7:00 – 10:00, 12:00 – 4:00 VTX
- Got the official word that I should be charging the project for research. Saved the email this time.
- Continuing to work on the papers list
- And in the process of looking at Daniele Quercia‘s work, I found Auralist: introducing serendipity into music recommendation which was cited by
An investigation on the serendipity problem in recommender systems. Which has the following introduction:
- In the book ‘‘The Filter Bubble: What the Internet Is Hiding from You’’, Eli Pariser argues that Internet is limiting our horizons (Parisier, 2011). He worries that personalized filters, such as Google search or Facebook delivery of news from our friends, create individual universes of information for each of us, in which we are fed only with information we are familiar with and that confirms our beliefs. These filters are opaque, that is to say, we do not know what is being hidden from us, and may be dangerous because they threaten to deprive us from serendipitous encounters that spark creativity, innovation, and the democratic exchange of ideas. Similar observations have been previously made by Gori and Witten (2005) and extensively developed in their book ‘‘Web Dragons, Inside the Myths of Search Engine Technology’’ (Witten, Gori, & Numerico, 2006), where the metaphor of search engines as modern dragons or gatekeepers of a treasure is justified by the fact that ‘‘the immense treasure they guard is society’s repository of knowledge’’ and all of us accept dragons as mediators when having access to that treasure. But most of us do not know how those dragons work, and all of us (probably the search engines’ creators, either) are not able to explain the reason why a specific web page ranked first when we issued a query. This gives rise to the so called bubble of Web visibility, where people who want to promote visibility of a Web site fight against heuristics adopted by most popular search engines, whose details and biases are closely guarded trade secrets.
- Added both papers to the corpus. Need to read and code. What I’m doing is different in that I want to add a level of interactivity to the serendipity display that looks for user patterns in how they react to the presented serendipity and incorporate that pattern into a trustworthiness evaluation of the web content. I’m also doing it in Journalism, which is a bit different in its constraints. And I’m trying to tie it back to Group Polarization and opinion drift.
- Also, Raz Schwartx at Facebook: , Editorial Algorithms: Using Social Media to Discover and Report Local News
- Working on getting all html and pdf files in one matrix
- Spent the day chasing down a bug where if the string being annotated is too long (I’ve set the number of wordes to 60), then we skip. THis leads to a divide by zero issue. Fixed now