A MEMORY DEVICE INCLUDING AN ASSOCIATIVE MEMORY FOR THE STORAGE OF DATA BELONGING TO A PLURALITY OF CLASSES

• US 20020038295A1
• Filed: 02/12/1998
• Published: 03/28/2002
• Est. Priority Date: 02/12/1997
• Status: Active Grant

• Current Assignee: SGS-Thomson Microelectronics, S.R.L. (STMicroelectronics NV)
• Original Assignee: SGS-Thomson Microelectronics, S.R.L. (STMicroelectronics NV)
• Inventors: KRAMER, ALAN, GOZZINI, GIOVANNI, CANEGALLO, ROBERTO, CHINOSI, MAURO, ROLANDI, PIERLUIGI, NAVONI, LORIS
• Granted Patent: US 6,415,293 B1
• Time in Patent Office: Days
• US Class Current: 707/1
• CPC Class Codes: G06F 16/90344 by using string matching techniques
• CPC Class Codes: G11C 15/046 using non-volatile storage elements
• CPC Class Codes: Y10S 707/961 Associative
• CPC Class Codes: Y10S 707/99945 Object-oriented database structure processing

Abstract

A memory device includes an associative memory for the storage of data belonging to a plurality of classes. The associative memory comprises a plurality of memory locations aligned along rows and columns for the storage of data along the rows. Each memory row comprises a plurality of groups of memory locations, each storing a respective datum, wherein groups of memory locations adjacent along one and the same row store data belonging to different classes. Groups of memory locations adjacent in the direction of the columns and disposed on different rows store data belonging to one and the same class. Each class comprises data having a different maximum lengths. The device is particularly suitable for the storage of words belonging to a dictionary for automatic recognition of words in a written text.

Claims

1. A memory device including an associative memory for the storage of data belonging to a plurality of classes, said associative memory comprising a plurality of memory locations aligned along a first and a second direction for the storage of the data along lines of memory extending along said first direction, each line of memory of said associative memory comprising a plurality of groups of memory locations, each group of locations of a line storing a respective datum, wherein groups of memory locations adjacent in said first direction and belonging to one and the same line store data belonging to different classes, and groups of memory locations adjacent in said second direction and belonging to different lines store data belonging to one and the same class.

2. A device as claimed in claim 1 wherein each of said classes comprises data having a same maximum length and adjacent groups of locations of a line store data having different maximum lengths.

3. A device as claimed in claim 1 wherein said data comprise sequences of values codifying characters forming words of a dictionary.

4. A device as claimed in claim 1 wherein said lines comprise rows of memory and the data belonging to one and the same class are stored in one and the same group of columns.

7. A device as claimed in claim 1 wherein said associative memory comprises a flash memory for storing analog signals.

8. A device as claimed in claim 1 wherein said associative memory comprises:

a first group of sections for storing data comprising analog weights used for recognition of words; and

a second group of sections for storing data comprising analog weights used for recognition of characters, wherein said first and second sections are mutually exclusive.

6. A device as claimed in claim 6, comprising storage means for storing a correspondence between said classes of data and the addresses of said groups of memory locations.

5. A device as claimed in claim 6, comprising selectively enabling means for said groups of memory locations which are adjacent in said second direction.

9. A method comprising:

parsing input data to provide a portion delimited by predetermined characteristics;

determining a length of said portion;

comparing said length to a table of lengths of data stored in an associative analog memory;

providing said portion to a section of said associative analog memory containing data having lengths comparable to said portion;

identifying a closest match between a datum stored in said section and said portion; and

storing an address from said associative analog memory corresponding to said datum in a second memory.

10. A method as claimed in claim 9 wherein said step of storing an address from said associative analog memory comprises a step of storing a binary address in a digital memory, wherein said binary address is from said associative analog memory and corresponds to said datum.

11. A method as claimed in claim 9 wherein said step of parsing input data includes a step of parsing input data comprising a sequence of analog values to provide a portion delimited by predetermined characteristics.

12. A method as claimed in claim 9 wherein said step of parsing input data includes a step of parsing input data comprising sequences of analog values individually representing alphanumeric characters to provide a portion delimited by one or more elements chosen from a group consisting of:

spaces, commas, periods, question marks, exclamation points, parentheses, brackets, colons, semicolons and quotation marks.

13. A method as claimed in claim 9 wherein said step of identifying a closest match includes steps of:

calculating a Manhattan distance between said portion and each datum stored in said section;

determining a least Manhattan distance between said portion and at least one datum stored in said section; and

providing binary address data for said at least one datum; and

wherein;

said step of storing an address includes a step of storing said binary address from said associative analog memory corresponding to said datum in a digital memory.

14. A method as claimed in claim 13, further comprising steps of:

providing said binary address data to a priority code generation device to derive priority code data corresponding to said binary address data; and

storing one binary address of said binary address data in a digital memory based on said priority code data.

15. A method as claimed in claim 9 wherein said step of parsing input data includes a step of parsing input data comprising a sequence of analog values into either a first group representing alphanumeric data or a second group representing character recognition data.

16. A method as claimed in claim 9 wherein said step of identifying a closest match includes steps of:

supplying output signals from said associative analog memory to a winner-take-all circuit; and

determining a least Manhattan distance between said portion and at least one datum from said section.

18. A method as claimed in claim 17 wherein said step of parsing a dataset includes steps of:

parsing a first dataset into a first ensemble representing weights used for recognizing characters; and

parsing a second dataset into a second ensemble representing weights used for recognizing words.

19. A method as claimed in claim 18 wherein said step of writing each group of data includes steps of:

writing data comprising said first ensemble to a first area of said associative analog memory; and

writing data comprising said second ensemble to a second area of said associative analog memory, wherein said first and second areas are mutually exclusive.

17. A method comprising:

parsing a dataset comprising sequences of analog values representing words to determine a length associated with each word of said dataset;

collecting sequences of analog values representing a word from said dataset having comparable lengths into groups; and

writing each group of said groups of data to a separate section of an associative analog memory.