how to cluster sentences in matlab

I want to cluster 4 sentences. For example s1,s2,s3,s4.

I have custom distance (from my own similarity function) of these sentences from each other. For example:

s1,s2:0.75 s1,s3=0.80 s2,s3:0.50 and ...

I just want to cluster these sentences.

How could i do it in matlab?

[Andrew Reibold]

I don't know what you mean by clustering.. :(

Not sure if its a vocabulary thing, like maybe there is another word, or if I just am totally ignorant of what clustering is.

Can you provide a short example of input AND desired output?

[Image Analyst]

The given example is not totally clear. Did you see Andrew's comment/request?

[rahman]

thanks for your answer.

I have 4 sentences: s1,s2,s3,s4. distances between each two sentences are as follow: dis(s1,s2)=0.75 , dis(s1,s3)=0.10 , dis(s3,s4)=0.60 and ... .

I want to put sentencs have least distance into one cluster.

for example: cluster1: s1, s2 and cluster2: s3,s4

I know about matlab clustring function. but this function uses the standard distance metrics like cosine or euclidean, but i have my own similarity matrix and i want to use this matrix as distances.

how could i do it?

[Oleg Komarov]

If you have the stats toolbox you can use hirarchical clustering. Try to read the tutorial, and post again if you get stuck.

[rahman]

yes. i want to use hirarchical clustering. but i want to use my own distance matrix. i dont want use standard distance in matlab. how can i use my own distances matrix as distances in linkage function?

[rahman]

my distance matrix is:

how should I cluster these sentences into 2 clusters?

for example:

cluster1=s1,s2

cluster2=s3,s4

please help me!! :(

it is very important for me!

[Oleg Komarov]

You can use squareform(y,'tovector') to generate the output in the way pdist(), then continue with linkage and eventually visualize with dendrogram.

[rahman]

can u write a simple code for me?? i am new user in matlab and it is very necessary for me!!

thanks a lot!!

[rahman]

please!!!!!!!!!!!!!!! :(

[rahman]

it's done :) thnak u very much oleg. :*