Machine learning issue for negative instances

I had to build a concept analyzer for computer science field and I used for this machine lea开发者_运维知识库rning, the orange library for Python. I have the examples of concepts, where the features are lemma and part of speech, like algorithm|NN|concept. The problem is that any other word, that in fact is not a concept, is classified as a concept, due to the lack of negative examples. It is not feasable to put all the other words in learning file, classified as simple words not concepts(this will work, but is not quite a solution). Any idea?

Thanks.

The question is very unclear, but assuming what you mean is that your machine learning algorithm is not working without negative examples and you can't give it every possible negative example, then it's perfectly alright to give it some negative examples.

The point of data mining (a.k.a. machine learning) is to try coming up with general rules based on a relatively small samples of data and then applying them to larger data. In real life problems you will never have all the data. If you had all possible inputs, you could easily create a simple sequence of if-then rules which would always be correct. If it was that simple, robots would be doing all our thinking for us by now.