Information filtering method and apparatus for preferentially taking out information having a high necessity

• US 6,901,392 B1
• Filed: 02/09/2001
• Issued: 05/31/2005
• Est. Priority Date: 09/04/1995
• Status: Expired due to Term

• Current Assignee: Panasonic Intellectual Property Corporation of America (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Kindo, Toshiki
• Primary Examiner(s): Knight, Anthony
• Assistant Examiner(s): Holmes, Michael B.
• Application Number:
  US09/779,838
• Time in Patent Office: 1,572 Days
• Field of Search: 706/25
• US Class Current: 706/25
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

A user initially judges whether each of pieces of information input as learning information is necessary or unnecessary, matrix elements of an affirmative metric signal indicating the records of the necessary information and matrix elements of a negative metric signal indicating the records of the unnecessary information are calculated in a learning unit from a plurality of keywords attached to the necessary information and the unnecessary information. Thereafter, a plurality of keywords attached to each piece of information data input to be estimated are converted into a vector in a vector generating unit, and an affirmative score signal and a negative score signal are calculated from the vector and the affirmative and negative metric signals in a score calculating unit. A value of the affirmative score signal is increased when many of keywords attached to a corresponding piece of information data are attached to the necessary information, and a value of the negative score signal is increased when many of keywords attached to a corresponding piece of information data are attached to the unnecessary information. Thereafter, necessity of each piece of information data is calculated from the affirmative and negative score signals, and the pieces of information data are stored in an unread data storing unit in order of necessity. Accordingly, information having a high necessity for the user can be easily retrieved from a large volume of information.

Claims

1. A metric learning apparatus which reflects, into a metric signal, a relationship between a user'"'"'s response to provided information and a keyword attached to the information, the information including information data and one or more keywords made up of a character stream and attached to the information data, the apparatus comprising:

information providing means for providing the user with the information data;

inputting means for enabling the user to input a signal indicating that the provided information data is necessary or unnecessary;

vector converting means for converting a keyword group signal to a vector signal, the keyword group signal consisting of at least one keyword signal attached to each piece of information;

metric learning means for assigning to the metric signal a value to predict a user'"'"'s necessity degree to the keyword on the basis of both of the signal showing that each information data is necessary or unnecessary and the vector signal; and

metric memorizing means for memorizing the metric signal.

2. A metric learning apparatus according to claim 1, wherein the metric signal is calculated on the basis of whether or not the provided information data is necessary or unnecessary and the vector signal, the metric signal comprising both an affirmative signal based on information obtained when the signal inputted with the inputting means shows that the provided information data is necessary and a negative signal based on information obtained when the signal inputted with the inputting means shows that the provided information data is unnecessary.

3. A metric learning apparatus according to claim 2, wherein the affirmative metric signal is made up of an auto-correlation matrix of the vector signal in cases where the signal inputted by the inputting means shows that the provided information data is necessary, and the negative metric signal is made up of an auto-correlation matrix of the vector signal in cases where the signal inputted by the inputting means shows that the provided information data is unnecessary.

4. A metric learning apparatus according to claim 2, wherein each of the affirmative and negative metric signals is made up of a matrix of elements in which an element (i, j) is calculated on a first frequency of the information judged to be necessary, a second frequency of the information judged to be unnecessary, a third frequency of the information simultaneously including both of an i-th keyword signal and a j-th keyword signal, and judged to be necessary, and a fourth frequency of the information simultaneously including both of an i-th keyword signal and a j-th keyword signal and judged to be unnecessary.

5. A metric learning apparatus according to claim 4, wherein the (i, j) element of the matrix is made up of a signal quantitatively estimate a difference between a first probability distribution indicating that the information is necessary or unnecessary, and a second probability distribution indicating that the information simultaneously including both of the i-th keyword signal and the j-th keyword signal is necessary or unnecessary.

6. A metric learning apparatus according to claim 1, wherein the keyword signal includes a classification symbol.

7. A metric learning method which reflects, into a metric signal, a relationship between a user'"'"'s response to provided information and a keyword attached to the information, the information including information data and one or more keywords made up of a character stream and attached to the information data, the method comprising the steps of:

providing the user with the information data;

enabling the user to input a signal indicating that the provided information data is necessary or unnecessary;

converting a keyword group signal to a vector signal, the keyword group signal comprising at least one keyword signal attached to each piece of information;

assigning to the metric signal a value to predict a user'"'"'s necessity degree to the keyword on the basis of both of the signal showing that each information data is necessary or unnecessary and the vector signal; and

memorizing the metric signal.

8. A metric learning method according to claim 7, the metric signal comprises both of an affirmative metric signal including information obtained when the signal inputted in the inputting step shows that the provided information data is necessary, and a negative metric signal including information obtained when the signal inputted in the inputting step shows that the provided information data is unnecessary.

9. A metric learning method according to claim 8, wherein the affirmative metric signal is made up of an auto-correlation matrix of the vector signal in cases where the signal inputted in the inputting step shows that the provided information data is necessary, and the negative metric signal is made up of an auto-correlation matrix of the vector signal in cases where the signal inputted in the inputting step shows that the provided information data is unnecessary.

10. A metric learning method according to claim 8, wherein each of the affirmative and negative metric signals of a matrix of elements of which an element (i, j) is calculated on a first frequency of the information judged to be necessary, a second frequency of the information judged to be unnecessary, a third frequency of the information simultaneously including both of an i-th keyword signal and a j-th keyword signal and judged to be necessary, and a fourth frequency of the information, simultaneously including both of an i-th keyword signal and a j-th keyword signal, and judged to be unnecessary.

11. A metric learning method according to claim 10, wherein the (i, j) element of the matrix is made up of a signal quantitatively estimating a difference between a first probability distribution indicating that the information is necessary or unnecessary and a second probability distribution indicating that the information simultaneously including both of the i-th keyword signal and the j-th keyword signal is necessary or unnecessary.