JPH03225389A - Device for practicing language - Google Patents

Abstract

PURPOSE:To perform a practice corresponding to the level of ability for language of a learner and to improve a learning effect by detecting the level of the ability for language of the learner based on an inputted sentence and deciding an outputted sentence in accordance with the semantic expression of the inputted sentence and the level of the ability for language. CONSTITUTION:An inputted sentence understanding part 6 performs the syntax analysis of the sentence inputted by the learner so as to generate a syntax analysis tree, and information with which the level of the ability for language of the learner can be detected is extracted. The dictionary which is used in such a case and the rule of grammar are stored in a grammar dictionary knowledge 9. Next, the semantic expres sion is generated from the understanding part 6 and the syntax tree and transferred to a conversation control part 7. The control part 7 calculates the synthetic level of the ability for language of the learner and decides the contents of speech by a device in accordance with a conversation control rule stored in a conversation control knowledge 10, then gives them to an outputted sentence generation part 8. The genera tion part 8 receives the decided contents of speech and information on the level of the ability for language, generates the outputted sentence based on an outputted sen tence generation knowledge 11 and outputs it to display.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a language training device that uses a computer system centered on conversation simulation, and in particular, to a language training device that uses a computer system that mainly performs conversation simulation. This is an attempt to properly train learners.

[Prior art] It has been pointed out that language education that applies grammar and translation-reading teaching methods is insufficient in terms of conversation, and to compensate for this, audio tapes, CDs, Videotapes and the like are commercially available. However, although these methods are effective for memorizing many example sentences and patterned conversations, the most important elements in real conversations are responding according to the situation, and answering questions such as the most important element in real conversation. In many cases, it is not very useful for developing the ability to

To solve these inconveniences, language training devices based on conversation simulation and using computer systems have been proposed (for example, "Intelligent CAI System for Language Training Based on Conversation Simulation"). "Composition", Hideki Yamamoto et al., Information Processing Society of Japan Transactions No. 30
Vol. No. 7, pp. 908-917. July 1989).

[Problems to be Solved by the Invention] However, conventional language training devices centered on conversation simulations perform uniform conversation control regardless of the language proficiency level of the learner. If the training level provided by the language training device is low, learners with high language proficiency levels will be dissatisfied, and if the training level provided by the language training device is high, learners with low language proficiency levels may not be able to keep up with the learning. In other words, there is a problem in that language training is not provided with appropriate guidance.

If such inconveniences are to be solved using conventional language training devices, it is necessary to separately prepare language training devices with different training levels in consideration of the learner's level.

The present invention has been made in consideration of the above points, and provides language training that is highly flexible with respect to training levels, and allows dialogue training of an appropriate level of difficulty to be performed through conversation control that corresponds to the language level of the learner. The aim is to provide a device for

[Means for Solving the Problem] In order to solve the problem, the present invention provides a language training device centered on a conversation simulation that forms an output sentence according to an input sentence from a learner. The language proficiency level detecting means detects the language proficiency level of the learner based on the language proficiency level of the learner, and the output sentence determining means determines the output sentence according to the detected language proficiency level in addition to the semantic expression of the input sentence.

[Operation] In the present invention, the language proficiency level detection means detects the language proficiency level of the learner based on the input sentence and provides it to the output sentence determination means. The output sentence determining means determines an output sentence according to the detected language proficiency level in addition to the semantic expression of the input sentence.

In this way, the training takes into account the language proficiency level of the learner.

[Example] Hereinafter, an example in which the present invention is applied to English language training will be described with reference to the drawings.

Here, Fig. 1 is a flowchart explaining the operation in this embodiment, Fig. 2 is a configuration diagram when the language training device of this embodiment is viewed as a computer system, and Fig. 3 is a functional block diagram of the embodiment device. , FIG. 4 is an explanatory diagram of an example of conversation control rules of the embodiment, FIG. 5 is an explanatory diagram of an example of dialogue between a language training device and a learner, and FIG. 6 is a partial content of grammar dictionary knowledge of the embodiment. FIG. 7 is an explanatory diagram showing this example of syntactic analysis according to the embodiment.

In Figure 2, this language teaching device using a computer system, like other computer systems,
The computer includes a central processing unit (CPU) 1, a main memory 2, a display device 3, a keyboard device 4, and a hard disk device 5.

The hard disk device 5 includes a natural language processing program for understanding input sentences, generating output sentences, etc., a conversation control program for determining output sentences according to input sentences, and various other programs used for processing at each stage. knowledge is stored. For example, conversation control rules shown in FIG. 4, which will be described later, a word dictionary, grammar rules, etc. shown in FIG. 6 are stored in the hard disk device 5. A natural language processing program and a conversation control program are loaded into the main memory 2 and executed by the CPU 1. As is well known, the display device 3 and the keyboard device 4 are for a man-machine interface between the learner and the training device, and dialogue is executed via these devices 3 and 4.

This embodiment is characterized in that the conversation is controlled in accordance with the language proficiency level of the learner. Next, we will implement conversation control processing that corresponds to the language proficiency level of the learner.
This will be explained with reference to the functional configuration shown in FIG. 3 and the flowchart shown in FIG.

Note that FIG. 1 shows the processing when one input sentence is given, and is repeated every time an input sentence is given.

When an input sentence is given, the input sentence understanding section 6 uses the grammar dictionary knowledge 9 to perform processing to understand the meaning of the input sentence. That is, the input sentence understanding unit 6 parses the sentence input by the learner to generate a parse tree (see Figure 7), and extracts information that can detect the learner's language proficiency level (step 12). ).

At that time, the dictionary and grammar rules to be used are stored in the grammar dictionary knowledge 9. In the case of this embodiment, each word in the dictionary has a level written therein representing the degree of difficulty of that word, and each grammar rule has a level written thereon representing the degree of difficulty of that rule. The input sentence understanding unit 6 performs syntactic analysis of the input sentence, and at the same time stores all the words used and the level described in the grammar rules in variables in order to detect the language proficiency level of the learner. Additionally, knowledge regarding misspellings and grammatical errors in words is also stored in the grammar dictionary knowledge 9.
If the input sentence understanding unit 6 discovers an error during syntax analysis, it presents an educational message and, in the same way as described above,
The number of errors is assigned to a variable in order to detect the language proficiency level of the learner.

Furthermore, the number of words in the input sentence is also substituted into a variable in order to detect the language proficiency level of the learner.

Next, the input sentence understanding unit 6 generates a semantic expression from the parsed syntax tree and passes it to the conversation control unit 7 (step 13
〉.

The conversation control unit 7 first calculates the learner's overall language proficiency level based on a predetermined formula using variables into which the learner's vocabulary level, grammar rule level, number of errors, and number of words are substituted. Calculate (step 14). Conversation control knowledge 10
Conversation control rules (see FIG. 4) are stored in .

This conversation control rule is a rule according to the 1f-then format, for example, which defines the next utterance of the language training device based on the semantic expression of the learner's input sentence and the learner's language proficiency level. When the language proficiency level of the learner is detected by calculation as described above, the conversation control unit 7 determines the next content to be uttered by the device according to the conversation control rules (
Step 15). The determined utterance content is provided to the output sentence generation unit 8 together with the learner's language proficiency level information.

The output sentence generation unit 8 receives the determined utterance content and language proficiency level information, generates an output sentence based on the output sentence generation knowledge 11, and displays and outputs it (step 16). When determining and generating output sentences, if the language proficiency level of the learner is high, sentences using words and structures with high difficulty are selected.

In this way, conversation control is realized according to the language proficiency level of the learner.

Next, a specific example of conversation control according to the embodiment will be explained using the drawings. FIG. 5 shows an example of a dialogue conducted by applying the conversation control rules shown in FIG. A specific example will be explained.

Note that the conversation control rules 21 to 28 shown in FIG. 4 are rules according to the 1f-then format as described above, and each consists of an if part (condition part) 17 and a then part 18. In the case of this embodiment, unlike the conventional example, the if section 17 includes not only the meaning expression 19 of the input sentence but also the language proficiency level information 20 of the learner. Therefore, the then part 20 that satisfies the conditions of the semantic expression 19 of the input sentence and the language proficiency level information 20 of the learner, that is, the sentence uttered by the device (output sentence) is selected.

In FIG. 5, the speaker 29 represents the provider of the sentence, ``Shiva'' represents the training device, ``Gaku'' represents the learner, and the numbers represent the order of utterances. An example of a guest who would like to stay overnight is a receptionist at a hotel or a receptionist at a hotel.

These are the system's display output and the learner's key human interaction. When the speaker 29 is a learner, the utterance content 30 includes variables 31 to 33 used to detect the learner's language proficiency level.
are associated. The first variable *1evel* is
The first element (the first bracket white information in the figure) lists the level of each word used in the input sentence, and the second element (the second bracket white information in the figure) lists the levels of each word used in the input sentence. It shows a list of levels for each grammar rule. Second variable *
error* is the number of grammatical errors, the third variable *w□rd
S* indicates the number of words in the sentence.

FIG. 6(B) shows the contents 51 to 61 of the grammar dictionary knowledge 9 regarding some of the words used in the learner's utterance content (input sentence) 30 in this dialogue example. (A> is the grammatical rule used in the learner's input sentence in this example dialogue (phrase structure rule in word plan functional grammar)
The contents 41 to 49 of the grammar dictionary knowledge 9 are shown.

These dictionary contents are used to determine the first variable *1evel*.

7(A) to 7(C) each show a syntactic parse tree for the learner's utterance content (input sentence) 30 in the dialog example of FIG. 5, and by creating this parse tree, the first variable * 1e
The second element of vel* is determined.

First, in the state where the conversation has started, the training device starts the first utterance "Ma".
y I help you? ” (Figure 5, C1).

In response, the learner said, “I want 5tay he
Suppose that the student responded “ree,” (Study 2).

The input sentence understanding unit 6 understands that the level of each word of this input sentence is (2222) from the dictionary contents 51 to 54 regarding the words shown in FIG. 6(B). That is, the first
Assign (2222) to the first element of the variable *1evel*. In addition, the input sentence understanding unit 6 forms a parse tree for this input sentence as shown in FIG. A＞
Contents regarding the grammar rules shown in 45.43.49.48
Take it out. For example, S (sentence) related to this input sentence is n
The phrase structure rule consisting of p (noun phrase) and vp (verb phrase) is associated with level 2 from the rule with reference numeral 45 in FIG. 6(B). By performing similar processing, (2222> is assigned to the second element of the first variable *]evel*.

The second variable *error* contains the infinitive t
There is an error in missing o and a misspelling of here, so 2 is substituted. Third variable *w.

Assign the number of words 4 to rds*.

The input sentence understanding unit 6 uses the variables 31 to 3 obtained in this way.
3 to the conversation control unit 7. Of course, the input sentence “I want”
t 5tay here,” is the input semantic expression of “
The party who wants to stay the night is also given to the conversation control unit 7.

The conversation control unit 7 first detects the language proficiency level of the learner from the given variables. Calculation for detection is performed using the following formula ((sum of elements of *1evel*) (*error*X2) + (*word*-5)) ' (number of elements of *1eve1*) - (1) . The language proficiency level of the learner is detected by detecting the range to which the value obtained by equation (1) (1.67 in this case) belongs.

Furthermore, learners with a high level of language proficiency also have a higher level of vocabulary and grammar rules than learners with a low level of language proficiency.
Equation (1) is determined taking into consideration that the number of words that make up a sentence is large and errors are thought to be small.

First input sentence “I Want 5tay here,”
”, the learner level is 1.

The conversation control unit 7 uses the rule 22 that has the meaning expression of the input sentence "I want to stay" and the language proficiency level 1 in the if part 7, and then outputs the next output sentence as "5in (Jle, Or double,
Or twin? Determine ′°. Although described here in natural language, the actual situation is different, and the output sentence generation section 8 to which decision information is given from the conversation control section 7 utters in natural language (FIG. 5, C3).

In response to this, assume that the learner responds with "Twin," (Figure 5, Science 4).In response to this, the input sentence understanding unit 6
The conversation control unit 7 performs the same processing as described above. 1st
The first element of the variable *1evel* is (2), the second element of the first variable *1evel* is (12), and the second variable *
error* becomes 0, the third variable *words* becomes 1, and the value according to equation (1) becomes 1. In this way, the language proficiency level of the learner is determined to be 1, the same as before, and based on rule 25, which also includes this language proficiency level as a condition, the device
you have a reservation? ”
(Fig. 5, C5).

In response, the learner may say “Yes I have
It is assumed that the response is ``a reservation,'' (Figure 5, Science 6).

In response to this, the input sentence understanding unit 6 and conversation control unit 7 perform the same processing as described above. The first element of the first variable *1evel* is (232223>, the first variable *1eve
The second element of l* is (2223), the second variable *erro
r* is 0, the third variable *Words* is 5, and (1
) is 2.5. Thus, the language proficiency level is determined to be 2, which is 1 higher than the previous one, and this language proficiency level 2 is
The device makes the next utterance (see Fig. 5) based on rule 27 which also has as one condition.

Therefore, according to the above-described embodiment, since the language training device determines the next utterance by detecting the language proficiency level of the learner, training can be appropriately performed in accordance with the language proficiency level of the learner. It is possible to improve the learning effect.

Furthermore, appropriate training can be provided to learners with different language proficiency levels using the same training device.

Furthermore, according to the above embodiment, since the language proficiency level is detected every time an input sentence is given, high training is performed for situations in which the learner is good at, and low training is performed for situations in which the learner is weak. This allows for highly flexible training.

In the above description, the present invention has been applied to a language training system for English conversation, but it goes without saying that the present invention can also be applied to other language training systems.

Further, the present invention can also be applied to a language training device in which the input sentences and output sentences are audio.

Furthermore, as viewpoints for detecting language proficiency level, other viewpoints may be used in addition to the word level, grammatical rule level, number of errors, and number of words used, and all of the above four viewpoints may be used. It is also possible to determine the language proficiency level by using only a part of the viewpoint without using the above. The calculation formula for determining the language proficiency level is also not limited to formula (1).

In the above embodiment, the language proficiency level is reviewed for each input sentence, but the language proficiency level may be reviewed in units of multiple input sentences.

[Effects of the Invention] As described above, according to the present invention, the device prepares utterances with the same content at different levels, and when an input sentence is given, the language ability of the learner as well as the semantic expression of the input sentence are prepared. Since the level is detected and the utterances of the device are determined based on these semantic expressions and language proficiency level, training can be performed in accordance with the language proficiency level of the learner, and the learning effect can be improved.

[Brief explanation of drawings]

FIG. 1 is a flowchart explaining the operation of an embodiment of the language training device according to the present invention, FIG. 2 is a configuration diagram of the language training device of this embodiment as a computer system, and FIG. 3 is an embodiment of the language training device according to the present invention. A functional block diagram of the device, FIG. 4 is an explanatory diagram of an example of conversation control rules of the embodiment, FIG. 5 is an explanatory diagram of an example of dialogue between the language training device and a learner, and FIG. 6 is a grammar dictionary of the embodiment. Explanatory diagram showing part of the content of knowledge, Part 7
The figure is an explanatory diagram showing this example of syntactic analysis according to the embodiment. 6... Input sentence understanding unit, 7... Conversation control unit, 8...
Output generation unit, 9... Grammar dictionary knowledge, 1o... Conversation control knowledge, 11... Output sentence generation knowledge, 19... Input semantic expression, 20... Language proficiency level.

Claims (1)

[Claims] In a language training device centered on a conversation simulation that forms an output sentence according to an input sentence from a learner, there is provided a language proficiency level detection method that detects a learner's language proficiency level based on the input sentence. 1. A language training device comprising: a means for determining an output sentence according to a detected language proficiency level in addition to a semantic representation of an input sentence.