DAS extraction issues with new page

[lfoppiano]

In this issue we further collect DAS extraction issues which seems to be hard to cure with more training data. They happen in Plos articles and it's due to the page break together with the double column, we could check whether we can fix the segmentation parser at feature level.

journal.pwat.0000127.pdf

<div type="availability">
                <div
                    xmlns="http://www.tei-c.org/ns/1.0">
                    <p>The original contributions presented in the study are publicly available. The data can be found at the following: Service provider data can be accessed at: 
                        <ref type="url" target="https://database.ib-net.org/countries_results?ctry=">https:// database.ib-net.org/countries_results?ctry=</ref> 29&amp;years=2018&amp;type=report&amp;ent=country&amp;mult= true&amp;report=1&amp;table=true&amp;chart= false&amp;chartType=column&amp;lang=EN&amp;exch=1. 2018 DHS data is available on application via the public repository: 
                        <ref type="url" target="https://dhsprogram.com/methodology/survey/survey-display-542.cfm">https://dhsprogram.com/methodology/ survey/survey-display-542.cfm</ref>. Regulator data is openly available and can be extracted from the
                    </p>
                </div>
            </div>

[lfoppiano]

additional examples:

• journal.pcbi.1011795.pdf
• 10.1371/journal.pbio.3002472

[lfoppiano]

• journal.pcbi.1011115.pdf DAS truncated
• journal.pone.0284609.pdf DAS missing

[lfoppiano]

Das is truncated by the new page: 10.1038_s41467-020-20536-y.pdf

[lfoppiano]

Another problem where a note was wrongly included in the DAS:

The role of value co-creation in building trust and reputation in the digital banking era-1.pdf

[lfoppiano]

Another example:

23311932.2024.2374940_training.pdf

[lfoppiano]

All these error cases have been added in the training data.