kleister-nda/run.ipynb

import csv
import re
from collections import Counter
from datetime import datetime
import datefinder
import word2number

def most_frequent(List, howmany=1):
    counter = Counter(List)
    return counter.most_common(1)

def get_jurisdiction(text):
    us_states= r"(Alabama|Alaska|Arizona|Arkansas|California|Colorado|Connecticut|Delaware|Florida|Georgia|Hawaii|Idaho|Illinois|Indiana|Iowa|Kansas|Kentucky|Louisiana|Maine|Maryland|Massachusetts|Michigan|Minnesota|Mississippi|Missouri|Montana|Nebraska|Nevada|New\sHampshire|New\sJersey|New\sMexico|New\sYork|North\sCarolina|North\sDakota|Ohio|Oklahoma|Oregon|Pennsylvania|Rhode\sIsland|South\sCarolina|South\sDakota|Tennessee|Texas|Utah|Vermont|Virginia|Washington|West\sVirginia|Wisconsin|Wyoming)"

    matches = re.findall(us_states, text, re.MULTILINE | re.IGNORECASE)
    result = most_frequent(matches)

    if result:
        return result[0][0].replace(" ", "_")

def get_parties(text):
    company_regex = r"(([A-Z][A-za-z]+,?\s)+(Inc\.|LLC|Ltd\.|Company|Corporation|INC\.|LTD\.|COMPANY|CORPORATION|Bank|Com|Council|Technology|Systems))"
    regex2 = r"([A-Z][a-z]+\s[A-Z]\.\s[A-Z][a-z]+)"

    matches = re.findall(company_regex, text, re.MULTILINE)
    matches = [m[0] for m in matches]
    
    regex2 = re.findall(regex2, text, re.MULTILINE)

    companies = []
    for m in matches:
        splitted = [w.capitalize().rstrip(",") for w in m.split()]
        companies.append(" ".join(splitted))

    result = most_frequent(companies,2)

    
    if len(result) < 2:
        result.extend(most_frequent(regex2, 2-len(result)))   

    if result:
        return [x[0].replace(" ", "_") for x in result]

def get_date(text):
    matches = []
    df_matches = datefinder.find_dates(text)
    
    while True:
        try:
            m = next(df_matches)
        except StopIteration:
            break
        except:
            continue
        matches.append(m)

    matches = filter(lambda x: 2022 > x.year > 1950, matches)

    dates = [x.strftime("%Y-%m-%d") for x in matches]

    result = most_frequent(dates)

    if len(result) == 0:
        return None
    else:
        return result[0][0]

def get_term(text):
    term_regex = r"\b([\w()]*)\s(months?|years?)\b"

    match = list(re.finditer(term_regex, text, re.MULTILINE))

    if match:
        number, unit = match[0].groups()
    else:
        return None

    if m := re.match(r"\d+", number):
        number = m.group()
    else:
        try:
            number = w2n.word_to_num(re.match(r"\b\w+\b", number).group())
        except:
            return None

    return str(number) + "_" + unit

def run(text, needed_info):
    jurisdiction, date, term, parties = None, None, None, None
    
    if "jurisdiction" in needed_info:
        jurisdiction = get_jurisdiction(text)
    if "effective_date" in needed_info:
        date = get_date(text)
    if "term" in needed_info:
        term = get_term(text)
    if "party" in needed_info:
        parties = get_parties(text)

    result_dict = {}

    if date:
        result_dict["effective_date"] = date
    if jurisdiction:
        result_dict["jurisdiction"] = jurisdiction
    if term:
        result_dict["term"] = term

    result_str = " ".join([f"{k}={v}" for k,v in result_dict.items()])

    if parties:
        for p in parties :
            result_str += f" party={p}"

    return result_str

filenames=[('dev-0/in.tsv',"dev-0/out.tsv"), ('train/in.tsv', "train/out.tsv"), ('test-A/in.tsv', 'test-A/out.tsv')]
for filename in filenames:
    with open(filename[0], 'r', encoding="utf-8") as in_file,\
                open(filename[1], "w") as out_file:
            reader = csv.reader(in_file, delimiter='\t', quoting=csv.QUOTE_NONE)
            for item in reader:
                needed_info = item[1].strip().split()
                text = item[2].replace("\\\\n", " ").replace("\\\\f", " ").replace("\\\\t", " ").strip()
                extracted = run(text, needed_info).replace(":", "_")
                out_file.write(extracted + "\n")