Multilines cells not well detected with guess mode

Hello,

I’m trying to use your fantastic piece of software to analyse the table provided in page 1 of this pdf : input.pdf

I’m using the last version of tabula-java. I compiled it thought git and mvn (as specified in readme.md).

I execute this command line : java -jar target/tabula-1.0.2-SNAPSHOT-jar-with-dependencies.jar -g input.pdf > output.csv

I’ve got the following output : output.txt

But, I expect this : output_expected.txt

Do you think is that possible ?

[jazzido]

Tabula itself! Get it here: http://tabula.technology — The script export format contains the area of interest:

[My3VM]

Hello there, I’ve come across similar scenario to detect the tables followed by MultiLine cells. I’ve tried the above idea, though it works very well using the tool (on Mac OSX), I’m unable to reproduce the exact results using tabula.jar.

Tried the one freshly built from source as well as the one that came with Tabula.app. (Tabula.app/Contents/Java). In both these cases the observation is consistent, that it misses the last row.,while rest of the results do come in well.

Tried to achieve similar workflow by tweaking code to handle both -g -l option which first guesses the rect and then passes the rectangles with SpreadsheetDetection algorithm, but the observation is still the same…,last rows get skipped.

Please suggest how could I augment both guess option and lattice way of extraction.