Device dependent rendering of PDF content
The technology disclosed relates to systems and methods for device-dependent display of an article from a PDF file. The article can have multiple columns. The system can use a library to render the article from the PDF file. The rendering can include bounding boxes positioned at on-page coordinates that can include one or more images and multiple text blocks of glyphs. The system can partition the text blocks and images in two or more columns using dynamically adjusted valleys between columns. The system can set a reading order of the article after rendering. The system can merge and split text blocks to form paragraphs of text. The system includes logic to infer semantic information about typographic roles of the paragraphs from at least the font information. The system can cause display of the article in a device-dependent format using the semantic information and the reading order.
1. A method of device-dependent display of an article from a PDF file that has multiple columns in at least parts of the article, the method including:
using a library to render the article from the PDF file, including rendering of a plurality of bounding boxes, positioned at on-page coordinates, that contain one or more images and multiple text blocks of glyphs, with font information for the glyphs;
setting a reading order of the article after the rendering, including pulling out text blocks spanning more than half of a width of a page and pulling out images, then reflowing the text blocks to produce the reading order;
merging the text blocks as they appear in the reading order into one or more paragraphs of text using the font information and using starting and ending positions of horizontally arranged text elements in the text blocks to delimit the paragraphs;
inferring semantic information about typographic roles of the paragraphs in the merged text blocks from at least the font information, including font name and font size distribution for sequences of the glyphs; and
causing display of the article in a device-dependent format, including the merged text blocks, using the semantic information and the reading order.
2. The method of claim 1 , wherein the font information of the glyphs further includes offsets and advances for characters indicating locations of characters in the bounding box.
3. The method of claim 1 , further including dividing some of the text blocks into separate paragraphs.
4. The method of claim 1 , further including using variation in vertical spacing between the horizontally arranged text elements during the merging of the text blocks into the paragraphs.
5. The method of claim 1 , wherein inferring the semantic information about the typographic roles of the paragraphs further includes distinguishing among at least headline text, subtitle text, body text, byline text, image attribution text, and image caption text and annotating the paragraphs with the inferred semantic information.
6. The method of claim 1 , further including displaying the images with the article.
7. The method of claim 1 , further including partitioning the text blocks and images into two or more columns and one or more sections of the article using dynamically adjusted valleys between the columns, wherein the dynamically adjusted valleys have sizes responsive to the font information for the glyphs in adjoining text blocks, and wherein dynamically adjusting the dynamically adjusted valleys further includes increasing a threshold for valley width, used when detecting a valley separating characters, responsive to large font sizes.
8. A system including a memory and one or more processors coupled to the memory, the memory loaded with computer instructions to generate device-dependent display of an article from a PDF file that has multiple columns in at least parts of the article, which computer instructions, when executed on the processors, implement actions comprising:
using a library to render the article from the PDF file, including rendering of a plurality of bounding boxes, positioned at on-page coordinates, that contain one or more images and multiple text blocks of glyphs, with font information for the glyphs;
setting a reading order of the article after the rendering, including pulling out text blocks spanning more than half of a width of a page and pulling out images, then reflowing the text blocks to produce the reading order;
merging the text blocks as they appear in the reading order into one or more paragraphs of text using the font information and using starting and ending positions of horizontally arranged text elements in the text blocks to delimit the paragraphs;
inferring semantic information about typographic roles of the paragraphs in the merged text blocks from at least the font information, including font name and font size distribution for sequences of the glyphs; and
causing display of the article in a device-dependent format, including the merged text blocks, using the semantic information and the reading order.
9. The system of claim 8 , wherein the font information of the glyphs further includes offsets and advances for characters indicating locations of characters in the bounding boxes.
10. The system of claim 8 , further implementing actions comprising, dividing some of the text blocks into separate paragraphs.
11. The system of claim 8 , further implementing actions comprising, using variation in vertical spacing between the horizontally arranged text elements during the merging of the text blocks into the one or more paragraphs.
12. The system of claim 8 , wherein inferring the semantic information about the typographic roles of the one or more paragraphs further includes distinguishing among at least headline text, subtitle text, body text, byline text, image attribution text, and image caption text and annotating the one or more paragraphs with the inferred semantic information.
13. The system of claim 8 , further implementing actions comprising, displaying the images with the article.
14. The system of claim 8 , further including using dynamically adjusted valleys between the columns, partitioning the text blocks and images into two or more columns and one or more sections of the article, wherein the dynamically adjusted valleys have sizes responsive to the font information for the glyphs in adjoining text blocks, and wherein dynamically adjusting the dynamically adjusted valleys further includes increasing a threshold for valley width, used when detecting a valley separating characters, responsive to large font sizes.
15. A non-transitory computer readable storage medium impressed with computer program instructions to generate device-dependent display of an article from a PDF file that has multiple columns in at least parts of the article, which computer program instructions, when executed on a processor, implement a method comprising:
using a library to render the article from the PDF file, including rendering of a plurality of bounding boxes, positioned at on-page coordinates, that contain one or more images and multiple text blocks of glyphs, with font information for the glyphs;
setting a reading order of the article after the rendering, including pulling out text blocks spanning more than half of a width of a page and pulling out images, then reflowing the text blocks to produce the reading order;
merging the text blocks as they appear in the reading order into one or more paragraphs of text using the font information and using starting and ending positions of horizontally arranged text elements in the text blocks to delimit the paragraphs;
inferring semantic information about typographic roles of the paragraphs in the merged text blocks from at least the font information, including font name and font size distribution for sequences of the glyphs; and
causing display of the article in a device-dependent format, including the merged text blocks, using the semantic information and the reading order.
16. The non-transitory computer readable storage medium of claim 15 , wherein the font information of the glyphs further includes offsets and advances for characters indicating locations of characters in the bounding box.
17. The non-transitory computer readable storage medium of claim 15 , implementing the method further comprising, dividing some of the text blocks into separate paragraphs.
18. The non-transitory computer readable storage medium of claim 15 , implementing the method further comprising, using variation in vertical spacing between the horizontally arranged text elements during the merging of the text blocks into the one or more paragraphs.
19. The non-transitory computer readable storage medium of claim 15 , wherein inferring the semantic information about the typographic roles of the paragraphs further includes distinguishing among at least headline text, subtitle text, body text, byline text, image attribution text, and image caption text and annotating the paragraphs with the inferred semantic information.
20. The non-transitory computer readable storage medium of claim 15 , implementing the method further comprising, displaying the images with the article.
21. The non-transitory computer readable storage medium of claim 15 , further including using dynamically adjusted valleys between the columns, partitioning the text blocks and images into two or more columns and one or more sections of the article, wherein the dynamically adjusted valleys have sizes responsive to the font information for the glyphs in adjoining text blocks, and wherein dynamically adjusting the dynamically adjusted valleys further includes increasing a threshold for valley width, used when detecting a valley separating characters, responsive to large font sizes.