|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
/** |
|
4
|
|
|
* (c) Kitodo. Key to digital objects e.V. <[email protected]> |
|
5
|
|
|
* |
|
6
|
|
|
* This file is part of the Kitodo and TYPO3 projects. |
|
7
|
|
|
* |
|
8
|
|
|
* @license GNU General Public License version 3 or later. |
|
9
|
|
|
* For the full copyright and license information, please read the |
|
10
|
|
|
* LICENSE.txt file that was distributed with this source code. |
|
11
|
|
|
*/ |
|
12
|
|
|
|
|
13
|
|
|
namespace Kitodo\Dlf\Format; |
|
14
|
|
|
|
|
15
|
|
|
/** |
|
16
|
|
|
* Fulltext ALTO format class for the 'dlf' extension |
|
17
|
|
|
* |
|
18
|
|
|
* ** This currently supports only ALTO 2.x ** |
|
19
|
|
|
* |
|
20
|
|
|
* @author Sebastian Meyer <[email protected]> |
|
21
|
|
|
* @package TYPO3 |
|
22
|
|
|
* @subpackage dlf |
|
23
|
|
|
* @access public |
|
24
|
|
|
*/ |
|
25
|
|
|
class Alto implements \Kitodo\Dlf\Common\FulltextInterface |
|
26
|
|
|
{ |
|
27
|
|
|
/** |
|
28
|
|
|
* This extracts the fulltext data from ALTO XML |
|
29
|
|
|
* |
|
30
|
|
|
* @access public |
|
31
|
|
|
* |
|
32
|
|
|
* @param \SimpleXMLElement $xml: The XML to extract the raw text from |
|
33
|
|
|
* |
|
34
|
|
|
* @return string The raw unformatted fulltext |
|
35
|
|
|
*/ |
|
36
|
|
|
public function getRawText(\SimpleXMLElement $xml) |
|
37
|
|
|
{ |
|
38
|
|
|
$rawText = ''; |
|
39
|
|
|
$xml->registerXPathNamespace('alto', 'http://www.loc.gov/standards/alto/ns-v2#'); |
|
40
|
|
|
// Get all (presumed) words of the text. |
|
41
|
|
|
$words = $xml->xpath('./alto:Layout/alto:Page/alto:PrintSpace//alto:TextBlock/alto:TextLine/alto:String/@CONTENT'); |
|
42
|
|
|
if (!empty($words)) { |
|
43
|
|
|
$rawText = implode(' ', $words); |
|
44
|
|
|
} |
|
45
|
|
|
return $rawText; |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
/** |
|
49
|
|
|
* This extracts the fulltext data from ALTO XML and returns it in MiniOCR format |
|
50
|
|
|
* |
|
51
|
|
|
* @access public |
|
52
|
|
|
* |
|
53
|
|
|
* @param \SimpleXMLElement $xml: The XML to extract the raw text from |
|
54
|
|
|
* |
|
55
|
|
|
* @return string The unformatted fulltext in MiniOCR format |
|
56
|
|
|
*/ |
|
57
|
|
|
public function getTextAsMiniOcr(\SimpleXMLElement $xml) |
|
58
|
|
|
{ |
|
59
|
|
|
$xml->registerXPathNamespace('alto', 'http://www.loc.gov/standards/alto/ns-v2#'); |
|
60
|
|
|
|
|
61
|
|
|
// get all text blocks |
|
62
|
|
|
$blocks = $xml->xpath('./alto:Layout/alto:Page/alto:PrintSpace//alto:TextBlock'); |
|
63
|
|
|
|
|
64
|
|
|
if (empty($blocks)) { |
|
65
|
|
|
return ''; |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
|
|
$miniOcr = new \SimpleXMLElement("<ocr></ocr>"); |
|
69
|
|
|
|
|
70
|
|
|
foreach ($blocks as $block) { |
|
71
|
|
|
$newBlock = $miniOcr->addChild('b'); |
|
72
|
|
|
foreach ($block->children() as $key => $value) { |
|
73
|
|
|
if ($key === "TextLine") { |
|
74
|
|
|
$newLine = $newBlock->addChild('l'); |
|
75
|
|
|
foreach ($value->children() as $wordKey => $word) { |
|
76
|
|
|
if ($wordKey == "String") { |
|
77
|
|
|
$attributes = $word->attributes(); |
|
78
|
|
|
$newWord = $newLine->addChild('w', $this->getWord($attributes)); |
|
79
|
|
|
$newWord->addAttribute('x', $this->getCoordinates($attributes)); |
|
80
|
|
|
} |
|
81
|
|
|
} |
|
82
|
|
|
} |
|
83
|
|
|
} |
|
84
|
|
|
} |
|
85
|
|
|
|
|
86
|
|
|
$miniOcrXml = $miniOcr->asXml(); |
|
87
|
|
|
if (\is_string($miniOcrXml)) { |
|
88
|
|
|
return $miniOcrXml; |
|
89
|
|
|
} |
|
90
|
|
|
return ''; |
|
91
|
|
|
} |
|
92
|
|
|
|
|
93
|
|
|
/** |
|
94
|
|
|
* This extracts and parses the word from attribute |
|
95
|
|
|
* |
|
96
|
|
|
* @access private |
|
97
|
|
|
* |
|
98
|
|
|
* @param \SimpleXMLElement $attributes: The XML to extract the word |
|
99
|
|
|
* |
|
100
|
|
|
* @return string The parsed word extracted from attribute |
|
101
|
|
|
*/ |
|
102
|
|
|
private function getWord($attributes) |
|
103
|
|
|
{ |
|
104
|
|
|
return htmlspecialchars((string) $attributes['CONTENT']) . ' '; |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
/** |
|
108
|
|
|
* This extracts and parses the word coordinates from attributes |
|
109
|
|
|
* |
|
110
|
|
|
* @access private |
|
111
|
|
|
* |
|
112
|
|
|
* @param \SimpleXMLElement $attributes: The XML to extract the word coordinates |
|
113
|
|
|
* |
|
114
|
|
|
* @return string The parsed word coordinates extracted from attribute |
|
115
|
|
|
*/ |
|
116
|
|
|
private function getCoordinates($attributes) |
|
117
|
|
|
{ |
|
118
|
|
|
return (string) $attributes['HPOS'] . ' ' . (string) $attributes['VPOS'] . ' ' . (string) $attributes['WIDTH'] . ' ' . (string) $attributes['HEIGHT']; |
|
119
|
|
|
} |
|
120
|
|
|
} |
|
121
|
|
|
|
Erikmitk /
kitodo-presentation
