Paper Reading #22: Personalized News Recommendation Based on Click Behavior (IUI 37)

Comments
http://stuartjchi.blogspot.com/2011/04/paper-reading-19-personalized-news.html 
http://shennessy11.blogspot.com/2011/04/paper-reading-21.html

Reference Information

   Title: Personalized News Recommendation Based on Click Behavior

   Author: Jiahui Liu, Peter Dolan, Elin Rønby Pedersen

   Publisher: IUI '10, February 7-10, 2010 Hong Kong

Summary

Online news outlets have become a very popular way for people to access the news from millions of sources around the world. One challenge these news delivery organizations face is helping their users find interesting articles to read. The voluminous amount of articles available can be overwhelming to users.

Content-based recommendation is a response to this information overload. It plays a central role in recommendation systems. Some systems require users to manually create and update profiles, similar to Google News. Few users may be unwilling or too burdened to take on this extra step.

The researchers in this article have developed a way to automatically construct a recommendation system based on profiles learned from user activity in Google News. They first conducted a large-scale anonymous analysis of Google News users' click logs. Based on this analysis, they constructed a Bayesian framework for predicting users' current interests and trends. They combined the content-based recommendation mechanism which uses learned user profiles with an existing collaborative filtering mechanism to generate personalized recommendations.

Experiments on live traffic demonstrated that the hybrid method improved the quality of news recommendations and increased traffic to the web site.

Discussion
I thought this was interesting because it was something that we kind of talked about in class last week. This however is a little more about discovery of things that people like versus things that they may not know they liked. I use Google News a lot and there really only is a couple of sections that I actively follow. I could see myself using a system like this to recommend things I like.

Book Reading #47: Why We Make Mistakes Microblog

Reference Information

   Title: Why We Make Mistakes

   Author: Joseph T. Hallinan

   Editors: Broadway Books (2009)


Summary
Chapter 8: We Like Things Tidy (16 pages)
In this chapter, Hallinan describes how we organize things within our memory. He talks about the hierarchical nature in which we like things organized. In addition, he gives a few examples about how people remember things. One of them was how people drew the Seine River much straighter than it actually was. He also goes on to talk about how people remember things. A lot of people will rationalize memories and change them, leaving out or making up details as they go. These added details cause them to remember events differently from how they happened.


Chapter 9: Men Shoot First (15 pages)
Hallinan compares and contrasts men and women. There is a relationship between overconfidence and perceived risk. Women are much less confident than men in several areas. One of the examples he cites is driving and fixing bugs in Excel. He goes on to talk about how boys tinker and explore further than girls.


Discussion
In regards to chapter 8, this reminds me of the "Lost in the Mall" experiment talked about in Opening Skinner's Box. I know that I have certainly done that before...left out facts that would've made my story less interesting or not applicable to what is being currently talked about. Chapter 9 had a lot of similarities to what ends up getting discussed in my Sociology class. We talk a lot about gender gaps and things like that. 

Paper Reading #21: Addressing the Problems of Data-Centric Physiology-Affect Relations Modeling (IUI 36)

Comments
http://pfrithcsce436.blogspot.com/2011/04/paper-reading-21-supporting-exploratory.html
http://shennessy11.blogspot.com/2011/04/paper-reading-22.html


Reference Information

   Title: Addressing the Problems of Data-Centric Physiology-Affect Relations Modeling

   Author: Roberto Legaspi, Ken-ichi Fukui, Koichi Moriyama, Satoshi Kurihara, Masayuki Numao, Merlin Suarez

   Publisher: IUI '10, February 7-10, 2010 Hong Kong


Summary

The researchers of this paper used a data centric approach to come up with a way to use machine learning to define "affective states". They used an EEG helmet, the researchers attempted to classify and define emotions of the wearer. A problem they ran into was the enormous size of the data set.

Emotion sensing algorithms are generally O(n^2) or O(n^3) complexity which obviously slows the entire process to a crawl. Another problem they encountered was sensing changes in emotion. Emotions can change rapidly, perhaps before analysis can be completed. Several changes to the algorithms were suggested which looked to improve the performance of the machine.

Discussion
This paper was very technical. The thoughts behind it were interesting but there are so many complex theories and equations behind all of it it's hard to understand. The thought of classifying and defining emotions seems very foreign.

Paper Reading #20: Automatically Identifying Targets Users Interact with During Real World Tasks

Comments
http://shennessy11.blogspot.com/2011/04/paper-reading-19.html 
http://stuartjchi.blogspot.com/2011/04/paper-reading-20-automatically.html

Reference Information

   Title: Automatically Identifying Targets Users Interact with During Real World Tasks

   Author: Amy Hurst, Scott E. Hudson, Jennifer Mankoff

   Publisher: IUI '10, February 7-10, 2010 Hong Kong

Summary
Information about location and size of targets on a screen has been a popular area of research. The ability to analyze user actions in that environment is essential for answering questions about usability, performance and daily use. Knowing the target size and location is necessary in assessing the pointing performance of impaired individuals.

The researchers in this article attempted to develop a slightly new Accessibility API. They relied on Microsoft's existing API (MSAA API) combined with a hybrid solution of their own that relied on machine learning and computer vision. They found that their hybrid approach resulted in a 75% success rate. They looked at 8 popular applications: MS Outlook, web browsers, MS Word, Windows Explorer, Media Player and MS PowerPoint.

The applications of their work include improved computer accessibility,  support for automatic extraction of a task sequence, and automatically scripting common actions. CRUMBS was used to capture information about the interaction. This is currently only limited to Microsoft machines.


Discussion
I'm glad they included pictures in this because I don't think I would've been able to understand it otherwise. That said, the pictures really help to illustrate their computer vision and machine learning algorithms. I didn't understand a ton of what they talked about but I did understand their general idea.

Full Blog: Things That Make Us Smart

Reference Information

   Title: Things That Make Us Smart

   Author: Donald A. Norman

   Editors: Broadway Books (1993)


Summary

The beginning of this book talks about how technology should be more human-centric. Technology can aid people but it can also make them dumb and dependent. Hallinan talks about how some technology is designed to aid people and how that same technology can confuse and interfere with workflows. He also talks about hard and soft sciences and the different types of cognition, reflective and experimental.


Norman goes on to talk about museums and how little they actually teach us because our attention span is so low. He expands more on experimental and reflective cognition, talking about how we must find balance between the two. He goes on to describe three types of learning: accretion, turning and restructuring.


In the last chapter Norman discusses artifacts and things to consider when fitting an artifact to a person. Surface artifacts are everything on the surface, or rather, what we see is all there is and internal artifacts are information that is represented internally. He gives a few examples such as the Tower of Hanoi to demonstrate that problems that are the same can appear different due to the amount of information present in an environment.


Discussion
I actually really liked this book. It was repetitive but not in the way the other Norman books were. I felt like he gave a lot of good examples to illustrate his point. I really liked what he had to say regarding the three puzzles and how the same problems get interpreted differently by different people.

Book Reading #46: Why We Make Mistakes Microblog

Reference Information

   Title: Why We Make Mistakes

   Author: Joseph T. Hallinan

   Editors: Broadway Books (2009)


Summary
Chapter 6: We're in the Wrong Frame of Mind (18 pages)
In this chapter, Hallinan talks about framing. Framing is essentially how we view something. Hallinan gives a lot of examples in this chapter. He talks about the time that we take to make decisions can affect the outcome (immediate or future), multiple-unit ricing, wine buying based on the music and store tags. Customers will key in on the first part of tag and will have an "anchor" as to how many to buy.


Chapter 7: We Skim (9 pages)
In this chapter, Hallinan talks about how we skim things. He gives a lot of examples of how we skim material and the trade-off along with it. Hallinan talks about how we miss a lot of important details and cites a rookie piano player that noticed an error that went unnoticed for several years. We only read the first few letters of a work and decide to assume the rest. Context is important when recognizing and remembering information.


Discussion
I really liked what he talked about in regards to skimming. I skim a lot nowadays, mostly due to the insane amount of readings we have. I really like what we read but I wish we could have a little more time and have a little less to do. I feel like I would be able to focus more on the book at hand if we weren't reading 3 at a time.

Book Reading #45: Things That Make Us Smart Microblog

Reference Information

   Title: Things That Make Us Smart

   Author: Donald A. Norman

   Editors: Broadway Books (1993)


Summary
Chapter 3: The Power of Representation (34 pages)
In this chapter Norman discusses the use of external aids and how they make us smart. He deems the most important of these the paper and pencil. He discusses cognitive artifacts and how people use them to keep track of complex events. There are two parts of a representational system: the represented world and representing world. Examples he uses are: getting flight information, representing numbers, filling medical prescripts and tic tac toe. He talks about how a person represents something makes the task related to the representation easier/more difficult to do.


Chapter 4: Fitting the Artifact to the Person (38 pages)
Norman continues to discuss artifacts and how an artifact fits a person. He talks about the difference between surface artifacts and internal artifacts. Surface artifacts are everything on the surface, or rather, what we see is all there is and internal artifacts are information that is represented internally. He gives a few examples such as the Tower of Hanoi to demonstrate that problems that are the same can appear different due to the amount of information present in an environment.


Discussion
I liked what Norman talked about when he discussed representing numbers. Also, the examples he used in chapter 4 were really good. I also liked when he talked about the different ways information is represented and how the best way depends on the information and task performed.