KR20080021771A - Sentence evaluating device and sentence evaluating program - Google Patents

Abstract

A sentence evaluating device (14) comprises a marking unit (21) for comparing the answer sentence made by the user and the right answer sentences stored in a right answer sentence database (18) and corresponding to the answer sentence, marking the answer sentence for each right answer sentence by a mark deduction method using a below-mentioned criterion, and adopting the height score as the score of the answer sentence. The marking a unit (21) has work group evaluating means (25) for marking the answer sentence in respect of the difference between the word group in the answer sentence and that in the correct answer sentence, a part-of-speech evaluating means (27) for making the answer sentence in respect of the parts of speech of the words in the answer sentence and the correct answer sentence, and final evaluating means (29) for finally marking the answer sentence after the markings by the evaluating means (25 to 28). ® KIPO & WIPO 2008

Description

Sentence evaluation device and sentence evaluation program {SENTENCE EVALUATING DEVICE AND SENTENCE EVALUATING PROGRAM}

The present invention relates to a sentence evaluating apparatus and a sentence evaluating program, and more particularly, to a sentence evaluating apparatus and a sentence evaluating program capable of objectively and simply evaluating an answer sentence translated into a predetermined language with respect to an original sentence. It is about.

In the field of English education such as junior high school and high school, the scoring of the test questions of the English writing questioned to the student is often performed based on the teacher's own scoring standard. That is, when a teacher grades an answer written by a student, the answer is compared with the correct answer, and subjective judgment is made according to the degree of difference between them. By the way, since there are several English that are the correct answer to the question, it is very time consuming to compare all of the correct answers with the student's answer and it is impossible to carry out in the actual education field. Therefore, it is common for the grading teacher to use the correct answer of one sentence or several sentences as the best answer and to score the limited best answer in comparison with the student's answer. On the other hand, there is little that is in complete agreement with the model solution. In addition, the difference between the model answer and the answer is difficult to pattern because it varies, such as word difference, word position difference, word spelling, word missing, word utilization type, and tense difference. For these reasons, if several teachers are assigned to share the test questions, the evaluation of the differences between the best answer and the answer is left to the subject of each teacher, and the grading criteria will not be completely consistent among the teachers. easy. In addition, even when one teacher has scored all the students, the scoring criteria are often changed at the beginning and the end of the scoring. Therefore, in general, it is currently difficult to secure objectivity in scoring English writing by a person.

By the way, the translation ability evaluation system which evaluates the translated sentence which the predetermined translator and the automatic translation apparatus produced is known (refer patent document 1), and it is also conceivable to apply this system to the scoring of English writing. This translation ability evaluation system stores a large number of English (correct answer) questions that are candidates for correct answers to Japanese test sentences, and contrasts each correct answer with a translation written in response to a Japanese test question. An indicator called the translation correct rate is obtained, and the highest translation correct rate is adopted. This translation correct answer rate is computed based on whether the same word as the word which comprises a translation exists in the same position as a correct answer.

[Patent Document 1]: Japanese Patent Application Laid-Open No. 2002-140326

However, in the above-described translation ability evaluation system, since the evaluation of the translation is performed only by the identicalness of the mutually matching words between the correct answer and the translation, there is a fear that the evaluation of the translation which does not correspond to the actual situation is performed. For example, if the same language group exists between the correct answer and the translation, but the positions of these fish groups are different, the ability to derive the same fish group is not considered at all, and is equivalent to the case where a completely different fish group is arranged. There is a problem to be dealt with.

The present invention has been devised in view of such a problem, and an object thereof is to provide a sentence evaluation device and a sentence evaluation program which can objectively perform an evaluation of an answer sentence translated into a predetermined language with respect to the original sentence. It is in doing it.

(1) A sentence evaluating apparatus for evaluating a solution sentence translated into a predetermined language with respect to a question sentence to be the original text in order to achieve the above object,

A correct answer database in which a correct answer to be a translation of the correct answer for the question is stored, and a scoring unit configured to perform a scoring process on the answer by contrast between the correct answer and the answer,

The scoring unit includes a fish group evaluating means for paying attention to the difference between the correct answer and the fish group in the answer.

The fish group evaluating means extracts the same or approximate fish group between the correct answer and the answer sentence, and makes the corresponding or approximate fish group the assessment target, and differs in the position of each evaluation target word between the correct answer and the answer sentence. Or the structure which makes a difference in a score according to the difference of the arrangement | sequence order of the said evaluation target language is employ | adopted.

(2) Moreover, the said fish group evaluation means employ | adopts the structure which makes a difference in a score by the said fish group and a fish group approximated.

(3) In addition, the fish group evaluation means is a fish group that is approximate when the difference between the fish group between the correct answer and the answer question is only a difference between the utilization type between the words constituting the fish group and / or a spelling miss. We adopt constitution to say.

(4) further, having a dictionary database in which a predetermined word is stored together with the utilization type,

The scoring unit includes utilization type evaluation means for focusing on the difference between the utilization type of the word in the correct answer and the answer sentence based on the dictionary database.

The utilization type evaluating means extracts, as a new evaluation target word, a word having a different utilization type between the correct answer and the answer sentence among words excluding the evaluation target words that have already been evaluated, and the correct answer and the answer for the corresponding new evaluation target word. It is possible to employ a configuration in which the score is different depending on the difference in position between the doors.

(5) Moreover, the said scoring part is equipped with the spelling evaluation means which focuses on the spelling miss of the word in the said answer sentence, and scores it,

The spelling evaluation means extracts a word spelled between the correct answer and the answer sentence as a new evaluation target word from among words excluding the evaluation target word that has already been evaluated, and, for the new evaluation target word, the correct answer and the It is possible to employ a configuration in which the score is different depending on the positional difference between the answer questions.

(6) further, having a dictionary database in which a predetermined word is stored with the kind of part-of-speech,

The scoring unit includes a part-of-speech evaluation means for scoring based on the identity of the parts of speech in the correct answer and the word in the answer based on the dictionary database,

The part-of-speech evaluation means extracts, as a new evaluation target word, a word whose kind of part-of-speech matches between the correct answer and the answer sentence among words excluding the evaluation target word that has already been evaluated, and the correct answer sentence is applied to the corresponding new evaluation target word. And a configuration in which the score is different depending on the difference in position between the answer questions and / or the type of the part of speech.

(7) In addition, the scoring unit may employ a configuration including final evaluation means for varying the score according to the number of remaining words except for the evaluation target word in the correct answer and the answer.

(8) Here, it is preferable that the said final evaluation means employ | adopts the structure which makes a difference in a score according to whether the said remaining word exists in the position corresponding to the said correct answer and the answer.

(9) Moreover, the said final evaluation means can employ | adopt the structure which makes a difference in a score according to the kind of parts of speech of the said remaining words.

(10) or more, in the correct answer database, a plurality of correct answers are stored,

In the scoring unit, each of the correct answers is subjected to a scoring process for each of the answers by each evaluation means, and a configuration may be adopted in which the best evaluation among the scoring results obtained for each correct answer is the evaluation of the answers.

(11) The present invention also provides a sentence evaluation for causing the computer to evaluate the answer sentence by comparing the correct answer sentence to be the correct translation of the original question and the answer sentence translated into a predetermined language with respect to the question sentence. As a program,

The same or approximate fish group is extracted as the evaluation target word between the correct answer and the answer, and the score is different depending on the difference in the position of each evaluation target or the arrangement order of the evaluation target between the correct and the answer. A configuration for causing the computer to execute a process of placing a message is adopted.

(12) In addition, the present invention, the correct answer to the translation of the correct answer to the question that becomes the original text,

As a sentence evaluating program for evaluating the answer sentence in a computer by preparing an answer sentence translated into a predetermined language with respect to the question sentence,

The same or approximate fish group is extracted as the evaluation target word between the correct answer and the answer, and the score is different depending on the difference in the position of each evaluation target or the arrangement order of the evaluation target between the correct and the answer. Fish evaluation step to put,

From the words excluding the evaluation target words in the fish group evaluation step, a word having the same kind of parts of speech between the correct answer and the answer sentence is extracted as a new evaluation target word, and the correct answer and the answer for the corresponding new evaluation target word. A part-of-speech evaluation step in which scores differ according to differences in position between doors and / or types of parts of speech,

In the final evaluation step, the score is varied according to the number of words except the target word in each step.

In order to implement, the computer is executed.

(13) Here, each step may be performed for each of a plurality of correct answers, and the computer may be adopted so as to determine the best evaluation among the scoring results obtained for each correct answer as the evaluation of the answer.

In addition, "a group of fish" is called a chunk, and uses a set of a plurality of consecutive words as a generic meaning in one word or a predetermined sentence.

In addition, "the same fish group" points out that the word before and after adjacent to the said fish group differs among the comparison object sentences among the fish groups which mutually match between the comparison object sentences. For example, referring to FIG. 5, in addition to the three fish groups surrounded by the square in FIG. 5 as matching fish groups between the answer and the correct answers, "eating", "and", "more", "than", The combination of seven fish groups of "two", "more than", and "than two" is mentioned. Among these, the fish group enclosed by the rectangle of FIG. 5 turns into "the same fish group" as the word before and behind the fish group being different between the comparison target sentences.

In addition, "approximate fish group" refers to including the word in the same fish group when the word before and after adjacent to the same fish group differs on predetermined conditions.

According to the above inventions (1) to (13), various evaluations including elements other than the determination of the identity of words at the same position are performed between the correct answer and the answer, so that the evaluation of the answer can be objectively performed according to the situation. Can be.

1 is a schematic system diagram of a writing composition learning system according to a first embodiment.

Fig. 2 is a diagram of question and answer questions used in the explanation examples.

3 is a flow chart illustrating a scoring process.

4 is an explanatory diagram of the contents of storage of the correct answer database.

5 is a diagram for explaining scoring in a fish group evaluation step.

6 is a diagram for explaining scoring in the utilization type evaluation step.

7 is a diagram for explaining scoring in the part-of-speech evaluation step.

8 is a diagram for explaining scoring in the spelling evaluation step.

9 is a diagram for explaining scoring in the final evaluation step.

10 is a diagram for explaining scoring in the final evaluation step.

11 is a diagram for explaining scoring in the final evaluation step.

12 is a schematic system configuration diagram of a writing composition system according to a second embodiment.

13 is a chart for explaining scoring in a fish group evaluation step.

14 is a diagram for explaining scoring in the part-of-speech evaluation step.

15 is a chart for explaining scoring in the final evaluation step.

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

Example 1

1, a schematic system configuration diagram of a writing composition learning system according to the present embodiment is shown. In this figure, the English composition learning system 10 is a system which evaluates the English composition which the user made about the predetermined | prescribed problem sentence (original text) by Japanese. The English writing learning system 10 transmits and receives various types of information to and from the terminal 11 owned by a user registered in advance and the computer 11 through a computer network 13 such as the Internet, and evaluates the English writing. It consists of the structure provided with the server 14 as a sentence evaluation apparatus.

When the question sentence is distributed from the server 14 to each terminal 11 of the user, and the English-language answer written by the user is transmitted from the terminal 11 to the server 14 side, In the server 14, scoring of the answer questions, correction, and determination of the learning level are automatically performed by a computer, and these results are transmitted to the user terminal 11.

 The terminal 11 is not particularly limited, and various devices can be employed as long as the terminal 11 can transmit and receive information required by the present invention, such as a personal computer, a portable information terminal, a mobile phone, and the like.

 The server 14 is comprised by software and / or hardware, and is comprised by several program modules and / or processing circuits, such as a processor. The server 14 includes a transceiver 16 which transmits and receives various types of information to and from the terminal 11, a question sentence database 17 in which a plurality of question sentences are stored, and a question sentence database 17. A correct answer database 18 storing an English correct answer which is a correct answer for each question, a dictionary database 19 in which a predetermined word is stored together with each utilization type and part of speech, and a transceiver 16 Is a scoring process which performs a scoring process on the questionnaire 20 which extracts any problem sentence transmitted from the questionnaire to the terminal 11 from the questionnaire database 17, and the user's answer sentence which was received by the transceiver 16. A scoring result processing section 22 which stores and counts the section 21 and the results scored by the scoring section 21, and transmits the relevant information such as correct answers and the like to the terminal 11 from the transmitting and receiving section 16 together with the scoring results. ). Incidentally, the server 14 is provided with a function for providing various contents, etc., but it is not the gist of the present invention. Thus, illustration and explanation are omitted here.

 In the correct answer database 18, all correct answers that can be considered for each question stored in the question database 17 are stored.

 The scoring unit 21 compares the answers prepared by the user with each of the correct answers in the correct answer database 18 corresponding to the corresponding answers, and scores the scores by the deduction method based on the criteria described below for each correct answer. And the highest score among them is used as the score of the answer.

The scoring unit 21 focuses on and scores the fish group evaluation means 25 which focuses on the difference between the correct answer and the fish group in the answer question, and the utilization type of the words in the correct answer and the answer question. Part-of-life evaluation means 26, part-of-speech evaluation means 27, which focuses on the identification of the parts of speech in the correct answer and the word in the answer sentence, and scores the spelling miss in the correct answer and the words in the answer sentence. And the final evaluation means 29 which performs final scoring after scoring by each of these evaluation means 25-28.

Next, the scoring process of the writing composition learning system 10 comprised as mentioned above is demonstrated, giving an example.

First, when the user requests the server 14 from the terminal 11, a predetermined question is asked from the questionnaire 20, and the question is transmitted from the transceiver 16 to the terminal 11. FIG. Then, the user writes the English text to the question, and transmits the answer to the server 14 from the terminal 11. Here, for the explanation by the following example, it is assumed that the question sentence shown in FIG. 2 is asked and the answer question shown in FIG. 2 is created by the user.

And in the scoring part 21 of the server 14, the answer question which a user made is graded. The scoring here is compared with each of a number of correct answers stored in the correct answer database 18, and the score of the answer is calculated for each correct answer as described later. Then, the highest score among the scores obtained for each correct answer becomes the user's score, and is sent to the scoring result processing unit 22 along with the correct answer that is the basis of the score, and the learning level of the user is determined, and with the level, Various contents such as correction and explanation of correct answers are sent to the terminal 11.

Here, scoring in the scoring part 21 is performed in the procedure shown by the flowchart of FIG.

First, the user's answer is accepted by the scoring unit 21 (S101). And, among the correct answers corresponding to the question asked, one correct answer is extracted from the correct answer database 18. Here, for example, as shown in FIG. 4, a plurality of correct answer databases are stored in the correct answer database 18. Below, the scoring by contrast with the correct answer of "No. 1" of FIG. 4 is demonstrated as an example among these correct answers.

First, in the fish group evaluating means 25, scoring which focuses on the difference of a fish group which is an aggregate of one word or a plurality of consecutive words is performed (fish group evaluation step: S103). That is, here, the same fish group is extracted between the answer question and the correct answer question by contrast of the answer question and the correct answer question, and the fish group becomes the evaluation target by the fish group evaluation means 25. In the above-mentioned example, as shown in FIG. 5, the fish group enclosed by the rectangle in each sentence is the same fish group between an answer sentence and a correct answer sentence, and these fish groups become a target object for evaluation.

Next, for these evaluation target words, it is determined whether or not the position of the word from the beginning coincides between the answer and the correct answer. In the above-described example, the evaluation target word "Jack" coincides with the position, while the other evaluation target words "eating and" and "more than two" do not coincide in position.

Then, one deduction point is given to each group of identical fish groups having different positions between the answer question and the correct answer question, and by adding them up, a deduction point in the fish group evaluation means 25 is obtained. In the above-described example, since the two groups of fish "eating and" and "more than two" do not coincide with each other between the answer and the correct answer, one point is deducted, and the fish group evaluation means (25 The deduction point in) becomes two points in total.

Next, in the utilization type evaluating means 26, the scores focusing on these utilization types are scored for each word except for the fish group which is the evaluation target language in the fish group evaluation means 25 (utilization type evaluation step: S104). That is, here, with respect to the remaining words that were not to be evaluated by the fish group evaluation means 25, the utilization type difference between the answer question and the correct answer statement based on the dictionary database 19 by the contrast of the answer question and the correct answer question. Only a word is extracted, and the word is subjected to evaluation by the utilization type evaluation means 26. In the above-described example, as shown in Fig. 6, the words "have" and the word "has" enclosed in the rectangles are words that differ only in utilization type between the answer and the correct answer, and these words are the evaluation target words here. do. In each of the subsequent drawings, a word marked with "×" means an evaluation target word that has been evaluated before that. In addition to the above-mentioned verbs, the difference in the utilization form here is, for example, in the case of nouns, singular and plural forms, and in the case of adjectives or adverbs, there are differences in the basic, comparative, and superlative forms.

In addition, with respect to these evaluation target words, it is determined whether or not the positions from the beginning coincide with each other between the answer and the correct answer. In the above-described example, these positions coincide with the word "have" in the answer sentence and the word "has" in the correct answer sentence.

Here, as a deduction point for the utilization type evaluation, if the position does not differ between the answer and the correct answer, only one point is given for the difference of the utilization type, while if the position is different between the answer and the correct answer, the utilization type and the position Two points are given for the difference of. By summing these deduction points, the deduction points in the utilization type evaluation means 26 are obtained. In the above-mentioned example, since the word "have" in the answer and the word "has" in the correct answer are different in utilization type but coincide in position, one deduction point is given and the deduction point in the utilization type evaluation means 26 is given. Is one point in total.

Next, in the part-of-speech evaluation means 27, for each word except for the evaluation target words in the fish group evaluation means 25 and the utilization type evaluation means 26, the parts of speech are paid attention to and the scoring is performed (part of speech evaluation step: S105). That is, here, the parts of speech are discriminated based on the data of the dictionary database 19 with respect to the remaining words that have not been evaluated by the evaluation means 25 and 26, and by contrast between the answer and the correct answer. The words whose parts of speech are the same are extracted between them, and the words become the evaluation targets in the parts of speech evaluation means 27.

The same part-of-speech here is determined based on a pre-set part-of-speech system. This part-of-speech system is not specifically limited, The thing by the big classification of "part-of-speech", the "adjective", etc., and the thing by the classification | segmentation which subdivided the major classification, such as "normal noun" and "unique noun" is used. In this embodiment, a part-of-speech system based on the latter classification is used.

In the above-described example, as shown in Fig. 7, a pair of words "in" and "for" surrounded by a square and a pair of "minutes" and "hours" are groups of words with the same parts of speech between the answer and the correct answer. Each of these words becomes an evaluation target word here.

Incidentally, although not in the above-described example, when the parts of speech of one sentence are plural in the other sentence, a word close to the head of the other sentence is selected.

In addition, it is determined whether or not the positions from the front doors coincide with each other between the answer and the correct answer sentences with respect to the pair of words used as the evaluation target words. In the above-mentioned example, the group of the word "in" in the answer and the word "for" in the correct answer is inconsistent in these positions, and the group of the word "minutes" in the answer and the word "hours" in the correct answer are also these positions. Does not match.

Here, if the position is not different between the answer and the correct answer as a deduction point for this part-of-speech evaluation, one point is given for the difference of the word, while if the position is different between the answer and the correct answer, the difference between the word and the position is different. 2 points are given. By summing these deduction points, the deduction points in the part-of-speech evaluation means 27 are obtained. In the above-described example, the word "in" in the answer and the word "for" in the correct answer have the same parts of speech but different words, and their positions do not coincide, so two points are deducted. In addition, two points of deduction points are provided similarly to the group of the word "minutes" in the answer and the word "hours" in the correct answer, and the deduction points in the part-of-speech evaluation means 27 are four points in total.

Next, in the spelling evaluation means 28, for each word except the evaluation target words in the fish group evaluation means 25, the utilization type evaluation means 26, and the part-of-speech evaluation means 27, the spelling miss of the word of the answer question is Focused scoring is performed (spelling evaluation step: S106). That is, here, with respect to the remaining words which were not evaluated by the evaluation means 25 to 27, the words in which the spellings are approximated between them are extracted by the contrast between the answer and the correct answer, and the word is evaluated here. It becomes a target.

Here, the determination of whether the spelling is approximated between each word of the answer and the correct answer is made using a known Levenshtein distance. The Levenstein distance is determined by the number of steps (coast) which is the minimum of the steps of matching the correct answer with respect to the words to be compared with each other, taking into account the omission of the letter, the letter coming, and the insertion of the letter.

For example, when the word "pzzel" is used in the answer to the word of the correct answer of "puzzle" by the spelling miss,

For "pzzel" in the answer, to change to "puzzle" in the correct answer:

(1) Insert the letter "u" between the letter "p" and the letter "z",

(2) Insert the letter "l" between the letter "z" and the letter "e",

(3) delete the last character "l",

Since three stages of are required, the "coast 3" here becomes the Levenstein distance between "puzzle" and "pzzel".

Then, between the answers and the correct answers, when comparing each word to be evaluated, the pair of words whose Levenstein distance is equal to or less than a predetermined threshold value is judged to be close to each other, and the words in the answer are misspelled. Is determined to be a word.

In the above-described example, as shown in Fig. 8, the word "dlinking" in the solution enclosed in the rectangle and the word "drinking" in the correct answer are words in which spelling is approximated between the answer and the correct answer. It becomes the evaluation target language here, and it is judged that the word "dlinking" in an answer is a word of misspelling.

In addition, with respect to these evaluation target words, it is determined whether or not the positions from the beginning coincide with each other between the answer and the correct answer. In the above-mentioned example, these positions are inconsistent between the word "dlinking" in the answer and the word "drinking" in the correct answer.

As a deduction point here, if the position is the same between the answer and the correct answer, one point is given for the misspelling, while if the position is different between the answer and the correct answer, two points are given for the difference between the spelling miss and the position. . By summing these deduction points, the deduction points in the spelling evaluation means 28 are obtained. In the above-mentioned example, since the word "dlinking" in the answer and the word "drinking" in the correct answer do not coincide with each other, two deduction points are given, and the deduction point in the spelling evaluation means 28 is It is two points in total.

Next, in the final evaluation means 29, scoring is performed on the remaining words that have not been the evaluation target words in the respective evaluation means 25 to 28 (final evaluation step: S107). That is, here, the deduction point is calculated | required about the remaining words as follows.

First, among words which have not yet been evaluated, it is determined whether a word exists at the same position between the answer and the correct answer. In the above-mentioned example, as shown in FIG. 9, there is no word which exists in the same position, and the deduction point here is O point.

Among the remaining words, it is determined whether or not a pair of words can be written between the answer and the correct answer, and if the pair can be written, two points for the difference between the word and the position are given as a deduction point for each pair. . In the above-mentioned example, as shown in FIG. 10, the word "thirty" in the answer and the word "been" in the correct answer are the evaluation targets, and as a set, two points are given as deduction points.

Finally, one deduction point is given to each word for words left in either of the answer and the correct answer. In the above-mentioned example, as shown in FIG. 11, the word "and", the word "a", and the word "half" remain in a correct answer, and three points of three words are given as a deduction point.

Next, the deduction points obtained by each of the above means are added together, and an approximation of the answer to the correct answer is obtained by the following equation (1) (S108).

Here, if the approximation is N, the sum of the deduction points is P, the number of words in the answer is W1, and the number of words in the correct answer is W2,

N = 1-(P / (W1 + W2))... Formula (1)

In the above-described example, as a result of Figs. 5 to 11, the total number of deduction points is 14 points, and the number of words in the answer is 11 words and the number of words in the correct answer is 14 words.

1-(14 / (11 + 14)) = 0.44

The above scoring process is performed for each correct answer (S109), the highest approximation of which is the score of the answer (S110), and the correct answer corresponding to the approximation is the corrected answer.

In addition, although the deduction point in each means is set to 1 point or 2 points, the score of the deduction point provided can be changed according to each evaluation item, and the importance of each item can also be given.

Therefore, according to such an embodiment, the answer can be scored by various items, and the objective scoring can be performed easily.

Next, another Example of this invention is described. In the following description, the same or equivalent components as those in the first embodiment will be denoted by the same reference numerals, and the description thereof will be omitted or simplified.

Example 2

The second embodiment is characterized in that the configuration of the scoring unit 21 is changed with respect to the first embodiment, and the scoring method of the answer question, that is, the calculation method of the deduction point and the approximation degree N, is changed.

That is, as shown in Fig. 12, the scoring unit 21 of the present embodiment includes a fish group evaluation means 25 which focuses on the difference between the correct answer and the fish group in the answer question, and the correct answer and the answer question. Part-of-speech evaluation means 27 which focuses on the identification of the parts-of-speech of the words in the inside, and the final evaluation means 29 which performs final scoring after scoring by each of these evaluation means 25 and 27 are provided, A scoring process is performed in order of the fish group evaluation means 25, the part-of-speech evaluation means 27, and the last evaluation means 29. FIG.

The scoring method in the scoring part 21 of a present Example is demonstrated below as an example.

Also in this embodiment, the same question, correct answer, and answer as in Example 1 are used, and a scoring process for the answer is described.

First, in the fish group evaluating means 25, scoring which focuses on the difference of a fish group which is an aggregate of one word or a plurality of consecutive words is performed (fish group evaluation step). That is, here, the same or approximate fish group is extracted between the answer question and the correct answer question by contrast of the answer question and the correct answer question, and the fish group becomes the evaluation target in the fish group evaluation means 25. Specifically, the same fish group is determined by mutual matching or the like between the correct answer and the answer. For words before and after adjacent to the same fish group, it is determined whether the utilization type is different or spelling miss between the correct answer and the answer sentence, and when it is determined that the utilization type is different or spelling miss, The word of is included in the same fish group to form a cool fish group. On the other hand, when the word before and after adjacent to the same fish group is not judged to be different in the utilization type or misspelling, the same fish group is included as it is without including the word.

In the above-mentioned example, the fish group enclosed by the rectangle in each sentence of FIG. 13 is the same or approximate between the answer and the correct answer, and these fish groups are the evaluation target words.

That is, the fish group A "Jack have" in the answer and the fish group A "Jack has" in the correct answer have the same "Jack" part, and the word "have" of the subsequent word is utilized as the "has" of the correct answer. It is judged that this is wrong, and becomes a nice fish group.

In addition, the fish group B "eating and dlinking" in the answer and the fish group B "eating and drinking" in the correct answer have the same "eating and" portion, and the word "dlinking" in the answer is "drinking". It is judged to be a missed spelling and becomes a nice fish group.

In addition, since the fish group C "more than two" in the answer and the correct answer is completely identical to each other, it becomes the same fish group.

The determination of the error of the utilization type and the spelling miss here is performed by the same method as the utilization type evaluation means 26 and the spelling evaluation means 28 described in the first embodiment.

Subsequently, one point for each word is given a deduction point for the word to be evaluated for the correct answer, and the word for the spelling miss, to the evaluation target word composed of the same or approximate fish group. In the above-described example, the fish group A `` Jack have '' in the answer sentence includes a word `` have '' having a wrong use type for the fish group A in the correct answer sentence, and the fish group B `` eating and dlinking '' in the answer sentence is the correct answer sentence. Since the word "dlinking" with the misspelled word "dlinking" is included with the fish group B, the deduction points for the fish group A and the fish group B are each 1 point. In addition, since the fish group C "more than two" of the answer question has no utilization type or spelling miss with respect to the fish group C of the correct answer question, the deduction point is zero.

Next, with respect to the same or approximate fish group, attention is paid to the difference in the arrangement order, and it is determined whether to give a deduction point. In the above-described example, the answers are arranged in the order of fish group A, fish group B and fish group C, and the correct answers are also arranged in order of fish group A, fish group B and fish group C. In such a case, the arrangement order is the same and there is no deduction point, that is, a deduction point O point.

On the other hand, if the correct answer is in the order of fish group A, fish group B and fish group C, the answer is arranged in the order of fish group C, fish group A and fish group B, and the same as the arrangement of fish group in the correct answer To do this, one step of moving fish C of the answer back to fish B is required. This required number of steps is multiplied by five deduction points to calculate deduction points paying attention to the arrangement order of fish groups. As described above, when the required number of steps is one step, the deduction point is five points.

Next, in the part-of-speech evaluation means 27, the scoring which pays attention to the identity of the parts-of- speech is performed with respect to each word except the evaluation target word in the fish group evaluation means 25 (part of speech evaluation step). That is, here, as in the first embodiment, parts of speech are discriminated based on the data in the dictionary database 19 with respect to the remaining words that have not been evaluated by the fish group evaluation means 25. By inquiring, a word in which the parts of speech are the same among them is extracted, and the word becomes the object of evaluation in the parts of speech evaluation means 27. In the above-described example, as shown in FIG. 14, the words "in" and "for" surrounded by a rectangle, and the words "minutes" and "hours" are words in which the parts of speech are the same between the answer and the correct answer, respectively. These words become the evaluation target here.

Here, the deduction point for this part-of-speech evaluation is determined according to the kind of part-of-speech which became the same between the answer and the correct answer. Specifically, if the same parts of speech are articles of article, three points of deduction points and prepositions, seven points of deduction points and ten points of deduction points are given. In the above-described example, since the word "in" in the answer and the word "for" in the correct answer are prepositions, the deduction point is 7 points. On the other hand, since the word "minutes" in the answer and the word "hour" in the correct answer are other parts of speech (nouns) other than the preposition, the deduction point is 10 points, and the deduction point in the parts of speech evaluation means 27 is 17 points. do.

And as with Example 1, with respect to the evaluation target word, it can also determine whether the position from the head | line coincides with an answer sentence and a correct answer sentence, and may change the score of the deduction point by the matching and inconsistency.

Next, in the final evaluation means 29, scoring is performed on the remaining words which have not been the evaluation target words in the respective evaluation means 25 and 27 (final evaluation step), and a deduction point is obtained as follows.

In other words, for each word in the answer and correct answers that have not yet been evaluated, a deduction point is calculated for each of the parts of speech, and the points are totaled to obtain a deduction point in the final evaluation means 29. Specifically, similarly to the part-of-speech evaluation means 27 of this embodiment, three points are deducted points if it is an article, seven points are deducted if it is a preposition, and ten points are deducted if it is other parts of speech. Here too, the above-described dictionary database 19 is used for the determination by the part-of-speech type.

In the above-described example, as shown in Fig. 15, the word "thirty" in the answer, the word "been", the word "and", and the word "a" and the word "half" in the correct answer are five words in total. Deduction points by parts of speech are given. In this case, since the word "a" is an article, three points are deducted, and four other words are not articles and prepositions, and thus the points are each deduction point, and the points in the final evaluation means 29 are 43 points in total.

In addition, the deduction point in each said means is not specifically limited, Various scores can be employ | adopted arbitrarily.

Next, the deduction points obtained by each of the above means are added together, and an approximation of the answer to the correct answer is obtained by the following equation (2).

Here, the approximation is N, the sum of the deduction points is P, the number of words between the number of words in the solution and the number of words in the correct answer is Wmax, and the maximum deduction point Pmax is set. 10 points)

N = 1-(P / (Wmax x Pmax))... Formula (2)

And in the said Formula (2), when approximation degree N becomes a negative value, approximation degree N is set to O.

In the above-described example, as a result of Figs. 13 to 15, the total number of deduction points is 62 points, so the number of words in the answer is 11 and the number of words in the correct answer is 14, so Wmax is 14,

Approximation N is

1-(62 / (14 x 10)) = 0.56.

Thereafter, as in Example 1, the above scoring process is performed on each of the correct answers, and the highest approximation N is the score of the answer, and the correct answer corresponding to the approximation N is the target of the corrective answer. do.

Therefore, also in Example 2, the same effect as Example 1 can be acquired.

Here, the following aspects are mentioned as a modification of Example 2. As shown in FIG. That is, by spelling miss of the word in the answer sentence, the word determined as a spelling miss is replaced with the correct word in the correct answer for the fish group which becomes a closest fish group, and then the part-of-speech evaluation step by the above-described part-of-speech evaluation means 27 It can also be done. According to the above-described example shown in FIG. 13, for the fish group B judged by the spelling miss as the approximate fish group, the word "dlinking" in the answer is replaced with the word "drinking" in the corresponding answer, and then the fish group evaluation means ( Regarding the remaining words that are not to be evaluated in 25), parts of speech are determined in consideration of parts of speech before and after the word based on the data in the dictionary database 19.

According to such a modification, the precision at the time of discriminating parts of speech can be improved. Specifically, when the part-of-speech is discriminated against words having a function of plural kinds of parts of speech, the parts of speech are specified in consideration of the parts of speech before and after the word. By the way, when the part-of-spell missed word (the word "dlinking" in the example of FIG. 13) is to determine the parts of speech after the word (the word "in" in FIG. 13), when the misspelled word does not exist Because of the large number of parts of speech, the part of speech of the misspelled word cannot be discriminated and the part of speech of the word cannot be specified after considering the part of speech of the word. In this case, from a statistical point of view based on the frequency of use, etc., it is inevitably limited to one part-of-speech, and it is not possible to determine an accurate part-of-speech in accordance with the mode of use. According to this modified example, even when the word has a plurality of parts-of-speech functions when specifying a part-of-speech word following the misspelled word by substituting the spelled-miss word with the correct word, the answer The part of speech can be more accurately specified according to the intention of the author (user).

Incidentally, in each of the above embodiments, the distance learning type English writing learning system using a computer network has been illustrated and described. However, the sentence evaluation apparatus of the present invention is not limited thereto, and a program capable of executing the process of the present invention is installed. It may be a computer. In addition, the terminal 11 owned by the user is distributed by distributing a program capable of executing the process of the present invention to each terminal 11 via the Internet or by distributing a recording medium such as a CD-ROM in which the program is stored. It is also possible to execute all the above-described processes within.

In each of the above embodiments, the questionnaire written in English is evaluated for the Japanese questionnaire. However, the present invention is not limited thereto, and the questionnaire for questionnaire written in another language is evaluated for the questionnaire of Japanese or another language. You may do so.

Moreover, this invention can also be applied to evaluation of the translated sentence which the translator or automatic translator made. In the evaluation of the translation capability of the automatic translator, a series of systemizations are also possible by interlocking a system to which the present invention is applied with an automatic translator.

In addition, the structure of each part of the apparatus in this invention is not limited to the structural example shown, It can be variously changed as long as it can have substantially the same effect | action.

Claims (13)

A sentence evaluating apparatus for evaluating an answer sentence translated into a predetermined language with respect to a question sentence to be an original text, A correct answer database in which a correct answer to be a translation of the correct answer for the question is stored, and a scoring unit configured to perform a scoring process on the answer by contrast between the correct answer and the answer, The scoring unit includes a fish group evaluation means for focusing on the difference between the correct answer and the fish group in the answer. The fish group evaluating means extracts the same or approximate fish group between the correct answer and the answer sentence, and makes the corresponding or approximate fish group the assessment target, and differs in the position of each evaluation target word between the correct answer and the answer sentence. Or the sentence evaluation apparatus which makes a difference in a score according to the difference of the arrangement | sequence order of the said evaluation target word. The method of claim 1, The said fish group evaluation means is a sentence evaluation apparatus which makes a difference in score by the same fish group and a fish group approximated. The method according to claim 1 or 2, The said fish group evaluation means makes a fish group an approximate fish group, when the difference of a fish group between a correct answer and an answer sentence is only a difference of utilization type and / or a spelling miss between the words which comprise the said fish group, Sentence evaluation device. The method of claim 1, A dictionary database in which a predetermined word is stored with the utilization type, The scoring unit includes utilization type evaluation means for focusing on the difference between the utilization type of the word in the correct answer and the answer sentence based on the dictionary database. The utilization type evaluating means extracts, as a new evaluation target word, a word having a different utilization type between the correct answer and the answer sentence among words excluding the evaluation target words that have already been evaluated, and the correct answer and the answer for the corresponding new evaluation target word. The sentence evaluation apparatus which puts a difference in a score according to the difference of the position between doors. The method according to claim 1 or 4, The scoring unit includes spelling evaluation means for paying attention to the spelling miss of the word in the answer sentence, and scoring the score. The spelling evaluation means extracts a word spelled between the correct answer and the answer sentence as a new evaluation target word from among words excluding the evaluation target word that has already been evaluated, and, for the corresponding new evaluation target word, the correct answer and the The sentence evaluation apparatus which puts a difference in a score according to the difference of the position between answer questions. The method according to any one of claims 1 to 5, Having a dictionary database in which a predetermined word is stored with the kind of part of speech, The scoring unit includes a part-of-speech evaluation means for scoring based on the identity of the parts of speech in the correct answer and the word in the answer based on the dictionary database, The part-of-speech evaluating means extracts, as a new evaluation target word, a word whose kind of part-of-speech matches between the correct answer and the answer sentence among words except for the evaluation target word that has already been evaluated, and the correct answer sentence for the corresponding new evaluation target word. And a sentence evaluating apparatus for placing a difference in scores according to differences in positions between the answer questions and / or kinds of parts of speech. The method according to any one of claims 1 to 6, The scoring unit includes a final evaluation means for placing a difference in the score according to the number of remaining words excluding the evaluation target words in the correct answer and the answer. The method of claim 7, wherein And said final evaluation means makes a difference in score depending on whether or not said remaining words exist in a position corresponding to said correct answer and said answer. The method of claim 7, wherein And said final evaluation means makes a difference in score according to the kind of parts of speech of said remaining words. The method according to any one of claims 1 to 9, A plurality of correct answers are stored in the correct answer database, In the scoring unit, each of the correct answers is evaluated for each of the answers by the evaluation means, and the sentence evaluation device which makes the best evaluation among the scoring results obtained for each correct answer as the evaluation of the answers. As a sentence evaluating program for evaluating the answer to the computer by comparing the correct answer to the original question and the answer translated into a predetermined language for the question, The same or approximate fish group is extracted as the evaluation target word between the correct answer and the answer, and the score is different depending on the difference in the position of each evaluation target or the arrangement order of the evaluation target between the correct and the answer. A sentence evaluation program for causing a computer to execute a process of placing a message. As a sentence evaluating program for evaluating the answer to the computer by comparing the correct answer to the original question and the answer translated into a predetermined language for the question, The same or approximate fish group is extracted as the evaluation target word between the correct answer and the answer, and the score is different depending on the difference in the position of each evaluation target or the arrangement order of the evaluation target between the correct and the answer. Fish evaluation step to put, From the words excluding the evaluation target words in the fish group evaluation step, a word having the same kind of parts of speech between the correct answer and the answer sentence is extracted as a new evaluation target word, and the correct answer and the answer for the corresponding new evaluation target word. A part-of-speech evaluation step in which scores differ according to differences in position between doors and / or types of parts of speech, Final evaluation step of varying the score according to the number of words except the evaluation target words in each step Sentence evaluation program for executing the computer in sequence. The method of claim 12, Wherein each step is executed for each of a plurality of correct answers and causes the computer to execute the best evaluation among the scoring results obtained for each correct answer as the evaluation of the answer.

Images