Method and apparatus of intelligent guidance in natural language

• US 4,736,296 A
• Filed: 12/24/1984
• Issued: 04/05/1988
• Est. Priority Date: 12/26/1983
• Status: Expired due to Fees

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Yoshiura, Hiroshi, Hiraoka, Ryosei, Nakanishi, Kunio, Katayama, Yasunori
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): Tbui, Kim Thanh
• Application Number:
  US06/685,993
• Time in Patent Office: 1,198 Days
• Field of Search: 364/419, 364/900 MS File, 364/200 MS File, 364/513
• US Class Current: 704/8
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/9017 using directory or table look-up use of a directory or look-up table in file systems G06F16/13
• CPC Class Codes: G06F 3/018 Input/output arrangements for oriental characters
• CPC Class Codes: G06F 3/0489 using dedicated keyboard keys or combinations thereof
• CPC Class Codes: G06F 3/16 Sound input; Sound output speech processing G10L
• CPC Class Codes: G10L 25/15 the extracted parameters being formant information
• CPC Class Codes: Y10S 706/902 Application using ai with detail of the ai system
• CPC Class Codes: Y10S 706/908 Electronic or computer, internal or network, circuit

Abstract

A method and apparatus of intelligent guidance in a natural language in which the user of an information terminal apparatus is only required to input the content of a job to be executed, in a natural language which is used in usual conversation, and not in the form of the so-called command train, and in which the terminal apparatus automatically analyzes and understands the inputted content, and displays a command train required to execute the job to the user, thus, even when the user not familiar with the manner of manipulating the information terminal manipulates the terminal, he can converse with the information terminal in the natural language.

Claims

1. An intelligent guidance system in a natural language having an information terminal comprising an input device, a processing device responsive to an input from said input device for performing a predetermined processing, and a display unit for displaying at least one of an identification of the processing operation and a processing result of said processing operation, wherein said processing device includes:

input/output control means for controlling input of a natural language input sentence from said input device and output of data to said display unit;

natural language analysis means for dividing said natural language input sentence composed of a character train into individual vocabularies, and for transforming each of the resultant individual vocabularies into an expression understandable by said processing device by analyzing the meaning of the input sentence;

rule base means having a rule base for storing beforehand a state of said information terminal before execution and after execution of various input sentences;

terminal state memory means for storing an operating state of said information terminal;

deduction means connected to receive an output from said natural language analysis means for outputting a goal state of said information terminal, as a command train, based on the contents of said rule base; and

plan memory means connected to said deduction means for storing said command train outputted from said deduction means.

2. A system according to claim 1, wherein said deduction means includes means for processing said goal state in accordance with a predetermined command, such that when a first intermediate goal does not coincide with a current state of said information terminal, the processing is changed to a second intermediate goal and when said second intermediate goal coincides with the current state of said information terminal, said deduction means outputs a command train to change said first and second intermediate goals.

3. A system according to claim 1, wherein said processing device further includes a history memory means for sequentially storing the state of said information terminal before a change thereof, when the content of said terminal state memory means changes by executing said command train, and wherein said plan memory stores said command train.

4. A system according to claim 3, wherein said deduction means includes:

goal generation means for producing the goal state of said information terminal from the meaning structure of the input sentence;

state update means for producing an intermediate goal state by adding to or depleting from said goal state an add formula and a precondition of said rule base; and

match means for comparing said intermediate goal state with the current state of said information terminal stored in said terminal state memory means, and for outputting the add formula used to change said goal state as said command train when said intermediate goal state coincides with said goal state, and for actuating said state update means when said intermediate goal state does not coincide with said goal state, by regarding said intermediate state as a new goal state.

5. A system according to claim 3, wherein

said rule base means stores a rule base having a precondition defining a condition required to execute said command train of said information terminal, a delete list describing information of said information terminal which disappears after the execution of said command train, and an add formula describing information which is added, and wherein

said deduction means includes means for retrieving a state from said delete list, when a request is received for deleting a state of said information terminal for the purpose of deleting the state.

6. A system according to claim 1, wherein said deduction means includes:

goal generation means for generating the goal state of said information terminal based on a meaning structure of the input sentence;

state update means for generating an intermediate goal state by adding to or deleting from said goal state an add formula and a precondition of said rule base; and

match means for comparing said intermediate goal state with the current state of said information terminal stored in said terminal state memory means, and for outputting the add formula used to change said goal state as said command train when said intermediate goal state coincides with said goal state, and for actuating said state update means when said intermediate goal state does not coincide with said goal state, by regarding said intermediate state as a new goal state.

7. A system according to claim 1, wherein

said rule base means stores a rule base having a precondition defining a condition required to execute said goal state, a delete list describing information of said information terminal which disappears after the execution of said goal state, and an add formula describing information which is added;

and wherein

said deduction means includes means for retrieving a state from said delete list, when a request is received for deleting at least part of a current state of said information terminal.

8. A system according to claim 1, wherein said natural language analysis means includes:

a morphemic analysis device for dividing said character train into individual vocabularies, said morphemic analysis device including an input character train memory for storing said character train;

a dictionary means for storing a plurality of keywords corresponding to said plurality of individual vocabularies and for storing related information;

a start position memory for storing a start position of said character train for cutting out a particular one of said plurality of vocabularies from said character train;

a character train comparison section for comparing contents of said dictionary memory with said character train from the start position of said character train;

a control section for controlling contents of said start position memory;

a character position memory for storing a position of said character train to be compared with the contents of said dictionary, based on an n-th character from said start point;

an upper limit memory for indicating a first address among the addresses in which characters from the first to (n-1)-the character in a keyword from the dictionary are equal to characters from the start position to the (n-1)-th character in said character train;

a lower limit memory indicating a final address; and

vocabulary cut out means for storing in said upper limit memory and said low limit memory a range of addresses of characters equal to an input character indicated by said character position memory and for cutting out a vocabulary.

9. A system according to claim 8, further including first means for detecting when the contents of said upper limit memory and said lower limit memory coincide with each other and for producing an output indicating that the current output of the vocabulary has finished, and second means responsive to the output of said first means for updating said start position memory.

10. A system according to claim 8 wherein with respect to a vocabulary indicated by the upper limit memory, when no corresponding next character is contained in said character position memory, said vocabulary is outputted.

11. A system according to claim 1 wherein an algorithm for retrieving vocabularies is based on a binary search method.

12. A method of intelligent guidance in a natural language for use in an intelligent guidance system having an information terminal including an input device, a processing device responsive to an input from said input device for performing a predetermined processing, and a display unit for displaying at least one of an identification of the processing operation and a processing result of said processing operation, wherein said processing device includes an input/output control device for controlling the input from said input device and the output to said display unit, and a natural language analysis means for dividing the natural language composed of a character train into a plurality of vocabularies, and by analyzing the characters and the meaning of the input sentence, for transforming said plurality of vocabularies into a machine expression understandable by said processing device, said method comprising the steps of:

(a) analyzing the inputted natural language composed of a character train;

(b) recognizing a goal state of said information terminal desired by an operator;

(c) obtaining an intermediate goal state by processing said goal state by a predetermined command;

(d) ascertaining that said intermediate goal state is equal to a current state of said information terminal; and

(e) changing said goal state to said intermediate goal state, and executing the predetermined command reversely.

13. A method according to claim 12, wherein said method further comprises a step of storing a state change of said information terminal, and storing the command executed at a given time so that this command and the state change are stored in a corresponding relationship with each other, and a history relating to said information terminal is outputted.

14. A method of dividing a natural language character train into unit words in an intelligent guidance system in a natural language, said intelligent guidance system having an information terminal including an input device, a processing device responsive to an input of a natural language character train forming an input sentence from said input device for performing a predetermined processing, and a display unit for displaying at least one of an identification of the processing operation and a processing result of said processing operation, said processing device including an input/output control device for controlling the input from said input device and an output to said display unit, and a natural langauge analysis means for dividing the natural language character train into a plurality of vocabularies, and by analyzing the meaning of the input sentence, for transforming the vocabularies into an internal expression understandable by said processing device, wherein the inputted natural language character train including an undefined word is character encoded, and the character encoded natural language character train is analyzed in a forward direction making reference to a unit word dictionary consisting of a key character train for a natural language unit word and its parts of speech and making reference to a conjunction matrix indicating the propriety of conjunction between natural language parts of speech, and the natural language character train is divided into unit words, said method comprising the steps of:

(a) preparing a reverse unit word dictionary in which the order of the key character train in said unit word dictionary is reversed, and a reverse conjuction matrix consisting of a transposed matrix of said conjunction matrix is provided;

(b) analyzing in a reverse direction an unanalyzed portion of said character encoded natural language character train making reference to said reverse unit word dictionary and said reverse conjunction matrix when said analysis in the forward direction becomes impossible; and

(c) extracting the undefined word in said natural language character train based on the result of the analysis in the forward direction and in the reverse direction.