Assignment-01.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Lesson-01 Assignment"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "`各位同学大家好，欢迎各位开始学习我们的人工智能课程。这门课程假设大家不具备机器学习和人工智能的知识，但是希望大家具备初级的Python编程能力。根据往期同学的实际反馈，我们课程的完结之后 能力能够超过80%的计算机人工智能/深度学习方向的硕士生的能力。`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 本次作业的内容"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1. 复现课堂代码\n",
    "\n",
    "在本部分，你需要参照我们给大家的GitHub地址里边的课堂代码，结合课堂内容，复现内容。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. 请回答以下问题\n",
    "\n",
    "回答以下问题，并将问题发送至 mqgao@kaikeba.com中：\n",
    "```\n",
    "    2.1. what do you want to acquire in this course？\n",
    "    2.2. what problems do you want to solve？\n",
    "    2.3. what’s the advantages you have to finish you goal?\n",
    "    2.4. what’s the disadvantages you need to overcome to finish you goal?\n",
    "    2.5. How will you plan to study in this course period?\n",
    "```"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. 如何提交\n",
    "代码 + 此 jupyter 相关，提交至自己的 github 中(**所以请务必把GitHub按照班主任要求录入在Trello中**)；\n",
    "第2问，请提交至mqgao@kaikeba.com邮箱。\n",
    "#### 4. 作业截止时间\n",
    "此次作业截止时间为 2019.7.6日"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5. 完成以下问答和编程练习"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 基础理论部分"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 0. Can you come up out 3 sceneraies which use AI methods? "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans: 智能问答，智能推荐，决策系统"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1. How do we use Github; Why do we use Jupyter and Pycharm;"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans: Github:1、提交作业，互相分享与协助   Jupyter and Pycharm:两者之前各有特色，Jupyter 适合调试并有很好的展示与演示效果  Pycharm 具有更好的性能。更适合大规模的开发应用"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. What's the Probability Model?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:概率模型是指某一个数据/文本 模型F(x) 计算输出为某种概率值"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. Can you came up with some sceneraies at which we could use Probability Model?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:概率模型用来预估用户的喜好  在用于推荐系统中"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4. Why do we use probability and what's the difficult points for programming based on parsing and pattern match?"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:1、概率模型的好处我以为可以不用设计那么多规则，并且依据计算出的函数对其他数据也是可以的。\n",
    "    2、解析和匹配就相对比较死板，不能很好的应对多种情况"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 5. What's the Language Model;"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:我的理解是在NLP中语言模型是根据词之间出现的概率分布计算，从而得到一个相对合理或者最大可能性的句子"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 6. Can you came up with some sceneraies at which we could use Language Model?\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:字或词纠错、机器翻译、语言识别"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 7. What's the 1-gram language model;"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:1-gram是指 基于一个某一个词从而计算得到整个句子出现的概率"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 8. What's the disadvantages and advantages of 1-gram language model;"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:优点：优点是1-gram基于统计学的原理进而得到序列出现的概率  缺点：缺点为在语言环境中一个词不能很好的表征一个序列出现概率的真正准确性"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 9. What't the 2-gram models;"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:2-gram是指 基于一个某一组词（一组词中包含两个词）从而计算得到整个句子出现的概率"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 编程实践部分"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 1. 设计你自己的句子生成器"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "如何生成句子是一个很经典的问题，从1940s开始，图灵提出机器智能的时候，就使用的是人类能不能流畅和计算机进行对话。和计算机对话的一个前提是，计算机能够生成语言。\n",
    "\n",
    "计算机如何能生成语言是一个经典但是又很复杂的问题。 我们课程上为大家介绍的是一种基于规则（Rule Based）的生成方法。该方法虽然提出的时间早，但是现在依然在很多地方能够大显身手。值得说明的是，现在很多很实用的算法，都是很久之前提出的，例如，二分查找提出与1940s, Dijstra算法提出于1960s 等等。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "在著名的电视剧，电影《西部世界》中，这些机器人们语言生成的方法就是使用的SyntaxTree生成语言的方法。\n",
    "\n",
    "> \n",
    ">\n",
    "\n",
    "![WstWorld](https://timgsa.baidu.com/timg?image&quality=80&size=b10000_10000&sec=1561818705&di=95ca9ff2ff37fcb88ae47b82c7079feb&src=http://s7.sinaimg.cn/mw690/006BKUGwzy75VK46FMi66&690)\n",
    "\n",
    "> \n",
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "在这一部分，需要各位同学首先定义自己的语言。 大家可以先想一个应用场景，然后在这个场景下，定义语法。例如：\n",
    "\n",
    "在西部世界里，一个”人类“的语言可以定义为：\n",
    "``` \n",
    "human = \"\"\"\n",
    "human = 自己 寻找 活动\n",
    "自己 = 我 | 俺 | 我们 \n",
    "寻找 = 看看 | 找找 | 想找点\n",
    "活动 = 乐子 | 玩的\n",
    "\"\"\"\n",
    "```\n",
    "\n",
    "一个“接待员”的语言可以定义为\n",
    "```\n",
    "host = \"\"\"\n",
    "host = 寒暄 报数 询问 业务相关 结尾 \n",
    "报数 = 我是 数字 号 ,\n",
    "数字 = 单个数字 | 数字 单个数字 \n",
    "单个数字 = 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 \n",
    "寒暄 = 称谓 打招呼 | 打招呼\n",
    "称谓 = 人称 ,\n",
    "人称 = 先生 | 女士 | 小朋友\n",
    "打招呼 = 你好 | 您好 \n",
    "询问 = 请问你要 | 您需要\n",
    "业务相关 = 玩玩 具体业务\n",
    "玩玩 = 耍一耍 | 玩一玩\n",
    "具体业务 = 喝酒 | 打牌 | 打猎 | 赌博\n",
    "结尾 = 吗？\"\"\"\n",
    "\n",
    "```\n",
    "\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "请定义你自己的语法: "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "第一个语法："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "you_need_replace_this_with_name_you_given = '''\n",
    "老师:年龄 职称 姓名   动作 授课门类 \n",
    "姓名:张三 | 李四 | 王二\n",
    "年龄:40岁 | 50岁 | 60岁 \n",
    "职称:教授 | 副教授 | 讲师\n",
    "动作:教 | 任教 \n",
    "授课门类:数学 | 哲学 | 经济 | 计算机\n",
    "'''"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "第二个语法："
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "you_need_replace_this_with_name_you_given = '''\n",
    "学生:年龄 姓名 动作 年级  专业课程\n",
    "姓名:小二 | 小三 | 小四\n",
    "年龄:21 | 22 | 23\n",
    "动作: 学 | 学习 \n",
    "专业课程:数学 | 哲学 | 经济 | 计算机\n",
    "'''"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "TODO: 然后，使用自己之前定义的generate函数，使用此函数生成句子。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "TODO: 然后，定义一个函数，generate_n，将generate扩展，使其能够生成n个句子:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "def gereate(garm,target): # 自己另加的函数\n",
    "    if target not in garm:\n",
    "        return target\n",
    "    else:\n",
    "        expended = [gereate(garm,t) for t  in random.choice(garm[target])]\n",
    "        return ''.join([e for e in expended])\n",
    "\n",
    "def generate_n():\n",
    "    # you code here \n",
    "    ef gereate_n(garm,target,n = 10):\n",
    "    gen = []\n",
    "    for i in range(n):\n",
    "            exp = [gereate(garm,t) for t  in random.choice(garm[target])] # generate 为随机生成一句的函数\n",
    "            gen.append(exp)\n",
    "    return gen\n",
    "print([''.join(f) for f in gereate_n(creat_grams(you_need_replace_this_with_name_you_give),target=\"老师\")])\n",
    "    pass"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 2. 使用新数据源完成语言模型的训练"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "按照我们上文中定义的`prob_2`函数，我们更换一个文本数据源，获得新的Language Model:\n",
    "\n",
    "1. 下载文本数据集（你可以在以下数据集中任选一个，也可以两个都使用）\n",
    "    + 可选数据集1，保险行业问询对话集： https://github.com/Computing-Intelligence/insuranceqa-corpus-zh/raw/release/corpus/pool/train.txt.gz\n",
    "    + 可选数据集2：豆瓣评论数据集：https://github.com/Computing-Intelligence/datasource/raw/master/movie_comments.csv\n",
    "2. 修改代码，获得新的**2-gram**语言模型\n",
    "    + 进行文本清洗，获得所有的纯文本\n",
    "    + 将这些文本进行切词\n",
    "    + 送入之前定义的语言模型中，判断文本的合理程度"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 3. 获得最优质的的语言"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "当我们能够生成随机的语言并且能判断之后，我们就可以生成更加合理的语言了。请定义 generate_best 函数，该函数输入一个语法 + 语言模型，能够生成**n**个句子，并能选择一个最合理的句子: \n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "提示，要实现这个函数，你需要Python的sorted函数"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[1, 2, 3, 5]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted([1, 3, 5, 2])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "这个函数接受一个参数key，这个参数接受一个函数作为输入，例如"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(1, 4), (2, 5), (4, 4), (5, 0)]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted([(2, 5), (1, 4), (5, 0), (4, 4)], key=lambda x: x[0])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "能够让list按照第0个元素进行排序."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(5, 0), (1, 4), (4, 4), (2, 5)]"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted([(2, 5), (1, 4), (5, 0), (4, 4)], key=lambda x: x[1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "能够让list按照第1个元素进行排序."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(2, 5), (1, 4), (4, 4), (5, 0)]"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sorted([(2, 5), (1, 4), (5, 0), (4, 4)], key=lambda x: x[1], reverse=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "能够让list按照第1个元素进行排序, 但是是递减的顺序。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def generate_best(garm,target,n = 10): \n",
    "    gen = []\n",
    "    for i in range(n): #\n",
    "            expended = [gereate(garm,t) for t  in random.choice(garm[target])]\n",
    "            gen.append(expended)\n",
    "    return gen\n",
    "gereate_sentences = ([''.join(f) for f in gereate_best(creat_grams(you_need_replace_this_with_name_you_give),target=\"老师\")])\n",
    "\n",
    "p = []\n",
    "for s in gereate_sentences:\n",
    "    p2 = get_probablity(s) # get_probablity 为预测模型\n",
    "    p.append(p2)\n",
    "h = [(x, y) for x, y in zip_longest(gereate_sentences, p)]\n",
    "print(sorted(h, key=lambda x: x[0]))\n",
    "    pass"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "好了，现在我们实现了自己的第一个AI模型，这个模型能够生成比较接近于人类的语言。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Q: 这个模型有什么问题？ 你准备如何提升？ "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Ans:1、依赖数量大的训练级  2、不能很好的区别语义  3、若两句话区分很小，分数基本相同"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### 以下内容为可选部分，对于绝大多数同学，能完成以上的项目已经很优秀了，下边的内容如果你还有精力可以试试，但不是必须的。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### 4. (Optional) 完成基于Pattern Match的语句问答\n",
    "> 我们的GitHub仓库中，有一个assignment-01-optional-pattern-match，这个难度较大，感兴趣的同学可以挑战一下。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "\n",
    "#### 5. (Optional) 完成阿兰图灵机器智能原始论文的阅读\n",
    "1. 请阅读阿兰图灵关于机器智能的原始论文：https://github.com/Computing-Intelligence/References/blob/master/AI%20%26%20Machine%20Learning/Computer%20Machinery%20and%20Intelligence.pdf \n",
    "2. 并按照GitHub仓库中的论文阅读模板，填写完毕后发送给我: mqgao@kaikeba.com 谢谢"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "> "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "各位同学，我们已经完成了自己的第一个AI模型，大家对人工智能可能已经有了一些感觉，人工智能的核心就是，我们如何设计一个模型、程序，在外部的输入变化的时候，我们的程序不变，依然能够解决问题。人工智能是一个很大的领域，目前大家所熟知的深度学习只是其中一小部分，之后也肯定会有更多的方法提出来，但是大家知道人工智能的目标，就知道了之后进步的方向。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "然后，希望大家对AI不要有恐惧感，这个并不难，大家加油！"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    ">"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![](https://timgsa.baidu.com/timg?image&quality=80&size=b9999_10000&sec=1561828422005&di=48d19c16afb6acc9180183a6116088ac&imgtype=0&src=http%3A%2F%2Fb-ssl.duitang.com%2Fuploads%2Fitem%2F201807%2F28%2F20180728150843_BECNF.thumb.224_0.jpeg)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}