concept word in category recommender system

This is an excerpt from Manning's book Practical Recommender Systems.
If the words Django web framework sound strange to you, take a look at the Django documentation at www.djangoproject.com/start/overview/.
Previously we talked about actors, but we could approach film features even more generally and say that everyone worth mentioning who’s in the production of the content could be a feature. But while words that occur only once in a document are good to save if you’re creating a search, you need to economize when you talk about recommendations.
Descriptions are full of filler words, in the sense that there are many words needed to make a human-readable description. But if you’re deconstructing the documents into an array of words, you lose the value of words like the or a. Because the word a by itself doesn’t give any descriptive information, it’s best to remove that and similiar words of that type. The words you don’t want in your model are the stop words.
The next step is to remove the stop words from the BoW. Stop words are, strangely enough, not synonyms of the word stop, but words that you’re going skip in your analysis. A stop words list is dependent on both your language and your domain. The one you’ll use here supports English but also many other languages (such as Danish). The package will be downloaded and installed executing the command line shown in listing 10.1 in the terminal window.