News articles data mining

[StranGhost]

I got a work and some parts confused me.

 

Here's the assignment:

 

You have about 100,000 Chinese news, with their news date and news sources (three kinds: 0 1 2).

Now, you got 20 query results (20 different key words).

Each results shows about 100 articles without their news date and sources. (They were sorted by the relevance with keywords)

Please use the 100,000 news to train a model, recognize these results' sources,

and sort them by published date.

 

--

 

My question is I know some training models could be used,

but I don't know how to transfer those Chinese terms into a number, coordinate or matrix.

I also want to know what kind of information is the most useful for classifying and sorting news in query results.

 

Anybody has relational experience?

Edited January 16, 2015 by str18000