BIDI_ML_INFO_EXTRACTION/BiddingKG/deepdive/udf/supervise_guest_BiddingAgency.py

#!/usr/bin/env python
#encoding:utf-8
from deepdive import *
import random
from collections import namedtuple
import logging
from commonutil import *

Label = namedtuple('Label', 'entity_id, label, type')

@tsv_extractor
@returns(lambda
        entity_id   = "text",
        label   = "int",
        rule_id = "text",
    :[])
# heuristic rules for finding positive/negative examples of transaction relationship mentions
def supervise(
        entity_id="text", entity_begin="int", entity_end="int",
        doc_id="text", sentence_index="int", sentence_text="text",
        tokens="text[]",  pos_tags="text[]", ner_tags="text[]",
    ):

    # Constants
    label = Label(entity_id=entity_id,label=None,type=None)
    Bidding = frozenset(["招标"])
    Agency = frozenset(["代理"])
    UNIT = frozenset(["单位","机构"])
    R_append = frozenset(["作为","以下","简称"])
    MAX_DIST = 15
    TYPE_MENTION = frozenset(["org","company"])

    # Common data objects
    if entity_begin>MAX_DIST:
        begin = entity_begin-MAX_DIST
    else:
        begin = 0
    if len(tokens)-entity_end>MAX_DIST:
        end = entity_end+MAX_DIST
    else:
        end = -1
    front_tokens = tokens[begin:entity_begin]
    end_tokens = tokens[entity_end:end]
    front_ner = ner_tags[begin:entity_begin]
    end_ner = ner_tags[end:entity_end]
    
    logging.info(sentence_text)

    # Rule: Candidates that are too far apart
    
    
    inter_bidd_front = Bidding.intersection(front_tokens)
    inter_unit_front = UNIT.intersection(front_tokens)
    inter_agent_front = Agency.intersection(front_tokens)
    inter_bidd_end = Bidding.intersection(end_tokens)
    inter_unit_end = UNIT.intersection(end_tokens)
    inter_agent_end = Agency.intersection(end_tokens)
    
    if len(inter_bidd_front)>0 and len(inter_unit_front)>0:
        if len(inter_agent_front)>0:
            if len(R_append.intersection(end_tokens))==0:
                index = find_index(combine(combine(list(inter_bidd_front),list(inter_agent_front)),list(inter_unit_front)),''.join(front_tokens))
                if index>=0:
                    if len(TYPE_MENTION.intersection(front_ner[index:entity_begin]))>0:
                        yield label._replace(label=-1,type="pos:third entity between")
                    else:
                        yield label._replace(label=1,type="招标代理机构:entity")
                else:
                    yield label._replace(label=-1,type="pos:no match")
        else:
            yield label._replace(label=-1,type="pos:非代理机构")
    elif len(inter_bidd_end)>0 and len(inter_unit_end)>0:
        if len(inter_agent_end)>0:
            if len(R_append.intersection(end_tokens))>0:
                index = find_index(combine(combine(list(inter_bidd_end),list(inter_agent_end)),list(inter_unit_end)),''.join(end_tokens))
                if index>=0:
                    if len(TYPE_MENTION.intersection(end_ner[entity_end+1:index]))>0:
                        yield label._replace(label=-1,type="pos:third entity between")
                    else:
                        yield label._replace(label=1,type="entity:招标代理机构")
                else:
                    yield label._replace(label=-1,type="pos:no match")
        else:
            yield label._replace(label=-1,type="pos:非代理机构")
    
    else:
        yield label._replace(label=-1,type="no match")