, ADJECTIVE and
NUMERALS, and the numbers in decimal
format on the edges connecting the vertices of the graph indicate the probability of transition from one state to another state
(State1→State2). In this graph, the probability that a NOUN word group will be followed by group of words belong to an
ADJECT 0.1. Table 1 below shows all the components of a Markov chain:
Table 1
.All components of Markov chain Q = Q
1
,Q
2
,…Q
n
A group of N cases
A = a
11
,a
12
,…a
n1
…a
nn
𝐴 is a transition matrix of probability, representing the probability of transition
from each state 𝑎𝑖𝑗 – 𝑖 to another state j.
∑
π = π
1
, π
2
,… π
n
The initial probability distribution for S cases.
pi – 𝑖 represents the probability that the Markov chain starts in a certain
state.
∑
HMM HMM is a very powerful statistical modeling tool whi is used for
speech recognition, handwriting recognition and other
NLP tasks. Hidden Markov Model (HMM) is an unsupervised machine learning algorithm that is part of graphical models. If
the training data is available, the HMM is trained using a supervised learning method [17;18]. It is important to understand
where the Hidden Markov Model algorithm actually fits or is used. Today, the HMM graphical model is used to predict
(hidden) states using sequential data such as weather, text, speech, etc. [19;20;21].
In some cases, it was necessary to predict the sequence of events that cannot be directly observed in the environment.
But while we are given a sequence of other observable states in the environment, these hidden states depend on the
observable states. That is, the most important point established by HMM is that the future state/event depends only on the
current state/event and not on other past states.
Arithmetically, the probability of a state at time t depends only on the time step t-1. In other words, the probability of
s(t) given s(t-1) is p(s(t) ┤| s(t-1)). This is used as a first order Markov model. If the probability of state s at time t depends on
time steps t-1 and t-2, this is called a 2nd order Markov model. That is, when dependence on past events increases, order
increases accordingly. A second-order Markov model can be written as p(s(t)┤|s(t-1), s(t-2)). The probability of successive
occurrence of events s1, s2 and s3 is calculated as follows [22]:
The figure below is a schematic of the simple Markov model that we have defined in the above equation.