CN110895787B

CN110895787B - Method for dynamically matching English corpus difficulty and student ability analysis

Info

Publication number: CN110895787B
Application number: CN201811050464.8A
Authority: CN
Inventors: 周刚
Original assignee: Individual
Current assignee: Individual
Priority date: 2018-09-10
Filing date: 2018-09-10
Publication date: 2021-06-22
Anticipated expiration: 2038-09-10
Also published as: CN110895787A

Abstract

The invention discloses a method for dynamically matching English corpus difficulty and student ability analysis, which comprises the following steps: (1) evaluating the corpus difficulty to obtain a corpus index ELP value and form a corpus; (2) evaluating the ability of the user to obtain an ELP value of the ability index of the user; (3) and according to the index of the user ability, the system recommends the linguistic data matched with the user ability. The method for dynamically matching the English corpus difficulty and the student ability analysis comprises the steps of firstly calculating the ELP value of each corpus to form a corpus, then carrying out ability evaluation by a user to obtain an ability index, recommending the matched corpus according to an ability index system, and carrying out good matching on the corpus difficulty and the user ability; in the process of corpus formation, various factors such as the type, word number and answer time of the corpus are fully considered, so that the corpus difficulty index in the corpus is more scientific.

Description

Method for dynamically matching English corpus difficulty and student ability analysis

Technical Field

The invention relates to a method for dynamically matching English corpus difficulty with student ability analysis, and belongs to the field of education.

Background

The method realizes the evaluation of the grading of the reading ability of students and has great significance for the students with the long-term reading ability. Through such evaluation, students can learn the stage of individual reading ability and even the development level of each subdivision ability dimension. Meanwhile, based on the assessment, a reference basis can be provided for the students to select reading books, and then reading interest and reading capacity are effectively improved by reading the books with proper difficulty. The status quo is evaluated by grading Chinese reading ability of foreign students.

At present, the reading ability grading standard systems commonly used in European and American countries mainly comprise the following systems: a blues (Lexile) graded reading evaluation system; developmental Reading Assessment hierarchies (development Assessment Levels); instructional Reading hierarchies (Guided Reading Levels); a hierarchy of Reading ability (hierarchy of Reading powers; DRP for short); ar (accessed reader) rating system, etc.

And the blues comprehensively analyze the reading ability of the readers according to the answers of the readers to the reading questions of different grades. The aspects analyzed include: the theoretical difficulty of the subject, the actual difficulty of the subject, the quality of the subject, the average answering accuracy, the grade of students, the total number of students taking part in the test and the specific number of the selected options of the subject, and then the data are analyzed by Rasch model software to finally obtain the reading ability level of the reader. The DRA test is divided into 4 parts: reading Engagement (Reading Engagement); read Fluency (Oral Reading Fluency); comprehension (Comprehension); continuity and concentration (Continuum/Focus). For student examinations with lower DRA ratings, these 4 parts must be completed in one examination; for students with higher grades, different parts can be completed at different times. The DRP test is in the form of providing a plurality of articles with different topics for students, wherein part of words in the articles are set as blank filling choices, the students select words meeting the context, and all related word options are high-frequency words familiar to the students. The AR rating system employs the STAR (standardized Test for the Assessment of reading) evaluation system developed by Renaissance. The evaluation system gives questions in a Computer Adaptive mode (Computer Adaptive Test), and automatically adjusts the difficulty according to the reading ability level of students. The method is characterized by comprising the following steps of considering five aspects such as vocabulary cognition and comprehension capacity, analysis comprehension capacity of article contents and basic structures, deep analysis capacity of article contents, plots and roles, analysis comprehension capacity of author-based construction and manipulation, thinking capacity of judgment, reasoning and the like. The STAR evaluation system gives a reading level guidance and an advanced tutoring suggestion according to the testing level of the student, and schools and teachers make a targeted advanced learning plan for the student on the basis of the testing result and objectively and accurately check the learning effect of independent reading of the student through the test regularly.

But the defects are that the Chinese students are developed aiming at the group with the native language of English, and the ability and the level of the Chinese students cannot be objectively reflected without considering the Chinese students as the bilingual acquirers; starting from the text corpus, the matching interaction problem between the text corpus and readers is rarely considered, and the level of the text corpus is readjusted according to the feedback of a large number of readers by using big data analysis.

Professor Wang Gong Ji of Beijing university publishes "Standard for reading and grading English in middle and primary schools of China" (experimental draft) in 2018; southern wuzun professor and its team have also invested in relevant research. In addition, on 12.4.2018, 86 competency forms are provided, wherein the Chinese English competency rating scale is issued by the ministry of education and the national language committee and formally implemented with 1.6.8 for the English learner in China. The defects are as follows: most of the descriptions are written in language and words, which provides compendial requirements, but lacks good practical operability. Such as: to understand and grasp various language materials accurately and thoroughly (extracted from the general table of Chinese English ability grade Scale); time efficiency has not been taken into account. The accuracy of the questions or the achievement rate of the tasks set for checking the reading comprehension rate cannot fully reflect the ability level of the reader, and the time length for the reader to finish reading and answering or completing the tasks needs to be considered.

Disclosure of Invention

The purpose of the invention is as follows: in order to overcome the defects in the prior art, the invention provides a method for dynamically matching English corpus difficulty and student ability analysis.

The technical scheme is as follows: in order to solve the technical problem, the method for dynamically matching the difficulty of the English corpus and the ability analysis of the students is characterized by comprising the following steps of:

(1) evaluating the corpus difficulty to obtain a corpus index ELP value and form a corpus;

(2) evaluating the ability of the user to obtain an ELP value of the ability index of the user;

(3) and according to the index of the user ability, the system recommends the linguistic data matched with the user ability.

Preferably, the step (1) includes the steps of:

(21) the system acquires English texts and marks the English texts, namely the texts are a (m), wherein m represents the sequence of acquiring the texts, the word number, the genre and the question type of the texts are acquired, and the number of answering persons is set;

(22) calculating the corpus difficulty K _1 of the text, wherein K _1 is 0.39 (the total word number/the total sentence number of the English text) +11.8 (the total syllable number/the total word number) -15.59;

(23) calculating local grading vocabulary difficulty K _2 in an English text;

(24) calculating an ELP index D _0 ═ k _3 of the initial difficulty of the corpus topics of the platform;

(25) the system obtains the difficulty index elp (m) of english text 0.4 × K _1+0.3 × K _2+0.3 × D _0, and stores the index elp (m)) in the system;

(26) the answerer selects the text to answer, the number P of answerers, the number Q of wrong answers, the answering time t of the answerer are recorded, and the definition of the answer pair is as follows: the corpus accuracy rate is more than 80%;

(27) when the number of P is the same as the set value, D _0 ═ K _3 ═ Q/P, ELP (m) ═ ELP ═ 0.4 ═ K _1+0.3 ═ K _2+0.3 ═ D _0, the answering time is the average time for the number of answerers P, P, Q is cleared to zero, and (a), (m), ELP (m) are stored in the system; when the number of P is different from the set value, repeating the step (6)

(28) And (5) repeating the steps (6) and (7).

Preferably, in the step (23), K _2 ═ W1+ W2+. Wn)/n, W represents a word, n represents the number of words in the sample, one word is repeated,

LEXISn＝Sum(Z_n1+Z_n2+Z_n3+Z_n4+Z_n5+Z_n6)，Z_n1is the Z absolute value of the COCA word frequency of the word; z_n2Is the Z absolute value of the BNC word frequency of the word; z_n3Is the Z absolute value of the COCA spoken indicator of the word; z_n4Is the Z absolute value of the COCA academic index for the word; z_n5Is the Z absolute value of the BNC spoken language index of the word; z_n6Is the Z absolute value of the BNC academic index for the word, Ymax is the maximum value of the Y values in the sample, Ymin is the minimum value of the Y values in the sample.

Preferably, in step (24), K _3 is 0.1 × P _1+0.1 × K _2+0.1 × P _3+0.1 × P _4+0.2 × P _5+0.2 × P _6, where P _1 is a value corresponding to the corpus difficulty K _1 converted into an integer, P _2 is a value corresponding to the word number, P _3 is a value corresponding to the answer time of the answerer, P _3 has an initial value of the value applied to the average time +2 minutes of the english expert making the answer, P _4 is a value corresponding to the genre, P _5 is a value corresponding to the topic type, P _6 is a value corresponding to the year, and values of P _1, P _2, P _3, P _4, P _5, and P _6 are listed as follows:

the value of P _1 is as follows:

the value of P _2 is as follows:

number of words in language	200 and below	201--300	301--400	401--500	501 and above
						P_2	60	70--80	81--90	91--100	100

The value of P _3 is as follows:

the value of P _4 is as follows:

the value of P _5 is as follows:

the value of P _6 is as follows:

grade of year	Gao Yi	High two	Gao III
				Value taking	80	90	100

Preferably, the step (2) includes the steps of:

(21) a preliminary estimated static individual ELP initial value L _0,

l _0 ═ L _0a + L _0b +. L _0s)/s, L _0s ═ s corpus ELP value (longest answer time/actual answer time), corpus is corpus text in the test library;

(22) first capacity evaluation: and pushing 10 corpora of corresponding levels from a test library according to a primary evaluation link L _0 to test, so as to obtain the capability level L _1 (static personal ELP value) of the user, wherein L _1 is 0.5L _1_ a + 0.5L _1_ b, L _1_ a is [ Q1+ Q2+ … Q10 ]/10, Qc represents the ELP value of the c-th corpus and the correctness of the corpus, L _1_ b is (K1+ K2+ K3+ K4)/4, wherein K1 is the average value of Q of grammatical corpora in ten corpora, K2 is the average value of Q of lexical corpora in ten corpora, K3 is the average value of Q of complete corpora in ten corpora, and K4 is the average value of Q of reading corpora in ten corpora.

Preferably, in the step (2), the user ability evaluation is updated once per month, L _ k is 0.4L _ { k-1} +0.6 { [ L _ { k-1} _1+ L _ { k-1} _2+ … L _ { k-1} _ q }/q }, L _ k represents the ELP value of the person in the month, L _ { k-1} represents the ELP value of the person in the previous month, and L _ { k-1} _ q is the personal ELP value calculated by the system at the time of the qth question answering in the previous month.

Preferably, in the step (3), the corpus index ELP value of the first corpus recommended by the system is 5 greater than the user ability index ELP value, and then the corpus index ELP values of the corpuses recommended by the system to the user present increasing distribution.

Has the advantages that: the method for dynamically matching the English corpus difficulty and the student ability analysis comprises the steps of firstly calculating the ELP value of each corpus to form a corpus, then carrying out ability evaluation by a user to obtain an ability index, recommending the matched corpus according to an ability index system, and carrying out good matching on the corpus difficulty and the user ability; in the process of corpus formation, various factors such as the type, word number and answer time of the corpus are fully considered, so that the corpus difficulty index in the corpus is more scientific.

Drawings

FIG. 1 is a flow chart of corpus difficulty calculation according to the present invention.

Detailed Description

The present invention will be further described with reference to the accompanying drawings.

As shown in fig. 1, a method for dynamically matching difficulty of english corpus with student ability analysis includes the following steps:

(3) according to the index of the user ability, the system recommends the linguistic data matched with the user ability, the matching can refer to a principle of 'i + 5', namely the ELP of the linguistic data pushed to the student by the system is higher than the current personal dynamic ELP value of the student, and the highest value is 5 greater than the personal ELP value of the student; the increasing principle is that 3-5 corpora pushed to students each time are in the sequence of increasing ELP values of the corpora, and the increasing value can be between 0.5 and 2; according to the multiple principle, under the condition that the student selects, the system pushes different linguistic data to the student for multiple times according to the 'i + 5' principle and the increasing principle; unknown principles, at least one topic (or background knowledge) in the corpus pushed to the student is not currently involved or is rarely involved by the student.

In the present invention, the step (1) comprises the steps of:

(23) calculating local grading vocabulary difficulty K _2 in an English text;

(26) the answerer selects the text to answer, and records the number P of answerers, the number Q of wrong answerers and the answering time t of the answerer;

(28) And (5) repeating the steps (6) and (7).

Wherein, in the step (23), K _2 ═ W1+ W2+. Wn)/n, W represents a word, n represents the number of words in the sample, and the number of repeated words in the text is multiple,

LEXISn＝Sum(Z_n1+Z_n2+Z_n3+Z_n4+Z_n5+Z_n6)，Z_n1is the Z absolute value of the COCA word frequency of the word; z_n2Is the Z absolute value of the BNC word frequency of the word; z_n3Is the Z absolute value of the COCA spoken indicator of the word; z_n+Is the Z absolute value of the COCA academic index for the word; z_n5Is the Z absolute value of the BNC spoken language index of the word; z_n6Is the Z absolute value of the BNC academic index of the word, Ymax is the maximum value of the Y value in the sample, and Ymin is the minimum value of the Y value in the sample; the samples are all words needed to be learned in high school, the word number specified in Shanghai high school is 7760 words, the value of COCA _ F, BNC _ F, COCA _ S, COCA _ A, BNC _ S, BNC _ A can be obtained in the samples, COCA _ F is COCA word frequency, BNC _ F is BNC word frequency, COCA _ S is COCA spoken language indicator, COCA _ A is COCA academic indicator, BNC _ S is BNC spoken language indicator, BNC _ A is BNC academic indicator,

x is sample column characteristic value, mu is sample column average value, sigma is sample total amount (total word number with index), sigma is 7760 in this sample, index value before partial word calculation

Index value calculated by partial word

Words	ZCOCA_F	ZBNC_F	ZCOCA_S	ZCOCA_A	ZBNC_S	ZBNC_A
							be	39.77	43.94	44.42	36.63	40.57	44.22
of	33.07	36.45	25.26	44.80	19.96	45.89
							and	34.20	31.28	30.68	35.60	29.82	27.40
to	32.41	31.05	31.87	29.83	26.76	28.08
							a(an)	32.66	29.28	28.28	26.32	24.80	25.58
in	23.10	23.12	19.10	26.45	15.82	26.00

Examples are: complexity calculation of word be:

in the step (24), K _3 is 0.1 × P _1+0.1 × K _2+0.1 × P _3+0.1 × P _4+0.2 × P _5+0.2 × P _6, where P _1 is a value corresponding to the corpus difficulty K _1 converted into an integer, P _2 is a value corresponding to the word number, P _3 is a value corresponding to the answer time of the answerer, an initial value of P _3 is a value applied to the average time +2 minutes of the english trial, P _4 is a value corresponding to the genre, P _5 is a value corresponding to the topic type, P _6 is a value corresponding to the year, and values of P _1, P _2, P _3, P _4, P _5, and P _6 are listed as follows:

the value of P _1 is as follows:

the value of P _2 is as follows:

The value of P _3 is as follows:

the value of P _4 is as follows:

the value of P _5 is as follows:

the value of P _6 is as follows:

grade of year	Gao Yi	High two	Gao III
				Value taking	80	90	100

The P _3 initial value is a value corresponding to the average time +2 minutes for the english specialist to try the text, for example: when 8 minutes are spent on average by experts of a corpus, the initial maximum answering time of the corpus is 10 minutes, and the corresponding value is 70. When 100 people use the corpus on a large data platform, the platform will calculate the average usage of the 100 people, generating P3_ 1. By analogy, P3_2 … P3_ n was generated per 100 persons.

In the present invention, the step (2) includes the steps of:

(21) a preliminary estimated static individual ELP initial value L _0,

l _0 ═ L _0a + L _0b +. L _0s)/s, L _0s ═ th corpus ELP value · individual accuracy rate (longest answer time/actual answer time), corpus is corpus text in the test library, longest answer time is a set value for corpus preparation, and the following is an example of a corpus including the following:

[ PROPERTIES ]

1、Flesch-Kincaid Grade Level：12.7

2. The number of words: 366

3. Subject matter: interpersonal relationship (Interpersonal relationships)

4. Sub-topic: peoples (parent, brother, site, other family members, friend, neighbor, teacher, etc.)

5. Key words: generation gap, conflict, differences

6. And (3) cutting: description text

7. Topic type: filling in the blank

8. ELP index: 59.19

9. The longest answering time is as follows: 14 minutes

10. The corpus questioner:

11. and (3) corpus auditor: chen indulge Yu

12. Taking out: quality investigation paper for high three English words in second school period of 2013 school year in northgate district

13. Grade and teaching material: gaoshanjin

14. The school period is as follows: first school term

15. A unit: unit 2 Society and Change

16. Title:

(22) first capacity evaluation: according to the primary evaluation link L _0, pushing 10 corpora of corresponding levels from the test library to test, and obtaining the capability level L _1 (static personal ELP value) of the user, wherein L _1 is 0.5L _1_ a + 0.5L _1_ b, L _1_ a is Q1+ Q2+ … Q10/10, Qc represents the ELP value of the c-th corpus and the correct rate of the corpus, and L _1_ b is (K1+ K2+ K3+ K4)/4, wherein K1 is the average value of Q of grammar corpus, K2 is the average value of Q of vocabulary corpus, K3 is the average value of Q of finish corpus, K4 is the average value of Q of reading corpus, namely, the average value of the sum of the ELP values and the accuracy of the reading corpuses in each corpus, namely, the average value of the sum of the ELP values and the accuracy of the eight reading corpuses in ten corpuses.

In step (2), the user ability evaluation is updated once every month, L _ k is 0.4L _ { k-1} +0.6 { [ L _ { k-1} _1+ L _ { k-1} _2+ … L _ { k-1} _ q }/q }, L _ k represents the ELP value of the person in this month, L _ { k-1} represents the ELP value of the person in the previous month, L _ { k-1} _ q is the personal ELP value calculated by the system at the time of the q-th answer in the previous month, b corpus exists at the time of the q-th answer, and L _ { k-1} _ q is the sum of the ELP values of each corpus and the correct rate/b.

For example: the dynamic personal ELP value of A in 3 months of study is marked as L _3, and the dynamic personal ELP value of A in 2 months of study is marked as L _ 2. In the month time from 2 months to 3 months, if a classmates complete 4 tasks (each task includes several words), 4 pieces of data L _2_1, L _2_2, L _2_3, and L _2_4 are generated, and then L _3 is 0.4L _2+0.6 { (L _2_1+ L _2_2+ L _2-3+ L _2_4)/4 }. Each task is determined by the teacher or by adaptive push decisions selected by the student. How many pieces at a time are not specified uniformly, but are averaged. L _2_1 is the average of the student's first task in month 2. Such as: if the first task in month 2 has 5 corpora, the value of L _2_1 is Q1+ Q2+ … Q5/5, Qc indicates the ELP value of the corpus in the c-th set and the accuracy of the corpus, the ELP values of the first five corpora are 32, 33, 34, 35 and 36, respectively, and the accuracy is 0.8, 0.85, 0.75, 0.85 and 0.75, respectively, and L _2_1 is (32: 0.8+ 33: 0.85+ 34: 0.75+ 35: 0.85+ 36: 0.75)/5: 27.2.

Claims

1. A method for dynamically matching English corpus difficulty with student ability analysis is characterized by comprising the following steps:

(3) according to the index of the user ability, the system recommends the corpus matched with the user ability;

the step (1) comprises the following steps:

(11) the system acquires English texts and marks the English texts, namely the texts are a (m), wherein m represents the sequence of acquiring the texts, the word number, the genre and the question type of the texts are acquired, and the number of answering persons is set;

(12) calculating the corpus difficulty K _1 of the text, wherein K _1 is 0.39 (the total word number/the total sentence number of the English text) +11.8 (the total syllable number/the total word number) -15.59;

(13) calculating local grading vocabulary difficulty K _2 in an English text;

(14) calculating an ELP index D _0 ═ k _3 of the initial difficulty of the corpus topics of the platform;

(15) the system obtains the difficulty index elp (m) of english text 0.4 × K _1+0.3 × K _2+0.3 × D _0, and stores the index elp (m)) in the system;

(16) the answerer selects the text to answer, and records the number P of answerers, the number Q of wrong answerers and the answering time t of the answerer;

(17) when the number of P is the same as the set value, D _0 ═ K _3 ═ Q/P, ELP (m) ═ ELP ═ 0.4 ═ K _1+0.3 ═ K _2+0.3 ═ D _0, the answering time is the average time for the number of answerers P, P, Q is cleared to zero, and (a), (m), ELP (m) are stored in the system; when the number of P is different from the set value, repeating the step (16)

(18) And (5) repeating the steps (16) and (17).

2. The method for dynamically matching difficulty of english corpus with student ability analysis according to claim 1, wherein: in the step (13), K _2 ═ W1+ W2+. Wn)/n, W represents a word, n represents the number of words in the sample,

LEXISn＝Sum(Z_n1+Z_n2+Z_n3+Z_n4+Z_n5+Z_n6)，Z_n1is the Z absolute value of the COCA word frequency of the word; z_n2Is the Z absolute value of the BNC word frequency of the word; z_n3COCA port for wordsConverting the Z absolute value of the indicator; z_n4Is the Z absolute value of the COCA academic index for the word; z_n5Is the Z absolute value of the BNC spoken language index of the word; z_n6Is the Z absolute value of the BNC academic index for the word, Ymax is the maximum value of the Y values in the sample, Ymin is the minimum value of the Y values in the sample.

3. The method for dynamically matching difficulty of english corpus with student ability analysis according to claim 1, wherein: in the step (14), K _3 is 0.1 × P _1+0.1 × K _2+0.1 × P _3+0.1 × P _4+0.2 × P _5+0.2 × P _6, where P _1 is a value corresponding to the corpus difficulty K _1 converted into an integer, P _2 is a value corresponding to the word number, P _3 is a value corresponding to the answer time of the answerer, an initial value of P _3 is a value applied by the english expert in the average time +2 minutes, P _4 is a value corresponding to the genre, P _5 is a value corresponding to the topic type, P _6 is a value corresponding to the year level, and values of P _1, P _2, P _3, P _4, P _5, and P _6 are listed as follows:

the value of P _1 is as follows:

the value of P _2 is as follows:

number of words in language 200 and below 201—300 301—400 401—500 501 and above P_2 60 70—80 81—90 91—100 100

The value of P _3 is as follows:

the value of P _4 is as follows:

the value of P _5 is as follows:

the value of P _6 is as follows:

grade of year Gao Yi High two Gao III Value taking 80 90 100

。

4. The method for dynamically matching difficulty of english corpus with student ability analysis according to claim 1, wherein: the step (2) comprises the following steps:

(21) initially estimating a static personal ELP initial value L _0, where L _0 is (L _0a + L _0b + … L _0s)/s, and L _0s is an s-th corpus ELP value, i.e., personal accuracy (longest answer time/actual answer time), and the corpus is a corpus text in the test library;

(22) first capacity evaluation: and pushing 10 sets of linguistic data of corresponding layers from a test library according to a primary evaluation link L _0 to test, wherein each set of linguistic data at least comprises three of a grammar, a vocabulary, an intact form or a reading to obtain the ability level L _1 of the user, the static personal ELP value L _1, L _1 is 0.5, L _1_ a +0.5, L _1_ b, L _1_ a is [ Q1+ Q2+ … Q10 ]/10, Qc represents the ELP value of the c-th set of linguistic data and the accuracy of the set of linguistic data, L _1_ b is (K1+ K2+ K3+ K4)/4, K1 is the average value of Q of the grammatical data in the ten sets of linguistic data, K2 is the average value of Q of the ten sets of linguistic data, K3 is the average value of the ten sets of Q of the ten sets of linguistic data, and K4 is the average value of the ten sets of Q of the ten sets of the Q in the ten sets of linguistic data.

5. The method of claim 4, wherein said method comprises: in the step (2), the user ability evaluation is updated once every month, L _ k is 0.4 × L _ { k-1} +0.6 { [ L _ { k-1} _1+ L _ { k-1} _2+ … L _ { k-1} _ q }, L _ k represents the ELP value of the person in the month, L _ { k-1} represents the ELP value of the person in the last month, and L _ { k-1} _ q is the personal ELP value calculated by the system at the time of answering the question q times in the last month.

6. The method for dynamically matching difficulty of english corpus with student ability analysis according to claim 1, wherein: in the step (3), the corpus index ELP value of the first corpus recommended by the system is 5 greater than the user ability index ELP value, and then the corpus index ELP values of the corpuses recommended by the system to the user present increasing distribution.