Apparatus for recognizing address appearing on mail article

• US 5,050,218 A
• Filed: 02/19/1991
• Issued: 09/17/1991
• Est. Priority Date: 08/26/1986
• Status: Expired due to Fees

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Egami, Kazunari, Ikeda, Osamu
• Primary Examiner(s): Razavi, Michael
• Application Number:
  US07/657,612
• Time in Patent Office: 210 Days
• Field of Search: 382/1, 382/58, 382/9, 382/63, 382/40, 382/19, 382/48, 209/583, 209/584, 364/900, 364/419, 364/400
• US Class Current: 382/101
• CPC Class Codes: G06F 18/00 Pattern recognition
• CPC Class Codes: G06F 40/157 using dictionaries or tables
• CPC Class Codes: G06V 30/242 Division of the character sequences into groups prior to recognition; Selection of dictionaries

Abstract

An address recognition apparatus and method is disclosed in which keyword and/or postal code detection is initially performed in order to limit the number of address dictionaries needed for final address detection. According to one embodiment, detected keywords are associated with property codes which may be designated for city names, street names, etc. The address dictionary memory is or can be separated into a plurality of segments, particular ones of which are selected on the basis of the property code of a given keyword.

Claims

1. An apparatus for recognizing an address appearing on an article, comprising:

character reading means for reading an address description appearing on an article to produce character data;

word extracting means responsive to said character data for extracting words from said character data, to produce a string of words;

keyword detecting means for detecting a keyword contained in said string of words by comparing each word of said string of words outputted from said word extracting means with a plurality of predetermined keyword dictionaries, to produce keyword information, when said string of words contains at least one keyword;

word combining means, responsive to said keyword information, for combining remaining non-keyword words contained in said string to produce a plurality of word combinations which are to be subjected to address recognition;

an address dictionary memory for storing a plurality of address dictionaries; and

address comparing means for comparing selected ones of said address dictionaries stored in said address dictionary memory with each word combination of said plurality of word combinations produced by said word combining means.

2. An apparatus according to claim 1, wherein said address description includes a plurality of lines each containing at least one word;

said word extracting means extracting a string of words for each line, and wherein said keyword information includes a property code attached to a keyword detected by said keyword detecting means, and said word combining means being responsive to said property code in combining the words of each of said lines to produce said plurality of word combinations.

3. An apparatus for recognizing an address appearing on an article, comprising:

character reading means for reading an address description appearing on an article to produce character data;

word extracting means responsive to said character data for extracting words from said character data, to produce a string of words;

keyword detecting means for detecting a keyword contained in said string of words by comparing said string of words outputted from said word extracting means with a plurality of predetermined keyword dictionaries, to produce keyword information when said string of words contains at least one keyword;

word combining means responsive to said keyword information for combining words contained in said string to produce a word combination which is to be subjected to address recognition;

an address dictionary memory for storing a plurality of address dictionaries;

address comparing means for comparing selected ones of said address dictionaries stored in said address dictionary memory with said word combination produced by said word combining means;

wherein said address description includes a plurality of lines each containing at least one word;

said word extracting means extracting a string of words for each line, and wherein said keyword information includes a property code attached to a keyword detected by said keyword detecting means, and said word combining means being responsive to said property code in combining the words of each of said lines; and

wherein said property code designates that the associated keyword is associated with one of a P.O. Box number, a country name, a city name, a street or avenue name, a building name and a predetermined number sequence.

4. An address recognizing apparatus, comprising:

character read means for reading a string of characters contained in an address description;

word extracting means for extracting words from said string of characters by editing said string of characters to form words;

address dictionary memory means for storing word dictionaries, each containing a string of characters; and

means for comparing the extracted words delivered from said word extracting means with said word dictionaries read out from said address dictionary memory means to thereby recognize an address, wherein said apparatus further comprises keyword dictionary memory means for storing keyword dictionaries, keyword detecting means for detecting at least one keyword contained in said address description by comparing said words contained in said address description with said keyword dictionaries, and means for determining appropriate dictionary portions of said word dictionaries to be compared with non-keyword ones of said extracted words on the basis of the keyword detection result obtained by said keyword detecting means.

5. An address recognizing apparatus for recognizing an address description on a mail article, said address description including a string of characters constituting at least one word and a postal code, comprising:

means for reading said address description to deliver word data corresponding to said words and postal code data corresponding to said postal code;

dictionary storage means for storing word dictionaries;

means for storing linkage information indicating predetermined linkages between said word dictionaries stored in said dictionary storage means and postal codes;

means for reading out predetermined at least one word dictionary from said dictionary storage means selected on the basis of said linkage information associated with said postal code; and

means for comparing said word data with said at least one word dictionary.

6. An address recognizing apparatus for recognizing an address description on a mail article, said address description including a string of characters constituting at least one word and postal code, comprising:

means for reading said address description to produce word data corresponding to said words and code data corresponding to said postal code;

means for storing a plurality of address dictionaries and linkage information, said plurality of address dictionaries being classified into groups on the basis of postal codes and the number of characters in the words constituting said address dictionary, respectively, and said linkage information indicating connections between said groups and appropriate ones of said plurality of address dictionaries for comparison being determined on the basis of the postal code and the number of characters constituting said words;

means for reading out said appropriate ones of said address dictionaries from said storing means in response to said linkage information designated by the postal code and the number of characters constituting the words contained in said address description to deliver a read-out address dictionary; and

means for comparing said word data with said read-out address dictionary to recognize said address description.

7. A method of recognizing an address description comprising:

reading character data of an address description to produce a character string;

extracting words from said character string, to produce a word string;

comparing said words with at least one keyword dictionary, to detect at least one keyword in said word string;

combining non-keywords words in the vicinity of the keyword to produce non-keyword word combinations which are to be subjected to address recognition, said combining being conducted in dependence upon at least the type of keyword detected; and

comparing said word combination of said non-keyword word combinations with predetermined address dictionaries which are selected from a dictionary memory on a basis of at least said type of said keyword.

8. A method as claimed in claim 7, further comprising the step of determining the number of characters contained in each extracted word, and selecting said at least one keyword dictionary on the basis of said number of characters.

9. A method as claimed in claim 8, further including the step of selecting the number of keyword dictionaries to be used for keyword comparison on the basis of the length of the extracted word.

10. A method as claimed in claim 7, wherein said word combining step comprises the steps of extracting from a keyword dictionary memory a property code associated with each detected keyword, and limiting the number of adjacent words to be combined to form said non-keyword word combination on the basis of said property code.

11. A method as claimed in claim 10, wherein the address dictionaries used for examining each non-keyword word combination are selected on the basis of the content of the property code associated with the keyword of said non-keyword word combination.

12. A method as claimed in claim 7 further including the step of extracting a postal code from said word string, and restricting said address dictionaries on the basis of said postal code.