Question

我一直在进行一个简单的英语测验。这样，我们需要验证答案，这些答案由用户在输入字段中输入。在当前的实现中，我正在将正确答案与用户答案进行精确比较。就像

HTML

<input type="text" id="answer" />
<button onclick="validate()">Validate</button>

JavaScript

var question = "Do you like movies?",
    answer = "No, I don't like movies.";

function validate() {
    var userInput = document.getElementById('answer').value;
    if(answer == userInput) {
      console.log("correct");
    } else {
      console.log("wrong");
    }
}

但是我不想完全验证。例如，忽略大小写，逗号，撇号等。例如，如果用户输入，

我不喜欢电影

答案可能是正确的。我不知道如何开始以及从哪里开始。任何人都可以帮忙。

Answer 1

一种选择是删除所有非单词字符和空格，并比较每个替换字符串的小写形式：

var question = "Do you like movies?",
  answer = "No, I don't like movies.";

const normalize = str => str
  .replace(/[^\w ]/g, '')
  .toLowerCase();
function validate(userInput) {
  const noramlizedInput = normalize(userInput)
  const noramlizedAnswer = normalize(answer);
  if (noramlizedInput == noramlizedAnswer) {
    console.log("correct");
  } else {
    console.log("wrong");
  }
}
validate('No i dont like movies');
validate("NO!!!!! I DON''t like movies.");

另一种选择是遍历userInput的所有可能的子字符串，找出哪个与所需的answer重叠最多，但这要复杂得多。

一个更简单的选择是检查有多少重叠的单词：

var question = "Do you like movies?",
  answer = "No, I don't like movies.";

const normalize = str => str
  .replace(/[^\w ]/g, '')
  .toLowerCase()
  .split(/\s+/)
function validate(userInput) {
  const noramlizedInputArr = normalize(userInput);
  const noramlizedAnswerArr = normalize(answer);
  const overlapCount = noramlizedInputArr.reduce((a, word) => (
    a + Number(noramlizedAnswerArr.includes(word))
  ), 0);
  console.log(overlapCount);
  if (overlapCount >= 4) {
    console.log("correct");
  } else {
    console.log("wrong");
  }
}
validate('No i dont like movies');
validate("NO!!!!! I DON''t like movies.");
validate("i dont like movies.");
validate("Yes I like movies.");

Answer 2

如果您只想捕捉拼写错误和小变化，则将标准度量标准称为编辑距离或Levenshtein distance。这是将一个文本更改为另一个文本所需的最小删除，插入或替换数量的计数。诸如“不，我不喜欢电影”和“不，我不喜欢电影”之类的字符串具有较小的编辑距离。

这是一个快速而肮脏的递归编辑距离功能，它将使您有个想法：

function validate(text, pattern) {
  // some simple preprocessing
  let p = pattern.toLowerCase().replace(/[^a-z]+/ig, '')
  let t= text.toLowerCase().replace(/[^a-z]+/ig, '')

  // memoize recursive algorithm
  let matrix  = Array.from({length: t.length + 1}, () => [])

  function editDistance(text, pattern, i = 0, j = 0){
    if(i == text.length && j == pattern.length) return 0
    if(i == text.length) return pattern.length - j
    if(j == pattern.length) return text.length - i   
    let choices = [
        (matrix[i+1][j+1] || (matrix[i+1][j+1] = editDistance(text, pattern, i+1, j+1))) + (text[i].toLowerCase() === pattern[j].toLowerCase() ? 0 : 1),
        (matrix[i+1][j]   || (matrix[i+1][j]   = editDistance(text, pattern, i+1, j))) + 1,
        (matrix[i][j+1]   || (matrix[i][j+1]   = editDistance(text, pattern, i, j+1))) + 1
    ]
    return  Math.min(...choices)
  }
  return editDistance(t, p)
}

// similar strings have smaller edit distances
console.log(validate("No I dont lik moves","No i dont like movies"))

// a little less similar
console.log(validate("Yes I like movies","No i dont like movies"))

// totally different
console.log(validate("Where is the bathroom","No i dont like movies"))

// careful -- small edit distance !== close meaning
console.log(validate("I do like tacos","I don't like tacos"))

匹配最小错别字的字符串，拾取最小可接受距离效果很好。当然，如果您要评估用户的意图，那么这些简单的色调都无法使用。像“我爱炸玉米饼”和“我讨厌炸玉米饼”之类的字符串具有较小的编辑距离，如果您不了解该语言，就无法确定它们的含义相反。如果您需要进行此级别的检查，则可以尝试使用Watson Conversation之类的服务，该服务将返回用户意图进行输入。

Javascript-如何知道另一个字符串中匹配了多少个字符串？

2 个答案: