diff --git a/Makefile b/Makefile index 51a0afe..790cbf0 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)/workdir/outputfolder" +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/src/pkg/corpusparser/corpusparser/Parser.py b/src/pkg/corpusparser/corpusparser/Parser.py index 847fa69..476a87c 100644 --- a/src/pkg/corpusparser/corpusparser/Parser.py +++ b/src/pkg/corpusparser/corpusparser/Parser.py @@ -10,7 +10,7 @@ logging.basicConfig(level=logging.INFO) # Create an iterator that outputs resulting sentences (python dict format). class Parser(): - def __init__(self, corpus, infiles, logger=None): + def __init__(self, corpus, infiles, output=None, outdir=None, dbaddr=None, logger=None): if corpus == "kres": self.kres_folder = Path(infiles[0]) @@ -20,11 +20,22 @@ class Parser(): else: raise ValueError("Argument corpus should be 'ssj' or 'kres'.") + self.output = output # None | file | db + if self.output == "file": + self.outdir = Path(outdir) + self.outdir.mkdir(parents=True, exist_ok=True) + elif self.output == "db": + self.dbaddr = "TODO" + 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": @@ -67,7 +78,7 @@ 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: @@ -84,22 +95,23 @@ class Parser(): # find the correspointing json file with srl links 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 @@ -155,7 +167,7 @@ class Parser(): sentence_text += el.text sentence_tokens += [{ "word": False, - "tid": int(el_id), + "tid": (int(el_id) if self.corpus == "kres" else -1), "text": el.text, }] elif el.tag in self.S_TAGS: @@ -166,23 +178,26 @@ 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 - if srl_links is None: - srl_links_parsed = None + + 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, srl_links) - res_dict[sentence_id] = { + 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 } - if srl_links is None: - self.logger.info("srl_links missing:{}:{}".format( - sentence_id, res_dict[sentence_id]["text"])) - return res_dict + self.stats["parsed_count"] += 1 + yield (xml_file, sentence_entry) diff --git a/src/pkg/corpusparser/corpusparser/main.py b/src/pkg/corpusparser/corpusparser/main.py new file mode 100644 index 0000000..cb23c8c --- /dev/null +++ b/src/pkg/corpusparser/corpusparser/main.py @@ -0,0 +1,58 @@ +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. + +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], + output=args.output, + outdir=args.outdir, + ) + res = [x[1]["sid"] for x in ssj_parser.sentence_generator()] + logger.info("Parsed {} sentences (ssj500k)".format(len(res))) + + # parse kres + logger.info("Parsing Kres: {}".format(args.ssj_file)) + kres_parser = Parser( + corpus="kres", + infiles=[args.kres_folder, args.kres_srl_folder], + output=args.output, + outdir=args.outdir, + ) + res = [x[1]["sid"] for x in kres_parser.sentence_generator()] + logger.info("Parsed {} sentences (kres)".format(len(res))) + + +## 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 +def handle_output(self, sent_ent, xml_file): + if self.output is None: + pass + if self.output == "file": + outfile = Path(self.outdir) / Path(xml_file.name.split(".")[0]).with_suffix(".json") + with outfile.open("a") as fp: + print(sent_ent) + json.dumps(sent_ent, fp) + 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()