System for generating natural language information from information expressed by concept and method therefor

• US 6,023,669 A
• Filed: 12/28/1995
• Issued: 02/08/2000
• Est. Priority Date: 12/28/1994
• Status: Expired due to Term

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Jeyachandran, Suresh, Suda, Aruna Rohra
• Primary Examiner(s): Cosimano, Edward R.
• Application Number:
  US08/580,311
• Time in Patent Office: 1,503 Days
• Field of Search: 395/751, 395/752, 395/753, 395/754, 395/755, 395/756, 395/757, 395/758, 395/759, 395/760, 395/1, 395/10, 395/12, 395/50, 395/902, 704/2, 704/4, 704/8, 704/9, 707/536, 706/45, 706/47
• US Class Current: 704/2
• CPC Class Codes: G06F 40/35 Discourse or dialogue representation
• CPC Class Codes: G06F 40/56 Natural language generation

Abstract

When information expressed by concept is input to a natural language processing system, the form of expressing the input concept is selected by referring to a knowledge base having the knowledge of languages, general knowledge, and the knowledge of specific areas. In addition, the information expressed by the concept is split into individual concepts, one natural language expression is selected from among one or more natural language expressions with respect to the split individual concepts, and the selected natural language expression is output. In this way, information for natural language appropriate for the situation is generated from the same concept.

Claims

1. A natural language processing system for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, comprising:

input means for inputting information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated from the author to the reader as a message embodying the concept;

a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains;

selecting means for selecting the expression form in a natural language of a message embodying the concept which is input from said input means by referring to the knowledge base, said selecting means determining the relationship between the author and the reader and determining the expression form of the message to reflect the relationship between the author and the reader by using data in said knowledge base, wherein said selecting means selects the expression form of the message from a plurality of possible expression forms embodying the concept in accordance with the determined relationship between the author and the reader;

modulation means for modulating the concept input from said input means on the basis of the expression form selected by said selecting means; and

conversion means for converting the concept modulated by said modulation means into the message embodying the concept in the expression form in the natural language selected by said selecting means.

2. A system according to claim 1, wherein said knowledge base has the knowledge of the rank relation between a speaker and a listener, and said selecting means selects the expression form in the natural language on the basis of the rank relation between the speaker and the listener.

3. A system according to claim 1, wherein said selecting means selects the tense of a verb.

4. A system according to claim 1, wherein said selecting means selects the mood of a verb.

5. A natural language processing system for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language from an author to a reader in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

input means for inputting information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains;

determination means for determining information to be added to a concept input from said input means by referring to the knowledge base in consideration of precedent concepts and for selecting a message embodying the concept with the added information, said determining means determining the expression form of the message preferred by the person intended to receive the message by using data in said knowledge base and by determining the information to be added to the input concept, wherein said determining means selects the expression form of the message from a plurality of possible expression forms embodying the concept and the information to be added to the concept in accordance with the determined, preferred expression form of the message;

modulation means for modulating the concept input from said input means by adding the information determined by said determination means to the concept; and

conversion means for converting the concept modulated by said modulation means into the message embodying the concept in a natural language.

6. A system according to claim 5, wherein said determination means determines a specific word and/or phrase corresponding to the input concept to be information to be added.

7. A system according to claim 5, wherein said determination means determines a specific concept to be mentioned in duplication from among input concepts to be information to be added.

8. A natural language processing system for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

input means for inputting information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains;

connecting means for connecting a plurality of concepts input from said input means and outputting the connected concepts as one concept in a message embodying the one concept by referring to the knowledge base, said connecting means determining the expression form of the message preferred by the person intended to receive the message by using data in said knowledge base and by connecting the plurality of concepts by adding a connection word or phrase, wherein said connecting means selects the expression form of the message embodying the one concept from a plurality of possible expression forms embodying the one concept in accordance with the determined, preferred expression form of the message; and

conversion means for converting the concept output from said connecting means into the message embodying the one concept in a natural language.

9. A natural language processing system for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

input means for inputting information expressed by a complex concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a knowledge base having knowledge of natural languages, general knowledge, and knowledge of specific domains;

splitting means for splitting the complex concept input from said input means into individual concepts while avoiding the repetition of common data by referring to the knowledge base;

selecting means for selecting one natural language expression of a message embodying the individual concepts split by said split means from one or more natural language expressions with respect to individual concepts split by said splitting means by referring to the knowledge base, said selecting means determining the expression form of a message preferred by the person intended to receive the message by using data in said knowledge base and the individual concepts split by said splitting means, wherein said selecting means selects the natural language expression form of the message from a plurality of possible expression forms embodying the individual concepts split by said splitting means in accordance with the determined, preferred expression form of the message; and

conversion means for converting the individual concepts split by said splitting means into the message embodying the individual concepts in a natural language expression selected by said selecting means.

10. A natural language processing system for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

input means for inputting information expressed by a list of concepts each comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains;

linguistic concept generation means for generating a linguistic concept dependent upon language to be added to individual concepts in the list of concepts input from said input means;

parameter determination means for determining a parameter for the linguistic concept generated by said linguistic concept generation means;

selecting means for determining the expression form of a message embodying the individual concepts preferred by the person intended to receive the message by using data in said knowledge base, the individual concepts and the determined parameter and selecting one natural language expression form of the message from a plurality of possible expression forms embodying the individual concepts in accordance with the determined, preferred expression form of the message; and

conversion means for converting the individual concepts into the message embodying the individual concepts in a natural language expression selected by said selecting means.

11. A system according to claim 10, wherein said parameter determination means includes gender determination means for determining the gender of a noun in a selected natural language expression selected by said selecting means.

12. A system according to claim 10, wherein said parameter determination means includes person determination means for determining the person mentioned in a selected natural language expression selected by said selecting means.

13. A system according to claim 10, wherein said selecting means selects a pronoun corresponding to duplicate nouns included in a selected natural language expression selected by said selecting means.

14. A system according to claim 10, wherein said selecting means selects an expression such that, when actors of a plurality of actions are mentioned more than once in a selected natural language expression selected by said selecting means, the description of one of the actors is omitted in the selected natural language expression selected by said selecting means.

15. A system according to claim 10, wherein said selecting means selects an interrogative sentence using an interrogative as an interrogative sentence.

16. A system according to claim 10, wherein said selecting means includes verb expression determination means for determining the form of expressing a verb.

17. A system according to claim 10, wherein said parameter determination means includes number determination means for determining the number of noun in a selected natural language expression selected by said selecting means.

18. A system according to claim 17, wherein said number determination means determines the number of noun by counting in the selected natural language expression selected by said selecting means.

19. A natural language processing method for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, comprising:

an input step of inputting information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated from the author to the reader as a message embodying the concept;

a selecting step of selecting and storing the expression form in a natural language of a message embodying the concept input from said inputting step by referring to a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains, said selecting step determining the relationship between the author and the reader and determining the expression form of the message to reflect the relationship between the author and the reader preferred by the reader by using data in the knowledge base, wherein said selecting step selects the expression form of the message from a plurality of possible expression forms embodying the concept in accordance with the determined relationship between the author and the reader; and

modulating the concept input by said input step on the basis of the expression form selected in said selecting step; and

a conversion step of converting the concept modulated by said modulating step into the message embodying the concept in the expression form in the natural language selected in said selecting step.

20. A method according to claim 19, wherein the knowledge of the rank relation between a speaker and a listener is prestored is the knowledge base, and, in said selecting step, the expression form in the natural language is selected on the basis of the rank relation between the speaker and the listener.

21. A method according to claim 19, wherein in said selecting step, the tense of a verb is selected.

22. A method according to claim 19, wherein in said selecting step, the mood of a verb is selected.

23. A natural language processing method for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

an input step of inputting information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a determining step of determining information to be added to the concept input in said input step by referring to a knowledge base in consideration of precedent concepts, the knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains and of selecting a message embodying the concept with the added information, said determining step determining the expression form of the message preferred by the person intended to receive the message by using data in the knowledge base and the information to be added to the concept, wherein said determining step selects the expression form of the message from a plurality of possible expression forms embodying the concept to which the determined information is to be added in accordance with the determined, preferred expression form of the message; and

a modulation step of modulating the concept input in said input step by adding the information determined in said determining step to the concept; and

converting the concept modulated in said modulation step into the message embodying the modulated concept in a natural language.

24. A method according to claim 23, wherein in said determining step, a specific word and/or phrase corresponding to the input concept is determined to be information to be added.

25. A method according to claim 23, wherein in said determining step, a specific concept to be mentioned in duplication from among the input concepts is determined to be information to be added.

26. A natural language processing method for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

an input step of inputting information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a connecting step of connecting a plurality of concepts input in said input step by adding a connection word or phrase and outputting the connected concepts as one concept in a message embodying the one concept by referring to a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains, said connecting step determining the expression form of the message preferred by the person intended to receive the message by using data in the knowledge base and the connected concepts, wherein said connecting step selects the expression form of the message embodying the one concept from a plurality of possible expression forms embodying the one concept in accordance with the determined, preferred expression form of the message; and

a conversion step of converting the one concept output in said connecting step into the message embodying the concept in a natural language.

27. A natural language processing method for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

an input step of inputting information expressed by a complex concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a splitting step of splitting the complex concept input in said input step into individual concepts while avoiding the repetition of common data by referring to a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains;

a selecting step of selecting one natural language expression of a message embodying the individual concepts split in said splitting step from one or more natural language expressions with respect to individual concepts split in said splitting step, said selecting step determining the expression form of a message preferred by the person intended to receive the message by using data in the knowledge base and the individual concepts split by said splitting step, wherein said selecting step selects the natural language expression form of the message from a plurality of possible expression forms embodying the individual concepts split by said splitting step in accordance with the determined, preferred expression form of the message; and

a conversion step of converting the individual concepts split by said splitting means into a message embodying the individual concepts in the natural language expression selected in said selecting step.

28. A natural language processing method for processing internal conceptual expressions of a computer not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, comprising:

an input step of inputting information expressed by a list of concepts each comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

a linguistic concept generating step for generating a linguistic concept dependent upon language to be added to individual concepts in the list of concepts input in said input step;

a parameter determining step of determining a parameter for the linguistic concept generating in said linguistic concept generating step;

a selecting step of determining the expression form of a message embodying the individual concepts preferred by the person intended to receive the message by using the individual concepts, the determined parameter, and data in a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains, said selecting step selecting one natural language expression form of the message from a plurality of possible expression forms embodying the individual concepts in accordance with the determined, preferred expression form of the message; and

a conversion step of converting the individual concepts into the message embodying the individual concepts in a natural language expression selected in said selecting step.

29. A method according to claim 28, wherein said parameter determining step includes a step of determining the gender of a noun in a selected natural language expression selected by said selecting step.

30. A method according to claim 28, wherein said parameter determining step includes a person determining step for determining the person mentioned in the selected natural language expression selected in said selecting step.

31. A method according to claim 28, wherein said selecting step selects a pronoun corresponding to duplicate nouns mentioned in the selected natural language expression selected in said selecting step.

32. A method according to claim 28, wherein said selecting step selects an expression such that, when actors of a plurality of actions are mentioned more than once in a selected natural language expression selected by said selecting means, the description of one of the actors is omitted in the selected natural language expression selected by said selecting step.

33. A method according to claim 28, wherein said selecting step selects an interrogative sentence using an interrogative as an interrogative sentence.

34. A method according to claim 28, wherein said selecting step includes a verb expression determining step of determining the form of expressing a verb.

35. A method according to claim 28, wherein said parameter determining step includes a number determining step of determining the number of nouns in a selected natural language expression selected by said selecting step.

36. A method according to claim 35, wherein in said number determining step the number of noun is determined by counting in the selected natural language expression selected by said selecting step.

37. A natural language processing system for outputting a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, the expression form of the message reflecting the rank relation between the author and the reader, comprising:

input generating means for inputting information expressed by a concept that is to be communicated to a person as a message embodying the concept;

a knowledge base having the knowledge of languages, general knowledge, the knowledge of specific domains, and the knowledge of the rank relation between the person intended to receive the message and the author of the message;

selecting means for determining the relationship between the author and the reader and for selecting the expression form of a message embodying the concept which is input from said input means by referring to the knowledge base, wherein said selecting means determines the expression form of the message to reflect the relationship between the author and the reader by using data of the rank relation between the person intended to receive the message and the author of the message stored in said knowledge base, wherein said selecting means selects the expression form of the message embodying the concept in a natural language from a plurality of possible expression forms embodying the concept in accordance with the determined relationship between the author and the reader reflecting the rank relation between the person intended to receive the message and the author of the message; and

output means for outputting the message having the expression form selected by said selecting means.

38. A natural language processing system for outputting a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, comprising:

input means for inputting information expressed by a concept that is to be communicated from an author to a reader to a person intended to receive as a message embodying the concept;

a knowledge base having the knowledge of languages, general knowledge, and the knowledge of specific domains including knowledge of the mood of verbs;

selecting means for selecting the expression form in a natural language of a message embodying the concept which is input from said input means by referring to the knowledge base, wherein said selecting means determines the relationship between the author and the reader and determines the expression form of the message to reflect the relationship between the author and the reader and determines the mood of the verb in the message preferred by the reader by using data stored in said knowledge base, wherein said selecting means selects the mood of the verb to be used in the message embodying the concept in a natural language from a plurality of possible expression forms embodying the concept in accordance with the determined, preferred mood of the verb; and

output means for outputting a message having the expression form selected by said selecting means.

39. A natural language processing system for outputting a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, comprising:

input means for inputting information expressed by a concept that is to be communicated from the author to the reader as a message embodying the concept;

a knowledge base having the knowledge of languages, general knowledge, and the knowledge of specific domains;

determination means for determining relationship between the author and the reader and determining the expression form of the message to reflect the relationship between the author and the reader by using data in said knowledge base and by determining information to be added to input information from said input means by referring to the knowledge base, wherein said determination means determines a specific concept to be mentioned in duplication from among input concepts to be information to be added, wherein said determination means selects the expression form of the message from a plurality of possible expression forms embodying the specific concept to be mentioned in duplication in accordance with the determined relationship between the author and the reader; and

output means for adding the information determined by said determination means to said input information and outputting a message embodying the information.

40. A natural language processing method for outputting a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, the expression form of the message reflecting the rank relation between the person intended to receive the message and the author of the message, comprising:

an input step of inputting information expressed by a concept that is to be communicated from the author to the reader as a message embodying the concept;

a selecting step of selecting and storing the expression form in a natural language of a message embodying the concept input from said inputting step by referring to a knowledge base having the knowledge of languages, general knowledge, the knowledge of specific domains, and the knowledge of the rank relation between the person intended to receive the message and the author of the message, wherein said selecting step determines the relationship between the author and the reader and determines the expression form of the message to reflect the relationship between the author and the reader by using data of the rank relation between the person intended to receive the message and the author of the message stored in the knowledge base, wherein said selecting step selects the expression form of the message embodying the concept in a natural language from a plurality of possible expression forms embodying the concept in accordance with the determined relationship between the author and the reader reflecting the rank relation between the person intended to receive the message and the author of the message; and

an output step of outputting the message having the expression form selected in said selecting step.

41. A natural language processing method for outputting a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, comprising:

an input step of inputting information expressed by a concept that is to be communicated from the author to the reader as a message embodying the concept;

a selecting step of selecting and storing the expression form in a natural language of a message embodying the concept input from said inputting step by referring to a knowledge base having the knowledge of languages, general knowledge, and the knowledge of specific domains including knowledge of the mood of verbs, wherein said selecting step determines the relationship between the author and the reader and determines the expression form of the message to reflect the relationship between the author and the reader and determines by referring to the knowledge base, wherein said selecting step selects the mood of the verb to be used in the message embodying the concept in a natural language from a plurality of possible expression forms embodying the concept in accordance with the determined, preferred mood of the verb; and

an output step of outputting the message having the expression form selected in said selecting step.

42. A natural language processing method for outputting a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, comprising:

an input step of inputting information expressed by a concept that is to be communicated from the author to the reader as a message embodying the concept;

a determining step of determining the relationship between the author and the reader and determining the expression form of the message to reflect the relationship between the author and the reader by using data in a knowledge base and by determining information to be added to the input information from said input step by referring to the knowledge base, the knowledge base having the knowledge of languages, general knowledge, and the knowledge of specific domains, wherein in said determining step a specific concept to be mentioned in duplication from among the input concepts is determined to be information to be added, wherein said determination step selects the expression form of the message from a plurality of possible expression forms embodying the specific concept to be mentioned in duplication in accordance with the determined, relationship between the author and the reader, embodying the information.

43. A computer usable medium having computer readable program code means embodied therein for causing a computer to process internal conceptual expressions thereof not in a natural language so that they are converted to a message in a natural language from an author to a reader in accordance with the relationship between the author and the reader, said computer readable program code means comprising:

first computer readable program code means for causing the computer to input information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated from the author to the reader as a message embodying the concept;

second computer readable program code means for causing the computer to select and store the expression form in a natural language of a message embodying the concept caused to be input by said first computer readable program code means by referring to a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains, said second computer readable program code means causing the computer to determine the relationship between the author and the reader and to determine the expression form of the message to reflect the relationship between the author and the reader by using data in the knowledge base, wherein said second computer readable program code means causes the computer to select the expression form of the message from a plurality of possible expression forms embodying the concept in accordance with the determined relationship between the author and the reader;

third computer readable program code means for causing the computer to modulate the concept caused to be inputted by said first computer readable program code means on the basis of the expression form caused to be selected by said second computer readable program code means; and

fourth computer readable program code means for causing the computer to convert the concept caused to be modulated by said third computer readable program code means into a message embodying the concept in the expression form in the natural language caused to be selected by said second computer readable program code means.

44. A computer usable medium having computer readable program code means embodied therein for causing a computer to process internal conceptual expressions thereof not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, said computer readable program code means comprising:

first computer readable program code means for causing the computer to input information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

second computer readable program code means for causing the computer to determine information to be added to a concept caused to be inputted by said first computer readable program code means by referring to a knowledge base in consideration of precedent concepts, the knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains, for causing the computer to select a message embodying the concept with the added information, for causing the computer to determine the expression form of the message preferred by the person intended to receive the message by using data in said knowledge base and by determining the information to be added to the input concept, and for causing the computer to select the expression form of the message from a plurality of possible expression forms embodying the concept and the information to be added to the concept in accordance with the determined, preferred expression form of the message;

third computer readable program code means for causing the computer to modulate the concept caused to be inputted by said first computer readable program code means by adding the information caused to be determined by said second computer readable program code means to the concept; and

fourth computer readable program code means for causing the computer to convert the concept caused to be modulated by said third computer readable program code means into a message embodying the modulated concept in a natural language.

45. A computer usable medium having computer readable program code means embodied therein for causing a computer to process internal conceptual expressions thereof not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, said computer readable program code means comprising:

first computer readable program code means for causing the computer to input information expressed by a concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

second computer readable program code means for causing the computer to connect a plurality of concepts caused to be inputted by said first computer readable program code means and to output the connected concepts as one concept in a message embodying the one concept by referring to a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains, said second computer readable program code means causing the computer to determine the expression form of the message preferred by the person intended to receive the message by using data in the knowledge base and by connecting the plurality of concepts by adding a connection word or phrase, wherein said second computer readable program code means causes the computer to select the expression form of the message embodying the one concept from a plurality of possible expression forms embodying the one concept in accordance with the determined, preferred expression form of the message; and

third computer readable program code means for causing the computer to convert the one concept caused to be outputted by said second computer readable program code means into a message embodying the concepts in a natural language.

46. A computer usable medium having computer readable program code means embodied therein for causing a computer to process internal conceptual expressions thereof not in a natural language so that they are converted to a message in a natural language in accordance with the expression form of the message determined to be preferred by a person intended to receive the message, said computer readable program code means comprising:

first computer readable program code means for causing the computer to input information expressed by a complex concept comprising an internal conceptual expression of a computer not in a natural language that is to be communicated to a person intended to receive a message embodying the concept;

second computer readable program code means for causing the computer to split the complex concept caused to be inputted by said first computer readable program code means into individual concepts while avoiding the repetition of common data, by referring to a knowledge base having the knowledge of natural languages, general knowledge, and the knowledge of specific domains;

third computer readable program code means for causing the computer to select one natural language expression form of a message embodying the individual concepts split by said splitting means from one or more natural language expressions with respect to the individual concepts caused to be split by said second computer readable program code means, said third computer readable program code means causing the computer to determine the expression form of a message preferred by the person intended to receive the message by using data in said knowledge base and the individual concepts caused to be split by said second computer readable program code means, wherein said third computer readable program code means causes the computer to select the natural language expression form of the message from a plurality of possible expression forms embodying the individual concepts caused to be split by said second computer readable program code means in accordance with the determined, preferred expression form of the message; and

fourth computer readable program code means for causing the computer to convert the individual concepts caused to be split by said second computer readable program code means into a message embodying the individual concepts in a natural language expression caused to be selected by said third computer readable program code means.