Question

Is there any dictionary based string matching algorithm in Java?

Something that will give the percentage of similarity between two strings based on the dictionary ?

Like

 public double getSimilarity(String str1, String str2);

for which an implementation like :

 getSimilarity("Professor", "Teacher")

will give a very high percentage ?

Thanks in advance 🙂

Answer 1

There is a great work done by Shaul Markovitch and Evgeniy Gabrilovich, described in their article: Wikipedia-based Semantic Interpretation for Natural Language Processing.

The idea is as follows: Index wikipedia (or other context source).

Creating a mapping for each term (word): term -> articles in which the term appears in.

Each term is basically represented by a vector – for simplicity, let’s say it is a binary vector – so for the term t the entry d will be ‘1’ if and only if the term t appears in the document d.

Now, given these vectors – to find if two terms t1, t2 are similar – all you have to do it take the vector similarity of the two vectors that represent t1 and t2.

---

Note: The binary vector is a simplification, in fact the article uses the tf-idf score, that the term t has in a document d.