JP6235373B2 - Language analysis method and system - Google Patents

Description

  The present invention relates to a language analysis method for automatically analyzing correctness of English sentences, and more particularly to a language analysis method and system for automatically analyzing correctness of prepositions in English sentences with a computer.

  For the purpose of educational support, etc., if the computer automatically analyzes the correctness or inaccuracies of freely-written English sentences, it is necessary to understand the meaning of the English sentences. Is used. First, the sentence is decomposed into words (morphemes), the syntax information and semantic information are added to each word by referring to the dictionary (morpheme analysis processing), and the phrase structure and dependency structure of the sentences are machined according to predetermined rules. Analyze automatically. The meaning of the sentence to be expressed can be analyzed by giving the semantic consistency between the verb and the noun from the information of the case tree for the obtained syntax tree and verb.

  For example, in Patent Document 1, when a word string of an English sentence that has been subjected to morphological analysis processing is input, the syntactic and semantic relationship is determined by sequentially focusing on the words at the beginning of the word string, and the phrase structure of the English sentence or A method for automatically analyzing a parsing process for determining a dependency structure by a computer is disclosed. Here, a method of determining a valid connection destination by a parsing process in an English sentence including a prepositional phrase having a plurality of connectable words is particularly described. According to this, it is possible to give an accurate grasp of the meaning that the sentence intends to express.

  By the way, the usage of prepositions which are English words (morphemes) is complicated, and it is difficult to select an appropriate preposition according to the context. For example, “He will go back Japan.” Is an error that the preposition of “He will go back to Japan.” Is missing, but because it is idiomatic, it can be judged relatively easily. . On the other hand, for example, “I walked with my dog in the morning.” Does not require the preposition “with”, but it is difficult to explain the reason for the error. In other words, “walk with a dog” reminds us of walking with a dog like a dog, so “walk a dog” is natural when it means walking the dog. ” Will give an explanation. Here, the correctness / incorrectness of the presence or absence of a preposition may differ depending on the meaning of the sentence.

  Even when the correctness of the preposition in such an English sentence is automatically analyzed by a computer, syntax analysis processing and semantic analysis can be used. In recent years, various corpora have been prepared, but Non-Patent Document 1 states that preposition errors do not occur randomly but tend to be in error according to the mother tongue. A corpus of English sentences created by a person who speaks native language should reflect the tendency of preposition errors peculiar to the specific language. Furthermore, for example, by using a method of automatically generating a case frame from a corpus as described in Non-Patent Document 2, it is possible to automatically analyze the correctness of a preposition in an English sentence by a computer.

JP 2005-134691 A

Alla Rozovskaya and Dan Roth, "Algorithm Selection and Model Adaptation for ESL Correction Tasks", Proc. Of the 49th Annual Meeting of the Association for Computational Linguistics, pp 924-933, Portland, Oregon, June 19-24, 2011 D. Kawahara and S. Kurohashi, "Acquiring reliable predicate-argument structures from raw corpora for case frame compilation", Proc. Of LREC, pp.1389-1393, 2010.

  As mentioned above, if there is a tendency for preposition errors according to the mother tongue, the case frame obtained from a corpus of English sentences created by a person whose mother tongue is a language other than English, and those who are native speakers of English By comparing with the case frame obtained from the corpus of the created English sentence, it should be possible to automatically analyze the correctness of the English sentence created by a person whose mother tongue is a specific language other than English.

  The present invention has been made in view of the above situation, and an object of the present invention is to provide a language analysis method for automatically analyzing the correctness of a preposition in an English sentence by a computer.

  The language analysis method according to the present invention is a computer-based language analysis method for automatically analyzing the correctness of a preposition in an English sentence, and (1) a reference English corpus consisting of an English sentence by an English native speaker and other than English In each of the specific English corpora consisting of English sentences by a native speaker of a specific language, a verb and a surface case taken by the verb, a basic case indicator corresponding to the type of case, and a word that is an element of the case A case frame comprising a set of a basic case including a corresponding basic case element, a preposition case indicator corresponding to a preposition, and a preposition case element corresponding to a word that is an element to which the preposition is given. And (2) for the case frame from the reference English corpus, the preposition case indicator of the verb, the basic case, and the preposition case. Integrating the preposition case elements into a case frame as a union of common preposition case elements, and (3) the preposition for the case frame from the specific English corpus not present in the case frame from the reference English corpus The preposition corresponding to the case mark is changed to correspond to a preposition that is probabilistically erroneous in a native language speaker of the specific language, and the preposition case mark is changed to match one of the case frames from the reference English corpus. Adding the prepositional case element of the case frame from the matched reference English corpus as a union, and (4) a preposition in an English sentence created by a native speaker of the specific language. Determining whether the error is true or false based on the error case frame.

  According to this invention, an error case frame can be created using a reference English corpus consisting of English sentences by an English native speaker and a specific English corpus consisting of English sentences by a native language speaker other than English, The computer can automatically analyze the correctness of prepositions in English sentences by native speakers of a specific language.

  In the above-described invention, (3) the preposition corresponding to the preposition case indicator is probabilistically erroneous in a native language speaker of the specific language, and the specific language corresponding to the preposition corresponding to the preposition case indicator It may be one of the English prepositions corresponding to the word. According to this invention, the correctness of the preposition in the English sentence as described above can be automatically analyzed with higher accuracy by the computer.

  The language analysis system according to the present invention is a computer-based language analysis system for automatically analyzing the correctness of a preposition in an English sentence, and (1) a reference English corpus consisting of an English sentence by an English native speaker and other than English In each of a specific English corpus consisting of English sentences by a native speaker of a specific language, a verb and a surface case taken by the verb, a basic case indicator corresponding to the type of case, and a word that is an element of the case And a preposition case including a preposition case indicator corresponding to a preposition case indicator corresponding to a preposition and a preposition case element corresponding to a word that is an element to which the preposition is given. Means for obtaining a frame; (2) for the case frame from the reference English corpus, the preposition of the verb, the basic case and the preposition case; Means for integrating the prepositional case elements into a case frame as a union of common signs, and (3) for the case frame from the specific English corpus not present in the case frame from the reference English corpus, The preposition corresponding to the preposition case marker and the preposition case marker corresponding to the preposition that is probabilistically erroneous in the native language speaker of the specific language is changed to change any of the preposition case markers from the reference English corpus. Means for adding a preposition case element of the case frame from the matched reference English corpus as a union with the case frame as an error case frame if they match.

  According to this invention, an error case frame can be created using a reference English corpus consisting of English sentences by an English native speaker and a specific English corpus consisting of English sentences by a native language speaker other than English, The computer can automatically analyze the correctness of prepositions in English sentences by native speakers of a specific language.

  In the above-described invention, it may include (4) means for determining whether a preposition is correct or incorrect in an English sentence created by a native speaker of the specific language based on the error case frame. According to this invention, the correctness of the preposition in the English sentence as described above can be automatically analyzed by the computer.

In the above-described invention, (3) the preposition corresponding to the preposition case indicator is probabilistically erroneous in a native language speaker of the specific language, and the specific language corresponding to the preposition corresponding to the preposition case indicator It may be one of the English prepositions corresponding to the word. According to this invention, the correctness of the preposition in the English sentence as described above can be automatically analyzed by the computer with higher accuracy.

It is a figure which shows the system configuration | structure of this invention. It is a figure which shows an error case frame. It is a flowchart of the error case frame production | generation which is the principal part of the method of this invention. It is a figure which shows a case frame. It is explanatory drawing of integration of a case frame. It is explanatory drawing of the preposition which is easy to mistake stochastically. It is a figure about determination of the correction information of a case frame. It is a figure about determination of the correction information of a case frame. It is a figure about determination of the correction information of a case frame.

  The details of a language analysis method for automatically analyzing the correctness of a preposition in an English sentence by a computer and a system therefor according to one embodiment of the present invention will be described with reference to FIGS.

  As shown in FIG. 1, the language analysis system 1 mainly includes a central control unit 30 that plays a central role in analysis processing, an error case frame generation unit 32 as a processing program that performs various processes together with the central control unit 30, and a correctness determination unit. 34. The central control unit 30 appropriately includes a native speaker corpus (reference English corpus) 102 made up of English sentences by an English native speaker of the corpus unit 100 and non-English sentences made up of native speakers of a specific language other than English. The information of the native speaker corpus (specific English corpus) 104 is connected so that it can be referred to. The corpus unit 100 may be outside the language analysis system 1 and can be referred to via a net line or the like. Furthermore, an input device 51 such as a keyboard or a scanner for inputting text data of an English sentence that is a target of correctness determination of the preposition, a printing machine (printer) 52 and a video device (monitor) 53 for outputting analysis results, etc. Are connected via the input / output interface unit 50.

  First, the configuration of the error case frame 10 created by the error case frame creation unit 32 will be described.

  As shown in FIG. 2, the error case frame 10 always includes a verb 17 that is the center of a sentence. In addition to the verb column 11 as a slot in which the verb 17 is written, the basic case column 12 as a slot in which the basic case is written. , A preposition case column 14 as a slot describing a preposition case, and a feedback message column 16 which is a portion describing an explanation of a preposition error. That is, the basic case column 12 and the preposition case column 14 are columns describing what surface case the verb 17 in the English sentence takes.

  Case indicators 18 such as “Subj:”, “Prt:”, “Prep_do:”, and “Prep_with:” in the basic case column 12 and the preposition case column 14 are labels for indicating the type of case such as the main case. In addition, the case elements 19 of “PERSON”, “back”, “tokyo”, and “japan” lined next to these case indicators 18 represent words to which the case indicator 18 is attached. Note that “PERSON” and curly braces {} representing people will be described later. Further, hereinafter, unless otherwise specified, “case” refers to a combination of the case indicator 18 and the case element 19.

  Here, the basic case column 12 is composed of at least one case, and the case indicators 18 that can enter the basic case column 12 include, for example, at least “Subj:” (Subject: main case), “Prt” (Particle : Verb) and “Com” (Complement). “Subj:” is indispensable.

  The preposition case column 14 is also composed of at least one case, and describes prepositions that can be taken by the verb. Specifically, the case indicator 18 in the preposition case column 14 is described as “Prep_x”. However, a preposition is entered in the x part. For example, if the preposition is “to”, it is described as “Prep_to”. Note that “Prep_do” for representing the direct object of the verb and “Prep_io” for representing the indirect object are also included in the preposition case for convenience. This is to cope with omission of prepositions or when no preposition is required.

  Further, in the preposition case column 14, “*” is given to a case with an error to indicate error information. For example, in Figure 2, “* Prep_do: {tokyo, japan}” is an incorrect case, and it is incorrect to take “Prep_do”, that is, “tokyo” or “japan” as the direct object. It means that some kind of preposition is necessary. On the other hand, correction information is described using “→” after the case that is an error. In other words, “* Prep_do: {tokyo, japan}” means that “Prep_to:” is correct as a preposition case.

  In the basic case column 12 and the preposition case column 14, two types of description methods commonly used are defined. One is an arbitrary case expressed by using parentheses “()”, for example, “(Prt: back)”. The other is that when there are a plurality of case elements 19, a plurality of case elements are separated by commas and enclosed in braces, for example, “* Prep_do: {tokyo, japan}”. For example, when an error case frame 10 that does not depend on a specific verb is defined, “ALL” may be entered in the verb column 11. In other words, the error frame 10 corresponds to any verb 17. Thus, error case frames 10 for different verbs having the same case and the same error can be collectively described. Similarly, also in the case indicator 18 and the case element 19, it is possible to give a description that does not depend on a specific thing so that the error case frames 10 can be described collectively.

  The feedback message column 16 is a part describing an explanation regarding the preposition error, and is mainly described by an operator by interpreting an error case frame 10 described later. Such an explanation can be used for feedback to a learner during error detection / correction.

  Next, a method for creating the error case frame 10 by the error case frame creation unit 32 will be described with reference to FIG.

  Incidentally, the basic idea for generating the error case frame 10 is that the case frame that exists in the non-native speaker corpus 104 and does not exist in the native speaker corpus 102 is the error case frame 10. However, with this alone, the correct case frame is extracted as the error case frame 10. Therefore, the following method is adopted.

(1) Generation of Case Frame from Corpus First, for each of the native speaker corpus 102 and the non-native speaker corpus 104, a sub-process for parsing each sentence is performed (FIG. 3, S1). Here, in this analysis, it is preferable to exclude in advance a sentence inappropriate for creating an error case frame 10 described later. This is, for example, processing for removing a sentence having a token length greater than or equal to a predetermined number or a sentence including a predetermined number or more of commas. In addition, this processing may be applied only to the native speaker corpus (reference English corpus) 102 that is larger than the non-native speaker corpus (specific English corpus) 104 as appropriate.

  Next, as shown in FIG. 4, a case frame is generated from the result of syntax analysis by filling the slots of the verb column 11, basic case column 12, and preposition case column 14 of the case frame 10a (FIG. 3, S2). . For example, for the English sentence “He will go back Japan with his son.” Shown in FIG. 4A, the verb “go” is entered in the verb column 17 as shown in FIG. Place the case in the corresponding place.

  Here, as the case element 19, a corresponding noun equivalent phrase head (lower) in lowercase and original form is used. For example, “Japan” is “japan”. However, since the suffix “-ing” may affect the determination of the preposition, the word ending with “-ing” is not made the original form. Some words are replaced with special words representing the corresponding meanings. Special words representing this meaning are written using only capital letters. For example, “he” and “his son” are replaced with “PERSON” representing a person. Such replacement can be done automatically based on a simple dictionary lookup. In addition, it is preferable to correct spelling errors by a spell checker for processing from the non-native speaker corpus (specific English corpus) 104.

  Hereinafter, a set of case frames 10a extracted from the native speaker corpus 102 and the non-native speaker corpus 104 will be referred to as a native speaker case frame 10b and a non-native speaker case frame 10c.

  By the way, when any of the following three conditions is satisfied, it can be considered according to circumstances that the case frame 10a is not generated as an exception. The first is a case where verbs are paralleled by a conjunction, for example, “go and get it”. This is because the usage of prepositions may be changed by parallelism. Second, “be”, “do”, and “have” are special verbs that are also used as auxiliary verbs, and can be exceptions. Third, it can be considered that the case element is “it”, “this”, “that”, “one” and a word that does not function as a normal noun, for example, “the”. This is because “it”, “this”, “that”, and “one” are considered to be used in different ways depending on what is specifically indicated. For others, there is a high possibility of syntax error.

Here, the identification of the preposition case that is an arbitrary case in the case frame is as follows: (i) The object is always an indispensable case (the object is also treated as a preposition case for convenience), (ii) To the left of the verb The preposition case that appears is always an arbitrary case. (Iii) The preposition cases that appear to the right of the verb are all arbitrary cases except the one closest to the verb. For example, with regard to (ii), “In the morning, he went shopping.” Is an arbitrary case because the preposition case of “In the morning” appears to the left of the verb. Regarding (iii), “He went to the market with his family.” Means that “with his family”, which is farther from the verb, is an arbitrary case.

In addition to the heuristics described above, two case frames may be compared for arbitrary case identification. For example, comparing “He went shopping” and “He went shopping at the market.”, It can be established as a sentence without “at the market”, and can be identified as an arbitrary case.

(2) Case Frame Integration An integration process is performed on the native speaker case frame 10b extracted from the native speaker corpus 102 (FIG. 3, S3). In the integration process, for the two case frames 10a of the native speaker case frame 10b, (i) the verb is the same, (ii) the basic case is the same, and (iii) the preposition case case mark is the same. In some cases, the case elements 19 in the preposition case column 14 are integrated for each case indicator 18.

  As shown in FIG. 5, for example, [Prep_to: tokyo] of the case frame 10b-1 and [Prep_to: japan] of the case frame 10b-2 are case elements 19 of “Prep_to:” “tokyo” and “ In order to make common except for “japan”, the case element 19 is integrated into [Prep_to: {tokyo, japan}] as in the case frame 10b by making the form of a union using curly braces {}.

  As described above, the integration process is performed only for the native speaker case frame 10 b extracted from the native speaker corpus 102. This is because the non-native speaker case frame 10c extracted from the non-native speaker corpus 104 includes both a correct case frame and an erroneous case frame. This is because the case element 19 is included. The non-native speaker case frame 10c may be integrated only when the verb, the basic case, and the preposition case are the same. Further, at the time of integration, the frequency of each case element 19 may be recorded, and the frequency information may be used for generating the error case frame 10.

(3) Acquisition of Error Case Frame Candidates The native speaker case frame 10b and the non-native speaker case frame 10c are compared to acquire the error case frame 10 candidates (FIG. 3, S4). Here, case frames that exist only in the non-native speaker case frame 10 c extracted from the non-native speaker corpus 104 are set as candidates for the error case frame 10.

(4) Determination of correction information Correction information is determined for the error case frame 10 candidate (FIG. 3, S5). This is determined by changing the case indicator 18 in the preposition case column 14 using an error set (confusion set) described later in consideration of the influence of the native language. In addition, when there are a plurality of case indicators 18, only one is changed and determined.

  By the way, in order to consider the influence of the native language, a non-native speaker corpus 104 composed of English sentences written by the native speaker may be used. For example, when a French speaker whose native language is French is used, an English sentence written by the French speaker is used for the non-native speaker corpus 104, so that the influence of the native language can be considered naturally.

  Here, as described in Non-Patent Document 1 described above, preposition errors do not occur randomly, but tend to be errors according to the mother tongue. For example, as shown in FIG. 6, the French preposition “` a ”corresponds to the English prepositions“ at ”,“ in ”,“ to ”, etc., so French speakers confuse these prepositions with each other. It can be expected that there is a tendency. Therefore, for example, [“at”, “in”] is an error set for “to”. An error set of each English preposition corresponding to such a native language is prepared and used for determination of correction information (S5 in FIG. 3).

  In this embodiment, an error set is automatically created using a statistical machine translation probability table. That is, intuitively, prepositions that are easily confused are specified based on the probability value.

  Referring to FIG. 6 again, specifically, the left column is a French (French) word and the right column is an English word. In this example, all the left and right words are prepositions, but they are not necessarily prepositions and may be corresponding words. Arrows in the figure represent English words in which each French word is easily translated. That is, when “e” represents an English word and “f” represents a French word, an arrow is given to a pair of words having a probability Pr (e | f) equal to or greater than a certain value. For example, “to” in English indicates a high probability of being translated from “` a ”in French. On the other hand, French “` a ”is easily translated into“ at ”and“ in ”in addition to“ to ”. In other words, “to” is easily confused with “at” or “in”. In this way, an error set is created by following the arrow twice.

  Finally, "Prep_do" and "Prep_io" are also added to the error set, respectively, in order to cope with missing prepositions and cases where prepositions are unnecessary. For example, {Prep_at, Prep_in, Prep_do, Prep_io} is obtained as an error set for “to”.

  In FIG. 6, when there are a plurality of arrows to be traced for the first time as in “in”, each arrow is traced, and the union of the obtained prepositions is set as an error set. That is, the error set for “in” is {Prep_to, Prep_at, Prep_of, Prep_do, Prep_io}.

  Determination of correction information of [Prep_do: Tokyo Prep_with: PERSON] shown in FIG. 7A will be described. For “Prep_do:”, an error set including this is selected, and a case indicator 18 including other prepositions, for example, “Prep_at” or “Prep_to:” of the error set {Prep_at, Prep_in, Prep_do, Prep_io} is selected. Change to For example, if you change to “Prep_to:”, you get [Prep_to: Tokyo Prep_with: PERSON]. If such a case frame is present in the native speaker case frame 10b, the case indicator 18 is determined to be correct and determined as correction information (see FIG. 7B). Further, as shown in FIG. 8, “*” indicating correction information is given, and what is described as the error case frame 10 candidate is determined as the error case frame 10.

(5) Expansion of case element Next, in order to improve the coverage of the error case frame 10, the case element 19 in the preposition case column 14 is expanded (FIG. 3, S6). With the correction information described above, the correct case frame corresponding to the error case frame 10 can be specified in the native speaker case frame 10b. As shown in FIG. 7B, for example, [* Prep_do: tokyo → Prep_to] in the error case frame 10 (see FIG. 8), [Prep_to: {tokyo, japan}] in the native speaker case frame 10b. Corresponds. In the native speaker case frame 10b by the integration process (FIG. 3, S3), the case elements 19 are integrated and described as a union. As shown in FIG. 9, the case element 19 of the error case frame 10 can be expanded by adding the information of the case element 19 to the corresponding case of the error case frame 10. That is, as shown in FIG. 2, “japan” is added to the case element 19 as [* Prep do: {tokyo, japan} → Prep_to]. In order to confirm whether or not this extension truly represents an error, the extension is allowed only when the newly obtained error case frame 10 does not exist in the native speaker case frame 10b.

(6) Output of error case frame The obtained error case frame 10 is output and constructed in a predetermined database (FIG. 3, S7). The information about each error frame 10 described above can be output, for example, in an XML format, constructed in a database, and used for correctness determination described later.

  As described above, the error case frame 10 can be automatically generated as long as the native speaker corpus 102 and the non-native speaker corpus 104 are present, and does not require the operation of adding error information that requires time and effort. In such a method, the validity of the error case frame 10 is checked by comparing two corpora twice instead of needing to give error information (FIG. 3, S4 and S5).

  It is also possible to generate the error case frame 10 using the non-native speaker corpus 104 to which error information is assigned. In that case, the error case frame 10 is selected and the correction information is determined based on the error information.

  Next, a description will be given of a method of determining whether an English sentence by a native speaker of a specific language other than English related to the non-native speaker corpus 104 is correct or incorrect by using the error case frame 10 described above.

  First, the correctness / incorrectness determination unit 34 generates a case frame after performing a syntax analysis on an English sentence for which correctness / incorrectness determination is performed. The case frame generation method is the same as the case frame generation from the corpus described above. Next, if such a case frame is matched with the error case frame 10 in the above-mentioned predetermined database and any of the verb field 17, the case indicator 18, and the case element 19 match, It is determined that there is an error in the preposition. In addition, when the correction information of the error case frame 10 that matches in the correctness determination is used, it is also possible to correct the error of the preposition.

  In the correctness determination, as described above, by using the database of the error case frame 10 obtained by using the non-native speaker corpus 104 corresponding to the native language of the creator of the English sentence for which the correctness determination is performed, the correctness determination is performed. Accuracy can be improved. For example, an English sentence by a Japanese speaker whose native language is Japanese uses a database of error case frames 10 obtained by using a non-native speaker corpus 104 of English sentences written by a Japanese speaker. In this case, there are no prepositions in Japanese, but particles are supported, error sets can be created in the same way, and preposition error sets that are easily confused by Japanese speakers are used, as in the example of French speakers described above. Thus, the error case frame 10 can be generated. In addition, correctness determination and error correction are possible.

  The correctness determination unit 34 may be provided in an attached system different from the language analysis system 1 including the error case frame generation unit 32. The error case frame 10 extracted by the case frame creation unit 32 can be accessed as a database, XML format, or the like from an attached system including the correct / incorrect determination unit 34, so that English sentences by native speakers of a specific language can be determined correctly.

    According to the above-described embodiment, an explanation regarding an error can be given as a feedback message as appropriate according to the purpose. For example, it can be provided in a form in which a human can intuitively understand why the correction candidate is selected.

  As mentioned above, although the Example by this invention and the modification based on this were demonstrated, this invention is not necessarily limited to this, A person skilled in the art will deviate from the main point of this invention, or the attached claim. Various alternative embodiments and modifications could be found without doing so.

10 Error case frame 10a Case frame 10b Native speaker case frame 10c Non-native speaker case frame 18 Case indicator 19 Case element

Claims (5)

A computer-based language analysis method for automatically analyzing the correctness of prepositions in English sentences,
(1) In each of a reference English corpus consisting of English sentences by an English native speaker and a specific English corpus consisting of English sentences by a native speaker of a specific language other than English,
Verbs and
About the surface case that the verb takes,
A basic case including a basic case indicator corresponding to a case type and a basic case element corresponding to a word which is an element of the case;
A preposition case that includes a preposition case indicator corresponding to the preposition and a preposition case element corresponding to a word that is an element to which the preposition is given;
Obtaining a case frame consisting of a set of
(2) For the case frame from the reference English corpus, integrating the common preposition case markers of the verb, the basic case, and the preposition case into the case frame as a union of the preposition case elements When,
(3) The case frame from the specific English corpus that does not exist in the case frame from the reference English corpus is probabilistically erroneous in the native language speaker of the specific language with the preposition corresponding to the preposition case indicator. The preposition of the case frame from the matched reference English corpus, if it matches any of the case frames from the reference English corpus, by changing the preposition case indicator in response to a preposition, if this matches any of the case frames from the reference English corpus Adding a case element as a union;
(4) A language analysis method comprising: determining whether a preposition is correct or incorrect in an English sentence created by a native speaker of the specific language using the error case frame. (3) Prepositions that are probabilistically erroneous in native speakers of the specific language due to the prepositions corresponding to the prepositional case markers further correspond to the words of the specific language corresponding to the prepositions corresponding to the preposition case indicators. The language analysis method according to claim 1, wherein the language analysis method is one of English prepositions. A computer-based language analysis system for automatically analyzing the correctness of prepositions in English sentences,
(1) In each of a reference English corpus consisting of English sentences by an English native speaker and a specific English corpus consisting of English sentences by a native speaker of a specific language other than English,
Verbs and
About the surface case that the verb takes,
A basic case including a basic case indicator corresponding to a case type and a basic case element corresponding to a word which is an element of the case;
A preposition case that includes a preposition case indicator corresponding to the preposition and a preposition case element corresponding to a word that is an element to which the preposition is given;
Means for obtaining a case frame consisting of a set of
(2) Means for integrating the common preposition case markers of the verb, the basic case, and the preposition case into one case frame as a union of the preposition case elements for the case frame from the reference English corpus When,
(3) The case frame from the specific English corpus that does not exist in the case frame from the reference English corpus is probabilistically erroneous in the native language speaker of the specific language with the preposition corresponding to the preposition case indicator. The preposition of the case frame from the matched reference English corpus, if it matches any of the case frames from the reference English corpus, by changing the preposition case indicator in response to a preposition, if this matches any of the case frames from the reference English corpus A language analysis system comprising: means for adding case elements as a union. (4) The language analysis system according to claim 3, further comprising: means for determining whether a preposition is correct or incorrect in an English sentence created by a native speaker of the specific language based on the error case frame. (3) Prepositions that are probabilistically erroneous in native speakers of the specific language due to the prepositions corresponding to the prepositional case markers further correspond to the words of the specific language corresponding to the prepositions corresponding to the preposition case indicators. The language analysis system according to claim 3, wherein the language analysis system is one of English prepositions.