Back to results
US 7,689,906 B2Granted
Technique for extracting data from structured documents
AssigneeAvaya Inc.
InventorsEric T. Bax, Charless C. Fowlkes, Louis Cisnero, Jr.
Life of this patent
Estimating the term…
Abstract
The present invention discloses a technique for extracting data from a file. In accordance with the present invention, a request to extract one or more data records from the file is received. The data records within the file are identified, without using prior knowledge of a structure of the file. The data records are then extracted.
Claims4
A method for extracting records from a structured text in a computer system, comprising:
- identifying potential locations of values of record fields in the text by identifying locations in the text of items in lists of known potential values for record fields,
- identifying a region of interest in the text by applying multiple candidate region partitioners, evaluating each to measure how well it isolates a region with a high density and a high amount of potential locations of values of record fields, selecting one that measures best, and applying it to produce a region of interest,
- segmenting the region of interest into record regions that each contain data for a single record by applying multiple candidate segmenters, evaluating each to measure how well it segments into regions such that each region has one field value per record field and such that different regions have similar numbers of field values for each record field, selecting one that measures best, applying it to produce record regions, extracting field values from record regions by identifying most likely locations of field values for each record field in each record region, and
- outputting records composed of extracted field values for record fields.
1 more element
An apparatus for extracting data from a file, comprising a computer and a computer program, performed by the computer, for:
- identifying potential locations of values of record fields in the text by identifying locations in the text of items in lists of known potential values for record fields,
- identifying a region of interest in the text by applying multiple candidate region partitioners, evaluating each to measure how well it isolates a region with a high density and a high amount of potential locations of values of record fields, selecting one that measures best, and applying it to produce a region of interest,
- segmenting the region of interest into record regions that each contain data for a single record by applying multiple candidate segmenters, evaluating each to measure how well it segments into regions such that each region has one field value per record field and such that different regions have similar numbers of field values for each record field, selecting one that measures best, applying it to produce record regions,
- extracting field values from record regions by identifying most likely locations of field values for each record field in each record region, and
- outputting records composed of extracted field values for record fields.
2 more elements
Dependencies are inferred from the claim text.