Page Model

ONE_HTML_PAGE ::= List( COMPANY )

COMPANY::= < Ticker, LastTradedAt, Change, PercentageChange,Volume, NmbTrades, Bid, Ask,
                              Low, High, PreviousClose, Open, 52Low, 52High, Eps, Div, Date, Time >

NOTE:

each HTML page is seen as a list of companies, where each company has the 18 attributes above. In database terms, an HTML page represents a table with eighteen columns.

Items to be extracted: all 18 items listed above.

NOTE: this source is particularly hard to wrapp because

• the same string "n/a" may denotes either one or two missing items (e.g., see the "VOL/#TRDS" and "52LOW/HIGH" columns in the first row of the table in the first sample document). As we did not find a convenient way to handle such inconsistencies, the wrapper induction task is extremely difficult given the ambiguous input.
• the difficulties above are multiplied by the fact that in some cases (e.g., see the date "May 02" from the first sample document) the content of a given item appears in the wrong column of the HTML table. As STALKER ignores the item's semantics, such misplaced data are confusing because of two reasons. First, such data generate "false positives" for the rules that extract the information from the column in which the data was wronfully stored (i.e., the information that SHOULD have been there). Second, we decided not to label the misplaced data because we are not interested in learning specialized rules for each piece of misplaced data (we fully realize that such a decission makes the learning task easier!).

  ---

  SAMPLE EXTRACTION OUTPUT

  • Sample Document 1
  • Sample Document 2

  ---