After reading the comments from the previous post on using Google Trends to find tendencies about data mining techniques, I think it is useful to make a few remarks (thanks Crawford for making me curious about that).
First, I remind that the results given by Google Trends only involve the Google search engine. Even if this now the most used search engine, it is far from being the only one.
Second, as stated on their webpage, “[Google Trends] analyzes a portion of Google web searches to compute how many searches have been done for the terms you enter relative to the total number of searches done on Google over time“. This means that the first bias concern the way this part is selected and its size.
Third, Google Trends results are normalized. This means that the value obtained, for example by regions, are normalized according to the total search volume in that region. This is the reason why you can see Sri Lanka or Iran as first regions for a term such as “neural network”. In brief, if we consider that Sri Lanka has a search volume which is a lot smaller than UK for example, the importance of “neural network” is much less in UK since it is hidden by several other searches on any topics. This is why a country such as UK may not appear as a top region for this term.
As written on Google Trends page, we should “Keep in mind that instead of measuring overall interest in a topic, Google Trends shows users’ propensity to search for that topic on Google on a relative basis“. I think that interpreting the basic results (i.e. the search volume) of Google Trends is already a challenge. This is, as written in the comments of the previous post, in part due to the fact that these terms can be found under several words and acronyms. For example, “artificial neural network” can also be searched under “neural network”, “neural net”, “ann”, etc. I think that reading Google Trends results in terms of regions and cities is even more difficult and interpreting them is definitely not a straightforward issue.