You can deploy this model and start earning money today!

ilsp/
justice

fasttext, flair

text-classification

eupl-1.1

Views

No views yet

ilsp/justice

Models for processing Greek court decisions

Paragraph classification

python

repo_id = "ilsp/justice"
model_path = hf_hub_download(repo_id=repo_id, filename="20250105-court_decisions_paragraph_classifier.ftz")
sample_decision = hf_hub_download(repo_id=repo_id, filename="sample_data/Α2485_2023.txt") # anonymized decision
model = load_model(model_path)
labels_map =  {
  'preamble': '__label__0', '__label__0': 'preamble',
  'panel': '__label__1', '__label__1': 'panel',
  'litigants': '__label__2', '__label__2': 'litigants',
  'justification': '__label__3', '__label__3': 'justification',
  'decision': '__label__4', '__label__4': 'decision',
  'post': '__label__5', '__label__5': 'post'}

with open(sample_decision) as inf:
    paras = [p for p in inf.read().split(NL) if p.strip()]
    random.shuffle(paras)
    text = NL.join(paras)

nchars = 150
for line in text.split(NL):
    pred = labels_map[model.predict(line.strip())[0][0]]
    if len(line) > nchars:
        line = line[0:nchars]
    print(f"{line} -> {pred}")    

Named entity recognition for anonymization in court decisions

python

from flair.data import Sentence, Token
from flair.models import SequenceTagger
from huggingface_hub import hf_hub_download

REPO_ID = "ilsp/justice"
MODEL_PATH = "decisions-ner-model.pt"
model_path = hf_hub_download(repo_id=REPO_ID, filename=MODEL_PATH)        
model = SequenceTagger.load(model_path)

text = "Για να δικάσει την από 30 Μαρτίου 2020 έφεση των 1) Νίκης Νικίδου του Νίκου , κατοίκου Νίκαιας ( Νεάπολης 1 ) , 2) Άννας Άννίδου του Άνθιμου , κατοίκου Αθήνας ( Αγράμπελης 1 ) και 3) Σοφίας Σοφίδου του Σοφοκλή , κατοίκου Στυλίδας ( Στρυμώνος 1 ) , οι οποίοι παρέστησαν με τον δικηγόρο Λυσία Λυσίου ( Α.Μ. 12341 ) , που τον διόρισαν με πληρεξούσιο ."
sentence = Sentence([Token(t) for t in text.split()]) # or use a sentence splitter
model.predict(sentence)
sentence.get_spans("ner")

[Span[11:13]: "Νίκης Νικίδου" → PERSON (1.0000),
 Span[14:15]: "Νίκου" → PERSON (1.0000),
 Span[19:21]: "Νεάπολης 1" → FAC (1.0000),
 Span[24:26]: "Άννας Άννίδου" → PERSON (1.0000),
 Span[27:28]: "Άνθιμου" → PERSON (1.0000),
 Span[32:34]: "Αγράμπελης 1" → FAC (1.0000),
 Span[37:39]: "Σοφίας Σοφίδου" → PERSON (1.0000),
 Span[40:41]: "Σοφοκλή" → PERSON (1.0000),
 Span[45:47]: "Στρυμώνος 1" → FAC (1.0000)]