments using some combination of scaffolds, which is evidence-based (e.g., relearning gains using evidence). One or more embodiments use a set of controlled experiments to determine which scaffolding option is most effective for a particular set of situations, a group of students, or a single student.

[0025] Knowledge about students

[0026] Typically, this knowledge is a set of measurements collected by the ITS from analysis of student behavior and performance. For example, the measurements may assess current answer gaps and the student's general level of understanding. One or more embodiments consider:

[0027] The size of the gap between the student answer and the corresponding reference answer (assertion match) (as a percentage of the total number of tokens in the reference answer) - how close are the answers?

[0028] Mastery level of students on the same topic - This indicates the general level of understanding prior to the current answer (cumulative score of previous answers related to the topic).

[0029] The amount of time students spend on a topic and the coverage of answers within that topic.

[0030] Methods for specifying brackets

[0031] One or more embodiments use (i) a set of rules, such as a map as in the following example, or

[0032] (ii) A machine learning model to determine which scaffolding options to show students after a partially correct answer. Learning gains can be assessed continuously through ITS (e.g., through answer gaps) or externally through other methods, and information gathered by associating gains with scaffolds fed back into the map or model to make changes based on evidence from that cohort of students. Learning gains can be assessed continuously through ITS or externally through other methods, and information gathered by associating gains with scaffolds fed back into the map or model. Increasing the number of scaffolding options allows for a more fine-grained approach and more combinations can be used.

[0033] Example

[0034] Figure 3An example scaffolding map is shown that relies on two inputs: a measure of answer gap (vertical axis) and the student's mastery level (horizontal axis). Other embodiments may have more than two dimensions. The ITS uses this map to select which scaffold to present to the student after a partially correct answer. The upper left quadrant 301 represents lower mastery of the subject matter and a larger gap between the student's answer and the correct answer to the current question; the upper right quadrant 303 represents higher mastery of the subject matter and a larger gap between the student's answer and the correct answer to the current question; the lower left quadrant 305 represents lower mastery of the subject matter and a smaller gap between the student's answer and the correct answer to the current question; and the lower right quadrant 307 represents higher mastery of the subject matter and a small gap between the student's answer and the correct answer to the current question. In the case of high / medium mastery (from Figure 3 ), ITS can show partial inflation or FITB (Fill in the Blank). As the gap size becomes very small ( Tokens, Figure 3 ), most will display partial cheerleading because the FITB is unlikely to coincide with a gap. (If a FITB whitelisted word is not in the answer gap, in one or more embodiments the tutor will choose partial cheerleading.) As the gap gets larger (moving up the vertical axis), the ITS is more likely to display the FITB, resulting in a 50 / 50 split between FITB and partial cheerleading. Consider a lower proficient, larger gap-size student (not understanding the current question, not understanding the subject overall, quadrant 301). Such a student should receive a lot of scaffolding; for example, through text links and / or asking primary achievement objective (EO) questions (the next level down in granularity; as indicated by the thick arrow, one can start with a text link and then continue by asking primary EO questions). Students in quadrant 305 may receive, for example, text links or partial cheerleading.

[0035] Note that in one or more embodiments, the model answer has multiple keywords, and multiple keywords are extracted from the student's answer; "tokens" or "gap takens" refer to keywords in the model answer that are missing from the student's answer.

[0036] Note that in quadrants 303 and 307, the double-headed arrows indicate the selection between partial inflation and blank filling.

[0037] One or more embodiments adopt an initial approach (guessing), such as Figure 3 Then, based on evidence from controlled experiments / conversations with many students, the approach was changed. For example, consider a student with high proficiency and a medium gap size. Figure 3In an initial approach, partial airflow or fill-in-the-blank (FITB) is used. By examining the learning gains, it can be determined that the ratio between partial airflow and FITB should be changed. For example, suppose 75% FITB and 25% partial airflow is initially used. Observed results / experiments may indicate that for moderate proficiency, 50% FITB and 50% partial airflow works better (i.e., achieves greater learning gains). In one or more embodiments, the changes are iterative and may be relatively small. However, these changes may accumulate over time or vary significantly, depending on the student population and the topics covered. Each topic may have a different map.

[0038] As mentioned above, in lower proficiency students ( Figure 3 In the case of a smaller gap size ( Marks, Figure 3 ), the ITS will display a related paragraph (text link) to help the student and then give partial encouragement, such as indicated by a thick arrow from "text link" to "partial encouragement." As the gap widens, lower proficient students may lose completely; therefore, the ITS should provide the simplest scaffolding, such as displaying a related paragraph (text link) and then asking a series of more basic questions to achieve the goal. Figure 3 The scaffold map shown represents an initial configuration. As students use scaffold options, ITS collects information about their mastery to assess their learning outcomes, which is then fed back to them to continually update the scaffold map. It is also possible to add additional inputs to the scaffold map. The same approach applies if the scaffold map is replaced with a machine learning model that can predict which scaffolds to use given a range of inputs (features).

[0039] Characteristics of Machine Learning Models

[0040] If you use a machine learning model (for example, Figure 4 If the predictive model 411 is used to specify scaffolds and their combinations, a set of features (inputs) that capture the student's behavior and performance may include: Examples but not limitations:

[0041] The size of the gap between the student's answer and the corresponding reference answer;

[0042] the student’s level of proficiency in the subject;

[0043] the amount of time students spend on a topic;

[0044] the percentage of answers that have been covered in the topic;

[0045] The number of times the student has revisited the topic; and

[0046] The number of questions students have asked in the topic.

[0047] In one or more embodiments, the input to the machine learning module includes the features / dimensions listed above. In one or more embodiments, when running a learning experiment, a training set of data (e.g., the training corpus 415 discussed below) is used. Consider which scaffolding options to use for each partially correct answer, and which scaffolds can maximize the learning effect. Then, revisit and adjust the machine learning model, for example, by adding or subtracting one or more dimensions and / or changing the way the scaffolding options are selected. In one or more embodiments, this is a supervised learning process using data annotated by human experts (to reflect the best gain); the system is trained based on these data to improve itself. Some embodiments do not require annotations from human experts, but simply work based on raw learning gains or other objective measures. Learning gains can be specified using any of a variety of techniques known to technicians. For example, consider a standard multiple-choice test. The results of a standard multiple-choice test can be used as an objective measure of learning gain that does not require manual annotation. In one or more embodiments, humans will typically still write multiple-choice tests.

[0048] One or more embodiments of the present invention select one or more scaffolding options after a partially correct answer based on knowledge about the student, which combines: (i) the student's cumulative understanding of the topic, and (ii) the student's ability to answer the current question in the topic. In one or more embodiments, the map (rule) or predictive model used to determine which scaffolding options to display is dynamic. Information about the learning gains of the student population and the scaffolding options is fed back to continuously update the map or model. Regarding the dynamic aspect, the map / model will change the experimental results, etc. based on the number of scaffolding options introduced. In a sense, many things are dynamic in the learning world. For example, a student may be learning a topic today, but six months later, the information about the topic has changed. Therefore, at least in some cases, a steady state may never be reached, but for most topics, the map / model may still remain dynamic. On the other hand, for well-known factual topics with little new research, a steady state may be reached in time (e.g., after a few months).

[0049] One or more embodiments are useful in the field of conversational ITS; that is, there is a dialogue between multiple turns. One or more embodiments involve partially correct answers, rather than correct or incorrect answers.

[0050] One or more embodiments provide an automated decision-making method that selects one or more scaffolding options after a partially correct answer in a session ITS based on: a set of measurements collected by the ITS from student behavior and performance analysis, current questions asked by the ITS, and past questions; and a set of rules or predictive models for specifying scaffolds and scaffold combinations after each partially correct answer. In one or more embodiments, the practical limitations of each type of scaffold are taken into account when selecting options. In one or more instances, learning gains from different types of scaffolds are evaluated within the ITS (or externally by other methods) (preferably but not necessarily continuously; this can also be evaluated at regular intervals, such as once a month). In one or more embodiments, the correlation of learning gains with scaffolds or combinations thereof is then continuously fed back to update a map or model for a group of students (e.g., students with similar characteristics). Learning outcomes can but need not be assessed in the same ITS; for example, they can be assessed externally through multiple-choice tests, etc. Differences may be noticed when dealing with different subjects (e.g., mathematics and art) and / or different groups of students.

[0051] Those skilled in the art of ITS know how to determine if something is a partially correct answer; for example, using gap tokens. Keyword matching is a common suitable method. Some embodiments enhance keyword matching by using word embeddings to determine words that are similar to each other (e.g., synonyms or words that are similar in the context of the task); in this case, both exact matches and synonym matches are scored. As used herein, a completely correct answer has a gap of 0%; an incorrect answer has a gap of 100% (no similarity at all); a partially correct answer is any answer that is not completely correct or incorrect; for example, from 1% to 99% without considering decimal places.

[0052] In some embodiments, starting with no data, a reinforcement learning algorithm (discussed below) is used for model 411, and goals are set on simple metrics about student learning. For reinforcement learning algorithms, there is typically no training corpus 415 or test corpus 417 (discussed below). Later, as data about the relationship between different types of scaffolds and student learning outcomes becomes available, one or more embodiments implement a supervised learning algorithm for model 411 (e.g., a Markov decision model) in which many different scaffold combinations are tried on a trial-and-error basis; some scaffold choices lead to better results, while others lead to worse results. This approach typically requires initial data to get started. Once enough data is available, a classical machine learning algorithm (e.g., a classifier) ​​can be trained, and test and training corpora 415, 417 can be used.

[0053] Now you should refer to Figure 4 The block diagram and Figure 5In view of the discussion thus far, it will be appreciated that, in general, an exemplary method according to one aspect of the present invention includes step 505, which is performed by at least one hardware processor implementing the conversational intelligent tutor system 401 (e.g., the processing unit 16 discussed below) from a student (e.g., Figure 4 403-N) obtains a partially correct natural language answer to a question posed by the conversational intelligent tutor system (for example, in step 503). Each user can communicate with the system 401 through an appropriate corresponding user interface 403-1, 403-2...403-N; optionally, the user interface is customized for the corresponding user. A further step 507 includes accessing contextual knowledge 407 related to the student by at least one hardware processor implementing the conversational intelligent tutor system. A further step 509 includes at least one hardware processor applying a rule set 409 and at least one of a prediction model 411 to select at least one scaffolding option for the student based on the partially correct answer and the contextual knowledge. A further step 510 includes causing at least one scaffolding option to be displayed to the student (for example, via a corresponding UI). The system 401 can then be operated to continue tutoring the student according to the selected scaffolding option.

[0054] Prediction model 411 may be supervised by using training corpus 415. A portion of the data may be reserved for testing corpus 417 to check the performance of the model.

[0055] In decision block 511, a determination is made as to whether it is time to evaluate the gains and update the rule set and / or model. This may be done continuously, and block 511 may be omitted; however, in one or more embodiments, the evaluation and updates 512, 513 are performed in batches for every predetermined number of questions. If it is not time to update (the "no" branch of 511), return to step 503. If it is time to update (the "yes" branch of 511), an additional step 513 includes at least one hardware processor iteratively updating the rule set and / or at least one of the rule sets based on a predictive model of an evaluation of learning gains associated with candidate scaffolds in a student population (e.g., evaluated in step 512). For example, the iterative update may include at least one hardware processor continuously evaluating the learning gains and updating at least one of the rule set and the predictive model. Alternatively, the iterative update may include an external entity that evaluates the learning gains, or a combination of external updates and continuous evaluation of the system may be employed. Block 413 represents two options.

[0056] Note that unless a static condition is reached, processing may be continuous, as indicated by decision block 515 , where the “no” branch is to step 503 and the “yes” branch is to terminator 517 .

[0057] In some cases, at least one hardware processor implementing the conversational intelligent tutor system collects contextual knowledge about the student as the student responds to questions posed and during at least one past interaction with the student (e.g., previous questions on the same topic).

[0058] In one or more embodiments, in the applying step, at least one bracket option includes a combination of brackets (see Figure 3 for example).

[0059] In some cases, the contextual knowledge collected by at least one hardware processor implementing the conversational intelligent tutor system includes: the size of the gap between the partially correct natural language answer and the corresponding reference answer; and a subject proficiency level corresponding to the student's proficiency level on the subject of the question asked by the conversational intelligent tutor system.

[0060] In some embodiments, the application includes applying a prediction model 411; for example, a machine learning model based on the following features: the size of the gap between a partially correct natural language answer and a corresponding reference answer; a topic proficiency level corresponding to the student's proficiency level in the topic of the question posed by the conversational intelligent tutor system; the time the student spends on the question posed by the conversational intelligent tutor system; the portion of correct answers to questions related to the topic of the question posed by the conversational intelligent tutor system that have been covered and / or the portion of incorrect answers to questions related to the topic of the question posed by the conversational intelligent tutor system that have been covered; the number of times the student has reviewed the topic of the question posed by the conversational intelligent tutor system; and the number of questions asked by the student within the topic of the question posed by the conversational intelligent tutor system.

[0061] Each user 405-1, 405-2 ... 405-N may have a computing device such as system 10, for example, as discussed below, and may access tutor system 401 through a suitable user interface 403-1, 403-2 ... 403-N as described elsewhere herein. Non-limiting examples of predictive models 411 including corpora 415, 417 have been discussed elsewhere herein. Rule set 409 may include, for example, Figure 3 , and compiled into executable code. Contextual knowledge 407 refers to knowledge related to the student; for example, proficiency level and gap size as discussed elsewhere herein, and / or other factors about the student (age, history, subject-related knowledge, etc.). Element 407 can be implemented, for example, via a database system such as a relational database management system, where knowledge about the student can be queried based on the student's identifier.

[0062] Learning gain assessment 413 can be internal or external; for example, multiple choice tests are administered to students periodically (weekly, monthly, or other desired intervals) and the before and after results are compared to assess the effectiveness of the scaffolding. Implementations may include, for example, custom programs in high-level languages, common statistical software packages along with database programs, such as those available from THEMATHWORKS INC., NATICK, MASSACHUSETTS USA The computational environment of the software. Comparison of before and after results quantitatively or only qualitatively (e.g., high or low scores).

[0063] According to the techniques disclosed herein, the conversational ITS 401 may be implemented, for example, by modifying a known commercially available conversational ITS ("conversational intelligent tutor package") (a non-limiting example is Watson Tutor available from International Business Machines Corporation, Armonk, New York, USA). Figure 5 The main routine / module ("scaffold selection framework") of the logic in the flowchart of the embodiment of the present invention is added to the known session ITS and interfaces with other components described herein to perform the various steps. This has the advantage of allowing flexible use of existing systems updated according to the technology disclosed herein.

[0064] As noted, in one or more embodiments, practical limitations of each type of scaffold are taken into account when selecting options. As will be appreciated by those skilled in the art, each type of scaffold has potential practical limitations. For example, a text link may encounter a situation where the textbook being used does not include a passage relevant to a particular question. In this case, even if a test link is the preferred form of scaffold, it cannot be used and other forms of scaffolding should be recommended, for example, by a predictive model. For FITB, a keyword is often missing, so it is best for the student to fill it in. However, what if the student has already mentioned the keyword? Prompting the student for the keyword would not be a useful strategy.

[0065] The skilled artisan will be familiar with many different types of stents; a set of illustrative, non-limiting examples have been discussed herein. One or more embodiments may be used with any type of stent and are not limited to the specific examples of stents disclosed herein.

[0066] On the other hand, a non-transitory computer-readable medium includes computer-executable instructions that, when executed by a computer implementing a conversational intelligent tutor system, cause the computer to perform or otherwise facilitate a method including any, some or all of the method steps.

[0067] On the other hand, reference is made to the Figure 6, a device includes: a memory 28; and at least one processor 16 coupled to the memory, and operable to: implement a conversational intelligent tutor system, the conversational intelligent tutor system obtains partially correct natural language answers to questions asked by the conversational intelligent tutor system from a student; and accesses contextual knowledge related to the student; based on the partially correct answers and the contextual knowledge, applies at least one of a rule set and a prediction model to select at least one scaffolding option for the student; and causes the at least one scaffolding option to be displayed to the student.

[0068] The at least one processor may further be operable to perform or otherwise facilitate any, some or all of the other method steps herein.As noted, a conversational ITS may be implemented, for example, by adding a scaffold selection framework as described herein to a conversational intelligent tutor package.

[0069] One or more embodiments of the present invention or elements thereof may be implemented in the form of an apparatus including a memory and at least one processor coupled to the memory, the processor being operable to perform the exemplary method steps. Figure 6 A computer system is described that can be used to implement one or more aspects and / or elements of the present invention, and a cloud computing node according to an embodiment of the present invention is also shown, it being understood that cloud, non-cloud and hybrid embodiments are possible. Now referring to Figure 6 , cloud computing node 10 is only one example of a suitable cloud computing node and is not intended to imply any limitation on the scope of use or functionality of the embodiments of the invention described herein. Regardless, cloud computing node 10 is capable of implementing and / or performing any of the functions set forth above.

[0070] In the cloud computing node 10 there is a computer system / server 12, which can operate with many other general or special purpose computing system environments or configurations. Examples of well-known computing systems, environments and / or configurations that may be suitable for use with the computer system / server 12 include, but are not limited to, personal computer systems, server computer systems, thin clients, fat clients, handheld or laptop devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputer systems, mainframe computer systems, and distributed cloud computing environments including any of the above systems or devices, etc.

[0071] Computer system / server 12 may be described in the general context of computer system executable instructions (e.g., program modules) executed by a computer system. Generally, program modules may include routines, programs, objects, components, logic, data structures, etc. that perform specific tasks or implement specific abstract data types. Computer system / server 12 may be practiced in a distributed cloud computing environment where tasks are performed by remote processing devices that are linked through a communications network. In a distributed cloud computing environment, program modules may be located in local and remote computer system storage media including memory storage devices.

[0072] like Figure 6 As shown, the computer system / server 12 in the cloud computing node 10 is presented in the form of a general-purpose computing device. The components of the computer system / server 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0073] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor or a local bus using any of a variety of bus architectures. By way of example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0074] Computer system / server 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by computer system / server 12, including volatile and nonvolatile media, removable and non-removable media.

[0075] The system memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer system / server 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write to a non-removable, non-volatile magnetic medium (not shown, commonly referred to as a "hard drive"). Although not shown, a magnetic disk drive for reading and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM or other optical medium) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data media interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of various embodiments of the present invention.

[0076] A program / utility 40 having a set (at least one) of program modules 42 may be stored in the memory 28, such program modules 42 including but not limited to an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0077] The computer system / server 12 may also communicate with one or more external devices 14 (e.g., keyboards, pointing devices, displays 24, etc.), one or more devices that enable a user to interact with the computer system / server 12, and / or any device that enables the computer system / server 12 to communicate with one or more other computing devices (e.g., network cards, modems, etc.). Such communication may be performed via an input / output (I / O) interface 22. In addition, the computer system / server 12 may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) via a network adapter 20. As shown, the network adapter 20 communicates with other modules of the computer system / server 12 via a bus 18. It should be understood that, although not shown in the figure, other hardware and / or software modules may operate with the computer system / server 12, including, but not limited to, microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0078] Thus, one or more embodiments may utilize software running on a general purpose computer or workstation. Figure 6As shown, such an embodiment may employ, for example, a processor 16, a memory 28, and an input / output interface 22 to a display 24 and external devices 14 such as a keyboard, a pointing device, etc. As used herein, the term "processor" is intended to include any processing device, such as a processing device including a CPU (central processing unit) and / or other forms of processing circuitry. In addition, the term "processor" may refer to more than one individual processor. The term "processor" is intended to include memory associated with a processor or CPU, such as RAM (random access memory) 30, ROM (read-only memory), fixed memory devices (such as hard disk drives 34), removable storage devices (such as floppy disks), flash memory, etc. In addition, the phrase "input / output interface" as used herein is intended to consider, for example, an interface for one or more mechanisms (such as a mouse) for inputting data to the processing unit, and an interface for one or more mechanisms. Provide results associated with the processing unit (such as a printer). The processor 16, the memory 28, and the input / output interface 22 may be interconnected, for example, via a bus 18, as part of the data processing unit 12. Suitable interconnections, such as via the bus 18, may also be provided to the network interface 20. An interface to interface with a computer network (eg, a network card) and a media interface to interface with appropriate media (eg, a floppy disk or CD-ROM drive) may be provided.

[0079] Therefore, as described herein, computer software including instructions or codes for executing the methods of the present invention may be stored in one or more associated storage devices (e.g., ROM, fixed or removable memory), and when ready to use, partially or completely loaded (e.g., loaded into RAM) and implemented by the CPU. Such software may include, but is not limited to, firmware, resident software, microcode, etc.

[0080] A data processing system suitable for storing and / or executing program code will include at least one processor 16 coupled directly or indirectly to memory elements 28 via a system bus 18. Program code, mass storage and cache memory 32 which provide temporary storage of at least some program code to reduce the number of times code must be retrieved from mass storage during implementation.

[0081] Input / output or I / O devices (including but not limited to keyboards, displays, pointing devices, etc.) can be coupled to the system either directly or through intervening I / O controllers.

[0082] Network adapters 20 may also be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks. Modems, cable modem and Ethernet cards are just a few of the currently available network adapters.

[0083] As used herein, including in the claims, a "server" includes a physical data processing system (e.g., Figure 6 System 12 is shown.) It will be appreciated that such a physical server may or may not include a display and keyboard.

[0084] As noted, one or more embodiments may be implemented at least partially in the context of a cloud or virtual machine environment, although this is exemplary and non-limiting. System 12 may function as cloud node 10, but may also function in non-cloud or hybrid implementations.

[0085] It should be noted that any of the methods described herein may include the additional step of providing a system comprising different software modules embodied on a computer-readable storage medium; these modules may include, for example, any or all of the appropriate elements depicted in the block diagrams and / or described herein; by way of example and not limitation, any, some or all of the modules / blocks and / or sub-modules / sub-blocks described. The method steps may then be performed using the different software modules and / or sub-modules of the system, as described above (e.g., in Figure 4 The software module and / or submodule is executed on one or more hardware processors (e.g., 16). The product may include a computer-readable storage medium, the code of which is suitable for performing one or more method steps described herein, including providing different software modules for the system.

[0086] One example of a user interface 403-1, 403-2 ... 403-N that may be used in some cases is Hypertext Markup Language (HTML) code provided by a server, etc. to a browser of a user's computing device. The HTML is parsed by a browser on the user's computing device to create a graphical user interface (GUI).

[0087] Exemplary System and Article Details

[0088] The present invention may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present invention.

[0089] A computer-readable storage medium may be a tangible device that can hold and store instructions used by an instruction execution device. A computer-readable storage medium may be, for example, but not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples of computer-readable storage media (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, such as a punch card or a raised structure in a groove on which instructions are stored, and any suitable combination of the foregoing. As used herein, a computer-readable storage medium is not to be interpreted as a transient signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., a light pulse through a fiber optic cable), or an electrical signal transmitted through a wire.

[0090] The computer-readable program instructions described herein can be downloaded from a computer-readable storage medium to each computing / processing device, or downloaded to an external computer or external storage device via a network, such as the Internet, a local area network, a wide area network, and / or a wireless network. The network can include copper transmission cables, optical fiber transmissions, wireless transmissions, routers, firewalls, switches, gateway computers, and / or edge servers. The network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in the computer-readable storage medium in each computing / processing device.

[0091] The computer program instructions for performing the operation of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages, such as Smalltalk, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. Computer-readable program instructions may be executed entirely on a user's computer, partially on a user's computer, as an independent software package, partially on a user's computer, partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., using an Internet service provider to connect via the Internet). In some embodiments, an electronic circuit, such as a programmable logic circuit, a field programmable gate array (FPGA), or a programmable logic array (PLA), may be personalized by utilizing the state information of the computer-readable program instructions, and the electronic circuit may execute the computer-readable program instructions, thereby realizing various aspects of the present invention.

[0092] Various aspects of the present invention are described herein with reference to the flow charts and / or block diagrams of the methods, devices (systems) and computer program products according to embodiments of the present invention. It should be understood that each box of the flow chart and / or block diagram and the combination of each box in the flow chart and / or block diagram can be implemented by computer-readable program instructions.

[0093] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processor of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0094] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operating steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0095] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present invention. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0096] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terms used herein are selected to best explain the principles of the embodiments, practical applications, or technical improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method comprising: Developing a machine learning model for scaffold selection by at least one hardware processor implementing a conversational intelligent tutor system by: Apply reinforcement learning models with objectives set on simple student learning indicators without using training data to obtain data on the relationship between different types of scaffolds and student learning outcomes; as well as training a supervised machine learning model using the data regarding the relationship between different types of scaffolds and student learning outcomes; obtaining, by the at least one hardware processor implementing the conversational intelligent tutor system, from a student a partially correct natural language answer to a question posed by the conversational intelligent tutor system; accessing, by the at least one hardware processor implementing the conversational intelligent tutor system, contextual knowledge related to the student, the contextual knowledge comprising a size of a gap between the partially correct natural language answer and a corresponding reference answer and a subject matter proficiency level corresponding to the student's proficiency level for the subject matter of the question posed by the conversational intelligent tutor system; Based on the partially correct answers and the contextual knowledge, the at least one hardware processor applies the supervised machine learning model to select at least one scaffolding option for the student; as well as The at least one scaffolding option is caused to be displayed to the student.

2. The method of claim 1, further comprising: The at least one hardware processor iteratively updates the supervised machine learning model based on an evaluation of learning gains associated with candidate scaffolds across a population of students.

3. The method of claim 2, wherein the iterative updating comprises: The at least one hardware processor continuously evaluates the learning gain and updates the supervised machine learning model.

4. The method of claim 2, wherein the iterative updating includes evaluating an external entity of the learning gain.

5. The method of claim 2, further comprising: The at least one hardware processor implementing the conversational intelligent tutor system collects the contextual knowledge about the student as the student answers the posed questions and during at least one past interaction with the student.

6. The method of claim 5, wherein in the applying step, the at least one bracket option comprises a combination of brackets.

7. The method of claim 5, wherein the applying of the supervised machine learning model comprises: The machine learning model is applied based on features, the features comprising: the size of the gap between the partially correct natural language answer and the corresponding reference answer; the subject matter mastery level corresponding to the student's mastery level of the subject matter of the question posed by the conversational intelligent tutor system; the time the student has spent on the subject of the question posed by the conversational intelligent tutor system; At least one of: a portion of covered correct answers to questions associated with the subject matter of the question posed by the conversational intelligent tutor system, and a portion of covered incorrect answers to questions associated with the subject matter of the question posed by the conversational intelligent tutor system; The number of times the student revisits the topic of the question posed by the conversational intelligent tutor system; and The number of times the student asked questions on the topic of the questions asked by the conversational intelligent tutor system.

8. The method of claim 1, further comprising the at least one hardware processor considering at least one practical scaffolding constraint when selecting the at least one scaffolding option for the student.

9. The method as claimed in claim 1 also includes implementing the at least one hardware processor of the conversational intelligent tutor system to determine the size of the gap between the partially correct natural language answer and the corresponding reference answer based on the number of keywords in the corresponding reference answer that are missing from the partially correct natural language answer.

10. A computer program product, comprising a computer-readable medium having computer-executable instructions embodied therein, wherein when the computer-executable instructions are executed by a computer implementing a conversational intelligent tutor system, the computer-executable instructions cause the computer to perform the steps included in the method as described in any one of claims 1 to 9.

11. An apparatus comprising: Memory; as well as at least one processor coupled to the memory, operative to: Development of a machine learning model for stent selection by: Apply reinforcement learning models with objectives set on simple student learning indicators without using training data to obtain data on the relationship between different types of scaffolds and student learning outcomes; as well as training a supervised machine learning model using the data regarding the relationship between different types of scaffolds and student learning outcomes; implementing a conversational intelligent tutor system that obtains from a student partially correct natural language answers to questions posed by the conversational intelligent tutor system; accessing contextual knowledge related to the student, the contextual knowledge comprising a size of a gap between the partially correct natural language answer and a corresponding reference answer and a subject matter proficiency level corresponding to the student's proficiency level for the subject matter of the question posed by the conversational intelligent tutor system; applying the supervised machine learning model to select at least one scaffolding option for the student based on the partially correct answers and the contextual knowledge; as well as The at least one scaffolding option is caused to be displayed to the student.

12. An apparatus as described in claim 11, wherein the at least one processor is also operative to iteratively update the supervised machine learning model based on an evaluation of learning gains associated with candidate scaffolds across student populations.

13. An apparatus as claimed in claim 12, wherein the iterative updating includes the at least one processor continuously evaluating the learning gain and updating the supervised machine learning model.

14. An apparatus as described in claim 12, wherein the at least one processor also operates to implement the conversational intelligent tutor system, which collects the contextual knowledge related to the student when the student answers the questions posed and during at least one past interaction with the student.

15. The apparatus of claim 14, wherein the at least one bracket option comprises a combination of brackets.

16. The apparatus of claim 14, wherein the applying of the supervised machine learning model comprises applying the machine learning model based on features, the features comprising: the size of the gap between the partially correct natural language answer and the corresponding reference answer; the subject matter mastery level corresponding to the student's mastery level for the subject matter of the question posed by the conversational intelligent tutor system; the time the student has spent on the subject of the question posed by the conversational intelligent tutor system; At least one of: a portion of covered correct answers to questions associated with the subject matter of the question posed by the conversational intelligent tutor system, and a portion of covered incorrect answers to questions associated with the subject matter of the question posed by the conversational intelligent tutor system; the number of times the student revisits the topic of the question posed by the conversational intelligent tutor system; as well as The number of times the student asked questions on the topic of the questions asked by the conversational intelligent tutor system.

17. The apparatus of claim 14, further comprising the at least one processor considering at least one practical scaffolding constraint when selecting the at least one scaffolding option for the student.

18. The apparatus of claim 11, wherein: The at least one processor is operative to implement the conversational intelligent tutor system by executing a conversational intelligent tutor package and a scaffold selection framework; the at least one processor operative to access the contextual knowledge associated with the student via the scaffold selection framework; The at least one processor is operative to apply the supervised machine learning model to select at least one scaffold option for the student via the scaffold selection framework based on the partially correct answer and the contextual knowledge; as well as The at least one processor operates to cause the at least one scaffold option to be displayed to the student via the conversational intelligent tutor package.

19. A computer system, comprising one or more modules, wherein the one or more modules are configured to implement the corresponding steps included in the method according to any one of claims 1 to 9.