OpenXML Filter: improve the segmentation quality and merge with complex script formatting

[Denis Konovalyenko]

The creation of this issue was inspired by issue #933.

We can leave only the meaningful formatting and boost the probability of consequential runs to be merged together on extraction thus improving the segmentation quality.

On extraction: if there is no complex script content ([\u0590-\u074F\u0780-\u07BF\u0900-\u109F\u1780-\u18AF\u200C-\u200F\u202A-\u202F\u2670-\u2671\uFB1D-\uFB4F]) specified in the original document and there is no w:cs specified, the complex script formatting can be removed (w:rFonts w:csTheme w:cs , w:iCs, w:bCs, w:szCs). Similar, if there is the complex script content only or the w:cs is specified, all non-complex script formatting can be removed (w:rFonts w:asciiTheme w:ascii w:hAnsiTheme w:hAnsi w:eastAsiaTheme w:eastAsia, w:b, w:i and w:sz).

On merge: if the translated content contains non-complex script characters and there is any complex script formatting present, related non-complex script formatting has to be specified (e.g. w:bCs will trigger adding w:b if true). Similar, if the translated run content contains complex script characters and there is any non-complex script formatting present, related complex script formatting has to be specified (e.g. w:b will trigger adding w:bCs).

By the way, there was some work done in relation to the w:rFonts run property already (for more information please refer to the corresponding part of the net.sf.okapi.filters.openxml.RunMerger).

[Denis Konovalyenko]

bPreferenceAggressiveCleanup conditional parameter does not take any effect on bold and italics run properties (w:b, w:I , w:iCs, w:bCs).

Also, if it is set to false, then size run properties (w:sz, w:szCs) are groomed if the run content is not empty. If this parameter is true, then the size properties are cleaned without other conditions (the check for emptiness).

[Denis Konovalyenko]

A related pull request #646 was opened.

[Denis Konovalyenko]

Pull request #646 was merged.