. Author manuscript; available in PMC: 2018 Jun 21.

Published in final edited form as: Proc ACM SIGMOD Int Conf Manag Data. 2018 Jun;2018:1301–1316. doi: 10.1145/3183713.3183729

Table 7.

Features from Fonduer’s feature library. Example values are drawn from the example candidate in Figure 1. Capitalized prefixes represent the feature templates and the remainder of the string represents a feature’s value.

Feature Type	Arity	Description	Example Value

Structural	Unary	HTML tag of the mention	`TAG_<h1>`
Structural	Unary	HTML attributes of the mention	`HTML_ATTR_font-family:Arial`
Structural	Unary	HTML tag of the mention’s parent	`PARENT_TAG_<p>`
Structural	Unary	HTML tag of the mention’s previous sibling	`PREV_SIB_TAG_<td>`
Structural	Unary	HTML tag of the mention’s next sibling	`NEXT_SIB_TAG_<h1>`
Structural	Unary	Position of a node among its siblings	`NODE_POS_1`
Structural	Unary	HTML class sequence of the mention’s ancestors	`ANCESTOR_CLASS_<s1>`
Structural	Unary	HTML tag sequence of the mention’s ancestors	`ANCESTOR_TAG_<body>_<p>`
Structural	Unary	HTML ID’s of the mention’s ancestors	`ANCESTOR_ID_l1`^b
Structural	Binary	HTML tags shared between mentions on the path to the root of the document	`COMMON_ANCESTOR_<body>`
Structural	Binary	Minimum distance between two mentions to their lowest common ancestor	`LOWEST_ANCESTOR_DEPTH_1`

Tabular	Unary	N-grams in the same cell as the mention^a	`CELL_cev`^b
Tabular	Unary	Row number of the mention	`ROW_NUM_5`
Tabular	Unary	Column number of the mention	`COL_NUM_3`
Tabular	Unary	Number of rows the mention spans	`ROW_SPAN_1`
Tabular	Unary	Number of columns the mention spans	`COL_SPAN_1`
Tabular	Unary	Row header n-grams in the table of the mention	`ROW_HEAD_collector`
Tabular	Unary	Column header n-grams in the table of the mention	`COL_HEAD_value`
Tabular	Unary	N-grams from all Cells that are in the same row as the given mention^a	`ROW_200_[ma]`^c
Tabular	Unary	N-grams from all Cells that are in the same column as the given mention^a	`COL_200_[6]`^c
Tabular	Binary	Whether two mentions are in the same table	`SAME_TABLE`^b
Tabular	Binary	Row number difference if two mentions are in the same table	`SAME_TABLE_ROW_DIFF_1`^b
Tabular	Binary	Column number difference if two mentions are in the same table	`SAME_TABLE_COL_DIFF_3`^b
Tabular	Binary	Manhattan distance between two mentions in the same table	`SAME_TABLE_MANHATTAN_DIST_10`^b
Tabular	Binary	Whether two mentions are in the same cell	`SAME_CELL`^b
Tabular	Binary	Word distance between mentions in the same cell	`WORD_DIFF_1`^b
Tabular	Binary	Character distance between mentions in the same cell	`CHAR_DIFF_1`^b
Tabular	Binary	Whether two mentions in a cell are in the same sentence	`SAME_PHRASE`^b
Tabular	Binary	Whether two mention are in the different tables	`DIFF_TABLE`^b
Tabular	Binary	Row number difference if two mentions are in different tables	`DIFF_TABLE_ROW_DIFF_4`^b
Tabular	Binary	Column number difference if two mentions are in different tables	`DIFF_TABLE_COL_DIFF_2`^b
Tabular	Binary	Manhattan distance between two mentions in different tables	`DIFF_TABLE_MANHATTAN_DIST_7`^b

Visual	Unary	N-grams of all lemmas visually aligned with the mention^a	`ALIGNED_current`
Visual	Unary	Page number of the mention	`PAGE_1`
Visual	Binary	Whether two mentions are on the same page	`SAME_PAGE`
Visual	Binary	Whether two mentions are horizontally aligned	`HORZ_ALIGNED`^b
Visual	Binary	Whether two mentions are vertically aligned	`VERT_ALIGNED`
Visual	Binary	Whether two mentions’ left bounding-box borders are vertically aligned	`VERT_ALIGNED_LEFT`^b
Visual	Binary	Whether two mentions’ right bounding-box borders are vertically aligned	`VERT_ALIGNED_RIGHT`^b
Visual	Binary	Whether the centers of two mentions’ bounding boxes are vertically aligned	`VERT_ALIGNED_CENTER`^b

All N-grams are 1-grams by default.

This feature was not present in the example candidate. The values shown are example values from other documents.

In this example, the mention is 200, which forms part of the feature prefix. The value is shown in square brackets.