Information recognition apparatus for recognizing recognition object information

• US 5,995,664 A
• Filed: 06/23/1997
• Issued: 11/30/1999
• Est. Priority Date: 06/21/1996
• Status: Expired due to Fees

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Shimomura, Hideki
• Primary Examiner(s): Bella, Matthew C.
• Application Number:
  US08/880,769
• Time in Patent Office: 890 Days
• Field of Search: 382/224, 382/221, 382/226, 382/227, 382/228, 382/229, 382/230, 382/231, 382/185, 382/215, 382/155, 704/251, 704/231, 704/240, 704/155, 704/256, 707/532, 707/533
• US Class Current: 382/229
• CPC Class Codes: G06F 40/103 Formatting, i.e. changing of presentation of documents automatic justification G06F40/189; automatic line break hyphenation G06F40/191
• CPC Class Codes: G06F 40/284 Lexical analysis, e.g. tokenisation or collocates
• CPC Class Codes: G06F 40/295 Named entity recognition

Abstract

The invention provides an information recognition apparatus for recognition of an address or the like which can recognize recognition object information, which is inputted in the form which does not have punctuations or element designations, at a high speed and with a high degree of accuracy. An element word recognition unit detects element word candidates of each information element of recognition element information and likelihoods of the element word candidates. A record number acquisition unit retrieves a record storage unit to acquire, for each element word candidate detected by the element word recognition unit, a record number of a record including the element word candidate. A likelihood calculation unit calculates likelihoods of the records using corresponding likelihood counters. A result discrimination unit discriminates a record to be determined as a recognition result of the recognition object information based on count values of the likelihood counters, and a result extraction unit extracts a record to be determined as a recognition result from the record storage section based on a result of the discrimination of the result discrimination unit.

Claims

1. An information recognition apparatus for recognizing recognition object information in the form of a series of pieces of information which is composed of a plurality of information elements for each of which a predetermined number of element words each of which can make the information element are determined, comprising:

a word storage section in which all element words which can make the information elements are stored;

a rule storage section in which rules representing a hierarchical relationship of the information elements are stored;

element word recognition means for recognizing words in recognition object information, detecting, for each of the information elements in the recognition object information, element word candidates based on a result of the recognition, the stored contents of said word storage section and the stored contents of said rule storage section and detecting likelihoods of the element word candidates;

a record storage section in which actually existing recognition object information which can be represented as combinations of element words is stored in the form of records each of which includes record items provided by the information elements of the recognition object information;

record number acquisition means for retrieving said record storage section using the element word candidates detected by said element word recognition means successively as a key to acquire, for each of the element word candidates, a record number of a record which includes the element word candidate;

likelihood calculation means for providing likelihood counters in a corresponding relationship to the individual record numbers acquired by said record number acquisition means and adding the likelihoods of the element word candidates detected by said element word recognition means to those of said likelihood counters which correspond to the record numbers of the records which include the element word candidates;

result discrimination means for discriminating a record to be determined as a recognition result based on the count values of said likelihood counters; and

result extraction means for extracting a record to be determined as a recognition result from said record storage section based on a result of the discrimination of said result discrimination means.

2. An information recognition apparatus as claimed in claim 1, further comprising a limited record storage section, and record limitation means for retrieving, only when the element word candidates detected by said element word recognition means include an element word candidate which has a high likelihood and corresponds to a record item in which a same word is not frequently included among different element word candidates, said record storage section using the element word candidate as a key and storing records which include the element word candidate into said limited record storage section, and wherein said record number acquisition means retrieves, when a record or records are stored in said limited record storage section, said limited record storage section using the element word candidates detected by said element word recognition means successively as a key to acquire, for each of the element word candidates, record numbers of records which include the element word candidate, but retrieves, when no record is stored in said limited record storage section, said record storage section using the element word candidates detected by said element word recognition means successively as a key to acquire, for each of the element word candidates, record numbers of records which include the element word candidate, and said result extraction means extracts a record to be determined as a recognition result from said record storage section or said limited record storage section based on a result of the discrimination of said result discrimination means.

3. An information recognition apparatus as claimed in claim 1, further comprising an element unification table section in which representative representations and deformation representations of the element words are stored in a corresponding relationship, and element word unification means for outputting those of the element word candidates detected by said element word recognition means which are represented in the representative representations as they are but outputting those of the element word candidates detected by said element word recognition means which are represented in the deformation representations after converted into element word candidates of the representative representations based on the stored contents of said element unification table section, and wherein said record number acquisition means retrieves said record storage section using the element word candidates outputted from said element word unification means successively as a key to acquire, for each of the element word candidates, a record number of a record which includes the element word candidate.

4. An information recognition apparatus as claimed in claim 1, further comprising a high order candidate storage section, and a high order candidate registration completion flag provided for each of said likelihood counters, and wherein said likelihood calculation means changes, when a likelihood is added to any of said likelihood counters, a corresponding one of the high order candidate registration completion flags so as to have a value representing completion of registration if the likelihood counter has a count value higher than a threshold value determined in advance and the high order candidate registration completion flag represents incompletion of registration, and stores a record number corresponding to the likelihood counter whose count value is higher than the threshold value into said high order candidate storage section, and said result discrimination means discriminates a record to be determined as a recognition result of the recognition object information based on the count values of the likelihood counters corresponding to the record numbers stored in said high order candidate storage section.

5. An information recognition apparatus as claimed in claim 1, wherein said element word recognition means recognizes each word in the recognition object information by character recognition.

6. An information recognition apparatus as claimed in claim 1, wherein said element word recognition means recognizes each word in the recognition object information by speech recognition.

7. An information recognition apparatus as claimed in claim 1, wherein the element words which can make elements of recognition object information stored in said word storage section include a zip code, a place name, a square, an organization name, a building name and an individual name, and the recognition object information stored in said record storage section is actually existing addresses each of which can be represented as a combination of element words.

8. An information recognition apparatus as claimed in claim 1, wherein the element words which can make elements of recognition object information stored in said word storage section include a customer and an individual name, and the recognition object information stored in said record storage section is actually existing customer transaction data each of which can be represented as a combination of element words.