Repository navigation
3. Word cloud
We have represented the hits of a query by a world cloud existing of 20-50 informative words. For example, when we search for "Iran contra affair" we get the following result:
As you can see these words seem very plausible. For example we see: president, Reagan, Iranian, Gulf, war. These are all words that you would expect to pop up when you are reading about the Iran contra affair. However the results are less good when we use a query that about a topic that played a smaller role in the news during the year of which the Reuters new articles exist. This is because of the fact that there are not many articles, if any at all, that cover the topic. For example the query "Mexican drug war" will produce the following word cloud:
We can immediately see that a lot of the words make less sense than in the first word cloud. Especially words like: the, he, would, but, it, also. These are all words that you could expect to come up in any query and are therefore not informative at all, they are just common. So as you can see, when there is not much information about the topic you are searching for, you would just get a word cloud with mostly common words. In order to get at least a world cloud with a lot of common words for a bad query, we removed all the punctuation from the articles because at first we would get a wordcloud containing a lot of: .,# etc.