Problem Statement

Prompt the user for the order

statistic n: 1, 2, 3, etc.

Read a file of tokens, building a map

(Map[List[Stringn]] -> List[String*])

from a list of n words to a list of the words in the text following these words: e.g., if n were 2, the map would contain a key for every pair of words in the text, and a value that is a list of all the words following the key (no matter where the pair occurs, with NO DUPLICATES allowed). Print all the associations, one per line, in any order (the n words followed by the list of words that follow them in the text).

Prompt the user for the number of random words to generate, and then prompt for the n words to start with. Build a

list (List[String*])

using the words to start with to generate a random next word, then use the previous n words (dropping the oldest word and adding the new word generated) to generate another random word; repeat. Note: you might have to stop prematurely if you generate the last n words in the text, if these words occur nowhere else. That is because in this case, there is no random word to generate following them! Print the list.

In simple words

This is given string

a b c b a d c b a d c a a b a a d

We pre populate a map with (x,y) -> (...) map like

[a, d] -> [c]
[a, b] -> [c, a]
[a, a] -> [b, d]
[b, c] -> [b]
[b, a] -> [d, a]
[c, b] -> [a]
[c, a] -> [a]
[d, c] -> [b, a]

We have to generate 5 length words taking first 2 input from user. We have to make other 3 letter by using above map.

Using above example

Enter # of words to generate: 10
Enter prefix word[0]: a
Enter prefix word[1]: d
Results = [a, d, c, a, a, b, a, d, c, a, a, d]

It is giving the correct output, but my questions are

  1. Can this be improved?
  2. Have I used collection classes efficiently?
  3. Is my algorithm efficient?
import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.StringTokenizer;

public class Application {

    public static void main(String[] args)
        String sample = "a b c b a d c b a d c a a b a a d";
        StringTokenizer st = new StringTokenizer(sample);
        List<String>    givenList = new ArrayList<String>();

         *     generate (x,y) -> (...) map
        Map<List<String>,List<String>> advMap = new HashMap<List<String>, List<String>>();
            String left1,left2,right = null;
            for(int listIndex = 0 ;listIndex < givenList.size() - 2; listIndex++)
            left1 = givenList.get(listIndex);
            left2 = givenList.get(listIndex+1);
            right = givenList.get(listIndex+2);
            List<String> leftList = new ArrayList<String>();
            //if key already exist, thne update right to the value list
                List<String> rightList = new ArrayList<String>();
                advMap.put(leftList, rightList);
        }// end for


        //Ask user for length of word, he wants to generate. Here I am hard coding it
        //Also, ask for first 2 letter he wants to start with
        int lenghtOfWord = 5;
        String x1 = "a";
        String x2 = "d";
        String[] genArray = new String[lenghtOfWord];
        genArray[0] = x1;
        genArray[1] = x2;
        for(int genIndex=0; genIndex <lenghtOfWord-2 ; genIndex++)
            List<String> leftKey = new ArrayList<String>();
            genArray[genIndex+2] = randomWord(advMap,leftKey);
        System.out.println("Generated Word");
        for(int i=0;i<genArray.length;i++)

           }// end main
            private  static String randomWord(Map<List<String>,List<String>> advMap, List<String> lKey)
            List<String> lVal = advMap.get(lKey);
            int rndIndx = (int) (Math.random() * lVal.size() );
            return lVal.get(rndIndx);
        {    //here is am expecting user will enter initial 2 letter which have mapping in the given inital advMap
            return null;
}//end class

This is not an assignment to me, found over net. I'm doing it for practice.

  • You suggest the problem is simple to read, but I am struggling to understand the problem. Two things.... this sounds suspiciously like NGrams. Also, linking to the original problem/challenge would help – rolfl Jun 4 '14 at 10:50
  • Sure this can be found under Program#4 in this link:… – anirban Jun 4 '14 at 10:53
  • Not sure why, but that is more understandable. Maybe because I read it again. OK, this is exactly the NGram problem. Get a body of text, build the N+1-grams, get a starting ngram, and use that to generate text. – rolfl Jun 4 '14 at 11:14
  • Ok thanks @rolfl, is my approach is correct in solving that problem? – anirban Jun 4 '14 at 11:42
up vote 3 down vote accepted

Let's go through a couple of things. Concept, then missing functionality (.... you have some).


An N-Gram is a sequence of N words that have been found in a span of text. You can have 1-grams, 2-grams, 3-grams, .... n-grams. You identify these n-grams by finding all possible n-wide spans of text, and storing them. In the sentence the quick brown fox, there are three 2-grams 'the quick', 'quick brown' and 'brown fox'. There are two 3-grams 'the quick brown' and 'quick brown fox'.

When processing natural languages, it is often statistically convenient to weigh the likelihood of a particular word happening 'next' after an existing sequence of words.

That is what this problem is about. given a span of 'n', find all the (n+1)-grams. Then, taking any n-width words, look for all the (n+1)-grams that start with those words. Randomly chose one to select the 'next' word. Then repeat the process until you run out of (n+1)-grams that match, or you hit the 'sentence' limit. You have just built a sentence that is statistically 'likely'. A smarter system will 'weight' the next word based on the frequency of the (n+1)-grams that were found in the text. I.e. If the original text has 'the white house' 10 times, and 'the white swan' just once, then it will 'randomly' choose 'house' 10 times more than 'swan'.

OK, that gives you some context for the problem.


The challenge/requirement was to take the n value as an input. You have hard-coded it as 2. In other words, you have 2-grams and 3-grams, when you are supposed to have n-grams and (n+1)-grams.

for(int listIndex = 0 ;listIndex < givenList.size() - 2; listIndex++)

That 2 should be an n, and all the logic changes needed to fix that.

Similarly, the x1 and x2 variables should be a list, or an array, because there could be more than 2.

You already have commented that the length of the output sentence is supposed to be user-input. You should make it user-input, as well as the first n words.

You have missed the point on the generating the sentence as well... You assume that there will always be a valid/matching random word to add to the sentence until you run out of words. This is not true. You may be part way through a sentence when you discover that the last n words in your sentence do not match any available (n+1)-grams, and there is thus nothing you can add to the sentence, and you have to stop short.

Code Style

Your code should be broken out in to more functions. You currently have just one which is used to get the next random word. You should have others to read the input file. You should probably have another that populates the map, etc.

Your main method is very heavy-weight, and should have function-extraction applied.

You also have indentation that is all over the place, and makes things hard to spot. It took me a while just to see that randomWord was a function.


Your code is only partially working, and some core functionality is missing. You are a good way along to getting a working solution. Hopefully the background on the problem will help you to understand what the problem is trying to solve.... basically: based on statistics from existing texts, randomly generate a new sentence that uses those statistics to predict what the next words in the sentence will be.

  • Yes I could have made user inputs and other functionality in separate function. Will update this question with modified code. I were more concerned in using the collection strategies effectively. Truly your effort is invaluable. – – anirban Jun 5 '14 at 7:38
  • @anirban - When you have updated your code, you should post it as a new question, not as an update to this question. – rolfl Jun 5 '14 at 16:31

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.