From d1401444dbfb515e62094f873d520a23ccbc29d9 Mon Sep 17 00:00:00 2001 From: Bryan Newbold Date: Mon, 7 May 2018 22:10:51 -0700 Subject: pig script to filter GWB CDX by SURT regexes --- pig/filter-cdx-paper-pdfs.pig | 41 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 41 insertions(+) create mode 100644 pig/filter-cdx-paper-pdfs.pig (limited to 'pig/filter-cdx-paper-pdfs.pig') diff --git a/pig/filter-cdx-paper-pdfs.pig b/pig/filter-cdx-paper-pdfs.pig new file mode 100644 index 0000000..6559066 --- /dev/null +++ b/pig/filter-cdx-paper-pdfs.pig @@ -0,0 +1,41 @@ + +-- Tries to filter down a large CDX file to a subset that is likely to be +-- journal article content, based on SURT regex patterns. +--- +-- Author: Bryan Newbold +-- Date: May 2018 + + +%default INPUT '' +%default OUTPUT '' + +set mapreduce.job.queuename default + +cdx = LOAD '$INPUT' AS cdxline:chararray; +cdx = FILTER cdx BY not STARTSWITH (cdxline, 'filedesc'); +cdx = FILTER cdx BY not STARTSWITH (cdxline, ' '); + +cdx = FOREACH cdx GENERATE STRSPLIT(cdxline,'\\s+') as cols, cdxline; +cdx = FOREACH cdx GENERATE (chararray)cols.$0 as surt, (chararray)cols.$1 as timestamp, (chararray)cols.$3 as mimetype, (chararray)cols.$4 as httpstatus, cdxline; +cdx = FILTER cdx BY not surt matches '-'; +cdx = FILTER cdx BY httpstatus matches '200'; +cdx = FILTER cdx BY mimetype matches '.*pdf.*'; + +-- This is the core regex +cdx = FILTER cdx + -- academic domains; personal (tilde) directories + BY surt matches '(edu,|..,edu|..,ac,).*\\).*\\/~.*' + + -- words in URL + OR surt matches '(?i).+\\).*/(pubs|research|publications?|articles?|proceedings?|papers?|fulltext)/.*' + + -- words in domains + OR surt matches '.*(,hal|,eprint|scielo|redalyc|revues|revistas|research|journal).*\\).*' + + -- DOI-like pattern in URL + OR surt matches '.*\\).*/10\\.\\d{3,5}/.*'; + +cdx = ORDER cdx by surt, timestamp PARALLEL 50; +cdx = FOREACH cdx GENERATE cdxline; +STORE cdx INTO '$OUTPUT' USING PigStorage(' '); + -- cgit v1.2.3