WO2015047074A1 - A system and method for creating learning objects from a document - Google Patents

Abstract

The present invention relates to a system and method for creating a Learning Object (LO) from a document. Particularly, the system (100) automatically generates the LO at sentence level, paragraph level and topic level in different modalities such as text, image, audio and video. The system (100) comprises a mapper (130), a ranker (140) and a Learning Object Builder (150). The method for creating a Learning Object from a document includes the steps of analysing and processing a document received from a user, matching a retrieved semantic structure by sentence index with a semantic structure from a multimedia repository (120) by the mapper (130), selecting and ranking projected graphs for each type of multimedia by the ranker (140) and producing the Sentence Learning Object, Paragraph Learning Object and Topic Learning Object by the Learning Object Builder (150).

Description

A SYSTEM AND METHOD FOR CREATING LEARNING OBJECTS FROM A

OOCUMENT

FIELD OF INVENTION

The presen invention relates to a system and method for creating Learning

Object (LO) from a document Particularly, the system automatically generates tbe LO at sentence level paragraph level and topic level in different modality such as text image, audio and video. BACKGROUND OF THE INVENTION

learning Object (LO) is an entity thai is used in learning, education or training system. The main idea of Learning Object is to break educational content down into small chunks thai can be reused in various learning environments, object-oriented programming. In other words, (earning object is a representation designed to afford uses in different educational contexts. There are a few examples of system and method that use the concept of Learning Object for education purpose>

An example of such system and method is disclosed in US Publication No. 2008/0 89684 A whereby the system and method of an e-feaming authorship Is based on meta-tagged media specific learning objects. In one embodiment, a method includes searching for learning objects in a repository based on a content query data publishing the learning objects to a content management system through applying a meta-data describing a context to the learning objects to such that the learning objects become readily accessible based on a search of the context and generating a set of executable files based on the learning objects and a number of auxiliary files associated with the set of executable files to create an e-iearning course of a specific media type, in addition, the method may include comparing a context query data and the meta-data of the each of the learning objects to perform a context search which is used to selec a subset of the learning object matching the context quety data. However, the prior art does not generate learning object at a sentence level, a paragraph level and a topic level.

Another example is disclosed in US Publication No, 2004/0205578 A1 whereby a system and method converts a document to create a reusable learning object. The reusable learning object is used to create a course of instruction for an online or virtoal classroom. The reusable learning object Is shared and modified by different users for each course. A document stored on a computing platform coupled to the system is made available for the reusable learning object. The document is created by an application and contains data that niay not be in a format useable by the reusable learning object. An autopilot and an export filter are invoked by a prompt or command from the application or another source. The export filter converts the data within the documen into metadata to support an extensible mark-up language f XML") format. The autopilot queries and receives information on the document to create the reusable learning object. The reusable learning object is created by the converted data and received information. Thus, the document is available for online training purposes despite not oeing in a format supported by the reusable learning object. However, the prior art generates learning object using a metadata and information wherein the metadata is converted from a data using an export filter which is manually defined by a user.

Hence, there Is a need to improve upon the existing system and method for automated creation of Learning Object (LO) from a document. The improvement shall expand the scope of Learning Object for education purpose. SUMMARY OF INVENTION

A system (100) for creating learning Object (LO) from a document comprises of a natural language processor (110) to analyse and process a document received from a user, a multimedia repository (120) to store multimedia items in different modality, and a learning object repository (180) to store Learning Object: wherein the system (100) is characterised in that the Learning Object Creation system (100) includes a mapper (130) to perform semantic structure matching between semantic structure of the multimedia items and semantic structure of the text of the document; a ranker (140) to select the matched semantic structures, project graphs and calculate the average score of matched semantic structures; and a Learning Object Builder (150) to create a Paragraph Learning Object, a Sentence Learning Object and a Topic Learning Object.

A method for creating teaming Object (LO) from a document by using the system comprises the steps of analysing and processing a document; matching a retrieved semantic structure by sentence inde with a semantic structure from the multimedia repository (120) by a mapper (130); selecting and ranking projected graphs for each type of multimedia by a ranker (140); and producing the Sentence Learning Object, Paragraph Learning Object and Topic Learning Object by 3 Learning Object Builder ( 50).

Preferably, matching the retrieved semantic structure by sentence index with a semantic structure from the multimedia repository (120) by the mapper {130} includes the steps of retrieving the semantic structure by sentence index by the mapper (130); selecting the multimedia type and the semantic structure from the multimedia repository {120} by the mapper (130); performing a matching of both semantic structures selected from the multimedia repositor (120} by the mapper (130} by using a projection technique; storing the results in a Learning Object Repository (160) if the semantic structures match each other; and determining whether there are more sentence indexes to be selected and completed by the mapper {130},

Preferably, selecting and ranking projected graphs for each type of multimedia by the ranker (140) includes the steps of selecting the projected graphs for each type of multimedia by the ranker (140); ranking the projected graphs for each type of multimedia by the ranker (140); selecting the top ranking graph by the ranker (140); selecting the matched semantic structure by the ranker (140}: and performing an average scoring by the ranker (140) by using different ranking methods.

Preferably, producing the Sentence Learning Object, Paragraph Learning Object and Topic Learning Object by the Learning Object Builder (150) includes the steps of aggregating the Sentence Learning Object by the system (100); creating the Paragraph Learning Object from the output of aggregation; and aggregating the Paragraph Learning Object to create the Topic Learning Object.

Preferably, aggregating the Sentence Learning Object by the system (100) includes the steps of selecting the sentence Index for aggregation; selecting the fop ranked projected graph for each of the multimedia type for the respective sentence index; arranging the projected graphs according to the order of the text, video, audio and image; producing the Sentence Learning Object from the arrangement; and storing the produced Sentence Learning Object in the Learning Object repositor (188).

Preferably, creating the Paragraph Learning Object from the output of aggregation includes the steps of selecting the paragraph index for aggregation: selecting and grouping related projected graph from each Sentence teaming Object into the type of multimedia accordingly; sorting the projected paragraphs by sentence order for each multimedia type: producing the Paragraph Learning Object from the sortation; and storing the produced Paragraph Learning Object in the Learning Object repository {160).

Preferably, aggregating the Paragraph Learning Object to create the Topic Learning Object includes the steps of selecting the topic inde for aggregation: selecting and grouping related projected graph from each Paragraph Learning Object into the type of multimedia accordingly; sorting the projected paragraphs by paragraph order for each multimedia; producing the Topic Learning Object from the sortation; and storing the produced Topic Learning Object in the Learning Object repository (160), Preferably, ranking methods include a Conceptual Similarity, a Relational

Similarity, a Conceptual and Relational Similarity, a Domain Similarity and/or a combination thereof.

BRIEF DESCRIPTION OF THE DRA INGS

The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

FIG, 1 illustrates a block diagram of a system {100} for creating Learning Object (LO) from a document according to an embodiment of the present invention.

FIG, 2 illustrates a flowchart of a method for creating Learning Object <LO) from a document according to an embodiment of the present invention. FIG, 3 illustrates a flowchart of sub-steps for matching a retrieved semantic structure by sentence index with a semantic structure of the method of G> 2.

FIG. 4 illustrates a flowchart of sub-steps for selecting and ranking projected graphs S for each type of multime lia of the method of FIG, 2.

FIG, 5 illustrates a flowchart of suh-steps for producing a Sentence Learning Object, a Paragraph Learning Object and a Topic Learning Object of the method of FIG, 2. 0 FIG. S illustrates a flowchart of sub-steps for aggregating Sentence Learning Object of the Learning Object Builder (150) of the method of FIG. 5.

FIG. 7 illustrates a flowchart of sub-steps for aggregating Paragraph Learning Object of the Learning Object Builder (150) of the method of FIG. 5,

δ

FIG. S illustrates a flowchart of sub-steps for aggregating Topic Learning Object of the Learning Object Builder (150) of the method of FIG. §.

DESCRIPTION OF THE PREFFERED EMBODIMENT

0 A preferred embodiment of the present invention will be described herein below with reference to the accompanying drawings, 3n the following description, well known functions or constructions are not described in detail since they would obscure th description with unnecessary detail. 5 FIG, 1 shows a system (100) for creating Learning Object (LO) from a document according to an embodiment of the present invention. The system {100} comprises of a mapper (130), a ranker (140) and a Learning Object Builder (150). The mapper {130} performs semantic structure matching between semantic structure of the multimedia items and semantic structure of the text of the document. The0 ranker {140} is used to select and project graphs, Moreover, the ranker (140) Is used to calculate the average score of matched semantic structures. The teaming Object Builder (1SS} is used to create a Paragraph Learning Object, a Sentence Learning Object and Topic Learning Object. The system {1 0} further comprises of a nafurai language processor (110) and a multimedia repository (120). The natural language5 processor {110} analyses and processes a document received from the user. The natural language processor {110} processes the document into a set of semantic structures which represent the document. The document can be in the form of text, video, audio and image. Each type of document is represented with a respective semantic structure. A multimedia repository (120) comprise of items in different modality such as text, video, audio and image. All the items in the multimedia repository (120) are indexed and semantical structured.

Fie. 2 shows a method for creating Learning Object {LO) from a document according to an embodiment of the present invention, The method includes the steps of analysing and processing a document received from the user as in step 201 , matching the semantic structure by sentence index with a semantic structure from the multimedia repository (120) by the mapper (130) as in step 202, selecting and ranking projected graphs for each type of multimedia by the ranker {140) as in step 203 and producing the Sentence Learning Object, Paragraph Learning Object and Topic Learning Object by the Learning Object Builder {160) as in step 204.

FIG, 3 shows a flowchart of sub-steps for matching a retrieved semantic structure by sentence index wit a semantic structure of the method of FIG. 2. In step 131, the mapper (130) retrieves the semantic structure b sentence index. In step 132, the mapper (130) selects the multimedia type and thereon, the mapper (130) selects the semantic structure from the multimedia repository {120} as in step 133. Later on, the mapper (130) performs a matching of both semantic structures selected from the multimedia repository (120) by using a projection technique as in step 134. The projection technique includes the matching of th available sentence semantic structure information with one of the multimedia semantic structure, wherein the selection of the multimedia semantic structures is done by iterating all the multimedia items in the multimedia repository (120) based on the sentence index. In decision 135, if the semantic structures do not match each other, the method returns to ste 134. If the semantic structures match each other, the matched results are stored in a Learning Object Repository (160). In decision 137, the mapper (130) determines whether there are more sentence indexes to be selected and completed, if there are more sentences indexes to be selected and completed, the method returns to step 131. If there is no more sentence index to be selected and completed, the method ends. FIG. 4 shows a flowchart of sub-steps for selecting and ranking projected graphs for each type of multimedia of the method of FIG, 2. The ranker (140) selects the projected graphs for each type of multimedia, rank them and selects the top ranking graph. The ranker (140) selects the matched semantic structures and then performs an average scoring by using different ranking methods such as Conceptual Similarity, Relational Similarity. Conceptual and Relational Similarity, Domain Similarity and or a combination thereof. Th Conceptual Similarity describes the similarity between two concepts by calculating the distance between them. The distance between the two concepts is calculated by their respective positions in the concept hierarchy. The Relational Similarity describes the similarity between two concepts wherein it defines the similarity between the two relations in the graph. The distance between the two relations is calculated by their respective positions in the relation hierarchy. The Domain Similarity expresses the similarity of concepts based on the overlapping of both graphs in terms of common concepts and relations to the domain context of the document.

FIG. S shows a flowchart of sub-steps for producing a Sentence Learning Object, a Paragraph Learning Object and a Topic learning Object of the method of FIG. 2. In step 300, the Sentence Learning Object is first aggregated by the system (100). From the output of aggregation, the method continues to create the Paragraph Learning Object as in step 400. The Paragraph Learning Object is further aggregated to create the Topic Learning Object as in step 500.

Fi<3. S shows a flowchart of sub-steps for aggregating Sentence Learning Object of the Learning Object Builder (160) of the method of FIG. 5. in step 301, the sentence index is selected for aggregation. In step 302, top ranked projected graph is selected for each of the multimedia type for the respective sentence index. Later on, the projected graphs are arranged according to the order of the text, video, audio a d image as in step 303. The Sentence Learning Object is produced from the arrangement and then stored in the Learning Object repository (160). In decision 304, if there are more sentence indexes to be selected, the method returns to step 301, if there is no more sentence index to he selected, the method ends.

FIG. 7 shows a flowchart of sub-steps for aggregating Paragraph Learning Object of the Learning Object Builder (150) of the method of FiG. 5. In step 401, the paragraph index is seiected for aggregation. In step 402, a!i related projected graph from each Sentence Learning Object are seiected and grouped into the type of multimedia accordingly. The projected paragraphs are sorted by sentence order for each multimedia type as in step 403. Later on. the Paragraph Learning Object is produced from the sedation and then stored in the Learning Object repository (160). In decision 404, if there are more paragraph indexes to be seiected, the method returns to step 401. if there is no more paragraph index to be selected, the method ends. FJG. 8 illustrates shows of sub-steps for aggregating Topic Learning Object of the Learning Object Buiider (150) of the method of FIG. S. In step 501. the topic index is selected for aggregation. In step S02, all related projected graph from each Paragraph Learning Object are seiected and grouped into the type of multimedia accordingly. The projected paragraphs are sorted by paragraph order for each multimedia type as in step S03. Later on, the Topic Learning Object is produced from toe softation and then stored in the Learning Object repository (1$0). in decision 604, if there are more topic indexes to be selected, the method returns to step S01. If there is no more topic index to be seleoted, the method ends. While embodiments of the invention have been illustrated and described, it is not intended that these embodiments illustrate and describe all possible forms of the invention. Rather, the words used i the specifications are words Of description rather than limitation and various changes may be made without departing from the scope of the invention.

Claims

A system (100) for creating Learning Object (LO) from a document comprises of:

a) a natural language processor (118) io analyse and process a document received from a user,

b) a multimedia repository (120) to store multimedia items in different modality, and

c} a learning object repository (160) to store Learning Object;

wherein the system (100) is characterised in that the Learning Object Creation system (100) includes:

a) a mapper {130} to perform semantic structure matching between semantic structure of the multimedia items and semantic structure of the text of the document;

h) a ranker (140) to select the matched semantic structures, project graphs and calculate the average score of matched semantic structures; and

o) a Learning Object Builder (150) to create a Paragraph Learning Object, a Sentence Learning Object and a Topic Learning Object.

A method for creating Learning Object (LO) from a document by using the system (100) as claimed in claim 1 comprises the steps of;

a) analysing and processing a document;

b) matching a retrieved semantic structure by sentence index with a semantic structure from the multimedia repository (120) by a mapper (130);

c) selecting and ranking projected graphs for each type of multimedia by a ranker (140); and

d) producing the Sentence Learning Object, Paragraph Learning Object and Topic Learning Object by a Learning Object Builder (ISO),

The method as claimed in claim 2, wherein matching the retrieved semantic structure by sentence index with a semantic structure from the multimedia repository (120) by the mapper (130) includes the steps of:

a) retrieving the semantic structure by sentence index by the mapper (130); b) selecting the multimedia type and the semantic structure from the multimedia repository {120} by the mapper (130);

c) performing a matching of both semantic structures selected from the multimedia repository (120) by the mapper (130) by using a projection technique;

d) storing the results in a Learning Object Repository (160) if the semantic structures match each other.

The method as claimed in claim 2, wherein selecting and ranking projected graphs for each type of multimedia by the ranker (140) includes the steps of: a) selecting the projected graphs for each type of multimedia by the ranker (140):

b) ranking the projected graphs for each type of multimedia by the ranker c) selecting the top ranking graph by the ranker (140);

d) selecting the matched semantic structures by the ranker (140): and e) performing an average scoring by the ranker (140) by using different ranking methods.

The method as claimed in claim 2, wherein producing the Sentence Learning Object, Paragraph Learning Object and Topic Learning Object by the Learning Object Builder (160) includes the steps of:

a) aggregating the Sentence Learning Object:

b) creating the Paragraph Learning Object from the output of aggregation; and

c) aggregating the Paragraph Learning Object to create the Topic Learning Object.

6. The method as claimed in claim 5, wherein aggregating the Sentence Learning Object includes the steps of:

a) selecting the sentence index for aggregation;

b) selecting the top ranked projected graph for each of the multimedia type for the respective sentence index;

c) arranging the projected graphs according to the order of the text, video, audio and image; d) producing the Sentence Learning Object from the arrangement; and e) storing the produced Sentence Learning Object in a Learning Object repository (160).

The method as claimed In claim 5, wherein creating the Paragraph Learning Object from the output of aggregation includes the steps of:

a) selecting the paragraph index for aggregation;

b) selecting and grouping related projected graph from each Sentence Learning Object into the type of multimedia accordingly;

c) sorting the projected paragraphs by sentence order for each multimedia type;

d) producing the Paragraph Learning Object from the sortatlon; and e) storing the produced Paragraph Learning Object in a Learning Object repository (160).

The method as claimed in claim 5 wherein, aggregating the Paragraph Learning Object to create the Topic Learning Object includes the steps of: a) selecting the topic index for aggregation;

b) selecting and grouping related projected graph from each Paragraph Learning Object into the type of multimedia accordingly;

c) sorting the projected paragraphs by paragraph order for each multimedia;

d) producing the Topic Learning Object from the sortatlon; and e) storing the produced Topic Learning Object in a Learning Object repositor (160).

The method as claimed in claim 4_; wherein the ranking methods include a Conceptual Similarity, a Relational Similarity, a Conceptual and Relational Similarity, a Domain Similarity and/or a combination thereof >