Mining the Dispatch

Topic modeling identifies topics, and it also provides a proportional breakdown of the often multiple topics that each individual article or advertisement addressed. This chart shows the changing percentage of print space for this topic in the paper as a whole—for each month the topic proportions for "for hire and wanted ads" in all articles containing that topic are added together and then divided by the total number of pieces published that month to calculate a percentage value.

You can also adjust the articles that the above chart takes into account. By default, articles and advertisements that have only a small proportion of their content in this category are included in the chart. Each individual article has a negligible effect, but small proportions in dozens or hundred of articles can add up. Use the controls below the chart to adjust the threshold of the articles charted. You can also choose to chart the summation of the topic proportion of all articles (which makes sense for a topic that's more of a theme) or you can choose to chart the raw count of all articles above your specified threshold (which might be preferable when looking at a topic that's more generic in nature). For the latter kind of chart, you'll likely want to adjust the threshold to something like 20% or 30%.

This list of words are those that the topic model identifies as most likely to appear in documents in this category.

In this column are pieces from the Daily Dispatch that best exemplified this topic; i.e. they had the highest topic proportions in this category. The pie chart to the right of each piece identifies the specific topic proportion for "for hire and wanted ads in the piece." You can view the complete topic proportion breakdown for an individual piece by clicking on the title. The handle on the bottom left of the chart can be adjusted to view articles and advertisements best exemplifying this topic for particular months.