Per Ola Kristensson | Research Projects

Blog
Publications
Software
Other Stuff

This page is intended to describe completed research projects. It is under construction.

Automating repetitive tasks in web applications

(with Zachary Bray)

Many tasks in web applications are highly repetitive. The research field that tries to automate tedious tasks for users is called Programming by Example (PBE). In this paper we contribute to this field by introducing a new set of algorithms to learn users' intended tasks within a web browser. Our technique is based on using ensembles of decision trees, in which each individual decision tree represents a particular hypothesis about the steps the user is trying to perform. The system is implemented as a Firefox web browser add-on. For two difficult use cases we have demonstrated that this system quickly converges to the solution, even in the presence of high noise.

Research paper:

Parakeet: Efficient Mobile Speech Recogniton

(with Keith Vertanen)

Speech recognition has undergone tremendous progress in the recent decades. However, an Achilles heel of speech recognition is the difficulty of correcting inevitable recognition errors. We developed Parakeet: a mobile continuous speech recognition system that enables users to correct recognition errors using a touch-screen interface. The system uses a word-confusion network to reveal the next best hypotheses in an efficient way to the user. There are also a number of other user experience enhancements in this system, for details see the publications below. In a formative study we found that novice users could enter text at a text entry rate of 13 wpm while simultaneously walking outdoors during a windy day. Ignoring rather long recognition delays, an expert user reached a text entry rate of 45 wpm for the same task. This indicates that mobile speech recognition holds tremendous potential as an efficient mobile text entry method, as long as users are provided with an efficient correction interface.

Press / media:

Most important research papers:

The efficacy of handwriting recognition

(with Leif Denby)

Research paper:

The efficacy of space time cube visualization

InfoTouch: Multitouch and Visualization

(with Olof Arnell, Anneli Björk, Nils Dahlbäck, Joakim Pennerup, Erik Prytz, Johan Wikman and Niclas Åström)

In this project we explored how a multitouch table could be used to encourage exploration in information visualization. Our application let users explore tagged images in Flickr by manipulating an interconnected tag cloud. Users could use both hands to push and pull tags together. When users pulled tags into a certain area the system showed the user a selection of photos with the corresponding tags. We discovered that users generally liked the system and that they used multitouch interaction about 10% of the time. This indicates that multitouch may be a useful complement to information visualization displays.

Research paper:

Gesture keyboard Tech (SHARK/ShapeWriter)

(with Shumin Zhai)

This technology enables users to write quickly on their mobile phones by sliding or swiping the finger over a touch-sensitive on-screen keyboard. For instance, to write the word "the" the user pushes down the finger on the T key, slides to the H and E keys, and then lifts up the finger on the E key. This input paradigm recognizes the shape of this gesture using a pattern recognizer. During practice, the shapes of words builds up in users' muscle memory which enables users to quickly recall the shapes for words without looking much at the keyboard (similar to how you remember your ATM code). The basic research was first published in 2002 and 2003. In 2004 we released a free public release (called SHARK Shorthand) that was available for download on the IBM alphaWorks website and hailed as "text entry epiphany" by jkOnTheRun, a top ten tech blog. In 2007 we founded ShapeWriter, Inc. to commercialize the technology. We made several public releases of ShapeWriter software for iPhone, Android and Windows Mobile devices. In 2010 the company was acquired by Nuance Communications, Inc.

Tech start-up:

Press / media:

Most important research papers: