JP6256317B2 - Answer scoring device and answer scoring program - Google Patents

Description

  The present invention relates to an answer scoring device and an answer scoring program for automatically scoring students' answers to test questions.

  Conventionally, various devices for automatically scoring answers have been made.

  For example, in the technique disclosed in Patent Document 1, in order to analyze whether or not an item desired to be learned is understood and a tendency thereof, first, a question sentence including a small item that the examinee should understand is created. To do. Each sub-item includes a collection of numerical values called “terms”, expressions, sentences, words, and the like.

  Then, create some free answer examples for the question sentence, create a checklist to check which terms are included in each answer example, and check which check boxes are checked. And create multiple answer patterns.

  Then, when scoring the actual answer by the student, the check box of the created list is checked. Thereby, it is determined which answer pattern the actual answer is the same as. Finally, the answer patterns of the answer patterns are aggregated to analyze the students' understanding tendency.

JP 2009-229606 A

  The point on the automatic scoring in Patent Document 1 described above is a collection of numerical values that will be included in the correct answer, an expression, a sentence, a word, or the like as a character string called “term”, and whether or not the answer is correct depending on the presence of this term This is the point of making a judgment.

  Thus, since the technique of Patent Document 1 performs processing using a specific character string, it is considered suitable for a test problem in which an answer such as an essay or a national language is written in text.

  At least for mathematical problems that would include graphs and figures in the answer, the answer could not be scored properly.

  In view of the circumstances as described above, an object of the present invention is to provide an answer scoring device and an answer scoring program capable of automatically scoring answers including graphs and figures.

  In order to achieve the above object, an answer scoring device according to an aspect of the present invention stores a model answer to a given question and an answer created for the question, and is stored in the storage unit. In the model answer, as a unit to be used for scoring, a statement object is extracted from the sentence part of the model answer, a table object is extracted from the table part of the model answer, and a graph part or a figure of the model answer The graph object is extracted from the part, and the extracted discussion object, the table object, and the graph object are stored in the storage unit as a model answer object table in a specific order, and the stored in the storage unit In the answer, the discussion object is extracted from the sentence part of the answer as the object. The table object is extracted from the table portion of the answer, the graph object is extracted from the graph portion or the graphic portion of the answer, and the extracted discussion object, the table object, and the graph object are extracted in the specific order. The answer analysis unit to be stored in the storage unit as an answer object table, and the statement object, the table object, and the graph stored in the specific order between the model answer object table and the answer object table An answer evaluation unit that compares objects and performs automatic scoring. Therefore, it is possible to automatically score answers including graphs and figures.

  In order to achieve the above object, in the answer scoring device according to one aspect of the present invention, when the answer analysis unit extracts the table object from the table part, each of the elements included in each table constituting the table May be extracted as an independent table object.

  In order to achieve the above object, in the answer scoring device according to one aspect of the present invention, when the answer analysis unit extracts the graph object from the graph part or the graphic part, the line constituting the graph or the graphic is displayed. In analyzing, analyzing characters included in the graph or the graphic, analyzing the relation between the line and the character based on the analysis result of the line and the analysis result of the character, and creating the answer A configuration may be employed in which a point portion is analyzed, and the graph objects obtained as a result of the analysis are collected to generate a graph object table.

  In order to achieve the above object, an answer scoring program according to an aspect of the present invention is an example of the above-mentioned model answer as an object to be used for scoring in a model answer to a question that is stored in a storage unit. Extract a statement object from a sentence part, extract a table object from a table part of the model answer, extract a graph object from a graph part or a graphic part of the model answer, extract the statement object, the table object, and the extracted object A graph object is stored in the storage unit as a model answer object table in a specific order, and in the answer to the question stored in the storage unit, as the object, the statement object from the sentence part of the answer And extract the table object from the table part of the answer The graph object is extracted from the graph part or graphic part of the answer, and the extracted discussion object, the table object, and the graph object are stored in the storage unit as an answer object table in the specific order. And the computer performs a procedure for automatic scoring by comparing the discussion object, the table object, and the graph object stored in the specific order between the model answer object table and the answer object table. Let

  As described above, according to the present invention, an answer including a graph and a figure can be automatically scored.

It is a block diagram in case an answer scoring apparatus is comprised with a common computer. It is a figure of the functional block implement | achieved when a program is run in CPU11. It is a block diagram of model answer DB17a and answer DB17b. It is a figure which shows process content list PCL showing the outline | summary of the process which the answer analysis part 11a extracts an object from an answer. It is a figure which shows the example of the question sentence given a question. It is a figure which shows the example of the answer with respect to the questioned question sentence. It is a figure which shows the text part which extracts a statement object among the examples of an answer. It is a figure of the statement attribute list SAL for determining the attribute of a statement object. It is a figure which shows the location extracted as an argument object among the examples of an answer. It is a figure which shows the example of the answer object table | surface AOT with which the statement object was described. It is a figure which shows the table | surface part which extracts a table | surface object among the examples of an answer. It is a figure which shows the table | surface object extracted from the increase / decrease table | surface TID. It is a figure which shows the example of the answer object table | surface AOT with which the discussion object and the table | surface object were described. It is a figure which shows the graph part which extracts a graph object among the examples of an answer. It is a figure which shows the example of the graph object extracted from the graph GRH. It is a figure which shows the example of the graph object table | surface GOT which collected all the graph objects extracted from the graph GRH. It is a figure which shows the example of the answer object table | surface AOT in which the argument object, the table object, and the graph object were described. It is a flowchart for demonstrating the flow of the whole process in the answer evaluation part 11b. It is a flowchart for demonstrating the flow of a comparison process with the model answer CA and the answer AS in the answer evaluation part 11b.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[Overview]
First, an outline of an answer scoring device according to an embodiment of the present invention will be described.

  The answer scoring device according to an embodiment of the present invention automatically scores an answer in a descriptive format including a graph and a table by comparing it with a prepared model answer. In order to properly compare the graph and table parts with the model answer, extract detailed objects from the graph and table, and compare them with the objects extracted from the model answer. Even so, automatic scoring can be performed appropriately.

  The outline of the answer scoring device according to the embodiment of the present invention has been described above.

[Constitution]
Next, the configuration of the answer scoring device according to an embodiment of the present invention will be described. The answer scoring device may be configured by dedicated hardware or software, or may be configured by a general computer. FIG. 1 shows a configuration diagram in the case where the answer scoring device is configured by a general computer.

  As shown in the figure, the answer scoring device 10 includes a central processing unit (CPU) 11, a read only memory (ROM) 12, a random access memory (RAM) 13, an operation input unit 14, a communication unit 15, a display unit 16, And a storage unit 17, and these blocks are connected via a bus 18.

  The ROM 12 stores a plurality of programs and data such as firmware for executing various processes. The RAM 13 is used as a work area for the CPU 11 and temporarily holds an OS (Operating System), various applications being executed, and various data being processed.

  The storage unit 17 is, for example, an HDD (Hard Disk Drive), a flash memory, or other nonvolatile memory. The storage unit 17 stores an OS, various applications, various data, a model answer DB (Database) 17a and an answer DB 17b described later.

  The communication unit 15 is connected to a network for exchanging information with an external device.

  The CPU 11 develops a program corresponding to an instruction given from the operation input unit 14 among the plurality of programs stored in the ROM 12 or the storage unit 17 in the RAM 13, and the display unit 16 and the storage unit according to the expanded program. 17 is appropriately controlled.

  The operation input unit 14 is, for example, a pointing device such as a mouse, a keyboard, a touch panel, and other operation devices.

  The display unit 16 is, for example, a liquid crystal display, an EL (Electro-Luminescence) display, a plasma display, or the like.

  Next, functional blocks realized by executing a program in the CPU 11 will be described. FIG. 2 is a functional block diagram realized by executing a program in the CPU 11.

  The functional blocks realized in the CPU 11 of the answer scoring device 10 are an answer analysis unit 11a and an answer evaluation unit 11b.

  The answer analysis unit 11a analyzes the model answer CA prepared in advance and the answer AS created by the student, and extracts an object which is a unit for comparing the model answer CA and the answer AS. The extracted objects are stored in the model answer object table CAOT of the model answer DB 17a and the answer object table AOT of the answer DB 17b.

  The answer evaluation unit 11b compares the answer AS of each student and the model answer CA, and evaluates and scores the answer AS. Specifically, the objects described in the model answer object table CAOT of the model answer DB 17a and the objects described in the answer object table AOT of the answer DB 17b are compared, and it is determined for each object whether or not they are the same. . The scoring result is stored in the answer scoring result SCR of the answer DB 17b.

  Next, the model answer DB 17a and the answer DB 17b stored in the storage unit 17 will be briefly described. FIG. 3 is a configuration diagram of the model answer DB 17a and the answer DB 17b.

  The model answer DB 17a includes one or more model answers CA prepared in advance and a model answer object table CAOT in which objects extracted from the model answer CA by the answer analysis unit 11a are stored in a table format.

  The answer DB 17b is an answer AS created by the student, an answer object table AOT in which objects extracted from the answer AS by the answer analysis unit 11a are stored in a table format, and a score result of each answer AS by the answer evaluation unit 11b. It includes an answer scoring result SCR.

  The configuration of the answer scoring device 10 has been described above.

[Outline of object extraction processing]
Next, an outline of processing in which the answer analysis unit 11a extracts an object from an answer or the like will be described. FIG. 4 is a diagram illustrating a processing content list PCL that represents an outline of processing in which the answer analysis unit 11a extracts an object from an answer or the like. In the figure, object extraction is called objectification.

  The procedure shown in the processing content list PCL shows a method of extracting a part (object) as a key point from sentences, tables, and graphs described in the answer AS and the model answer CA.

  In the procedure shown in the processing content list PCL, the answer analysis unit 11a first analyzes the sentence portion described in the answer AS (execution order 1).

  In the analysis in the procedure of the execution order 1, first, the sentence is divided into the optimum length for the comparison of the answer AS. Next, sentences relating to mathematical expressions and sentences referring to graphs and tables are extracted as objects. Next, we analyze the relationship and dependency between sentences.

  In addition, when the attention number ("Formula 1", "Formula 2", etc.) for showing the described numerical formula etc. is described in the sentence, the answer analysis unit 11a also extracts the number.

  Next, the answer analysis unit 11a clarifies the relationship between the analyzed sentence and the tables and graphs referenced in the sentence (execution order 2).

  In the processing in the order of execution order 2, it is confirmed whether the information expressed by the chart and the information described by the text are consistent.

  Next, the answer analysis unit 11a analyzes a graph (figure) (execution order 3 to 6).

  In the analysis of a figure, first, an analysis is performed on the lines constituting the figure (execution order 3).

  In line analysis, whether a target line is a figure component, a part of a graph, a base axis (X axis, Y axis, etc.), an auxiliary line, a figure or graph Analyze whether the line is for explanation.

  Next, the answer analysis unit 11a analyzes the characters described in the graphic (execution order 4).

  In character analysis, analyze whether a character or character string described in a figure explains the scale (size) or coordinates of a line, surface, or solid drawn in the vicinity. Analyzes whether it represents the name of a drawn line, surface, or solid.

  Next, the answer analysis unit 11a further analyzes the intersections, vertices, extreme points, etc. of the graphic based on the analysis result of the execution order 3, and extracts all objects in the graphic (execution order 5). Here, we analyze the point that is the key to answering the question.

  In the execution order 5, the similarity of the graphic and the parallelism of the graph may be analyzed, or the characteristics of the trajectory of the graph may be analyzed.

  At the end of the graphic analysis, the answer analysis unit 11a structures the objects in the extracted graphic (execution order 6).

  Specifically, the structuring aligns objects related to the extracted graphic. This is for improving convenience when comparing the answer AS and the model answer CA.

  Next, the answer analysis unit 11a analyzes the table (execution order 7).

  In the analysis of the table, an analysis is performed regarding what the table is, what the numbers and symbols described have, and what the order of description means.

  The above is the outline of the procedure for extracting an object from the answer AS or the model answer CA. Note that. Since the analysis of the graphic and the analysis of the table are analyzed in the order referred to in the sentence written in the answer AS, they are not always executed in the execution order described above.

  The outline of the process in which the answer analysis unit 11a extracts the object from the answer has been described above.

[Specific example of object extraction processing]
Next, a specific example of the object extraction process described above will be described. In the following description, it is assumed that the problem shown in FIG. 5 “find the local maximum / minimum value of the curve y = x ³ −29x ² −60x + 900 on the plane” is given, and the answer AS shown in FIG. Assume that it was obtained.

  The answer AS shown in FIG. 6 includes a part described by sentences, a part described by an increase / decrease table, and a part described by a graph.

<Extracting objects from text>
First, as shown in FIG. 7, an object is extracted for a portion described by a sentence. Note that an object extracted from a sentence is called a statement object.

  In order to extract a discussion object, it is necessary to analyze the relationship between each discussion object candidate (a character string of a length that can be an object) in the sentence and determine the attribute of each discussion object. Therefore, the discussion attribute list SAL shown in FIG. 8 is used.

In the discussion attribute list SAL, a specific example of the description for determining the attribute is associated with the attribute of the discussion object. For example, in the first line of the answer, there is a description “if f (x) = x ³ −29x ² −60x + 900”. Therefore, a specific example of the argument attribute list SAL includes “to be placed”. It can be seen that the attribute of the discussion object is “definition”.

  In this way, the answer analysis unit 11a extracts the statement object 1 to the statement object 6 as the statement object, as shown in FIG. 9, and refers to the statement attribute list SAL to the statement object 6 from the statement object 1 to the statement object 6. Determine the attributes of.

  In the statement attribute list SAL shown in FIG. 9, “reason”, “condition”, “conversion”, “parallel”, “insertion”, “substitution”, and “description” as attributes of the statement object. There is a pronoun.

  Further, as shown in FIG. 10, the answer analysis unit 11a generates a table (answer object table AOT) related to the extracted six discussion objects.

  In the answer object table AOT illustrated in FIG. 10, an attribute, an object to be used and a usage method, and a figure object / table object that refers to the used object are associated with each extracted object. For example, the attribute of the statement is “definition” with respect to the statement object 1, for example, with respect to the statement object 4, the attribute of the statement is “reason”, the object to be used is the statement object 3, and the table ( It can be seen that the change table is referenced.

  With the above procedure, the answer object table AOT is in a state where the descriptive object extracted from the sentence portion of the answer AS is described.

<Extracting objects from the table>
Next, as illustrated in FIG. 11, the answer analysis unit 11 a extracts an object for a portion expressed by a table (increase / decrease table TID). An object extracted from the table is called a table object.

  When extracting a table object, first, each cell is represented as (i, j) in order to distinguish each cell (one divided by vertical and horizontal lines) constituting the table. The upper left cell is set to (0, 0). When going to the right from there, i is increased, and when going downward, j is increased.

  In this way, the table object extracted from each cell can be expressed as shown in FIG.

  For example, since there is a description “x” in the upper left cell, “(0,0) = x”, and there is a description “minimum” in the cell in the third row from the top and the fifth column from the left. , “(4,2) = minimum”.

  The answer analysis unit 11a refers to the answer object table AOT. In the answer object table AOT, it can be seen that the table object extracted as described above is referred to by the argument object 4, so that the answer analysis unit 11a converts the table object into the answer object table AOT as shown in FIG. Insert in the line between the discussion object 4 and the discussion object 5.

  By the above procedure, the answer object table AOT is in a state in which the discussion object extracted from the sentence part of the answer AS and the table object extracted from the increase / decrease table part are described.

<Extracting objects from the graph>
Next, as shown in FIG. 14, the answer analysis unit 11a extracts an object for a portion expressed by the graph GRH. An object extracted from the graph is called a graph object.

  The extraction of the graph object is performed by the processes 3 to 6 in the processing content list PCL described above. FIG. 15 is a diagram illustrating an example of a graph object extracted from the graph GRH. As the extracted graph object, for example, the X axis is the base axis GO1, the letter “x” attached to the X axis is the description GO12 indicating the X axis, and the curve of f (x) is “line graph (N type ) ”As GO7, the point (20, f (20)) is extracted as the minimum value GO14.

  All the graph objects extracted from the graph GRH are aligned and collected in the graph object table GOT shown in FIG. 16 as information on f (x). From the graph object table GOT, it can be seen that the type of the graph is N, the maximum value is obtained with x = -1, and the minimum value is obtained with x = 20.

  The answer analysis unit 11a refers to the answer object table AOT. In the answer object table AOT, since the graph object extracted as described above indicates that the graph GRH is referred to by the statement object 5, the answer analysis unit 11a converts the graphic object as shown in FIG. It is inserted in the row between the statement object 5 and the statement object 6 in the answer object table AOT.

  Through the above procedure, the answer object table AOT includes a descriptive object extracted from the sentence part of the answer AS, a table object extracted from the increase / decrease table part, and a graph object extracted from the graph part in a specific order. It will be described and completed.

  The specific example of the object extraction process has been described above. The model answer object table CAOT is created in the same procedure.

[Flow of processing in answer evaluation unit 11b]
Next, the flow of processing in the answer evaluation unit 11b will be described. FIG. 18 is a flowchart for explaining the overall processing flow in the answer evaluation unit 11b, and FIG. 19 is for explaining the flow of comparison processing between the model answer CA and the answer AS in the answer evaluation unit 11b. It is a flowchart.

  It is assumed that the extraction of the objects from the model answer CA and the answer AS is completed before the process in the answer evaluation unit 11b is performed.

  Further, here, one answer AS is compared with the prepared n model answers CA, but when there are a plurality of answer AS, the process described below is repeated for each answer AS. Needless to say.

  In the following description, there are n types of model answers CA, the first model answer CA is expressed as a model answer (1), and the nth model answer CA is expressed as a model answer (n).

  Each model answer CA has two variables, a correct flag and an object match number, and each variable belonging to the nth model answer is expressed as a correct flag (n) and an object match number (n).

  Further, in one answer object table AOT or one model answer object table CAOT, an object described first is an object (1), and an object described mth is an object (m).

<Overall processing flow>
This will be described with reference to FIG.

  First, the answer evaluation unit 11b sets a variable n to 1 (step S1).

  Next, the answer evaluating unit 11b determines whether or not the model answer (n) exists (step S2).

  When the model answer (n) exists (YES in step S2), the answer evaluation unit 11b performs a comparison between the model answer (n) and the answer AS for each extracted object (step 3).

  Next, the answer evaluation unit 11b increases n by 1 and returns the process to step S2 (step S4).

  When the model answer (n) does not exist (NO in step S2), the answer evaluation unit 11b stores the scoring information of the answer AS in the storage unit 17, and ends the process (step S5).

<Flow of comparison between model answer (n) and answer AS>
This will be described with reference to FIG.

  First, the answer evaluation unit 11b sets a variable m to 1 (step S10).

  Next, the answer evaluating unit 11b sets the object coincidence number (n) to 0 (step S11).

  Next, the answer evaluation unit 11b sets the correct answer flag (n) to YES (step S12).

  Next, the answer evaluating unit 11b determines whether or not there is an mth object in the model answer object table CAOT of the model answer (n) and the answer object table AOT of the answer AS (Step S13).

  When there is no m-th object (NO in step S13), it is considered that the comparison with one model answer (n) is completed, and the number of m-1 is set as the number of object matches (n), and the entire process (Step S17).

  When there is the m-th object (YES in step S13), the answer evaluation unit 11b determines whether or not the m-th object in the model answer object table CAOT of the model answer (n) and the answer object table AOT of the answer AS are equal. Judgment is made (step S14).

  If the mth objects are equal (YES in step S14), the variable m is incremented by 1 to compare the next objects, and the process returns to step S13 (step S15).

  If the mth objects are not equal (NO in step S14), the answer evaluating unit 11b sets NG to the correct flag (n) (step S16), and sets the number of m-1 to the object match number (n). Then, the process returns to the entire process (step S17).

  The flow of processing in the answer evaluation unit 11b has been described above.

[Additional notes]
In addition, this invention is not limited only to the above-mentioned embodiment, Of course, various changes can be added within the range which does not deviate from the summary of this invention.

10 ... answer scoring device 11 ... CPU
11a ... answer analysis unit 11b ... answer evaluation unit 12 ... ROM
13 ... RAM
DESCRIPTION OF SYMBOLS 14 ... Operation input part 15 ... Communication part 16 ... Display part 17 ... Memory | storage part 18 ... Bus

Claims (4)

A storage unit for storing an exemplary answer to the question that has been presented and an answer created for the question;
In the model answer stored in the storage unit, as an object as a unit for scoring, extract a statement object from the sentence part of the model answer, extract a table object from the table part of the model answer, A graph object is extracted from the graph part or graphic part of the model answer, and the extracted description object, the table object, and the graph object are stored in the storage unit as a model answer object table in a specific order,
In the answer stored in the storage unit, as the object, the statement object is extracted from the sentence part of the answer, the table object is extracted from the table part of the answer, and from the graph part or the graphic part of the answer Extracting the graph object, and storing the extracted discussion object, the table object, and the graph object in the storage unit as an answer object table in the specific order;
An answer evaluator for automatically scoring by comparing the argument object, the table object, and the graph object stored in the specific order between the model answer object table and the answer object table Answer scoring device. The answer scoring device according to claim 1,
The answer analysis unit
When extracting the table object from the table portion, each of the elements included in each square constituting the table is extracted as an independent table object. The answer scoring device according to claim 1 or 2,
The answer analysis unit
When extracting the graph object from the graph part or graphic part,
Analyzing the lines that make up the graph or figure,
Analyzing characters included in the graph or the figure,
Based on the analysis result of the line and the analysis result of the character, the association between the line and the character is analyzed,
Analyzing the point that is the point in creating the answer,
An answer scoring device that generates a graph object table by collecting the graph objects obtained as a result of the analysis. In the model answer to the question that is stored in the storage unit, as the object to be used for scoring, a statement object is extracted from the sentence part of the model answer, and the table object is extracted from the table part of the model answer. Extract and extract a graph object from the graph part or graphic part of the model answer, and extract the statement object, the table object, and the graph object in a specific order as a model answer object table in the storage unit. Remember,
In the answer to the question stored in the storage unit, as the object, the statement object is extracted from the sentence part of the answer, the table object is extracted from the table part of the answer, and the graph part of the answer Alternatively, the graph object is extracted from the graphic part, and the extracted discussion object, the table object, and the graph object are stored in the storage unit as an answer object table in the specific order,
In order to cause the computer to execute a procedure for automatically scoring by comparing the discussion object, the table object, and the graph object stored in the specific order between the model answer object table and the answer object table No answer scoring program.

Images