From 1263ee33535d232d702324980e7ff69305ed8795 Mon Sep 17 00:00:00 2001 From: Bryan Newbold Date: Fri, 21 May 2021 17:41:41 -0700 Subject: ingest PDF extraction updates --- python/sandcrawler/html_metadata.py | 54 +++++++++++++++++++++++++++++++++++++ 1 file changed, 54 insertions(+) (limited to 'python/sandcrawler/html_metadata.py') diff --git a/python/sandcrawler/html_metadata.py b/python/sandcrawler/html_metadata.py index 1e58778..c805f0a 100644 --- a/python/sandcrawler/html_metadata.py +++ b/python/sandcrawler/html_metadata.py @@ -383,6 +383,60 @@ PDF_FULLTEXT_PATTERNS: List[dict] = [ "technique": "PDF URL link", "example_page": "https://www.degruyter.com/document/doi/10.1515/zaw-2021-0001/html", }, + { + "in_doc_url": "repositorio.unicamp.br/handle/", + "in_fulltext_url": "/bitstream/", + "selector": "table.panel-body a[target='_blank']", + "attr": "href", + "technique": "PDF URL link", + "example_page": "http://www.repositorio.unicamp.br/handle/REPOSIP/287750", + }, + { + "in_doc_url": "dlc.library.columbia.edu/durst/", + "selector": "dd.blacklight-lib_non_item_in_context_url_ssm a[href]", + "attr": "href", + "technique": "Access URL link", + "example_page": "https://dlc.library.columbia.edu/durst/cul:18931zcrk9", + }, + { + "in_doc_url": "fldeploc.dep.state.fl.us/geodb_query/fgs_doi", + "in_fulltext_url": "pdf", + "selector": "p a[href]", + "attr": "href", + "technique": "PDF URL link", + "example_page": "http://fldeploc.dep.state.fl.us/geodb_query/fgs_doi.asp?searchCode=IC29", + }, + { + "in_doc_url": "preprints.jmir.org/preprint/", + "selector": "a.pdf-download-button", + "attr": "href", + "technique": "PDF URL link", + "example_page": "https://preprints.jmir.org/preprint/22556", + }, + { + "in_doc_url": "bloomsburycollections.com/", + "in_fulltext_url": "pdf", + "selector": "li.download-item a[href]", + "attr": "href", + "technique": "PDF URL link", + "example_page": "https://www.bloomsburycollections.com/book/the-political-economies-of-media-the-transformation-of-the-global-media-industries/the-political-economies-of-media-and-the-transformation-of-the-global-media-industries", + }, + { + "in_doc_url": "emerald.com/insight/content/", + "in_fulltext_url": "pdf", + "selector": "a.intent_pdf_link", + "attr": "href", + "technique": "PDF URL link", + "example_page": "https://www.emerald.com/insight/content/doi/10.1108/RAMJ-11-2020-0065/full/html", + }, + { + "in_doc_url": "ingentaconnect.com/content/", + "in_fulltext_url": "pdf", + "selector": "a.pdf[data-popup]", + "attr": "data-popup", + "technique": "PDF URL link", + "example_page": "https://www.ingentaconnect.com/content/ista/sst/2021/00000049/00000001/art00007", + }, ] FULLTEXT_URL_PATTERNS_SKIP = [ -- cgit v1.2.3