python - 如何将扫描文本中的数据导入 Django 模型

Question

我有数百页的“测验”问题、多项选择以及相关的答案和解释。我正在尝试创建一个简单的 Django 应用程序来管理这些问题。我创建了一个简单但有效的 Python 解析器来将扫描的 OCR 页面解析为正确的对象。

我想要一个“实用程序”来让这个 Django 应用程序的管理员能够将 OCR 纸上的测验内容导入到相关的 Django 数据库表中。这将是一项罕见的任务，而且不一定适合包含在 Web UI 中。

我问过有关使用中间 JSON/YAML 固定装置的问题，并被告知更合适的方法是直接创建和保存我的模型 [1] 的实例。然后我尝试按照 [2] 和 [3] 建议的方式创建一个独立的脚本，但无法克服kwargs = {"app_label": model_module.__name__.split('.')[-2]}
IndexError:列表索引超出范围错误。

我还遇到了关于创建自定义 django-admin.py/manage.py 命令的 [4]。这似乎是处理任务的逻辑上合适的方式；但是，我很想听听那些有更多经验和头脑的人的意见(我已经把我的都吃光了:)。

引用资料:

1. Importing data from scanned text into Django as YAML fixture or SQL
2. what is the simplest way to create a table use django db api ,and base on 'Standalone Django scripts'
3. Standalone Scripts
4. Writing custom django-admin commands

例子:

• OCR 文本

Page 12 34. Hiedegger is a _____ . (a) philosopher (b) boozy beggar (c) both a and b (d) none of these 35. ...

• Django 模型

  class Question(models.Model):
      text = models.TextField()
  class Choice(models.Model):
      question = models.ForeignKey(Question)
      order = models.IntegerField(default=1)
      text = models.TextField()
  

• 目标，像这样...

  q = Question.objects.create(text="Hiedegger is a _____ .")
  q.save()
  c = Choice(text="philosopher", order=1, question=q.pk)
  c.save()
  

Answer 1

这是我想出的工作版本。肮脏，但有效。 @akonsu 和@Ivan Kharlamov 都提供了帮助。谢谢...

import os, re, Levenshtein as lev, codecs
from SimpleQuiz.quiz.models import Choice, Question
from django.core.management.base import BaseCommand, CommandError
import optparse

class Command(BaseCommand):
    args = '--datapath=/path/to/text/data/'
    can_import_settings = True
    help = 'Imports scanned text into Questions and Choices'
    option_list = BaseCommand.option_list + (
        optparse.make_option('--datapath', action='store', type='string',
                             dest='datapath',
                             help='Path to OCRd text files to be parsed.'),
        )
    requires_model_validation = True
    # Parser REs
    BACKUP_RE = re.compile(r'\~$|bak$|back$|backup$')
    QUEST_RE = re.compile(r'^[0-9]{1,3}[.][ ]')
    CHOICE_RE = re.compile(r'^[a-e][.][ ]')

    def handle(self, *args, **options):
        # get the data path
        try:
            os.path.exists(options['datapath'])
        except Exception as e:
            raise CommandError("None or invalid path provided: %s" % e.message)
        self.datapath = os.path.expanduser(options['datapath'])

        # generate list of text strings from lines in target files
        self.data_lines = []
        for fn in os.listdir(os.path.join(self.datapath, 'questions/')):
            if self.BACKUP_RE.search(fn):
                self.stderr.write("Skipping backup: %s\n" % (fn))
            else:
                for line in codecs.open(os.path.join(self.datapath, 'questions/', fn), 'r', encoding='latin-1'):
                    if not self.is_boilerplate(line):
                        if not line.strip() == '':
                            self.data_lines.append(line)

    #----------------------------------------------------------------------- 
    #--------------------- Parse the text lines and create Questions/Choices
    #----------------------------------------------------------------------- 
        cur_quest = None
        cur_choice = None
        cur_is_quest = False
        questions = {}
        choices = {}
        for line in self.data_lines:
            if self.is_question(line):
                [n, txt] = line.split('.', 1)
                qtext = txt.rstrip() + " "
                q = Question.objects.create(text=qtext)
                q.save()
                cur_quest = q.pk
                questions[cur_quest] = q
                cur_is_quest = True
            elif self.is_choice(line):
                [n, txt] = line.split('.', 1)
                num = self.char2dig(n)
                ctext = txt.rstrip() + " "
                c = Choice.objects.create(text=ctext, order=num, question=questions[cur_quest])
                c.save()
                cur_choice = c.pk
                choices[cur_choice] = c
                cur_is_quest = False
            else:
                if cur_is_quest:
                    questions[cur_quest].text += line.rstrip() + " "
                    questions[cur_quest].save()
                else:
                    choices[cur_choice].text += line.rstrip() + " "
                    choices[cur_choice].save()
        self.stdout.write("----- FINISHED -----\n")
        return None

    def is_question(self, arg_str):
        if self.QUEST_RE.search(arg_str):
            return True
        else:
            return False

    def is_choice(self, arg_str):
        if self.CHOICE_RE.search(arg_str):
            return True
        else:
            return False

    def char2dig(self, x):
        if x == 'a':
            return 1
        if x == 'b':
            return 2
        if x == 'c':
            return 3
        if x == 'd':
            return 4
        if x == 'e':
            return 5

    def is_boilerplate(self, arg_str):
        boilerplate = [u'MFT PRACTICE EXAMINATIONS',
                       u'BERKELEY TRAINING ASSOCIATES ' + u'\u00A9' + u' 2009',
                       u'BERKELEY TRAINING ASSOCIATES',
                       u'MARRIAGE AND FAMILY THERAPY',
                       u'PRACTICE EXAMINATION 41',
                       u'Page 0', u'Page 1', u'Page 2', u'Page 3', u'Page 4',
                       u'Page 5', u'Page 6', u'Page 7', u'Page 8', u'Page 9',
                       ]
        for bp in boilerplate:
            if lev.distance(bp.encode('utf-8'), arg_str.encode('utf-8')) < 4:
                return True
        return False