Turning Trajectory’s Book-Analysis Tech Loose On #SOTU
Sentiment Curves And Axises Of Evil
Update: The Trajectory team has added the text of Barack Obama’s 2016 (and final) State of the Union address to its #SOTU analysis program, so you can now see and compare the complete canon of these speeches, from 1790 forward.
We have deconstructed every sentence in every State of the Union address to analyze and summarize their common and unique themes.
Who else but Trajectory?
That’s Jim Bryant talking, CEO of the Boston-based “intelligent network” that’s working with publishers on their distribution challenges. It’s also the home of an ambitious Natural Language Processing (NLP) effort that “reads” the complete text of a book into its system. The idea is recommendations: when a consumer looks for a book like the one he or she just enjoyed, Trajectory’s digital comparisons are searching for such factors as matching “sentiment curves,” keywords, settings, characters—whatever can be used to make a positive match of one good story to another.
And all it took to get the attention of Bryant and his chief content officer Scott Beatty was the White House’s announcement that President Barack Obama’s last State of the Union on Tuesday (12th January) will be “non-traditional.”
Trajectory is ready with a non-traditional response.
As a walkup to the president’s address, Bryant and Beatty’s team has created a compilation of every State of the Union (#SOTU), from the first address by George Washington on 8 January 1790 to Barack Obama’s speech on 20 January of last year.
On this page, you see the collection, and on accompanying tabs, you can see some of the ways the Trajectory NLP breakdown can parse the texts of the presidential speeches. Keywords, language, content, mood and more can be gleaned form this type of “deep learning analysis,” as it’s called.
Then, by having the system compare various factors, or “vectors,” Trajectory’s technologists can create comparisons they use to chart intensity and sentiment, among other expressive qualities.
In fact, before comparisons are made, the data rendered by the compilation alone is interesting.
- Before President Obama says his first word in his final State of the Union on Tuesday evening, 1,631,219 words of presidential oration will have been given in #SOTU presentations.
- The speeches comprise 16,406 words in sentences averaging about 30 words each.
- Over the years, places have been mentioned in these addresses more than people, 1,410 places to 1,135 persons’ names.
- Be sure to note the “Trajectory China 300” words: there are 240 of them in the State of the Union addresses. The China 300 is a list of words that Beijing recommends Chinese citizens master as a basis for learning English. Here’s our recent story about English as the book world’s lingua franca. In it, consultant Thad McIlroy tells us that there are more people learning English in China than speaking it in the United States.
Beatty has been looking over some of the results of the system’s analysis.
“For example,” he says, “you can easily analyze the keywords and concepts of President James Monroe’s seventh State of the Union, which first outlined the Monroe Doctrine, a defining moment in the foreign policy of the United States. And then in the 2002 State of the Union address of George W. Bush, you find that phrase ‘an axis of evil.’
“We have compared each speech to all of the others,” Bryant says, “to identify the closest matches, and we’ve summarized each speech by identifying its unique use of language and sentiment.”
Beatty talks about more of the results the team is finding:
When we compared President Obama’s speech in 2015, looking for the closest “sentiment curve,” “we were surprised to discover that the address most similar in thought, attitude and opinion was President Grant’s in 1870.
“Another pattern we discovered was the similarity of the first addresses from John F. Kennedy in 1961″—a State of the Union delivered 10 days after his “Ask not what your country can do for you” inaugural address—”and Ronald Reagan’s 1982 State of the Union.”
Check It Out
You can do your own research at Trajectory’s site.
Start here for a running readout of State of the Union text from George Washington and a look at the elements being parsed by the system as it compares the various speeches.
Then use the tabs at the top to look at sentiment-curve and keyword comparisons. Amaze your friends with your #SOTU savvy on Tuesday night (9 p.m. ET / 2 a.m. Wednesday GMT) during the address at the U.S. Capitol.
When not moonlighting as a presidential speech slicer and dicer, Trajectory is busy servicing more than 300 international channels of book distribution, which together represent some 230,000 digital “endpoints” including ebook retailers, library distributors, school distributors and alternative digital sales channels.