Coocurrence of defined concepts

KNIME Extensions
1

Hello,

we have a set of texts where we would like to identify related concepts which are stated near to each other.

We want to find in one sentence a relevant product feature and the customer's opinion statement belonging to this.
Since we want only the the pairs [product] and [opinion] we need to detect which concepts belong together, e.g. by defining the occurence together.

We tried to tag by products and by opinion terms, filter the terms, but could not figure out how we can best get the right pairs (occuring near 5 terms to each other or something similar). We can not be sure that the remaining opinion statements are belonging to the products or how much terms are filtered out.

Thanks for any hints how we could approach this

2

Hi,

once you filter all terms except for the tagged once the proximity information (i.t.o. number of words in between) is lost, since only the tagged terms remain. With the Term Co-occurrence counter is it not possible to determine proximity. You would need to extract N-grams and filter these. Hoewever, this would be more a workaround.

Cheers, Kilian

3

Kilian,

Thanks for the answer!

Regards, Bernd

4

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.