Avoid simply relying on crawl depth, which: a) biases from the home page (when many people do not start from the home page) and b) biases toward recent or other content that happens to be highlighted at the time of the crawl.
Note that lower level fields may sometimes be needed to compute more useful fields. Also, sometimes the higher level fields may be more difficult to compute, so they are not always worth it.
This would usually be implemented by scraping.