System and method for recommending educational resource

This disclosure generally relates to systems and methods for recommending educational resources. Specifically, the present disclosure relates to applying a clustering algorithm to student data to recommend educational resources.

Grouping students by class, grouping students for specific educational activities, choosing the right materials, matching students to the best educational support, when and what educational for students Educational choices, such as deciding whether to use interventions, are based on information that is readily available to the educator making the decision.

<p><patcit num="1"><text>U.S. Patent Application No. 12 / 109,496</text></patcit></p>

<p> This disclosure is directed to recommender systems for encouraging student clustering into educational groups. Recommender systems include a processor to execute a set of programmable instructions to receive a request to cluster multiple current students into at least two educational groups according to multiple constraints. The processor also accesses student data that corresponds to multiple students, and the student data includes attribute data that describes at least one attribute for that student that corresponds to each student of the multiple students. The processor also clusters multiple current students into a selected number of clusters and outputs the results of the clustering to the user. Clustering is at least based on the homogeneity of attribute data corresponding to each current student of multiple current students and multiple constraints.</p>

<figref num="1">It is a block diagram of one exemplary educational recommender system according to the present disclosure.</figref><figref num="2">A flow chart of the methods used by the educational recommender system shown in Figure 1 to recommend educational resources to be used for one or more current students according to the present disclosure is shown.</figref><figref num="3">A flow chart of the methods used by the educational recommender system shown in Figure 1 to recommend educational resources to be used for one or more current students according to the present disclosure is shown.</figref><figref num="4">A flowchart of the method used by the educational recommender system shown in FIG. 1 for clustering a group of students into a cluster according to the present disclosure is shown.</figref><figref num="5">This disclosure illustrates a flow chart of methods used by the educational recommender system shown in Figure 1 to group students for individual learning activities and to recommend materials to use.</figref><figref num="6">This disclosure illustrates a flow chart of methods used by the educational recommender system shown in Figure 1 to group students for individual learning activities and to recommend materials to use.</figref><figref num="7">The present disclosure illustrates a flow chart of methods used by the educational recommender system shown in Figure 1 to group students for collaborative learning activities and to recommend materials to use.</figref><figref num="8">The present disclosure illustrates a flow chart of methods used by the educational recommender system shown in Figure 1 to group students for collaborative learning activities and to recommend materials to use.</figref>

First referring to Figure 1, one exemplary educational recommender system according to the present disclosure is shown, which is referred to as an educational recommender system 100 as a whole.

The recommender system 100 includes a server 102 that has access to the data storage facility 104. At least one multifunction device (MFD) 106 and at least one client computer device 108 over network 110 to request recommendations and to update the data stored in the data storage facility 104. Data communication with the server 102 and the data storage facility 104.

The network 110 can be the Internet, an intranet, Ethernet®, LAN, WAN, or any other means for providing data communication between multiple digital devices. It is also envisioned that data communication between any component of recommender system 100 can be by direct connection, which can be wired or wireless. In this example, the network 110 is the Internet.

The server 102 is a computer device having a processor 120 and all communication devices necessary to communicate with the data storage equipment 104, MFD 106 and / or the client computer device 108 directly or over the network 110. .. Server 102 includes a computer running software for providing services for collecting educational data and providing educational recommendations to client computer devices. For example, the server 102 can include a web server, a server, a mini computer, a mainframe computer, a personal computer, a mobile computer device, and the like. Processor 120 of server 102 executes recommender software modules for processing requests and providing recommendations. The functions of recommender software modules can be combined into a single module or distributed among different combinations of modules.

The data storage facility 104 includes at least one non-volatile storage device for storing the information that the server 102 needs to access to make the requested recommendations. In this example, the data storage facility 104 includes a first storage device, a student data warehouse 140 and a learning material repository 142. The software required to access the data in the data storage facility 104, including searching for the data, manipulating the data, and storing the data, may be contained within the server 102 and /. Alternatively, it may be included in the data storage facility 104. The server 102 and the data storage facility 104 can be configured as one component or as separate components for data communication.

The student data warehouse 140 stores student data about a plurality of students. Student data is virtually static, with little or rare updates, and attributes related to each student, such as personal data (eg, school enrolled, school currently enrolled, date of birth). , Gender, geographical location, socioeconomic information, health information, school districts, grades, classrooms, etc.) and empirical data that is dynamic and likely to be updated frequently.

The student data warehouse 140 can further store educator data that describes the attributes of a plurality of educators, such as teachers, professionals, and / or tutors. This data can include personal data and experience data such as teacher temperament, lesson style, experience, and evaluation by superiors.

Each of the attributes in the empirical data for students and educators can, where appropriate, be used for weighting purposes in determining homogeneity (discussed further below), eg, on a scale of 1-10. It can be associated with an intensity score that indicates the intensity of the attribute, represented by.

The learning material repository 142 contains teaching material data. Teaching material data includes, for example, physical teaching materials (eg, paper-based textbooks, worksheets, assessments, visual displays (eg, maps, graphs), etc.), and digital teaching materials (eg, computerized interactive). Includes lesson programs, examinations, games, visual displays, etc.), which are used to teach different skills and / or items at different difficulty levels. Such materials can include, for example, materials for teaching and examining subjects such as reading, mathematics, science, history, geography, and the like. The teaching material can include fixed data or dynamic data. An example of a teaching material with dynamic data is a teaching material using natural language processing (NLP), which is further described below. Teaching material data further includes data describing educational teaching material attributes such as, for example, course sections, units, state standards, recommended teaching methods, and so on.

The material data may include the actual material in the form of a document or other digital file, or may include references and / or links to the material. The material data also includes individual material or materials such as author or creator, publisher, educational style, length, difficulty level, subject or item, target skill, degree of diversity or homogeneity of the item, etc. Includes attributes associated with teaching methods. In one example, an attribute or part thereof is configured as metadata.

Student data, educator data and teaching material data are all searchable and interrelated. Student examination data is linked or associated with the student who was examined, the materials used, and the educator who taught the materials. Therefore, it is possible to associate the level of success (or the level of lack of success) with student attributes, educator attributes and teaching material attributes, which can be the overall level (based on the overall score) or the granular level (based on the overall score). (Based on the particle size data of the examination) can be included.

The MFD106 is a printer device 122, a scanner device 124, a processor 126, a user interface (UI) 128, and a data storage facility 104, a server 102 and / or a client computer device 108, either directly or via a network 110. Includes all communication devices needed to communicate. The printer device 122, scanner 124 and processor 126 can be integrated and housed in a single housing, or they can be separate components that interface with each other. The printer device 122 is any device that performs a marking output function for any purpose, such as a digital copier, a zero graph printing system, an inkjet printing system, a reprography printing system, a bookbinding machine, a facsimile machine, etc. Including the system. The marking style can include, for example, applying toner, ink, dye, etc. to the substrate or the like. The substrate can be a material such as paper, cardboard, transparent sheets, paper derivatives and the like.

The scanner device 124 of the MFD 106 includes hardware for imaging the document and generating the corresponding image data. Image data is stored with storage devices that are included with or accessible by processor 126. Processor 126 executes a scanner software module for processing image data. Processor 126 further processes the image data corresponding to the examination document, determines which markings contained in the image data correspond to human markings, and provides granularity examination data corresponding to human markings indicating the response of the examination. Run the Granularity Examination Data Generation Module (GADG) to generate. Processor 126 further provides that the processor 126 communicates with server 102 and / or data storage equipment 104, receives input via UI128 and produces output to the user, and is an educational recommendation as further described below. Run the Educational Recommender Interface (ERI) software module to allow you to process data to provide functionality.

UI128 includes user input devices such as keypads, touchpads, touch screens or equivalents thereof, and display devices such as indicator lights, display screens and the like. The UI128 can include a graphical user interface (GUI) through which the user can enter and receive information. The communication device may be integrated within the MFD 106 or may be provided as an independent unit. In addition to providing communication capabilities, communication devices can buffer data for the purpose of protecting it in the event of a failure, such as a power outage or network outage.

The client computer device 108 is a personal computer having a processor 130, a UI 132, a data storage facility 104, and all communication devices necessary to communicate with the MFD 106 and / or the server 102 directly or over the network 110. Or a computer device such as a mobile computer device. UI132 includes user input devices such as keyboards, keypads, touchpads, mice, touch screens or equivalents thereof, and display devices such as indicator lights, display screens and the like. The UI 132 can include a graphical user interface (GUI) through which the user can enter and receive information. The processor 130 is such that the processor 130 communicates with the server 102, the data storage facility 104 and / or the MFD 106, receives input via UI 132 and produces output to the user, and as described further below. Runs an ERI software module that allows you to process data to provide educational recommendation functionality. The ERI module executed by processor 130 can be an enhanced version compared to the ERI module executed by processor 126 of MFD106, both of which are described in more detail below.

Each software module, including the ERI and GAGD modules, contains a set of programmable instructions that can be executed by the corresponding processor 126 or 130. The functions of each software module can be combined into a single module, or can be distributed among different combinations of modules.

The operation of the recommender system 100 will be described below. The server 102 and the data storage facility 104, called the Educational Recommendation Service (ERS) Provider 150, serve the Client 160. Each client 160 gains access to the service provided by the ERS provider 150, for example by subscribing to the service and loading the required software on at least one MFD 106 and / or computer device 108. .. Client 160 can access the service to submit a request for educational recommendations or to submit data to be stored in the data storage facility 104. Of course, the data in the data storage facility 104 is accumulated over time by service managers who can store learning materials and attributes (eg, research-based) for non-client students. , Client 160 also accumulates by submitting data about learning materials and the students represented by that client 160. Data is associated with metadata, labels, other data fields, etc., so it can be categorized and retrieved when needed. Therefore, the data storage facility 104 may have a relatively small amount of data at the start, but accumulates a large amount of data over time.

The client 160 submits data to the data storage facility 104 via either the client computer device 108 or the MFD 106. In addition to subscribing to the service, the client 160 may also store data about the students it represents locally, for example on a local database, or partially or wholly with the ERS provider 150 for data storage. You can also rely on it. When client 160 receives or generates learning material or student data, which includes student attribute data and may also contain associated metadata, submits that data to the ERS provider 150 for storage. .. Client 160's MFD106 receives input in the form of a scanner-captured document and produces the corresponding output provided to RS Provider 150. The output can be image data such as .pdf or .tff format, or can be converted to another format, such as .doc or .wpd by Optical Character Recognition (OCR). It can be in text format.

The MFD106 also uses its GADG module to process the examination document and generate the corresponding examination data. All US Patent Application Nos. 12 / 237,692 and two additional patent applications, both filed at the same time as this application on December 19, 2008, both by De Young et al., All named "AUTOMATIC EDUCATIONAL ASSESSMENT SERVICE", Describe the system and method for processing the examination document and generating the examination data. The examination document is an answer sheet marked by the student being examined or the teacher examining the student and corresponds to the examination or examination. The answer sheet is originally provided by the data storage facility 104 and can be printed by the MFD 106.

The examination or examination will be conducted on the student and the answer sheet will be marked according to the student's response by the student or by the teacher conducting the examination. The examination can include multiple questions, including an answer input area corresponding to each question. An example of a test given to younger students to test their learned fluency in reading is for the teacher to show the student a chart similar to a visual acuity test chart containing various letters and ask the student to read each letter. It is an examination to instruct. The teacher puts a mark on each letter, which indicates speed and / or accuracy.

Therefore, the mark corresponding to the student's answer (or grade) entered by the educator or student is entered in the appropriate answer input area. The data storage facility 104 provides an unmarked version of the test template for the MFD 106, which scans the marked test. The MFD106 runs the GADG module to extract markings from marked examinations by comparing marked examinations with unmarked templates. The extracted markings corresponding to each answer input area are student-generated answers.

The extracted marking data is the annotation data extracted by either the MFD 106, the client computer device 108 or the server 102 and the annotation data corresponding to the examination performed by the data storage facility 104. Processed by comparison. The result of the comparison is the particle size examination data showing the student's grades for each question in the examination conducted. A student's performance index for a particular question can indicate whether the student answered the question correctly or incorrectly, and can also indicate the type of error the student made. Particle size examination data is stored and associated with the students who have undergone examination and the specific examination used. In addition, particle size examination data is associated with educational resources involved in teaching the subject to the examined students. The particle size examination data further covers other stored entities relating to the subject matter that falls within the scope of the examination used to generate the particle size examination data, such as the educator who conducted the examination and / or the subject of the examination. Can be associated with course sections, units or state standards included in. By providing granularity assessment data rather than a single total score, assessments can identify patterns of error, eg, to identify specific academic weaknesses.

Although the description of the particle size inspection data has been described with respect to collecting the particle size examination data via MFD106, the particle size examination data can be provided to the server 102 and / or the data storage facility 104 by any source. .. Regardless of the source, the granularity assessment data contains the results of the assessments conducted on the student, this assessment contains multiple questions to assess the current student, and the granularity assessment data includes each question of the multiple questions. Includes an independent assessment of each of the.

This illustrates the processing of a request by service provider 150 for an educational recommendation for a group of at least one current student with reference to FIGS. 2 and 3. At step 202, server 102 receives a request for recommendations on educational resources for use by one or more current students based on one or more student attributes associated with the current student. Educational resources can be actions (eg, how to group current students), teaching methods (eg, courses offered to current students), teaching materials (eg, textbooks for current students), or educators. (Current student or teacher or tutor assigned to the group). Examples of educational resources identify student risk factors such as clustering current students into groups for educational activities or assigning current students to specific teachers or tutors, and lacks in academic achievement or learning content. And suggest learning materials or educators to use.

In step 204, the server 102 accesses the data storage facility 104. In optional step 206, server 102 determines a predecessor student group that includes a plurality of predecessor students. Predecessor groups include students who are or have been in the same educational stage as the current student group (eg, in the same grade, have taken a specific exam or course related to that recommendation, etc.). Predecessor groups can be selected by determining homogeneity compared to multiple current students (see below for a detailed description of the determination of homogeneity). Leading groups include students who have achieved the success level selected within that group, such as students who score above (or below) the threshold for a particular examination, in this case indicating lack of success. Further selection can be made based on the selected criteria specified in the selection constraint, such as narrowing down to include. Selection constraints can be set by the administrator or specified in a request submitted by the user. Alternatively, the predecessor group may include all or most of the students whose student data associated with that student is stored in the data storage facility 104.

In step 208, server 102 executes a clustering algorithm for clustering either the predecessor student or the current student into at least two clusters, where two or more students are associated with them. It is clustered based on the homogeneity of at least one selected attribute of at least one attribute. One way to imagine and perceive homogeneity-based clustering is to envision a multidimensional space in which each dimension of the multidimensional space corresponds to one attribute. Each student is associated with a conceptual position in multidimensional space, which position is defined by the selected attribute of the student's student attributes. The determination of homogeneity of two students is based on the relationship between the positions associated with the two students. If two or more students are determined to meet the criteria for homogeneity, they are clustered. When clustered, at least one selected attribute associated with each student in each cluster of at least two clusters defines a conceptual area within a multidimensional space.

One type of relationship between positions is proximity, in which two students are clustered if their respective positions meet the conditions of the selected proximity. The degree of proximity and the method of determining the proximity are defined by cluster constraints or default values.

As mentioned above, homogeneity refers to the degree of similarity between students being compared, based on a combination of at least one student attribute. Different student attributes can be weighted differently to determine the degree of similarity between students. The clustering algorithm is based on "homogeneity", which involves grouping students with similarities in selected student attributes (eg, one or more student attributes) together, and further. , Differences in their attributes between clusters can be included, for example by maximizing the differences between the attributes of the average student in each cluster. Clustering algorithms can use hierarchical or non-hierarchical methods, all of which are within the scope of the present disclosure.

In this example, the clustering algorithm is a statistical clustering algorithm, such as the large-scale clustering algorithm used in Patent Document 1. As is well known to those skilled in the art, other statistical clustering algorithms useful for clustering, such as latent semantic indexing, k-means clustering, expectation maximization clustering, or other statistical clustering algorithms. The class of algorithms is also within the scope of this disclosure. In Patent Document 1, a sparse similarity matrix that encodes the similarity between clustered entities is calculated, and then a non-negative factorization is performed on the similarity matrix to perform soft cluster allocation (soft). cluster assignment) is generated.

The clustering algorithm can operate on a sparse or dense similarity matrix, depending on the type of problem being solved. For large-scale clustering problems (thousands of predecessor or current students, such as when using data from national students who took national exams one year), dense matrices are memory and computation. A sparse similarity matrix is useful because it is difficult to manage in terms of cost. Dense matrices can be useful for clustering problems involving smaller predecessors or groups of current students (eg, assigning current students to classes for the next year).

The clustering algorithm can use various non-negative matrix factorization methods. For example, probabilistic latent semantic analysis (PLSA) easily includes constraints in clustering, such as the fact that two students should or should not be in the same class the following year. It can be especially suitable for student clustering problems.

In step 210, the server 102 maps to the cluster formed in step 208 by mapping the non-clustered students to the clustered students. If the predecessor students were clustered in step 208, the current students are mapped to the cluster. Similarly, if the current student is clustered in step 208, the predecessor student is mapped to the cluster.

The mapping is based on the degree of homogeneity of the particular student attributes of each student being mapped, compared to the particular student attributes of the students in the cluster. The same method for determining homogeneity in the clustering step can be used to determine homogeneity during mapping. Students and clusters are mapped if the relationship between the location associated with the student and the area associated with the cluster meets the selected criteria given in the mapping constraint.

Mapping constraints indicate the relationships and conditions that must be met in order to map students to clusters or to map clusters to students. The mapping constraint may be the same as the clustering constraint. For example, a mapping constraint can indicate that a student is mapped to a region that includes or is closest to the location associated with that student. The closest determination can be based on the center of the region or the average position of the region. This condition is the degree of proximity (eg, preselected distance) that must be met in order to map the student to the area (or location) associated with the predecessor student. Alternatively, the proximity determination can be based on the boundaries of the region or some other feature of the region. In the case of overlapping clusters, the mapping constraint determines to which area the student should be mapped if the student's location is within the overlapping area associated with two or more clusters. Priority schemes or other factors can be included. Mapping constraints can be adjusted interactively by the user until the user is satisfied with the result. The algorithm used for mapping should follow a clustering algorithm.

The predecessor students involved in the mapping are predecessor students selected based on the level of grade that indicates the measure of success specified by the success constraint. Success is indicated by some measure, and according to success constraints, success in departments, department subsets, behavioral functions, emotional functions (eg, satisfaction), etc. can be indicated. Success constraints are included in the request or are determined by the administrator or default values. Success constraints can be adjusted interactively by the user until the user is satisfied with the result. Success constraints determine whether and / or when to select predecessors who will be included in the clustering and / or mapping steps, for example, based on the achievement of success in the academic, behavioral or emotional domain. You can specify whether to do it.

The specified success threshold may be low, and the success constraint specifies that the selected predecessor is selected based on showing a success below the specified threshold. It may be. It can be used for a variety of purposes, such as to identify and / or eliminate educational resources that have proven unsuccessful to similar students. In step 212, each of the current students is correlated with the resource associated with the predecessor student or cluster of predecessor students to which the current student is mapped. If the predecessor students are clustered in step 208 and the current students are mapped to the cluster in step 210, the current students can be correlated with more than one educational resource used by the students in the cluster. Yes, educational resources are ranked based on the success achieved by the students in the cluster. If current students are clustered in step 208 and predecessor students are mapped in step 210, each student in the cluster is correlated with the educational resources associated with the predecessor student mapped to that cluster, and if the mapping stage. If the ranking is made in, the correlation can specify the ranking of the preceding students.

At step 214, the server generates educational recommendations in response to the request. Educational recommendations include recommendations that encourage each current student of at least one current student to use one or more determined educational resources that correlate with that student. In step 216, the recommendation is communicated to the user. In step 218, the server 102 receives feedback from the user and determines whether the feedback includes acceptance of recommendations or adjustment of requirements.

The user submits feedback from the client side 160 by manipulating the corresponding UI 128 or 132 via either the MFD or the client computer device 108. UI128 or 132 may include a GUI that allows the user to adjust the request by visually adjusting the constraints and weights associated with the student attributes as well as the weights contained in the current request. it can.

Note that the current request is replaced by the tuned request, but all tuned versions of the original request may be accessible to the user. If the user chooses to coordinate the request, control goes to step 220. At step 220, adjustments to the request are received. At step 222, it is determined whether the adjustments to the requirements received in step 220 include criteria for selecting students from the predecessor group. If included, control is passed to step 206 for selecting from the predecessor groups, otherwise control is passed to step 208 for clustering the selected predecessor groups. If the user accepts the recommendation, the procedure ends and the server waits to receive another request.

Therefore, the recommendation process is interactive, allowing the user to make adjustments to constraints, constraints weights, etc., and to see the results. From each of the results produced during each iteration of the dialogue process, the user considers the best set of results for the user (eg, student grouping and / or educational recommendations for learning materials to use). ) Can be selected.

Adjustable constraints and their associated weights control how clustering and mapping are performed and the selection of students for whom clustering and mapping is performed. Initial and adjusted values for each of the constraints and associated weights are selected. The term "selected" in this sense means a user's choice, a choice based on the execution of an algorithm, a choice by a processor, or a choice according to a default value (eg, set by the manufacturer or administrator). Represents that.

This illustrates another example of how service provider 150 processes a request for educational recommendations for a group of at least one current student with reference to Figure 4. In step 402, server 102 receives a request for recommendations for clustering a group of current students into a selected number of clusters based on multiple clustering constraints. Such requirements can be used, for example, by an administrator performing class placement for next year's students, which may include assigning teachers and / or students to specific classes, or, for example, subjects. It can be used by educators to divide a class into individual groups to teach (eg, reading or math) in smaller groups, or to carry out group activities or projects.

Requests identify a group of current students and form the type of recommendation required (eg, classroom and / or assignment, group activity assignments that include or do not include requirements for materials that each group should use). Specifies the selected number of clusters to be and the clustering constraints. Clustering constraints include logistics constraints, current student attribute constraints, and / or educational resource constraints. The request may further include weights added to each of the clustering constraints. If no weights are added, the default weights will be used during the clustering process.

In step 404, the server 102 accesses the data storage facility 104 and accesses attributes relating to the current student. If the request specified an attribute for the educational resource, the server selects the educational resource that has the most similar or closest attribute to the attribute specified in the request, and clusters the selected educational resource. Used to do. When the request specifies an educational resource, but the server determines that the data related to the educational resource specified in the request is not stored in the data storage facility 104 (or not enough data is stored). Server 102 selects the surrogate educational resource that most closely matches the educational resource specified in the request. Deputies are selected based on the attributes of the surrogate and the attributes of the educational resource specified in the request. The educational resource attributes may be specified in the request, or the server 102 provides the attributes to the user when it discovers that data about the specified educational resource is not available in the data storage facility 104. May be urged.

In another situation, the server 102 is sufficient if it determines that the data on the success of the specified educational resource available at the data storage facility 104 is inadequate (very little or no data exists). Data or large amounts of data can be selected to act as a surrogate for educational resources stored in the data storage facility 104. Criteria for selecting a surrogate may be selected by the administrator or specified in the request. If the need for a proxy is determined after the request is first submitted, the user may be encouraged to provide criteria for selecting a proxy. Information submitted by the user through prompts after submitting the initial request can be considered contained in the request for the purposes of this example.

In step 406, server 102 executes a clustering algorithm for clustering a group of current students into a selected number of clusters (eg, at least two clusters) based on clustering constraints. Therefore, clustering is based on the homogeneity of student attributes, eg, student attributes specified in the student attribute constraint condition, and further satisfies the logistic constraint condition specified in the request. Each of the current students in a cluster is associated with a position in a multidimensional space that has the dimensions defined by the student attributes specified in the student attribute constraints. Clustering involves clustering current students whose relationships with each other satisfy the selected conditions according to selectable weighted clustering constraints, as described with respect to FIGS. 2 and 3.

Server 102 is a cluster of current students based on the criteria for matching students and resources given as matching constraints in the request (or the default criteria for matching if nothing is specified). Match the attributes of for each educational resource or associated surrogate attribute. Each cluster of current students is mapped to the best matching educational resource. Educational resources mapped to each cluster of current students are recommended for that cluster of current students.

If the request includes a request that uses historical information about the success of using an educational resource, the data storage facility has reviewed the data of the predecessor student and experienced each identified educational resource. Find a student. Predecessor student data can be filtered based on selection constraints, which are either provided in the request or selected by the administrator.

More specifically, in this example, the analysis of each educational resource identified in the request is, for example, either provided in the request or selected by the administrator. All successful (or unsuccessful) predecessors using their educational resources or associated surrogate according to some preselected threshold level specified in), eg, as indicated by the examination data. Implemented by selection. Predecessor students can choose based on successful grades indicated by above a relatively high threshold or unsuccessful grades indicated by falling below a relatively low threshold.

The selected predecessor students are clustered into the number of clusters specified in the request based on the clustering constraints. Clustering is based on the same student attribute homogeneity that was used to cluster current students. The current student cluster is mapped to the previous student cluster according to mapping constraints, based on the same student attribute homogeneity that was used to cluster the current student and the previous student. Correlate each cluster of current students with the educational resources successfully used by the preceding student cluster to which the cluster of current students is mapped (or on behalf of them educationally).

In step 408, a request-responsive recommendation is generated based on the result of clustering. Recommendations include clusters formed by current students. Recommendations can also include materials that correlate with each cluster. At step 410, the recommendation is communicated to the user to cluster the current student into the cluster generated in step 406.

In step 412, the server 102 waits for feedback from the user and determines whether the feedback includes acceptance of recommendations or adjustment of requirements. The user operates the corresponding UI 128 or 132 from either the client side 160 via either the MFD or the client computer device 108, eg, the constraints and weights associated with the student attributes, and are included in the request. Submit feedback by adjusting constraints. Adjustable constraints include clustering constraints, mapping constraints, selection constraints, success constraints, criteria for matching students to resources, and so on. In addition, constraints can be removed or added. Note that the current request is replaced by the tuned request, but all tuned versions of the original request may be accessible to the user.

If the user chooses to coordinate the request, control goes to step 414. At step 414, server 102 receives adjustments to the request, after which control is passed to step 408 for clustering current students. If the user accepts the recommendation, the procedure ends in step 416 and the server 102 waits to receive another request.

Therefore, the recommendation process is interactive, allowing the user to make adjustments to constraints, constraints weights, etc., and to see the results. From each of the results produced during each iteration of the dialogue process, the user considers the best set of results for the user (eg, student grouping and / or educational recommendations for learning materials to use). ) Can be selected.

The constraints and their associated weights control how clustering, mapping, and selection of students to perform clustering and mapping are performed. Initial and adjusted values for each of the constraints and associated weights are selected. The term "selected" in this sense means a user's choice, a choice based on the execution of an algorithm, a choice by a processor, or a choice according to a default value (eg, set by the manufacturer or administrator). Represents that. Throughout this application, the term "preselected" has the same meaning as "selected" as described herein.

100: Educational recommender system 102: Server 104: Data storage equipment 106: Multi-function device 108: Client computer device 110: network 120, 126, 130: Processor 122: Printing device 124: Scanner device 128, 132: User interface 140: Student Data Warehouse 142: Learning materials repository 150: Educational Recommender Service Provider 160: Client