Unsupervised Affix Identification Approach using Probabilistic Dependence

It is difficult to identify an affix letter from a root letter in a given word in many NLP applications. This challenge is even harder in morphologically complex languages, such as Arabic. This paper investigates to identify affix letters using probabilistic dependence concept with an unsupervised method. A relationship is shown when 90% of the negative dependence points appeared at the affix letters.