The dynamics of large populations—from social movements to natural disasters—now unfolds in the digital commons. My research is motivated by a desire to understand, and democratically enhance, collective behaviour.
Academically, I pursue that through research in social media, machine learning, and natural language analysis. Commercially, I build out data pipelines and reporting systems (business intelligence). I connect users to data and help them learn to love to use it. I seed or grow organizations' data-literate culture. I apply various statistical and machine learning techniques, including recommender systems, clustering, visualization, data mining, and natural language analysis.
I frequently give presentations, training courses, and teach in various formats in topics relating to computer science, social informatics, machine learning, and data usage training in organizations for various types of users.
It’s not too hard to randomly generate samples from a discrete set of possible outcomes, given you have the probability for each outcome. But what about when the number of outcomes is very large? A naive approach will take time linear in the number of outcomes, but the alias method gives samples in constant time! My implementation gives on-demand samples from a categorical distribution with better performance than numpy. Useful for generating random words out of a large vocabulary based on a unigram distribution.
I recently implemented word2vec in python. It's backed by unit tests, and you can pip install theano-word2vec. I found this helpful to learn the assumptions and workings of word2vec, and to gain more experience with Theano and Lasagne. Let me show you how I did it, and how you can import it into your deep learning projects.
As someone who works a lot with text, I often need to compare two text strings having some unknown number differences resulting from encoding artifacts, escaped characters, and random errors. The dynamic time warp algorithm can do this, but it's just too slow! What do I do?
The tree frog can sense the mood of other creatures because of its extraordinarily sensitive skin that detects minute quantities of airborne hormones. This keen sensitivity allows the tree frog to see things that go completely unnoticed by most other creatures. It is an impeccable judge of character, and chooses to spend its time near creatures of good heart. The tree frog makes no wasted movements—it finds advantage by sensing the world deeply and patiently waiting in the right place at the right time. Though it usually keeps away from danger, the tree frog is weak to physical and chemical insult, and may fall ill or die with little warning. Give the tree frog to someone who saw the potential in your soul and chose to invest in you, as a way to thank them, and assure them that they are wise because you will make good on their investments.
The sea turtle is solitary, tranquil, and caries a sense equanimity and peace. But, speak to one and you will soon understand why they spend their time alone: they have a rather off-putting sense of humour! Its sharp wit leaves no one unscathed, not even itself. Too wise to become caught up in trivialities, too jaded to be caught by false promises, the turtle is happy to make its own way under the waves. Give the sea turtle to a colleague who you admire for being principled, heterodox, clear-thinking, and unmoved by politics or social convenience.
The scorpion is tough, loyal, and persistent. Its poisonous sting is deadly. However, it moves rather slowly, and cannot defend itself from fast-moving enemies. Despite its deadly aspect, the scorpion seeks long term alliance and will defend its allies and their mission to the death. Give the scorpion to thank a loved one for their longstanding loyalty and service. It may also be given in acknowledgement when freeing them from the bond that they so dutifully kept.
The octopus is wiley and dangerous to its foes. Smart, strong, and fast, its only true weakness is sticking its beak into too much of other people's business. The octopus is a natural multitasker, and is comfortable juggling so many things that would boggle the mind of most creatures. Give the octopus as a sign of respect and awe for the ambitious go-getter who has for a time given you their precious, if fragmented, attention.
The little bird has a fleet spirit, and can reach far off places before nightfall. Sometimes capricious, the little bird will flee from danger, but will dutifully deliver any entrusted message. Give the little bird to one who must be far away, in world or spirit, that you may stay connected.
Edward Newell, Kian Kenyon-Dean, Jackie Chi Kit Cheung. Deconstructing and reconstructing word embedding algorithms. arXiv:1911.13280, 2019.
Edward Newell, Drew Margolin, Derek Ruths. An Attribution Relations Corpus for Political News. Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC), 2018.
Edward Newell, Jackie Cheung. Constructing a Lexicon of Relational Nouns. Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC), 2018.
Edward Newell, Ariane Schang, Drew Margolin, Derek Ruths. Assessing the Verifiability of Attributions in News Text. Proceedings of the Eighth International Joint Conference on Natural Language Processing (IJCNLP), 2017.
Edward Newell, Derek Ruths. Using Relational Framing to Characterize Geographical Differences in Views on Immigration. Workshop on Social Media in Demographics Research, 2016.
Edward Newell, Derek Ruths. How One Microtask Affects Another. ACM Conference on Human Factors in Computing Systems, 2016
Edward Newell, Stefan Dimitrov, Andrew Piper, Derek Ruths. To Buy or to Read: How a Platform Shapes Reviewing Behavior. AAAI Conference on Web and Social Media, 2016.
Edward Newell, David Jurgens, Haji Mohammad Saleem, Hardik Vala, Jad Sassine, Caitrin Armstrong, and Derek Ruths. User Migration in Online Social Networks: A Case Study on Reddit During a Period of Community Unrest. AAAI Conference on Web and Social Media, 2016.