Extracting ingredients from a recipe text is a very common activity especially for data scientists and developers who want to study recipes or want to make statistical representations about nutritive values of cuisine recipes. Ingredients is not the only useful information we want to extract, the quantity used for each ingredient and how they are prepared are also interesting informations that we can extract by the same method presented in this work. Hidden Markov Models are the first idea that came in my mind because there are previous successful works that used this method for information extraction ((Freitag & McCallum, 2000),(Freitag & McCallum, 1999),(Seymore, McCallum, Rosenfeld, et al., 1999),(Bikel, Miller, Schwartz, & Weischedel, 1998),(Leek, 1997)), and also because modeling sequences of words where we have to estimate the hidden state is typically a hidden Markov procedure. In this work we are concentrating on the external knowledge part deployed in what we called a simple double agent Viterbi algorithm.