basic JALC XML DOI metadata parser

2 files changed, 385 insertions, 0 deletions

diff --git a/python/parse_jalc_xml.py b/python/parse_jalc_xml.py
new file mode 100644
index 00000000..7df79421
--- /dev/null
+++ b/python/parse_jalc_xml.py
@@ -0,0 +1,209 @@
+
+import sys
+import json
+import datetime
+import unicodedata
+from bs4 import BeautifulSoup
+from bs4.element import NavigableString
+
+
+DATE_FMT = "%Y-%m-%d"
+
+def is_cjk(s):
+    if not s:
+        return False
+    return unicodedata.name(s[0]).startswith("CJK")
+
+class JalcXmlParser():
+    """
+    Converts JALC DOI metadata (in XML/RDF format) to fatcat release entity
+
+    NOTE: some JALC DOIs seem to get cross-registered with Crossref
+    """
+
+    def __init__(self):
+        pass
+
+    def parse_file(self, handle):
+
+        # 1. open with beautiful soup
+        soup = BeautifulSoup(handle, "xml")
+
+        # 2. iterate over articles, call parse_article on each
+        for record in soup.find_all("Description"):
+            resp = self.parse_record(record)
+            print(json.dumps(resp))
+            #sys.exit(-1)
+
+
+    def parse_record(self, record):
+        """
+        In JALC metadata, both English and Japanese records are given for most
+        fields.
+        """
+
+        #extra = dict()
+        #extra_jalc = dict()
+
+        titles = record.find_all("title")
+        title = titles[0].string.strip()
+        original_title = None
+        if title.endswith('.'):
+            title = title[:-1]
+        if len(titles) > 1:
+            original_title = titles[1].string.strip()
+            if original_title.endswith('.'):
+                original_title = original_title[:-1]
+
+        doi = None
+        if record.doi:
+            doi = record.doi.string.lower().strip()
+            assert doi.startswith('10.')
+
+        contribs = []
+        people = record.find_all("Person")
+        if people and (len(people) % 2 == 0) and is_cjk(people[1].find('name').string):
+            # both english and japanese names are included
+            for i in range(int(len(people)/2)):
+                # both english and japanese names are included for every author
+                eng = people[i*2]
+                jpn = people[i*2 + 1]
+                raw_name = eng.find('name')
+                orig_name = jpn.find('name')
+                if not raw_name:
+                    raw_name = orig_name
+                contrib = dict(
+                    raw_name=raw_name.string,
+                    role='author',
+                )
+                if raw_name and orig_name:
+                    contrib['extra'] = dict(original_name=orig_name.string)
+                contribs.append(contrib)
+        elif people:
+            for eng in people:
+                raw_name = eng.find('name')
+                contrib = dict(
+                    raw_name=eng.find('name').string,
+                    role='author',
+                )
+                contribs.append(contrib)
+
+        release_year = None
+        release_date = None
+        date = record.date or None
+        if date:
+            date = date.string
+            if len(date) is 10:
+                release_date = datetime.datetime.strptime(state['completed-date'], DATE_FMT).date()
+                release_year = release_date.year
+                release_date = release_date.isoformat()
+            elif len(date) is 4:
+                release_year = int(date)
+
+        pages = None
+        if record.startingPage:
+            pages = record.startingPage.string
+            if record.endingPage:
+                pages = "{}-{}".format(pages, record.endingPage.string)
+        volume = None
+        if record.volume:
+            volume = record.volume.string
+        issue = None
+        if record.number:
+            # note: number/issue transform
+            issue = record.number.string
+
+        issn = None
+        issn_list = record.find_all("issn")
+        if issn_list:
+            # if we wanted the other ISSNs, would also need to uniq the list.
+            # But we only need one to lookup ISSN-L/container
+            issn = issn_list[0].string
+
+        container = dict()
+        container_extra = dict()
+        container_name = None
+        if record.publicationName:
+            pubs = [p.string.strip() for p in record.find_all("publicationName")]
+            pubs = [p for p in pubs if p]
+            assert(pubs)
+            if len(pubs) > 1 and pubs[0] == pubs[1]:
+                pubs = [pubs[0]]
+            elif len(pubs) > 1 and is_cjk(pubs[0]):
+                # ordering is not reliable
+                pubs = [pubs[1], pubs[0]]
+            container_name = pubs[0]
+            container['name'] = container_name
+            if len(pubs) > 1:
+                orig_container_name = pubs[1]
+                container_extra['original_name'] = pubs[1]
+        publisher = None
+        if record.publisher:
+            pubs = [p.string.strip() for p in record.find_all("publisher")]
+            pubs = [p for p in pubs if p]
+            if len(pubs) > 1 and pubs[0] == pubs[1]:
+                pubs = [pubs[0]]
+            elif len(pubs) > 1 and is_cjk(pubs[0]):
+                # ordering is not reliable
+                pubs = [pubs[1], pubs[0]]
+            publisher = pubs[0]
+            container['publisher'] = publisher
+            if len(pubs) > 1:
+                container_extra['publisher_alt_name'] = pubs[1]
+        if container_extra:
+            container['extra'] = container_extra
+        if not container:
+            container = None
+
+        # the vast majority of works are in japanese
+        # TODO: any indication when *not* in japanese?
+        lang = "ja"
+
+        # reasonable default for this collection
+        release_type = "article-journal"
+
+        re = dict(
+            work_id=None,
+            title=title,
+            original_title=original_title,
+            release_type="article-journal",
+            release_status='submitted', # XXX: source_type?
+            release_date=release_date,
+            release_year=release_year,
+            #arxiv_id
+            doi=doi,
+            #pmid
+            #pmcid
+            #isbn13     # never in Article
+            volume=volume,
+            issue=issue,
+            pages=pages,
+            publisher=publisher,
+            language=lang,
+            #license_slug   # not in MEDLINE
+
+            # content, mimetype, lang
+            #abstracts=abstracts,
+
+            # raw_name, role, raw_affiliation, extra
+            contribs=contribs,
+
+            #   name, type, publisher, issnl
+            #   extra: issnp, issne, original_name, languages, country
+            container=container,
+
+            # extra:
+            #   withdrawn_date
+            #   translation_of
+            #   subtitle
+            #   aliases
+            #   container_name
+            #   group-title
+            #   pubmed: retraction refs
+            #extra=extra,
+        )
+        return re
+
+if __name__=='__main__':
+    parser = JalcXmlParser()
+    parser.parse_file(open(sys.argv[1]))
diff --git a/python/tests/files/jalc_lod_sample.xml b/python/tests/files/jalc_lod_sample.xml
new file mode 100644
index 00000000..3a9dd770
--- /dev/null
+++ b/python/tests/files/jalc_lod_sample.xml
@@ -0,0 +1,176 @@
+<rdf:RDF xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#"
+xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:dcterms="http://purl.org/dc/terms/"
+xmlns:foaf="http://xmlns.com/foaf/0.1/"
+xmlns:prism="http://prismstandard.org/namespaces/basic/2.0/">
+<rdf:Description rdf:about="http://doi.org/10.2497/jjspm.36.898">
+<prism:doi>10.2497/jjspm.36.898</prism:doi>
+<dcterms:title>New carbides in the Ni-Ti-Mo-C system.</dcterms:title>
+<dcterms:title>Ｎｉ－Ｔｉ－Ｍｏ－Ｃ系に出現する新炭化物相について</dcterms:title>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Hashimoto Yasuhiko</foaf:name>
+<foaf:familyName>Hashimoto</foaf:familyName>
+<foaf:givenName>Yasuhiko</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>橋本 雍彦</foaf:name>
+<foaf:familyName>橋本</foaf:familyName>
+<foaf:givenName>雍彦</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Koyama Koichiro</foaf:name>
+<foaf:familyName>Koyama</foaf:familyName>
+<foaf:givenName>Koichiro</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>香山 滉一郎</foaf:name>
+<foaf:familyName>香山</foaf:familyName>
+<foaf:givenName>滉一郎</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Suzuki Kenji</foaf:name>
+<foaf:familyName>Suzuki</foaf:familyName>
+<foaf:givenName>Kenji</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>鈴木 建次</foaf:name>
+<foaf:familyName>鈴木</foaf:familyName>
+<foaf:givenName>建次</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Takahashi Teruo</foaf:name>
+<foaf:familyName>Takahashi</foaf:familyName>
+<foaf:givenName>Teruo</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>高橋 輝男</foaf:name>
+<foaf:familyName>高橋</foaf:familyName>
+<foaf:givenName>輝男</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dc:creator>Hashimoto Yasuhiko</dc:creator>
+<dc:creator>橋本 雍彦</dc:creator>
+<dc:creator>Koyama Koichiro</dc:creator>
+<dc:creator>香山 滉一郎</dc:creator>
+<dc:creator>Suzuki Kenji</dc:creator>
+<dc:creator>鈴木 建次</dc:creator>
+<dc:creator>Takahashi Teruo</dc:creator>
+<dc:creator>高橋 輝男</dc:creator>
+<dcterms:publisher>Japan Society of Powder and Powder Metallurgy</dcterms:publisher>
+<dcterms:publisher>一般社団法人 粉体粉末冶金協会</dcterms:publisher>
+<dcterms:date rdf:resource="http://www.w3.org/2001/XMLSchema#date">1989</dcterms:date>
+<prism:volume>36</prism:volume>
+<prism:number>8</prism:number>
+<prism:startingPage>898</prism:startingPage>
+<prism:endingPage>902</prism:endingPage>
+<prism:issn>0532-8799</prism:issn>
+<prism:issn>0532-8799</prism:issn>
+<prism:issn>1880-9014</prism:issn>
+<prism:issn>1880-9014</prism:issn>
+<dcterms:publicationName>
+Journal of the Japan Society of Powder and Powder Metallurgy
+</dcterms:publicationName>
+<dcterms:publicationName>粉体および粉末冶金</dcterms:publicationName>
+</rdf:Description>
+<rdf:Description rdf:about="http://doi.org/10.2497/jjspm.36.903">
+<prism:doi>10.2497/jjspm.36.903</prism:doi>
+<dcterms:title>
+Effects of grain size on cutting performance of Al2O3-TiC ceramics tool.
+</dcterms:title>
+<dcterms:title>Ａｌ２Ｏ３－ＴｉＣセラミックス工具の切削性能におよぼすセラミックス粒度の影響
+</dcterms:title>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Katsumura Yuji</foaf:name>
+<foaf:familyName>Katsumura</foaf:familyName>
+<foaf:givenName>Yuji</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>勝村 祐次</foaf:name>
+<foaf:familyName>勝村</foaf:familyName>
+<foaf:givenName>祐次</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Sobata Kaoru</foaf:name>
+<foaf:familyName>Sobata</foaf:familyName>
+<foaf:givenName>Kaoru</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>蕎麦田 薫</foaf:name>
+<foaf:familyName>蕎麦田</foaf:familyName>
+<foaf:givenName>薫</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Uehara Yoshito</foaf:name>
+<foaf:familyName>Uehara</foaf:familyName>
+<foaf:givenName>Yoshito</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>上原 好人</foaf:name>
+<foaf:familyName>上原</foaf:familyName>
+<foaf:givenName>好人</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>Suzuki Hisashi</foaf:name>
+<foaf:familyName>Suzuki</foaf:familyName>
+<foaf:givenName>Hisashi</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dcterms:creator>
+<foaf:Person>
+<foaf:name>鈴木 寿</foaf:name>
+<foaf:familyName>鈴木</foaf:familyName>
+<foaf:givenName>寿</foaf:givenName>
+</foaf:Person>
+</dcterms:creator>
+<dc:creator>Katsumura Yuji</dc:creator>
+<dc:creator>勝村祐次</dc:creator>
+<dc:creator>Sobata Kaoru</dc:creator>
+<dc:creator>蕎麦田 薫</dc:creator>
+<dc:creator>Uehara Yoshito</dc:creator>
+<dc:creator>上原好人</dc:creator>
+<dc:creator>Suzuki Hisashi</dc:creator>
+<dc:creator>鈴木寿</dc:creator>
+<dcterms:publisher>Japan Society of Powder and Powder Metallurgy</dcterms:publisher>
+<dcterms:publisher>一般社団法人 粉体粉末冶金協会</dcterms:publisher>
+<dcterms:date rdf:resource="http://www.w3.org/2001/XMLSchema#date">1989</dcterms:date>
+<prism:volume>36</prism:volume>
+<prism:number>8</prism:number>
+<prism:startingPage>903</prism:startingPage>
+<prism:endingPage>907</prism:endingPage>
+<prism:issn>0532-8799</prism:issn>
+<prism:issn>0532-8799</prism:issn>
+<prism:issn>1880-9014</prism:issn>
+<prism:issn>1880-9014</prism:issn>
+<dcterms:publicationName>
+Journal of the Japan Society of Powder and Powder Metallurgy
+</dcterms:publicationName>
+<dcterms:publicationName>粉体および粉末冶金</dcterms:publicationName>
+</rdf:Description>
+</rdf:RDF>