corpusparser as standalone project

Makefile

@@ -0,0 +1,35 @@
+# All required components, to create and fill a database,
+# instantiate backend and frontend. 
+
+MAKE_ROOT = $(shell pwd)
+
+### Input data
+# I received ssj500k in one .xml file,
+# kres is composed of many .xml files
+# I generated srl tags for kres in separate .json files
+# (for each kres.xml file there is a kres.json file with srl tags)
+SSJ_FILE = "$(MAKE_ROOT)/data/samples/ssj_example/ssj500k-sl.body.sample.xml"
+KRES_FOLDER = "$(MAKE_ROOT)/data/samples/kres_example"
+KRES_SRL_FOLDER = "$(MAKE_ROOT)/data/kres_srl"
+OUTPUT = "file"
+OUTDIR = "/home/voje/workdir/test_out"
+DBADDR = ""
+
+DB_ADM_USER = testadmin
+DB_ADM_PASS = testadminpass
+DB_USR_USER = testuser
+DB_USR_PASS = testuserpass
+export
+
+.PHONY: cjvt-corpusparser-env
+
+all: cjvt-corpusparser-env install
+
+# prereq (environment)
+cjvt-corpusparser-env:
+	cd dockerfiles/cjvt-corpusparser-env; $(MAKE)
+
+# commands inside containers:
+install:
+	pip3 install -e .
+

README.md

@@ -0,0 +1,41 @@
+# corpusparser
+A tool for parsing ssj500k and Kres into a unified .json format. 
+
+## Quickstart
+Run `make`. You will get a container with python3 and this package installed. 
+
+## Input:
+### ssj500k
+To parse ssj500k, point to the monolythic `ssj500k-sl.body.xml` file (tested on ssj500k 2.1).  
+
+### Kres
+To parse Kres, point to folders:
+* Kres folder, containig several (around 20K) .xml files (`F00XXXXX.xml.parsed.xml`).  
+* Kres SRL folder, containing SRL links for the corresponding F00...xml files (`F00XXXXX.srl.json`).  
+
+## Internal data format
+This is the internal python dict data format. It can be stored to file as `.json` or stored into a database for application usage.  
+```python
+{
+	'sid': 'F0034713.5.0',
+	'text': 'Mednarodni denarni sklad je odobril 30 milijard evrov vredno posojilo Grčiji. ',
+	'tokens': [
+		{'text': 'Mednarodni', 'lemma': 'mednaroden', 'msd': 'Ppnmeid', 'word': True, 'tid': 1},
+		{'text': 'denarni', 'lemma': 'denaren', 'msd': 'Ppnmeid', 'word': True, 'tid': 2},
+		{'text': 'sklad', 'lemma': 'sklad', 'msd': 'Somei', 'word': True, 'tid': 3},
+		{'text': 'je', 'lemma': 'biti', 'msd': 'Gp-ste-n', 'word': True, 'tid': 4},
+		{'text': 'odobril', 'lemma': 'odobriti', 'msd': 'Ggdd-em', 'word': True, 'tid': 5},
+		{'text': '30', 'lemma': '30', 'msd': 'Kag', 'word': True, 'tid': 6},
+		{'text': 'milijard', 'lemma': 'milijarda', 'msd': 'Sozmr', 'word': True, 'tid': 7}, # ... 
+	]
+	'jos_links': [
+		{'to': 1, 'from': 3, 'afun': 'dol'},
+		{'to': 2, 'from': 3, 'afun': 'dol'},
+		{'to': 3, 'from': 5, 'afun': 'ena'}, # ...
+	]
+	'srl_links': [
+		{'to': 3, 'from': 5, 'afun': 'ACT'},
+		{'to': 7, 'from': 5, 'afun': 'PAT'}
+	]
+}
+```

corpusparser.egg-info/PKG-INFO

@@ -0,0 +1,10 @@
+Metadata-Version: 1.0
+Name: corpusparser
+Version: 0.0.1
+Summary: Parser for kres and ssj500k
+Home-page: UNKNOWN
+Author: Kristjan Voje
+Author-email: kristjan.voje@gmail.com
+License: MIT
+Description: UNKNOWN
+Platform: UNKNOWN

corpusparser.egg-info/SOURCES.txt

@@ -0,0 +1,5 @@
+corpusparser.egg-info/PKG-INFO
+corpusparser.egg-info/SOURCES.txt
+corpusparser.egg-info/dependency_links.txt
+corpusparser.egg-info/not-zip-safe
+corpusparser.egg-info/top_level.txt

corpusparser.egg-info/dependency_links.txt

@@ -0,0 +1 @@
+

corpusparser.egg-info/not-zip-safe

@@ -0,0 +1 @@
+

corpusparser.egg-info/top_level.txt

@@ -0,0 +1 @@
+

corpusparser/Parser.py

@@ -0,0 +1,196 @@
+from pathlib import Path
+import re
+import json
+from lxml import etree
+import logging
+
+logging.basicConfig(level=logging.INFO)
+
+# Read input file(.xml, .json; kres or ssj500k).  
+# Create an iterator that outputs resulting sentences (python dict format).  
+class Parser():
+
+    def __init__(self, corpus, infiles, logger=None):
+
+        if corpus == "kres":
+            self.kres_folder = Path(infiles[0])
+            self.kres_srl_folder = Path(infiles[1])
+        elif corpus == "ssj":
+            self.ssj_file = Path(infiles[0])
+        else:
+            raise ValueError("Argument corpus should be 'ssj' or 'kres'.")
+
+        self.corpus = corpus
+        self.W_TAGS = ['w']
+        self.C_TAGS = ['c']
+        self.S_TAGS = ['S', 'pc']
+        self.logger = logger or logging.getLogger(__name__)
+        self.stats = {
+            "parsed_count": 0,
+            "missing_srl": []
+        }
+
+    def parse_jos_links(self, sent_el):
+        if self.corpus == "kres":
+            return self.parse_jos_links_kres(sent_el)
+        else:
+            # 'syntax' is the linkgroup we're looking for
+            return self.parse_any_links_ssj(sent_el, "syntax")
+
+    def parse_jos_links_kres(self, sent_el):
+        lgrps = sent_el.findall(".//links")
+        if len(lgrps) < 1:
+            raise IOError("Can't find links.")
+        res_links = []
+        for link in lgrps[0]:
+            res_links += [{
+                "from": int(link.get("from").split(".")[-1]),
+                "afun": link.get("afun"),
+                "to": int(link.get("dep").split(".")[-1]),
+            }]
+        return res_links
+
+    def parse_ssj_target_arg(self, text):
+        # from: 0, to: 6
+        # <link ana="syn:modra" target="#ssj1.1.3 #ssj1.1.3.t6"/>
+        # from: 6, to: 7
+        # <link ana="syn:dol" target="#ssj1.1.3.t6 #ssj1.1.3.t7"/>
+        lst = [x.split(".")[-1] for x in text.split(" ")]
+        return [int(x[1:] if x[0] == "t" else 0) for x in lst]
+
+    def parse_any_links_ssj(self, sent_el, links_type):
+        lgrps = sent_el.findall(".//linkGrp")
+        links = [x for x in lgrps if x.get("type") == links_type][0]
+        res_links = []
+        for link in links:
+            tar = self.parse_ssj_target_arg(link.get("target"))
+            res_links += [{
+                "from": tar[0],
+                "afun": link.get("ana").split(":")[1],
+                "to": tar[1],
+            }]
+        return res_links
+
+    def parse_srl_links(self, sent_el, sent_srl_links=None):
+        if self.corpus == "kres":
+            return self.parse_srl_links_kres(sent_el, sent_srl_links)
+        else:
+            return self.parse_any_links_ssj(sent_el, "SRL")
+
+    def parse_srl_links_kres(self, sent_el, sent_srl_links):
+        res_links = []
+        for link in sent_srl_links:
+            res_links += [{
+                "from": int(link["from"]),
+                "afun": link["arg"],
+                "to": int(link["dep"]),
+            }]
+        # find the correspointing json file with srl links
+        return res_links
+
+    def sentence_generator(self):
+        # Using generators so we don't copy a whole corpu around in memory.
+        if self.corpus == "kres":
+            for xml_file in self.kres_folder.iterdir():
+                # self.parse_xml_file(xml_file)
+                yield from self.parse_xml_file(xml_file)
+        else:
+            yield from self.parse_xml_file(self.ssj_file)
+
+    def parse_xml_file(self, xml_file):
+        srl_from_json = {}
+        if self.corpus == "kres":  
+            # in case of kres, read the SRL links form a separate json file
+            file_id = xml_file.name.split(".")[0]
+            json_file = self.kres_srl_folder / Path(file_id).with_suffix(".srl.json")
+            with json_file.open("r") as fp:
+                srl_from_json = json.loads(fp.read())
+
+        with xml_file.open("rb") as fp:
+            # remove namespaces
+            bstr = fp.read()
+
+        utf8str = bstr.decode("utf-8")
+        utf8str = re.sub('\\sxmlns="[^"]+"', '', utf8str, count=1)
+        utf8str = re.sub(' xml:', ' ', utf8str)
+
+        root = etree.XML(utf8str.encode("utf-8"))
+
+        divs = []  # in ssj, there are divs, in Kres, there are separate files
+        if self.corpus == "kres":
+            divs = [root]
+        else:
+            divs = root.findall(".//div")
+
+        res_dict = {}
+
+        # parse divs
+        for div in divs:
+            f_id = div.get("id")
+
+            # parse paragraphs
+            for p in div.findall(".//p"):
+                p_id = p.get("id").split(".")[-1]
+
+                # parse sentences
+                for s in p.findall(".//s"):
+                    s_id = s.get("id").split(".")[-1]
+                    sentence_text = ""
+                    sentence_tokens = []
+
+                    # parse tokens
+                    for el in s.iter():
+                        if el.tag in self.W_TAGS:
+                            el_id = el.get("id").split(".")[-1]
+                            if el_id[0] == 't':
+                                el_id = el_id[1:]  # ssj W_TAG ids start with t
+                            sentence_text += el.text
+                            sentence_tokens += [{
+                                "word": True,
+                                "tid": int(el_id),
+                                "text": el.text,
+                                "lemma": el.get("lemma"),
+                                "msd": (el.get("msd") if self.corpus == "kres"
+                                    else el.get("ana").split(":")[-1]),
+                            }]
+                        elif el.tag in self.C_TAGS:
+                            # only Kres' C_TAGS have ids
+                            el_id = el.get("id") or "none"
+                            el_id = el_id.split(".")[-1]
+                            sentence_text += el.text
+                            sentence_tokens += [{
+                                "word": False,
+                                "tid": (int(el_id) if self.corpus == "kres" else -1),
+                                "text": el.text,
+                            }]
+                        elif el.tag in self.S_TAGS:
+                            # Kres' <S /> doesn't contain .text
+                            sentence_text += " "
+                        else:
+                            # pass links and linkGroups
+                            pass
+                    sentence_id = "{}.{}.{}".format(f_id, p_id, s_id)
+
+                    jos_links = self.parse_jos_links(s)
+
+                    if self.corpus == "kres":
+                        srl_links_raw = srl_from_json.get(sentence_id)
+                        if srl_links_raw is None:
+                            srl_links_parsed = None
+                            self.stats["missing_srl"] += [(sentence_id, sentence_text)]
+                        else:
+                            srl_links_parsed = self.parse_srl_links(s, srl_links_raw)
+                    else:
+                        srl_links_parsed = self.parse_srl_links(s)
+                        if len(srl_links_parsed) == 0:
+                            self.stats["missing_srl"] += [(sentence_id, sentence_text)]
+
+                    sentence_entry = {
+                        "sid": sentence_id,
+                        "text": sentence_text,
+                        "tokens": sentence_tokens,
+                        "jos_links": jos_links,
+                        "srl_links": srl_links_parsed
+                    }
+                    self.stats["parsed_count"] += 1
+                    yield (xml_file, sentence_entry)

corpusparser/__init__.py

@@ -0,0 +1 @@
+from corpusparser.Parser import Parser

corpusparser/main.py

@@ -0,0 +1,102 @@
+from pathlib import Path
+from corpusparser import Parser 
+import argparse
+import logging
+import json
+
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+## Main handles command line arguments and writing to files / DB.  
+
+def ssj_to_json_file(sentence_generator, outfolder):
+    # this funciton is based on the fact that files are parsed sequentially
+    outfolder = Path(outfolder)
+    outfolder.mkdir(parents=True, exist_ok=True)
+    outfile = outfolder / "ssj500k.json"
+
+    data_buffer = []
+    for s in sentence_generator:
+        sdata = s[1]
+        data_buffer += [sdata]
+
+    # outfile = Path(outfile)
+    with outfile.open("w") as fp:
+        logger.info("Writing to {}".format(outfile))
+        json.dump(data_buffer, fp)
+
+def kres_to_json_files(sentence_generator, outfolder):
+    outfolder = Path(outfolder) / "kres_json"
+    outfolder.mkdir(parents=True, exist_ok=True)
+
+    def write_buffer_to_file(outfile, outfile_buffer):
+        logger.info("Writing file: {}".format(outfile))
+        with outfile.open("w") as fp:
+            json.dump(outfile_buffer, fp)
+
+    outfile_buffer = None
+    current_outfile = None
+    for s in sentence_generator:
+        infile = s[0]
+        outfile = outfolder / Path(infile.name.split(".")[0]).with_suffix(".json")
+
+        # parser sequentially parses files; when we're done with a file, write it out
+        if current_outfile is None:
+            current_outfile = outfile
+            outfile_buffer = [] 
+        elif outfile != current_outfile:
+            write_buffer_to_file(current_outfile, outfile_buffer)
+            current_outfile = outfile
+            outfile_buffer = []
+
+        # update buffer
+        sdata = s[1]
+        outfile_buffer += [sdata]
+    write_buffer_to_file(current_outfile, outfile_buffer)
+
+
+
+def to_db():
+    return "TODO"
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Parsing corpora kres and ssj500k.")
+    parser.add_argument('--kres-folder', required=True)
+    parser.add_argument('--kres-srl-folder', required=True)
+    parser.add_argument('--ssj-file', required=True)
+    parser.add_argument('--output', required=False, default=None)
+    parser.add_argument('--outdir', required=False, default=None)
+    parser.add_argument('--dbaddr', required=False, default=None)
+    args = parser.parse_args()
+
+    # parse ssj
+    logger.info("Parsing ssj500k: {}".format(args.ssj_file))
+    ssj_parser = Parser(
+        corpus="ssj",
+        infiles=[args.ssj_file],
+    )
+    # res = [x[1]["sid"] for x in ssj_parser.sentence_generator()]
+    # logger.info("Parsed {} sentences (ssj500k)".format(len(res)))
+
+    # ssj to json
+    ssj_to_json_file(ssj_parser.sentence_generator(), args.outdir)
+
+    # parse kres
+    logger.info("Parsing Kres: {}".format(args.ssj_file))
+    kres_parser = Parser(
+        corpus="kres",
+        infiles=[args.kres_folder, args.kres_srl_folder],
+    )
+    # res = [x[1]["sid"] for x in kres_parser.sentence_generator()]
+    # logger.info("Parsed {} sentences (kres)".format(len(res)))
+
+    # kres to json
+    kres_to_json_files(kres_parser.sentence_generator(), args.outdir)
+
+
+## Handling output is situational --- implement it outside of Parser.
+## Parser returns tuples (orig_file, element)
+# 1. parse per-file and output to file (JSON)
+# 2. parse and save to DB
+
+# TODO

dockerfiles/cjvt-corpusparser-env/Dockerfile

@@ -0,0 +1,15 @@
+FROM ubuntu:16.04
+
+RUN apt-get update --fix-missing
+RUN apt-get install -y \
+vim \
+python3 \
+python3-pip \
+sshfs
+
+RUN pip3 install \
+	lxml \
+	argparse \
+	pathlib
+
+ENV PYTHONIOENCODING UTF-8

dockerfiles/cjvt-corpusparser-env/Makefile

@@ -0,0 +1,19 @@
+IMAGE_NAME="cjvt-corpusparser-env"
+
+all: build run
+
+build:
+	docker build . -t $(IMAGE_NAME)
+
+run:
+	docker run \
+    	-it \
+	-v /home/${USER}:/home/${USER} \
+	--user $(shell id -u):$(shell id -g) \
+	-v /etc/passwd:/etc/passwd \
+	-v /etc/group:/etc/group \
+    	-v $(MAKE_ROOT):/project \
+	-w /project \
+	$(IMAGE_NAME) \
+    /bin/bash
+

setup.py

@@ -0,0 +1,10 @@
+from setuptools import setup
+
+setup(name='corpusparser',
+  version='0.0.1',
+  description=u"Parser for kres and ssj500k",
+  author=u"Kristjan Voje",
+  author_email='kristjan.voje@gmail.com',
+  license='MIT',
+  zip_safe=False,
+)