en-ner-conll-2003/solution.py

import pandas as pd
import numpy as np
import csv
import os.path
import shutil
import torch
from tqdm import tqdm
from itertools import islice
from sklearn.model_selection import train_test_split
from torchtext.vocab import Vocab
from collections import Counter
from nltk.tokenize import word_tokenize
import gensim.downloader as api
from gensim.models.word2vec import Word2Vec

class NERModel(torch.nn.Module):

    def __init__(self,):
        super(NERModel, self).__init__()
        self.emb = torch.nn.Embedding(23628,200)
        self.fc1 = torch.nn.Linear(600,9)
        

    def forward(self, x):
        x = self.emb(x)
        x = x.reshape(600) 
        x = self.fc1(x)
        return x

def process_output(lines):
    result = []
    for line in lines:
        last_label = None
        new_line = []
        for label in line:
            if(label != "O" and label[0:2] == "I-"):
                if last_label == None or last_label == "O":
                    label = label.replace('I-', 'B-')
                else:
                    label = "I-" + last_label[2:]
            last_label = label
            new_line.append(label)
            x = (" ".join(new_line))
        result.append(" ".join(new_line))
    return result

def build_vocab(dataset):
    counter = Counter()
    for document in dataset:
        counter.update(document)
    return Vocab(counter, specials=['<unk>', '<pad>', '<bos>', '<eos>'])

def data_process(dt):
    return [ torch.tensor([vocab['<bos>']] +[vocab[token]  for token in  document ] + [vocab['<eos>']], dtype = torch.long) for document in dt]

def labels_process(dt):
    return [ torch.tensor([0] + document + [0], dtype = torch.long) for document in dt]

def predict(input_tokens, labels):

  results = []
  
  for i in range(len(input_tokens)):
    line_results = []
    for j in range(1, len(input_tokens[i]) - 1):
        x = input_tokens[i][j-1: j+2].to(device_gpu)
        predicted = ner_model(x.long())
        result = torch.argmax(predicted)
        label = labels[result]
        line_results.append(label)
    results.append(line_results)

  return results

train = pd.read_csv('train/train.tsv.xz', sep='\t', names=['a', 'b'])

labels = ['O','B-LOC', 'I-LOC','B-MISC', 'I-MISC', 'B-ORG', 'I-ORG', 'B-PER', 'I-PER'] 
train["a"]=train["a"].apply(lambda x: [labels.index(y) for y in  x.split()])
train["b"]=train["b"].apply(lambda x: x.split())

vocab = build_vocab(train['b'])

  tensors = []

  for sent in train["b"]:
    sent_tensor = torch.tensor(())
    for word in sent:
      temp = torch.tensor([word[0].isupper(), word[0].isdigit()])
      sent_tensor = torch.cat((sent_tensor, temp))

    tensors.append(sent_tensor)

device_gpu = torch.device("cuda:0")
ner_model = NERModel().to(device_gpu)
criterion = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(ner_model.parameters())

train_labels = labels_process(train['a'])
train_tokens_ids = data_process(train['b'])

train_tensors = [torch.cat((token, tensors[i])) for i, token in enumerate(train_tokens_ids)]

for epoch in range(5):
    acc_score = 0
    prec_score = 0
    selected_items = 0
    recall_score = 0
    relevant_items = 0
    items_total = 0
    ner_model.train()
    for i in range(len(train_labels)):
        for j in range(1, len(train_labels[i]) - 1):
            X = train_tensors[i][j - 1: j + 2].to(device_gpu)

            Y = train_labels[i][j: j + 1].to(device_gpu)

            Y_predictions = ner_model(X.long())

            acc_score += int(torch.argmax(Y_predictions) == Y)
            if torch.argmax(Y_predictions) != 0:
                selected_items += 1
            if torch.argmax(Y_predictions) != 0 and torch.argmax(Y_predictions) == Y.item():
                prec_score += 1
            if Y.item() != 0:
                relevant_items += 1
            if Y.item() != 0 and torch.argmax(Y_predictions) == Y.item():
                recall_score += 1

            items_total += 1
            optimizer.zero_grad()
            loss = criterion(Y_predictions.unsqueeze(0), Y)
            loss.backward()
            optimizer.step()

    precision = prec_score / selected_items
    recall = recall_score / relevant_items
    f1_score = (2 * precision * recall) / (precision + recall)
    print(f'epoch: {epoch}')
    print(f'f1: {f1_score}')
    print(f'acc: {acc_score / items_total}')

def create_tensors_list(data):
  tensors = []

  for sent in data["a"]:
    sent_tensor = torch.tensor(())
    for word in sent:
      temp = torch.tensor([word[0].isupper(), word[0].isdigit()])
      sent_tensor = torch.cat((sent_tensor, temp))

    tensors.append(sent_tensor)

  return tensors

dev = pd.read_csv('dev-0/in.tsv', sep='\t', names=['a'])
dev["a"] = dev["a"].apply(lambda x: x.split())

dev_tokens_ids = data_process(dev["a"])

dev_extra_tensors = create_tensors_list(dev)

dev_tensors = [torch.cat((token, dev_extra_tensors[i])) for i, token in enumerate(dev_tokens_ids)]

results = predict(dev_tensors, labels)
results_processed = process_output(results)

with open("dev-0/out.tsv", "w") as f:
  for line in results_processed:
    f.write(line + "\n")

test = pd.read_csv('test-A/in.tsv', sep='\t', names=['a'])
test["a"] = test["a"].apply(lambda x: x.split())

test_tokens_ids = data_process(test["a"])

test_extra_tensors = create_tensors_list(test)

test_tensors = [torch.cat((token, test_extra_tensors[i])) for i, token in enumerate(test_tokens_ids)]

results = predict(test_tensors, labels)
results_processed = process_output(results)

with open("test-A/out.tsv", "w") as f:
  for line in results_processed:
    f.write(line + "\n")

model_path = "seq_labeling.model"
torch.save(ner_model.state_dict(), model_path)