Context Aware Data Extraction From Textual Records

A computer-implemented method for extracting data from textual records, comprising:

• receiving, by a computer processor, a reference record;
• receiving a selection for a selected data field in the reference record;
• identifying, by the computer processor, the data type of the selected data field;
• for each other data field in the reference record from the first data field to the data field preceding the selected data field, and for the data field immediately following the selected data field:
• identifying, by the computer processor, the data type of the other data field;
• forming, by the computer processor, an extraction pattern, ExPat, by concatenating the identified data types in order of occurrence in the reference record;
• splitting the extraction pattern, ExPat, into a first part, Rp, and a second part, Rt, where the first part, Rp, comprises data types from the beginning of the extraction pattern to the data type of the data field preceding the selected data field, and the second part, Rt, comprises the remaining data types in the extraction pattern, ExPat;
• identifying the number of expressions in the first part, Rp, and assigning the number of expressions in the first part to n, n=|Rp|;
• if the first part, Rp, contains more expressions than the second part, Rt:
• reducing the length of the extraction pattern, ExPat, by iteratively performing:
• inserting the n-th element of the first part, Rp, at the beginning of the second part, Rt, thereby incrementing the length of Rt by 1;
• for all ordered subsets, Rc, of the first part, Rp, containing the same number of expressions as the second part, Rt:
• checking, by the computer processor, whether the subset, Rc, is disjoint to the second part, Rt;
• if at least one subset, Rc, of the first part, Rp, is not disjoint to the second part, Rt:
• decrement n by 1, n=n−1, and perform the next iteration, else exit the iteration;
• forming, by the computer processor, an optimized extraction pattern, ExPat*, by concatenating DATA*, ‘(’, Rt, and ‘)’, where DATA* is a first expression of arbitrary length including zero having the universal data type DATA, and Rt comprises the ordered sequence of data types in the second part, Rt;
• for each textual data record:
• extracting data from the textual data record by applying the optimized extraction pattern, ExPat*, to the textual data record.

A non-transitory computer-readable medium having computer-executable instructions that, upon execution of the instructions by a processor of a computer, cause the computer to

• a) receive a reference record;
• b) receive a selection of a data field in the reference record;
• c) identify data types of data fields comprising the reference record;
• d) form a preceding pattern of data types, where the preceding pattern is formed from the data types of the data fields preceding the selected data field in the reference record;
• e) form an extraction pattern of data types, where the extraction pattern is formed from the data type of the selected data field and data types for at least one additional data field following the selected data field in the reference record;
• f) determine whether the extraction pattern is disjoint to any segment of the preceding pattern, where each segment is comprised of sequential data types in the preceding pattern and has same length as the extraction pattern;
• g) insert the data type for last data field in the preceding pattern at the beginning of the extraction pattern to form a new extraction pattern;
• h) repeat steps f) and g) with the new extraction pattern until a determination is made that the new extraction pattern is disjoint to any segment comprising the preceding pattern;
• and
• i) form an optimized extraction pattern from the new extraction pattern in response to a determination that the new extraction pattern is disjoint to any segment comprising the preceding pattern.