Notes from the presentation:

Slide 1 - "The Atheists Who Believe In God" is a 'novel' I generated in November. It uses Pew's survey data from 2007 to construct narratives about specific self-proclaimed atheists who claim to also believe in God. This passage gives a lot of information: it tells you the age, the income level, education level, belief about separation of church and state, and his support of meditation.

This is an example of "natural language generation"; I was able to create a program that can turn a raw dataset into a story. There is no need for a legend to understand the data; just read the story. You may look at the repo of the code here: https://github.com/tra38/The-Atheists-Who-Believe-In-God

Slide 2 - Computers are good at understanding numbers and raw data, but humans have difficulty understanding them (hence the need for visualizations). On the other hand, humans love understanding stories, while computers get confused easily.

The solution is to program the computer to plug in the raw data into a "narrative" that humans can read. The jargon term for this process is 'natural language generation'.

Narrative Science and Automated Insights are the two main competitors in this industry. Already, they are responsible for producing news stories and finance reports and want to scale their reporting to other fields as well.

Slide 3 - These sample stories were all generated by Automated Insight's program: "Wordsmith"

"Bank of the Republic": Example finance report

"Homesnap": Example housing report

"Hired Goon": Example fantasy football story.

Slide 4 - The problem with existing robo-journalism is that they need raw, quantitative data and much of what we do in life is...qualitative. It can't be reduced to a number. Or CAN IT?

If you can turn what we say and do into a number, then a computer can understand it. When they understand it, they can plug that into a template and spit it out as a narrative.

The process of turning unstructured data like reports into 'structured data' is known as "natural language processing". This slide mentions "Freud", a program that Narrative Science uses to engage in natural language processing.

Slide 5 - "Quill Engage" is a free web application that Narrative Science has released as a demo of its 'Freud' analysis. It can look at your Twitter feed and determine what topics you tweeted, what is your tone in those tweets, and what topics your followers wants you to tweet.

This information is obviously useful if you are managing a Twitter account and want to make your followers happy.

Slide 6 - In October 2015, Automated Insights decides to make its "Wordsmith" platform available to the general public. It is currently in open beta, but it should be released this month.

I would like to try this out myself, but I don't really have any dataset to use or examine.

Slide 7 - This appears to be the main guiding principle of both Automated Insights and Narrative Science - produce stories that are geared towards a niche audience rather than generic and widespread.

Both Automated Insight and Narrative Science started their work by producing short newspaper stories about topics that only concerned a few people (earning reports, sports matches).

Both companies *claim* that their company wouldn't take away jobs. Instead their bots will write (a) tedious stories that journalists could never want to do, or (b) stories based on Big Data that journalists could NEVER do.

Slide 8 - This is an example of me taking numbers and converting them to words.

Slide 9 - I am also able to handle two different stats and combine them together to create 4 different possible narratives.