Dear Kapil Sethi!

Contextual feature: the result with use of it should be better than without it.

So, if you exlude the contextual feature the result will be worse.

It could be used for contextual feature detection.

In the numerous mining you can use also correlation matrix.

More precise definition you can find in the work

Robust Classification with Context-Sensitive Features by Peter David Turney

Research Example for Text Mining:

I start text mining from extracting features (e.g. words or n-grams).

For every text there is the vector of features inclusion

Example: text - "I robot"

features i you robot man

incl.vector 1 0 1 0

If I interested classification robotic non-robotis:

I will see that feature "robot" will have large normalized frequency of apearing than in non-robotic text.

So, the feature "robot" is contextual.

Also I will see, that there are many features that do not influence the result (the frequency in robotic and non-robotic texts will be similar).

Thanks Andrei for this answer

i agree with Andrei Zakharov answer

can you tell me how to implement on the dataset

Of course (see messages)