Making bigram features from a particular dataset

Question

I have a folder which has a number of files which have a format like these

madvise
write
write
write
write
read
read
madvise
ioctl
ioctl
getuid
epoll_pwait
read
recvfrom
sendto
getuid
epoll_pwait

that is it is a set of words which repeat.This is how all the files are like. Now I have created a feature vector table using unigram that is each word becomes a feature and each file becomes a row where I put the frequency of that word occuring in the respective columns.

Now I want to create a similar FVT using bigrams.I was wondering how to do that in this case.

Answer 1

Bigram is better to be used with sentences. In your case, files contain a list of words, as I could understand. Therefore using bigrams in your project might not yield expecting results. However, if you are still willing to do that, this is how you calculate bigrams:

Take the list of words and count the frequencies of adjacent words. Ex:

(madvise, write) - 1

(write, write) - 3

(write, read) - 1

(read, read) - 1

(read, madvise) - 1

.

.

.

.

(sendto, epoll_pwait) - 1