How to detect a table without many borders

[NoahPerkovich]

Hi all, I am wondering if you have advice for detecting a table that looks like this with your page.find_tables() function?

[NoahPerkovich]

Note, this table is in a larger PDF of an academic paper, I am hoping to scrape multiple tables from this document, and from other documents

[JorjMcKie]

There is not much hope here. Columns seem to be fairly recognizable, rows definitely are not:
Even a human eye has problems to find a suitable principle that tells it that e.g. this is meant to represent 2 rows and not 9:

You can specify additional, "virtual" lines in find_tables() via the add_lines parameter. This is a list of pairs of points. These lines will be treated as if they had actually been drawn.
A possible strategy may consist of:

1. Locate the header by searching for words "Country", "Reference", ... having the same bottom coordinate. The top coordinate of these words is the table top. The left coordinate designate the column boundaries. This information can be used to define 2 horizontal lines (above and below the header words) and 5 or more vertical lines, the identified column borders.
2. The union of all points we have identified so far for defining the lines are used to define a clip rectangle for the table.
3. When feeding this into find_tables(clip=clip, add_lines=line_list) we will get at least some table. Probably, the table rows will be wrong though ...