For this project, I ended up using ARI and FKGL indexes to compare the readability of each president's speeches. CLI is opted out because it would require dividing the raw text into 100-word-chunks before cleaning the text in order to calculate the number of sentences in each chunk. Therefore, for further improvement of this project, ways to count sentences after cleaning text may be desirable.
Also, more interesting implications can be drawn by combining the speech data with other factors such as satisfaction level, number of terms served of each president, etc. Visualization method besides line charts may help present the result in more meaningful way. Possible directions to consideration include interactive graphs, matrix, etc.