基于Python构建一个高效词汇表

from collections import counter
import pandas as pd
from tqdm import tqdm
import numpy as np

def process_line(line_vocab, len_size):
    """
    处理单行数据，构建局部词汇表
    
    参数:
        line_vocab (str): 输入文本行
        len_size (int): n-gram长度
        
    返回:
        counter: 词频统计对象
    """
    local_vocab = counter()
    
    # 单字统计
    if len_size == 1:
        local_vocab.update(counter(list(line_vocab)))
    
    # 双字组合统计    
    elif len_size == 2:
        vocab_data = np.array([list(line_vocab[:-1]), list(line_vocab[1:])])
        vocab_data = vocab_data[0,:] + vocab_data[1,:]
        local_vocab.update(counter(vocab_data.tolist()))
    
    # 三字组合统计    
    elif len_size == 3:
        vocab_data = np.array([list(line_vocab[:-2]), 
                              list(line_vocab[1:-1]), 
                              list(line_vocab[2:])])
        vocab_data = vocab_data[0,:] + vocab_data[1,:] + vocab_data[2,:]
        local_vocab.update(counter(vocab_data.tolist()))
    
    # 四字组合统计    
    elif len_size == 4:
        vocab_data = np.array([list(line_vocab[:-3]), 
                              list(line_vocab[1:-2]), 
                              list(line_vocab[2:-1]), 
                              list(line_vocab[3:])])
        vocab_data = vocab_data[0,:] + vocab_data[1,:] + vocab_data[2,:] + vocab_data[3,:]
        local_vocab.update(counter(vocab_data.tolist()))
    
    # 五字组合统计    
    elif len_size == 5:
        vocab_data = np.array([list(line_vocab[:-4]), 
                              list(line_vocab[1:-3]), 
                              list(line_vocab[2:-2]), 
                              list(line_vocab[3:-1]), 
                              list(line_vocab[4:])])
        vocab_data = vocab_data[0,:] + vocab_data[1,:] + vocab_data[2,:] + vocab_data[3,:] + vocab_data[4,:]
        local_vocab.update(counter(vocab_data.tolist()))

    del line_vocab  # 显式释放内存
    return local_vocab

# 加载预处理数据
lines = pd.read_pickle("pretrain_hq.pkl")

# 初始化全局词表
global_vocab = counter()

# 逐行处理
for line in tqdm(lines):
    global_vocab.update(process_line(line, 1))

# 保存结果
pd.to_pickle(global_vocab, "vocab_{}.pkl".format(1))

from multiprocessing import pool

def parallel_process(lines, len_size):
    with pool() as pool:
        results = pool.starmap(process_line, [(line, len_size) for line in lines])
    return sum(results, counter())

from flask import flask, request
import pandas as pd

app = flask(__name__)
vocab = pd.read_pickle("vocab_1.pkl")

@app.route('/analyze', methods=['post'])
def analyze():
    text = request.json['text']
    result = {word: vocab[word] for word in text.split()}
    return jsonify(result)