USC/ISI Automatic Data Extraction Demo

Step 2: Learning data prototype

We use machine learning algorithm to find the common patterns of tokens that describe the start and end of the training examples in Step 1. The tokens include general tokens, such as CAPITALized or ALPHAbetic words, as well as specific words like Smith. The legend appears below.

Source: amazon.com

Next

AUTHOR

STARTING PATTERNS	ENDING PATTERNS
_AL _CP _AL _AC . _CP _CP _CP _CP _AC . _CP	_CP _CP _AC . _CP

TITLE

STARTING PATTERNS	ENDING PATTERNS
_CP _CP _CP _CP _LC the _CP _CP ( _CP in The _CP _CP	_CP _AL _CP ' s _CP _CP _AL )

ISBN

STARTING PATTERNS	ENDING PATTERNS
_NGG	_NGG

PRICE

STARTING PATTERNS	ENDING PATTERNS
_N2D . 00 _N2D . 36 _N2D . 80 _N2D . 96 _N2D . 95 10 . 36	_N2D . _N2D 21 . _N2D 14 . _N2D 11 . _N2D 10 . _N2D 21 . 00 10 . 36

AVAILABILITY

STARTING PATTERNS	ENDING PATTERNS
_CP ships within 24 hours _CP title usually ships within 2 - 3 days Usually ships within 24 hours This title usually ships within 2 - 3 days	Usually ships within 24 _LC This title usually ships within 2 - 3 _LC Usually ships within 24 hours This title usually ships within 2 - 3 days

Legend:
_AC..............All caps words
_AL..............Alphabetic words
_AN.............Alphanumeric tokens
_CP..............Capitalized words
_LC..............Lowercase words
_NM.............Number
_N1D............1 digit number
_N2D............2 digit number
_N3D............3 digit number
_NLG............Large number