JP7296029B2 - Language teaching material generation system - Google Patents

Description

The present invention relates to a system for generating language teaching materials.

Conventionally, a method of generating language teaching materials using a computer is known. For example, Patent Literature 1 describes an electronic learning system that incorporates learner's operation information and reaction information and changes an initial learning story to another learning story. According to this electronic learning system, the learning effect can be dramatically improved by tailoring the learner's original learning story.

Japanese Patent Laid-Open No. 2006-308815

However, language learning materials generated by conventional methods do not necessarily reflect the situations in which learners actually use the foreign language. There was a problem of difficulty.

SUMMARY OF THE INVENTION The present invention has been made in view of such circumstances, and an object of the present invention is to generate a language teaching material that reflects a situation in which a learner actually uses a foreign language.

In order to solve the above problems, the present invention provides a speech analysis unit that analyzes the speech of a conversation conducted using a foreign language and identifies a plurality of utterances and the time point at which each of the plurality of utterances was made; a goal extracting unit for extracting, as goals, a plurality of utterances including words and phrases described in materials used when the conversation was conducted, from the plurality of specified utterances; and the plurality of extracted goals. For each of the above, from among the specified plurality of statements, a statement positioned between the goal and another goal following the goal and within a predetermined range from the other goal A result extracting unit that extracts the result of the goal, and for each of the extracted goals, the time between the time when the goal is said and the time when another goal following the goal is said a required time specifying unit for specifying the time required for the goal; and for each of the plurality of extracted goals, a position between the goal and another goal following the goal among the plurality of specified utterances. For each of the multiple utterances, determine the degree of relevance that indicates the degree to which words and phrases constituting the goal or the results extracted for the goal are included, and utterances of a predetermined number or less in descending order of the determined degree of relevance are considered to be related to the goal. a related utterance extraction unit for extracting utterances; and for each of the plurality of goals extracted, proficiency in the foreign language suitable for the goal based on the degree of difficulty of words and phrases constituting the related utterances extracted for the goal. a proficiency level determination unit that determines the degree of proficiency, and between the proficiency level determined for the goal from among the extracted goals and the proficiency level of the foreign language learner, and the level determined for the goal a goal selection unit that selects a plurality of goals that satisfy a predetermined condition between the proficiency level of the foreign language learning supporter and the proficiency level of the foreign language learning supporter; a goal grouping unit that performs grouping so that the total required time does not exceed a predetermined learning time; and a teaching material generating unit that generates a file indicating the required time for language learning as a teaching material.

In a preferred embodiment, the language learning material generation system performs, for each of the plurality of goals extracted, between the point in time when the goal is stated and the point in time when another goal subsequent to the goal is stated. a body movement identifying unit that identifies a predetermined body movement by analyzing a video recording of the scene of the conversation; and a supplementary item specifying unit that specifies supplementary items indicating points to note regarding the learner's body movement or learning environment, wherein the teaching material generation unit includes, for each of the plurality of grouped groups, A file indicating a plurality of goals that constitute the group, the required time for each of the plurality of goals, and supplementary items is generated as a teaching material.

According to the present invention, it is possible to generate a language learning material that reflects a situation in which a learner actually uses a foreign language.

Block diagram showing an example of functional configuration related to discourse community analysis Diagram showing an example of notation data Diagram showing an example of material data Diagram showing an example of goal/result list data Diagram showing an example of related message list data A diagram showing an example of a judgment condition table A diagram showing an example of body motion list data Diagram showing an example of word list data Block diagram showing an example of a functional configuration related to the generation of English conversation teaching materials A diagram showing an example of difficulty determination table data Diagram showing an example of proficiency level determination table data Diagram showing an example of goal level data Diagram showing an example of lesson information table data A diagram showing an example of teaching material pattern data Diagram showing an example of supplementary item determination table data A diagram showing an example of a teaching material file Flow diagram showing an example of discourse community analysis processing Flow diagram showing an example of goal/result list data generation processing Flow diagram showing an example of related utterance extraction processing Flow diagram showing an example of English conversation teaching material generation processing Flow diagram showing an example of proficiency level determination processing Flow chart showing an example of teaching material pattern identification processing Flow diagram showing an example of teaching material file generation processing

1. Embodiment A language learning material generation system 1 according to an embodiment of the present invention will be described with reference to the drawings. The language teaching material generation system 1 described below is a system for particularly generating English conversation teaching materials, and is composed of one or a plurality of servers.
1-1. Configuration When creating English conversation teaching materials using the language learning teaching material generation system 1, an ethnographic survey is conducted on the environment in which English conversation, which is the subject of learning, is actually conducted. For example, if a learner wants to learn English conversation in the business of a particular company, he or she conducts an ethnographic survey on the people who actually engage in the business. This ethnographic study records the audio and video of people who actually do the work. The language teaching material generation system 1 performs discourse community analysis based on the recorded voice and video. Here, discourse is the exchange of words between two speakers, and in linguistics it is translated as discourse. A discourse community sees this exchange as a collection of discourses made up of speakers with a common purpose (this is called a community). Therefore, discourse community analysis analyzes the discourse exchanged between speakers with multiple specific purposes, and the process of how the community achieves the purpose through discourse. Analyze. Specifically, by extracting "goals and results,""contents of remarks that contributed to goal achievement,""body movements and environmental elements that contributed to goal achievement," and "resources," the cultural characteristics of the environment can be identified. make clear. Then, English conversation teaching materials are generated based on the results of the discourse community analysis. The generated English conversation teaching materials are mainly used for lessons for learners at facilities for learning English conversation.
The discourse community analysis and the generation of English conversation teaching materials executed by the language teaching material generation system 1 will be described below.

FIG. 1 is a block diagram showing an example of a functional configuration related to discourse community analysis. Among the functions shown in the figure, a voice storage unit 2, a notation storage unit 4, a material storage unit 5, a goal/result list storage unit 7, a related utterance list storage unit 9, a video storage unit 10, and a body movement list storage unit 12 And the word list storage unit 14 is implemented by a memory. On the other hand, the speech analysis unit 3, the goal/result list generation unit 6, the related statement list generation unit 8, the body movement list generation unit 11, and the word list generation unit 13 are realized by the processor executing programs stored in the memory. be done. This program is a program that can be distributed via a network such as the Internet or a non-temporary recording medium. Each function will be described below.

The voice storage unit 2 stores voice data of actually conducted English conversations, which are English conversations to be learned. In the following description of this embodiment, English conversation during a meeting is assumed as the English conversation to be learned. The voice data stored in the voice storage unit 2 is generated for each participant in the conference using a pin microphone and stored in association with the participant's identification information. Alternatively, the voices of all conference participants are collected with one microphone, and voiceprint analysis is used to generate voice data for each participant.

The speech analysis unit 3 analyzes the speech data stored in the speech storage unit 2, specifies the content of the speech during the conference, and associates the specified content of the speech with the speaker and the time of the speech, thereby notating the score. Generate data. FIG. 2 is a diagram showing an example of this notation data. The musical notation data shown in the figure represents, for example, Mr. A's remark "there should be some pros and cons about the change of the logo design" at a point in time one minute and two seconds after the start of the meeting. showing. The notation data generated by the voice analysis unit 3 is stored in the notation storage unit 4 . Note that the musical notation data stored in the musical notation storage unit 4 may be manually corrected when the accuracy of voice recognition is low.

The material storage unit 5 stores data of materials used in the conference. In the following description of the present embodiment, it is assumed that the material used in the above conference is a material showing an agenda (in other words, the order of progress of proceedings). FIG. 3 is a diagram showing an example of this material data. The material data shown in the figure indicates, for example, that the first agenda item of the meeting (1) is self-introduction, and the required time is scheduled to be 5 minutes.

The goal/result list generation unit 6 identifies and associates goals, results, and required times based on the notation data stored in the notation storage unit 4 and the material data stored in the material storage unit 5, Generate goal/result list data. Here, the goal is the purpose of the discussion held during the meeting, for example, "I want to decide on the design of the logo." The result is the result (in other words, the conclusion) of discussing the goal, for example, the result of "decided on design C". Duration is the time spent discussing the goal.

In specifying goals among these three elements, the goal/result list generator 6 extracts, from among the utterances indicated by the notation data, utterances containing words and phrases constituting the proceedings indicated by the material data. For example, assuming the material data shown in FIG. 3 as the material data, utterances containing words such as "logo" and "recruitment" are extracted. In that case, you can ignore the verb to be and the article. In addition, the goal/result list generator 6 also extracts utterances including predetermined words and phrases indicating requests, such as "please" and "may I ask you," among utterances indicated by musical notation data.

Next, in specifying the results of the extracted goals, the goal/result list generation unit 6 selects the goal whose result is to be specified among the plurality of utterances indicated by the notation data as the following goal. Extract the utterance immediately before switching. More specifically, the utterances positioned between the goal whose result is to be specified and the next goal following it, and within a predetermined range from the next goal are extracted. . Here, the statement located within a predetermined range from the next goal is specifically the statement immediately before the next goal. If the statement extracted by the goal/result list generator 6 is inappropriate, it may be corrected manually.

Next, in specifying the required time of the extracted goal, the goal/result list generator 6 specifies the time until the goal whose required time is to be specified switches to the next subsequent goal. More specifically, the time between when the goal for which the required time is to be specified is stated and when the next subsequent goal is stated is specified. For example, if the goal for which the required time is to be specified is stated 5 minutes after the start of the meeting, and the next goal is stated 10 minutes after the start of the meeting, the required time will be 5 minutes. is identified.

FIG. 4 is a diagram showing an example of goal/result list data generated by the goal/result list generation unit 6. As shown in FIG. The goal/result list data shown in FIG. indicates that The goal/result list data generated by the goal/result list generation unit 6 is stored in the goal/result list storage unit 7 . The goal/result list data stored in the goal/result list storage unit 7 may be corrected manually if the accuracy of data extraction is low.

Next, the related statement list generation unit 8 generates statements that contributed to the achievement of each goal indicated by the goal/result list data stored in the goal/result list storage unit 7 (hereinafter referred to as "related statements"). The utterances indicated by the musical notation data stored in the musical notation storage unit 4 are extracted, and list data of the extracted related utterances is generated. When extracting related statements for each goal, first, for each statement located between the goal to be processed and the next goal that follows it, the degree of usage indicating the degree of relevance to the goal to be processed is calculated. . This usage degree is a value that indicates the degree to which an utterance includes words and phrases forming the goal to be processed or the result of the goal. For example, the goal to be processed and its result are ``Change the logo design'' and ``Design C fixed'', respectively, and a statement consisting of 14 words ``there should be some pros and cons about the change of the logo design''. Assuming the case of calculating the usage degree for, this statement includes three words "change", "logo", and "design" included in the goal or result, so the usage degree is "3/14" becomes. Note that the verb to be and the article may be ignored when calculating the Usage degree.

Once the Usage degree is calculated for each statement, a predetermined number of statements are extracted as related statements in descending order of the Usage degree. The number of related utterances extracted here shall be 30% or less of the total number of utterances located between the goal to be processed and the next goal following it. Note that this numerical value of "30%" is merely an example, and may be changed as appropriate according to the accuracy desired by the user of this system.

FIG. 5 is a diagram showing an example of related statement list data generated by the related statement list generation unit 8. As shown in FIG. The related remarks list data shown in FIG. be some pros and cons about the change of the logo design”. The related statement list data generated by the related statement list generation section 8 is stored in the related statement list storage section 9 .

Next, the moving image storage unit 10 stores moving image data in which scenes of the conference are recorded. The moving image data stored in the moving image storage unit 10 is generated by a camera installed so as to photograph all the participants of the conference.

The body motion list generation unit 11 generates body motions (hereinafter simply referred to as “body motions”) that contribute to the achievement of each goal indicated by the goal/result list data stored in the goal/result list storage unit 7 and the motions thereof. A person is identified by analyzing the moving image data stored in the moving image storage unit 10, and list data of the identified body movements is generated. In order to identify the body movements for each goal, the video data is analyzed for the prescribed body movements between the time when the goal to be processed is said and the time when the next goal following this is said. to specify. Here, the predetermined body motion is a body motion that satisfies a predetermined determination condition. FIG. 6 is a diagram showing an example of a determination condition table that defines this predetermined determination condition. For example, the determination condition table shown in FIG. 10 defines a determination condition that "the finger holding the pen was moved to the position where it touches the memo paper" for the body action of "taking notes". Machine learning (more preferably deep learning) is used to specify the body motion defined in this determination condition table. Specifically, for example, Language model-based approach, Sequence learning-based approach or LSTM (Long short-term memory)-E may be used (https://www.microsoft.com/en-us/research/ wp-content/uploads/2017/10/Tao-Mei-Intelligent-Video-Analysis-ACMMM-2017-Pub.pdf). Alternatively, use a model called LRCN (Long-term Recurrent Convolutional Networks), which combines CNN (Convolutional neural network) and RNN (recurrent neural network), or identify body movements by detecting optical flow. (See https://www.slideshare.net/YoshitakaUshiku/ss-123687543). The body motion list generator 11 uses face recognition technology to identify the performer of the body motion.

FIG. 7 is a diagram showing an example of body action list data generated by the body action list generation unit 11. As shown in FIG. In the body motion list data shown in the same figure, for example, the first goal of meeting (1) is "Change the logo design", the result is "Design C fixed", and one of the body motions is "Prepare materials". "Check and read" is indicated. The body action list data generated by the body action list generation section 11 is stored in the body action list storage section 12 .

Next, the word list generation unit 13 classifies each word indicated by the notation data stored in the notation storage unit 4 by word origin and part of speech using a dictionary database (not shown), thereby generating word list data. Generate. FIG. 8 is a diagram showing an example of word list data generated by the word list generating section 13. As shown in FIG. The word list data generated by the word list generation unit 13 is stored in the word list storage unit 14. FIG. Technical terms that are not listed in general dictionaries and words that are listed in general dictionaries but are used with special meanings (contextual) may be classified manually.

The functions described above generate goal/result list data, related statement list data, body action list data, and word list data used to generate English conversation teaching materials. These data sets are generated for each of a plurality of meetings in which English conversation to be learned is held.

Next, generation of English conversation teaching materials will be described. FIG. 9 is a block diagram showing an example of a functional configuration relating to generation of English conversation teaching materials. Among the functions shown in the figure, related statement list group storage unit 15, difficulty level determination table storage unit 16, proficiency level determination table storage unit 17, goal level storage unit 19, goal/result list group storage unit 20, lesson information table The storage unit 21, the teaching material pattern storage unit 23, the body action list group storage unit 24, the supplementary item determination table storage unit 25, the word list group storage unit 26, and the teaching material storage unit 28 are implemented by memories. On the other hand, the goal level determination unit 18, the teaching material pattern generation unit 22, and the teaching material generation unit 27 are implemented by a processor executing a program stored in the memory. This program is a program that can be distributed via a network such as the Internet or a non-temporary recording medium. Each function will be described below.

The related comment list group storage unit 15 stores related comment list data generated for each of a plurality of conferences.

The difficulty level determination table storage unit 16 is a table that associates words with their difficulty levels, and stores data of the table that is referenced to determine the difficulty levels of words. FIG. 10 is a diagram showing an example of the difficulty determination table data stored in the difficulty determination table storage unit 16. As shown in FIG.

The proficiency level determination table storage unit 17 is a table that associates the difficulty level of words with the proficiency level of English conversation, and stores table data that is referenced to determine the proficiency level of English conversation based on the difficulty level of words. Here, the level of CEFR (Common European framework of reference for languages) is used as the proficiency level of English conversation stored in this table. FIG. 11 is a diagram showing an example of proficiency level determination table data stored in the proficiency level determination table storage unit 17. As shown in FIG.

The goal level determination unit 18 generates goal level data by determining the proficiency level of English conversation suitable for each goal indicated by the related statement list data group stored in the related statement list group storage unit 15 . In determining the proficiency level for each goal, the proficiency level is determined based on the related utterances extracted for the goal to be processed. Specifically, the degree of difficulty of each word constituting the related statement extracted for the goal to be processed is determined by referring to the difficulty determination table stored in the difficulty determination table storage unit 16, and the determination is made. The proficiency level corresponding to the difficulty level is determined by referring to the proficiency level determination table stored in the proficiency level determination table storage unit 17 . Then, the highest proficiency level among the determined proficiency levels is specified as the proficiency level of the goal to be processed. The difficulty level of each word constituting the related statement extracted for the goal to be processed is determined by referring to the difficulty level determination table stored in the difficulty level determination table storage unit 16. The corresponding proficiency level is stored as the proficiency level of the relevant related utterance, and by repeating this for each related utterance, the average value or the mode of the stored proficiency levels of a plurality of related utterances is used as the goal to be processed. It may be specified as proficiency.

FIG. 12 is a diagram showing an example of goal level data generated by the goal level determination unit 18. As shown in FIG. The goal level data shown in the figure indicates, for example, that the proficiency level of English conversation suitable for the goal (1) of the meeting (1) is B1. The goal level data generated by the goal level determination section 18 is stored in the goal level storage section 19 .

Next, the goal/result list group storage unit 20 stores the goal/result list data generated for each of the multiple meetings.

The lesson information table storage unit 21 stores, for each class of an English conversation lesson, table data that associates the English conversation proficiency levels of learners and learning supporters who are participants with the teaching material patterns of the class. Here, the CEFR level is used as the proficiency level of English conversation stored in this table. The CEFR levels of learners and learning supporters are determined in advance by testing. FIG. 13 shows an example of lesson information table data stored in the lesson information table storage unit 21. As shown in FIG. For example, the lesson information table data shown in FIG. It indicates that it is "P1".

The teaching material pattern generation unit 22 generates teaching material pattern data suitable for each class indicated by the lesson information table data stored in the lesson information table storage unit 21 . In generating teaching material patterns for each class, first, a goal suitable for the class to be processed is selected. Specifically, with reference to the goal level data and the lesson information table data stored in the goal level storage unit 19, a predetermined condition is established between the learner participating in the class to be processed and the proficiency level of the learning supporter. Select a goal that has a proficiency that satisfies Here, the predetermined conditions are the first condition that the proficiency level of the goal is equal to or one step higher than the proficiency level of the learner, and the second condition that the proficiency level of the goal does not exceed the proficiency level of any learning supporter. are both conditions. For example, assuming the class G1 shown in FIG. 13 as the class to be processed, and the goal level data shown in FIG. 12 as the goal level data, the goal (2 ) and goals (1) and (5) of meeting (2) will be selected.

After selecting goals suitable for the class to be processed, the selected goals are grouped into lessons. Specifically, the goal/result list data group (particularly required time) stored in the goal level storage unit 19 is referred to, and the selected goal is set to the goal whose total required time of each group is one time. Teaching material pattern data is generated by grouping so as not to exceed the lesson time (for example, 50 minutes) and assigning a teaching material ID to each group. FIG. 14 is a diagram showing an example of this teaching material pattern data. In the teaching material pattern data shown in the figure, for example, the teaching material pattern P1 is composed of teaching materials A1-1, etc., and the teaching materials A1-1 correspond to the goals (2), (3), (6) and (8) of the meeting (1). ). The required times for these goals (2), (3), (6) and (8) are 5 minutes, 10 minutes, 20 minutes and 10 minutes respectively, and the total required time of 45 minutes is 50 minutes. It fits in the lesson time of The teaching material pattern data generated by the teaching material pattern generation unit 22 is stored in the teaching material pattern storage unit 23, and is associated with the class to be processed in the lesson information table data.

Next, the body action list group storage unit 24 stores the body action list data generated for each of the multiple meetings.

The supplementary item determination table storage unit 25 is a table that associates body motions with supplementary items that indicate points to note regarding the learner's body motions or the learning environment, and is referred to in order to determine supplementary items based on the body motions. store data for tables that FIG. 15 is a diagram showing an example of supplementary matter judgment table data stored in the supplementary matter judgment table storage unit 25. As shown in FIG.

The word list group storage unit 26 stores word list data generated for each of a plurality of meetings.

The educational material generation unit 27 generates a file of each educational material indicated by the educational material pattern data stored in the educational material pattern storage unit 23 . When generating each educational material file, first, the goal/result list data group stored in the goal/result list group storage unit 20 is referred to, and each goal constituting the educational material to be processed and the time required for the goal are specified. . Once the goal and required time are specified, next, for each goal constituting the teaching material to be processed, the body motion list data stored in the body motion list group storage unit 24 is referred to specify the body motion, and the specified body The supplementary item corresponding to the action is specified by referring to the supplementary item determination table data stored in the supplementary item determination table storage unit 25 . Once the supplementary items have been specified, next, the word list data stored in the word list group storage unit 26 is referred to for each goal that constitutes the teaching material to be processed, and technical terms and contextual are specified. Once technical terms and contextuals are specified, finally, a teaching material file is generated in which the specified goals, required time, supplementary items, technical terms and contextuals are described. FIG. 16 is a diagram showing an example of this educational material file. The teaching material file generated by the teaching material generating unit 27 is stored in the teaching material storage unit 28 .

By the functions described above, the teaching material pattern data indicating the curriculum of the English conversation lesson and each teaching material file used in the English conversation lesson are generated. Using these generated data, an English conversation lesson is given to the learner.

1-2. Operation Next, the flow of processing executed by the language learning material generation system 1 described above will be described. Specifically, the flow of discourse community analysis processing and English conversation teaching material generation processing will be described.

FIG. 17 is a flow diagram showing an example of discourse community analysis processing. When the discourse community analysis process shown in FIG. 1 is performed, the voice storage unit 2 stores voice data of an English conversation during a meeting to be learned. The material storage unit 5 also stores data of materials indicating the agenda used in the conference. Further, the moving image storage unit 10 stores moving image data in which scenes of the conference are recorded.

In this situation, the voice analysis unit 3 analyzes the voice data stored in the voice storage unit 2, specifies the content of the speech during the conference, and associates the specified content of the speech with the speaker and the time of the speech. notation data is generated (step Sa1). The generated musical notation data (see FIG. 2) is stored in the musical notation storage unit 4 .

Next, the goal/result list generation unit 6 identifies and associates goals, results, and required times based on the notation data stored in the notation storage unit 4 and the material data stored in the material storage unit 5. Thus, goal/result list data is generated (step Sa2). FIG. 18 is a flowchart showing an example of processing for generating this goal/result list data.

In the process shown in the figure, the goal/result list generation unit 6 first generates, from among the utterances indicated by the musical notation data, utterances including words and phrases constituting proceedings indicated by the material data and predetermined words and phrases indicating requests. Extract the utterance containing the goal as a goal (step Sa21).

Next, for each goal extracted in step Sa21, the goal/result list generator 6 converts the utterances indicated by the musical notation data immediately before the goal switches to the following goal as the result of the goal. (step Sa22). More specifically, the utterances located between the goal and the next goal following it and immediately before the next goal are extracted as the result of the goal.

Next, for each goal extracted in step Sa21, the goal/result list generation unit 6 specifies the time required for the goal to switch to the next succeeding goal as the time required for the goal (step Sa23). More specifically, the time between when the goal is stated and when the next goal is stated is specified as the time required for the goal.

Goal/result list data is generated through steps Sa21 to Sa23. The generated goal/result list data (see FIG. 4) is stored in the goal/result list storage unit 7. FIG.

After the goal/result list data is generated, the related statement list generation unit 8 next creates the notation storage unit 4 for each goal indicated by the goal/result list data stored in the goal/result list storage unit 7. Related utterance list data is generated by extracting related utterances from the utterances indicated by the notation data stored in (step Sa3). FIG. 19 is a flow diagram showing an example of processing for extracting related statements about a goal.

In the process shown in the figure, the related statement list generator 8 first calculates the usage degree of each statement positioned between the goal to be processed and the next goal following it (step Sa31).

Next, the related statement list generator 8 extracts a predetermined number of statements as related statements in descending order of the degree of usage calculated in step Sa31 (step Sa32). The number of related utterances extracted here shall be 30% or less of the total number of utterances located between the goal to be processed and the next goal following it.

Related statement list data is generated by performing steps Sa31 and Sa32 for each goal. The generated related statement list data (see FIG. 5) is stored in the related statement list storage unit 9. FIG.

After the related statement list data is generated, the body action list generation unit 11 stores each goal indicated by the goal/result list data stored in the goal/result list storage unit 7 in the video storage unit 10. The moving image data obtained is analyzed to specify a predetermined body motion and its performer, and list data of the specified body motion is generated (step Sa4). In identifying the body motion and its performer for each goal, a predetermined body motion performed between the time when the goal to be processed was said and the time when the next goal following this was said The operator is identified by analyzing the video data. The body action list data (see FIG. 7) generated by the body action list generation section 11 is stored in the body action list storage section 12 .

Next, the word list generation unit 13 classifies each word indicated by the notation data stored in the notation storage unit 4 by word origin and part of speech using a dictionary database (not shown), thereby generating word list data. Generate (step Sa5). The generated word list data (see FIG. 8) is stored in the word list storage unit 14. FIG.

As a result of the discourse community analysis processing described above, goal/result list data, related statement list data, body action list data, and word list data used to generate English conversation teaching materials are generated. This discourse community analysis process is executed for each of a plurality of meetings in which English conversation to be learned is held.

Next, FIG. 20 is a flowchart showing an example of the English conversation teaching material generation process. When the English conversation teaching material generation process shown in FIG. 1 is performed, the related statement list group storage unit 15 stores related statement list data generated for each of a plurality of conferences. Also, the goal/result list group storage unit 20 stores goal/result list data generated for each of a plurality of meetings. The body action list group storage unit 24 also stores body action list data generated for each of a plurality of meetings. The word list group storage unit 26 also stores word list data generated for each of a plurality of meetings.

In this situation, the goal level determination unit 18 determines the proficiency level of English conversation suitable for each goal indicated by the related statement list data group stored in the related statement list group storage unit 15, thereby determining the goal level. Data is generated (step Sb1). FIG. 21 is a flow chart showing an example of processing for determining a proficiency level of English conversation suitable for a goal.

In the process shown in the figure, the goal level determination unit 18 first determines the difficulty level of each word constituting the related statement extracted for the goal to be processed (step Sb11). The difficulty level determination table stored in the difficulty level determination table storage unit 16 is referred to when determining the difficulty level.

Next, the goal level determination unit 18 determines the proficiency level of English conversation corresponding to the difficulty level of each word determined in step Sb11 (step Sb12). In determining the proficiency level, the proficiency level determination table stored in the proficiency level determination table storage unit 17 is referred to.

Next, the goal level determination unit 18 identifies the highest proficiency level among the proficiency levels determined in step Sb12 as the proficiency level of the goal to be processed (step Sb13).

Goal level data is generated by performing the above steps Sb11 to Sb13 for each goal. The generated goal level data (see FIG. 12) is stored in the goal level storage unit 19. FIG.

After the goal level data is generated, the teaching material pattern generation unit 22 next generates teaching material pattern data suitable for each class indicated by the lesson information table data stored in the lesson information table storage unit 21 (step Sb2). FIG. 22 is a flow diagram showing an example of processing for identifying a teaching material pattern suitable for a class.

In the process shown in the figure, the teaching material pattern generator 22 first selects a goal suitable for the class to be processed (step Sb21). Specifically, with reference to the goal level data and the lesson information table data stored in the goal level storage unit 19, a predetermined condition is established between the learner participating in the class to be processed and the proficiency level of the learning supporter. Select a goal that has a proficiency that satisfies Here, the predetermined conditions are the first condition that the proficiency level of the goal is equal to or one step higher than the proficiency level of the learner, and the second condition that the proficiency level of the goal does not exceed the proficiency level of any learning supporter. are both conditions.

Next, the teaching material pattern generation unit 22 groups the goals selected in step Sb21 into lesson units (step Sb22). Specifically, the goal/result list data group (particularly required time) stored in the goal level storage unit 19 is referred to, and the selected goal is set to the goal whose total required time of each group is one time. Grouping is performed so as not to exceed the lesson time, and a teaching material ID is given to each group.

Teaching material pattern data is generated by performing the above steps Sb21 and Sb22 for each class. The generated teaching material pattern data (see FIG. 14) is stored in the teaching material pattern storage unit 23, and is associated with the class to be processed in the lesson information table data.

After the teaching material pattern data is generated, the teaching material generating section 27 next generates a file of each teaching material indicated by the teaching material pattern data stored in the teaching material pattern storage section 23 (step Sb3). FIG. 23 is a flowchart showing an example of processing for generating a teaching material file.

In the processing shown in the figure, the teaching material generation unit 27 first refers to the goal/result list data group stored in the goal/result list group storage unit 20, and determines each goal and its requirements that make up the teaching material to be processed. Specify the time (step Sb31).

Next, the teaching material generation unit 27 refers to the body motion list data stored in the body motion list group storage unit 24 to specify a body motion for each goal that constitutes the teaching material to be processed. The corresponding supplementary item is specified by referring to the supplementary item determination table data stored in the supplementary item determination table storage unit 25 (step Sb32).

Next, the educational material generation unit 27 refers to the word list data stored in the word list group storage unit 26 for each goal that constitutes the educational material to be processed, and specifies technical terms and contextual terms (step Sb33).

Next, the teaching material generation unit 27 generates a teaching material file in which the goals and required time specified in step Sb31, the supplementary matters specified in step Sb32, and the technical terms and contextual specified in step Sb33 are described (step Sb34). . The generated educational material file (see FIG. 16) is stored in the educational material storage unit 28 .

Each educational material file is generated by performing steps Sb31 to Sb33 for each educational material.

As a result of the English conversation teaching material generation processing described above, teaching material pattern data indicating the curriculum of the English conversation lesson and each teaching material file used in the English conversation lesson are generated. Using these generated data, an English conversation lesson is given to the learner.

According to the language teaching material generation system 1 described above, teaching materials are automatically created through discourse community analysis. can be created. In addition, teaching materials can be created without the help of teaching material development specialists.

2. Modifications The above embodiments may be modified as follows. Two or more modified examples described below may be combined with each other.

2-1. Modification 1
In the above embodiment, the English conversation to be learned is assumed to be English conversation during a meeting, but the English conversation to be learned is not limited to this. For example, English conversation conducted in other situations such as answering the phone or interviewing may be the object of learning.

2-2. Modification 2
When generating musical notation data, the speech analysis unit 3 does not necessarily have to associate the speaker with the specified speech content.

2-3. Modification 3
The material data stored in the material storage unit 5 does not necessarily have to be data of materials used during the conference. For example, minutes data may be used as long as the material describes the order of proceedings.

2-4. Modification 4
When extracting goals, the goal/result list generation unit 6 includes utterances that include words and phrases that constitute the proceedings shown in the material data, and that include predetermined words and phrases that indicate a request, in order to increase the accuracy of goal extraction. You may make it extract an utterance. Further, when extracting goals, the goal/result list generator 6 may refer to a synonym database to extract statements containing synonyms of the words forming the proceedings shown in the material data.

2-5. Modification 5
When extracting the result of a goal, the goal/result list generating unit 6 extracts a predetermined number (2 or more) of statements before the next goal instead of extracting the utterance immediately before the next goal of the goal. You may make it extract an utterance. Then, for each extracted utterance, a Usage degree indicating the degree of relevance to the goal may be calculated, and the utterance with the highest Usage degree may be extracted as the result of the goal. Here, the degree of usage is a value that indicates the degree to which the statement includes words and phrases that constitute the goal.

2-6. Modification 6
The number of related comments extracted by the related comment list generation unit 8 may be uniformly fixed without changing according to the number of comments between goals. Further, when calculating the Usage level, the related statement list generation unit 8 may refer to the synonym database to calculate the degree of inclusion of not only the words and phrases constituting the goal and its result, but also synonyms of the words and phrases. good.

2-7. Modification 7
Other indicators than CEFR may be used as indicators of proficiency in English conversation.

2-8. Modification 8
When determining the proficiency level suitable for the goal, the goal level determination unit 18 determines the proficiency level in which the number of appearances or the appearance ratio of the word exceeds a predetermined threshold, instead of specifying the highest proficiency level as the proficiency level of the goal. , may be identified as the goal proficiency. Alternatively, the proficiency level with the highest word appearance count or appearance ratio may simply be specified as the goal proficiency level.

2-9. Modification 9
When determining the proficiency level suitable for the goal, the goal level determination unit 18 determines the proficiency level based on the difficulty level of the words that make up the related statement, instead of determining the proficiency level based on the speaking speed at which the related statement was made. You may make it determine a proficiency level according to. Here, the speaking speed when the related statement is made is the number of words that make up the related statement divided by the time required to make the related statement (in other words, the time difference between the start and end of the related statement). It is obtained by The goal level determination unit 18 stores in advance a proficiency level determination table that associates speaking speeds with proficiency levels, and refers to this proficiency level determination table to convert the obtained speaking speed into proficiency levels.

2-10. Modification 10
When determining the proficiency level suitable for a goal, the goal level determination unit 18 determines the proficiency level between the goal and the following goal instead of determining the proficiency level based on the difficulty of the words that make up the related remarks. The proficiency level may be determined based on the difficulty level of all utterances positioned at . When this modified example is adopted, the related comment list generator 8, the related comment list storage unit 9, and the related comment list group storage unit 15 become unnecessary.

2-11. Modification 11
The condition based on which the learning material pattern generation unit 22 selects a goal is not limited to the set of the first condition and the second condition. The teaching material pattern generation unit 22 may select a goal based on only one of the first condition and the second condition. Or, choose a goal based on another condition, such as the goal's proficiency being equal to the learner's proficiency, or the goal being said by a particular speaker. good too. Note that when a condition is set that the goal is stated by a specific speaker, the teaching material pattern generation unit 22 refers to the notation data and selects the goal stated by the specific speaker. do.

2-12. Modification 12
When generating the teaching material file, the teaching material generating unit 27 may refer to the word list data to specify words other than technical terms and contextual terms and describe them in the teaching material file.

2-13. Modification 13
In the above embodiment, English conversation is assumed as a learning object, but conversation in other languages may be possible as long as the language is different from the native language of the learner.

1... Language learning material generation system 2... Voice storage unit 3... Voice analysis unit 4... Notation storage unit 5... Material storage unit 6... Goal/result list generation unit 7... Goal/result list storage unit 8 Related statement list generation unit 9 Related statement list storage unit 10 Moving image storage unit 11 Body motion list generation unit 12 Body motion list storage unit 13 Word list generation unit 14 Word list storage Part 15 Related statement list group storage unit 16 Difficulty level determination table storage unit 17 Proficiency level determination table storage unit 18 Goal level determination unit 19 Goal level storage unit 20 Goal/result list group Storage unit 21 Lesson information table storage unit 22 Teaching material pattern generation unit 23 Teaching material pattern storage unit 24 Body action list group storage unit 25 Supplementary item determination table storage unit 26 Word list group storage unit , 27... Teaching material generation unit, 28... Teaching material storage unit

Claims (2)

a voice analysis unit that analyzes the voice of a conversation conducted using a foreign language and identifies a plurality of utterances and the point in time at which each of the plurality of utterances was made;
a goal extracting unit for extracting, as goals, a plurality of utterances including words and phrases described in materials used when the conversation was conducted, from among the plurality of specified utterances;
For each of the plurality of extracted goals, from among the plurality of identified utterances, a utterance positioned between the goal and another goal following the goal, and which is a utterance that is positioned between the goal and another goal following the goal, and a result extracting unit that extracts utterances located within the range as a result of the goal;
A required time specifying unit for specifying, for each of the extracted goals, the time between when the goal is said and when another goal following the goal is said as the time required for that goal. and,
For each of the extracted goals, for each of the plurality of utterances positioned between the goal and another goal following the goal, among the specified utterances, for the goal or for the goal a related statement extracting unit that determines a degree of relevance indicating the degree of inclusion of phrases constituting the extracted result, and extracts a predetermined number or less of statements as related statements of the goal in descending order of the determined degree of relevance;
a proficiency level determination unit that determines, for each of the plurality of goals extracted, the proficiency level of the foreign language suitable for the goal based on the level of difficulty of words and phrases that form related remarks extracted for the goal;
Between the proficiency level determined for the goal and the foreign language learner 's proficiency level from among the plurality of goals extracted, and learning support for the foreign language between the proficiency level determined for the goal and the foreign language learner's proficiency level a goal selection unit that selects a plurality of goals that satisfy a predetermined condition between the proficiency level of the user;
a goal grouping unit for grouping the plurality of selected goals so that the total required time of the goals constituting each group does not exceed a predetermined learning time;
A language learning material generation system, comprising: a teaching material generation unit that generates, as teaching materials, a file indicating, for each of the plurality of grouped groups, a plurality of goals constituting the group and a required time for each of the plurality of goals. For each of the plurality of extracted goals, a predetermined body action performed between the point at which the goal is stated and the point at which another goal subsequent to the goal is stated is recorded in the scene of the conversation. a body movement identification unit that analyzes and identifies the recorded video;
A supplementary matter identification unit that identifies, for each of the plurality of goals extracted, a supplementary matter that is pre-associated with the body movement identified for the goal and that indicates a cautionary note regarding the learner's body movement or learning environment. and
The teaching material generating unit generates, for each of the plurality of grouped groups, a file indicating, as teaching materials, a plurality of goals constituting the group, the required time for each of the plurality of goals, and supplementary matters. 2. A language learning material generation system according to claim 1.

Images