tasks: wikipedia update

1 files changed, 86 insertions, 3 deletions

diff --git a/python/refcat/tasks.py b/python/refcat/tasks.py
index b4d200d..3489cc1 100644
--- a/python/refcat/tasks.py
+++ b/python/refcat/tasks.py
@@ -281,11 +281,56 @@ class WikipediaCitationsMinimalDataset(luigi.ExternalTask, Refcat):
           "ID_list": "{PMID=15944443, DOI=10.1093/molbev/msi185}"
         }
 
-
+    An updated version: wikipedia-citations-enwiki-20211201, with better ID extraction.
     """
     def output(self):
         return luigi.LocalTarget(path=os.path.join(settings.WIKIPEDIA_CITATIONS, "minimal_dataset.json"))
 
+class WikipediaCitations20211201(luigi.ExternalTask, Refcat):
+    """
+    Update wikipedia citations dataset: https://archive.org/details/wikipedia-citations-enwiki-20211201
+
+    Example line:
+
+    {
+      "revision_id": 991003499,
+      "refs": [
+	{
+	  "Authors": [
+	    {
+	      "first": "Liévin",
+	      "last": "Ndayizeye"
+	    },
+	    {
+	      "first": "Benoît",
+	      "last": "Nzigidahera"
+	    },
+	    {
+	      "first": "Abdelaziz Elamin",
+	      "last": "Gesmallah"
+	    }
+	  ],
+	  "CitationClass": "journal",
+	  "Date": "2019-03-27",
+	  "ID_list": {
+	    "DOI": "10.1007/s42690-019-00013-w",
+	    "ISSN": "1742-7592"
+	  },
+	  "Issue": "2",
+	  "Pages": "125-130",
+	  "Periodical": "International Journal of Tropical Insect Science",
+	  "PublisherName": "Springer Science and Business Media LLC",
+	  "Title": "Current distribution of Bactrocera latifrons Hendel in the different agro-ecological zones of Burundi",
+	  "Volume": "39"
+	}
+      ],
+      "site_name": "enwiki",
+      "page_title": "List of Bactrocera species"
+    }
+    """
+    def output(self):
+        return luigi.LocalTarget(path=os.path.join(settings.WIKIPEDIA_CITATIONS_20211201,
+                                                   "enwiki-20211201-pages-articles.citations.json"))
 
 class OpenLibraryEditions(luigi.ExternalTask, Refcat):
     """
@@ -1517,6 +1562,37 @@ class WikipediaDOI(Refcat):
     def output(self):
         return luigi.LocalTarget(path=self.path(ext="tsv.zst"), format=Zstd)
 
+class Wikipedia20211201DOI(Refcat):
+    """
+    Updated wikipedia citations dataset.
+
+    {"type_of_citation":"cite journal","page_title":"Abortion in
+    Alabama","Title":"Why We Should Stop Using the Term \"Elective
+    Abortion\"","ID_list":"{PMID=30585581, DOI=10.1001/amajethics.2018.1175}"}
+    """
+    def requires(self):
+        return WikipediaCitations20211201()
+
+    def run(self):
+        with self.output().open("w") as output:
+            with self.input().open() as handle:
+                for line in handle:
+                    doc = json.loads(line)
+                    for i, ref in enumerate(doc.get("refs", [])):
+                        if not "ID_list" in ref:
+                            continue
+                        if not "DOI" in ref["ID_list"]:
+                            continue
+                        doi = ref["ID_list"]["DOI"]
+                        reduced = doc
+                        reduced["refs"] = []
+                        reduced["index"] = i
+                        reduced["Title"] = ref.get("Title")
+                        fields = [doi, doc["page_title"], json.dumps(reduced)]
+                        output.write("\t".join(fields) + "\n")
+
+    def output(self):
+        return luigi.LocalTarget(path=self.path(ext="tsv"))
 
 class BrefZipWikiDOI(Refcat):
     """
@@ -1526,17 +1602,24 @@ class BrefZipWikiDOI(Refcat):
     """
     def requires(self):
         return {
-            "wiki": WikipediaDOI(),
+            # "wiki": WikipediaDOI(),
+            "wiki": Wikipedia20211201DOI(),
             "fatcat": FatcatDOI(),
         }
 
     def run(self):
         output = shellout("""
-                          skate-reduce -m wiki -W <(zstdcat -T0 {wiki}) -L <(zstdcat -T0 {fatcat}) |
+                          skate-reduce -m wiki -W {wiki} -L <(zstdcat -T0 {fatcat}) |
                           zstd -T0 -c > {output}
                           """,
                           wiki=self.input().get("wiki").path,
                           fatcat=self.input().get("fatcat").path)
+        # output = shellout("""
+        #                   skate-reduce -m wiki -W <(zstdcat -T0 {wiki}) -L <(zstdcat -T0 {fatcat}) |
+        #                   zstd -T0 -c > {output}
+        #                   """,
+        #                   wiki=self.input().get("wiki").path,
+        #                   fatcat=self.input().get("fatcat").path)
         luigi.LocalTarget(output).move(self.output().path)
 
     def output(self):