Programmatically determine whether to describe an object with "a" or "an"?

Question

I have a database of nouns (ex "house", "exclamation point", "apple") that I need to output and describe in my application. It's hard to put together a natural-sounding sentence to describe an item without using "a" or "an" - "a house is BIG", "an exclamation point is SMALL", etc.

Is there any function, library, or hack i can use in PHP to determine whether it is more appropriate to describe any given noun with A or AN?

Answer 1

I was also looking for such solution but in JavaScript. So I ported it over to JS, you can check out the actual project in github https://github.com/rigoneri/indefinite-article.js

Here is the code snippet:

 function indefinite_article(phrase) {

    // Getting the first word 
    var match = /\w+/.exec(phrase);
    if (match)
        var word = match[0];
    else
        return "an";

    var l_word = word.toLowerCase();
    // Specific start of words that should be preceeded by 'an'
    var alt_cases = ["honest", "hour", "hono"];
    for (var i in alt_cases) {
        if (l_word.indexOf(alt_cases[i]) == 0)
            return "an";
    }

    // Single letter word which should be preceeded by 'an'
    if (l_word.length == 1) {
        if ("aedhilmnorsx".indexOf(l_word) >= 0)
            return "an";
        else
            return "a";
    }

    // Capital words which should likely be preceeded by 'an'
    if (word.match(/(?!FJO|[HLMNS]Y.|RY[EO]|SQU|(F[LR]?|[HL]|MN?|N|RH?|S[CHKLMNPTVW]?|X(YL)?)[AEIOU])[FHLMNRSX][A-Z]/)) {
        return "an";
    }

    // Special cases where a word that begins with a vowel should be preceeded by 'a'
    regexes = [/^e[uw]/, /^onc?e\b/, /^uni([^nmd]|mo)/, /^u[bcfhjkqrst][aeiou]/]
    for (var i in regexes) {
        if (l_word.match(regexes[i]))
            return "a"
    }

    // Special capital words (UK, UN)
    if (word.match(/^U[NK][AIEO]/)) {
        return "a";
    }
    else if (word == word.toUpperCase()) {
        if ("aedhilmnorsx".indexOf(l_word[0]) >= 0)
            return "an";
        else 
            return "a";
    }

    // Basic method of words that begin with a vowel being preceeded by 'an'
    if ("aeiou".indexOf(l_word[0]) >= 0)
        return "an";

    // Instances where y follwed by specific letters is preceeded by 'an'
    if (l_word.match(/^y(b[lor]|cl[ea]|fere|gg|p[ios]|rou|tt)/))
        return "an";

    return "a";
}