diff --git a/.gitignore b/.gitignore index c57e05a..43fb148 100644 --- a/.gitignore +++ b/.gitignore @@ -1,3 +1,3 @@ data/samples/ -*/__pycache__/ *egg-info/ +*.pyc diff --git a/Makefile b/Makefile index 51a0afe..7734cb5 100644 --- a/Makefile +++ b/Makefile @@ -11,6 +11,9 @@ MAKE_ROOT = $(shell pwd) SSJ_FILE = "$(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 = "" export .PHONY: dev-env preflight @@ -30,5 +33,5 @@ data/samples: # when debugging, run this once, then run python3 ... by hand preflight: data/samples pip3 install -e src/pkg/corpusparser/. - python3 src/preflight/main_parse.py --kres-folder $(KRES_FOLDER) \ - --ssj-file $(SSJ_FILE) --kres-srl-folder $(KRES_SRL_FOLDER) + python3 src/pkg/corpusparser/corpusparser/main.py --kres-folder $(KRES_FOLDER) \ + --ssj-file $(SSJ_FILE) --kres-srl-folder $(KRES_SRL_FOLDER) --output $(OUTPUT) --outdir $(OUTDIR) --dbaddr $(DBADDR) diff --git a/data/debugging/dbg.txt b/data/debugging/dbg.txt new file mode 100644 index 0000000..2b6909d --- /dev/null +++ b/data/debugging/dbg.txt @@ -0,0 +1,6 @@ +F0034713.20.1": [{"dep": "7", "arg": "REC", "from": "9"}, {"dep": "10", "arg": "ACT", "from": "9"}, {"dep": "13", "arg": "MWPRED", "from": "12"}, {"dep": "18", "arg": "MANN", "from": "19"}, {"dep": "20", "arg": "LOC", "from": "19"}] + +Sentence: +F0034713.20.0 +1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 +Sodobni ali preprosto neosveščeni potrošnik vse prerad zavrže stvar, ki se je malenkostno pokvarila in bi se jo zlahka dalo popraviti. \ No newline at end of file diff --git a/data/kres_srl b/data/kres_srl index f1acfc4..465d987 120000 --- a/data/kres_srl +++ b/data/kres_srl @@ -1 +1 @@ -/home/kristjan/kres_srl/final_json/ \ No newline at end of file +/home/voje/work_data/final_json \ No newline at end of file diff --git a/data/kres_srl_ikt b/data/kres_srl_ikt new file mode 120000 index 0000000..465d987 --- /dev/null +++ b/data/kres_srl_ikt @@ -0,0 +1 @@ +/home/voje/work_data/final_json \ No newline at end of file diff --git a/data/kres_srl_t420 b/data/kres_srl_t420 new file mode 120000 index 0000000..f1acfc4 --- /dev/null +++ b/data/kres_srl_t420 @@ -0,0 +1 @@ +/home/kristjan/kres_srl/final_json/ \ No newline at end of file diff --git a/src/pkg/corpusparser/corpusparser/Parser.py b/src/pkg/corpusparser/corpusparser/Parser.py index fb8b408..d908d04 100644 --- a/src/pkg/corpusparser/corpusparser/Parser.py +++ b/src/pkg/corpusparser/corpusparser/Parser.py @@ -1,14 +1,16 @@ -from corpusparser import Sentence 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): + def __init__(self, corpus, infiles, logger=None): if corpus == "kres": self.kres_folder = Path(infiles[0]) @@ -22,6 +24,11 @@ class Parser(): 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": @@ -64,37 +71,40 @@ class Parser(): }] return res_links - def parse_srl_links(self, sent_el, sent_srl_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): - print("HA") - if len(sent_srl_links) == 0: - print("HI") - return [] - print(sent_srl_dict) + 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 [] + return res_links - def parse(self): + 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) - break # TODO dev break + # self.parse_xml_file(xml_file) + yield from self.parse_xml_file(xml_file) else: - self.parse_xml_file(self.ssj_file) + yield from self.parse_xml_file(self.ssj_file) def parse_xml_file(self, xml_file): - srl_dict = {} + 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_dict = json.loads(fp.read()) + srl_from_json = json.loads(fp.read()) with xml_file.open("rb") as fp: # remove namespaces @@ -112,7 +122,7 @@ class Parser(): else: divs = root.findall(".//div") - res_dict = [] # TODO: try making an iterator instead + res_dict = {} # parse divs for div in divs: @@ -150,7 +160,7 @@ class Parser(): sentence_text += el.text sentence_tokens += [{ "word": False, - "tid": el_id, + "tid": (int(el_id) if self.corpus == "kres" else -1), "text": el.text, }] elif el.tag in self.S_TAGS: @@ -161,22 +171,26 @@ class Parser(): pass sentence_id = "{}.{}.{}".format(f_id, p_id, s_id) - # make a generator instead of holding the whole corpus in memory - if sentence_id in res_dict: - raise KeyError("duplicated id: {}".format(sentence_id)) jos_links = self.parse_jos_links(s) - srl_links = srl_dict.get(sentence_id) if self.corpus == "kres" else None - srl_links_fixed = self.parse_srl_links(s, srl_links) - print(srl_links) - res_dict[sentence_id] = { + + 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": "BBBB", - "srl_links": "AAAAA", + "jos_links": jos_links, + "srl_links": srl_links_parsed } - - print(res_dict[sentence_id]) - print("------------------------------------------------- END") - return # TODO dev break - return res_dict + self.stats["parsed_count"] += 1 + yield (xml_file, sentence_entry) diff --git a/src/pkg/corpusparser/corpusparser/README.md b/src/pkg/corpusparser/corpusparser/README.md new file mode 100644 index 0000000..0d41d60 --- /dev/null +++ b/src/pkg/corpusparser/corpusparser/README.md @@ -0,0 +1,38 @@ +# corpusparser +A tool for parsing ssj500k and Kres into a unified .json format. + +## 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'} + ] +} +``` \ No newline at end of file diff --git a/src/pkg/corpusparser/corpusparser/Sentence.py b/src/pkg/corpusparser/corpusparser/Sentence.py deleted file mode 100644 index aba9a4a..0000000 --- a/src/pkg/corpusparser/corpusparser/Sentence.py +++ /dev/null @@ -1,3 +0,0 @@ -class Sentence(): - def __init__(): - print("Sentence __init__(): TODO") diff --git a/src/pkg/corpusparser/corpusparser/__init__.py b/src/pkg/corpusparser/corpusparser/__init__.py index 213c88a..d993b8c 100644 --- a/src/pkg/corpusparser/corpusparser/__init__.py +++ b/src/pkg/corpusparser/corpusparser/__init__.py @@ -1,2 +1 @@ -from corpusparser.Parser import Parser -from corpusparser.Sentence import Sentence \ No newline at end of file +from corpusparser.Parser import Parser \ No newline at end of file diff --git a/src/pkg/corpusparser/corpusparser/__pycache__/Parser.cpython-35.pyc b/src/pkg/corpusparser/corpusparser/__pycache__/Parser.cpython-35.pyc deleted file mode 100644 index e438b67..0000000 Binary files a/src/pkg/corpusparser/corpusparser/__pycache__/Parser.cpython-35.pyc and /dev/null differ diff --git a/src/pkg/corpusparser/corpusparser/__pycache__/Sentence.cpython-35.pyc b/src/pkg/corpusparser/corpusparser/__pycache__/Sentence.cpython-35.pyc deleted file mode 100644 index c0ae9f7..0000000 Binary files a/src/pkg/corpusparser/corpusparser/__pycache__/Sentence.cpython-35.pyc and /dev/null differ diff --git a/src/pkg/corpusparser/corpusparser/__pycache__/__init__.cpython-35.pyc b/src/pkg/corpusparser/corpusparser/__pycache__/__init__.cpython-35.pyc deleted file mode 100644 index a365c8e..0000000 Binary files a/src/pkg/corpusparser/corpusparser/__pycache__/__init__.cpython-35.pyc and /dev/null differ diff --git a/src/pkg/corpusparser/corpusparser/main.py b/src/pkg/corpusparser/corpusparser/main.py new file mode 100644 index 0000000..e0ba065 --- /dev/null +++ b/src/pkg/corpusparser/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 diff --git a/src/preflight/main_parse.py b/src/preflight/main_parse.py deleted file mode 100644 index 35bb2bb..0000000 --- a/src/preflight/main_parse.py +++ /dev/null @@ -1,25 +0,0 @@ -from corpusparser import Parser -import argparse - -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) - args = parser.parse_args() - - # parse ssj - """ - ssj_parser = Parser( - corpus="ssj", - infiles=[args.ssj_file] - ) - ssj_parser.parse() - """ - - # parse kres - kres_parser = Parser( - corpus="kres", - infiles=[args.kres_folder, args.kres_srl_folder] - ) - kres_parser.parse()