Hi everyone!
This week at the Applied Statistics Workshop we will be welcoming *Michelle
Torres*, a graduate student in Political Science and Statistics at
Washington University in St Louis. She will be presenting work
entitled *Understanding
visual messages: visual framing and the Bag of Visual Words*. Please find
the abstract below and on the Applied Stats website here
<https://projects.iq.harvard.edu/applied.stats.workshop-gov3009>.
As usual, we will meet at noon in CGIS Knafel Room 354 and lunch will be
provided. See you all there!
-- Dana Higgins
*Title:* * Understanding visual messages: visual framing and the Bag of
Visual Words *
*Abstract:* How should one perform matching in observational studies when
the units are text documents? The lack of randomized assignment of
documents into treatment and control groups may lead to systematic
differences between groups on high-dimensional and latent features of text
such as topical content and sentiment. Standard balance metrics, used to
measure the quality of a matching method, fail in this setting. We present
a framework for matching documents that decomposes matching methods into
two parts: (1) a text representation, and (2) a distance metric. We
consider various methods that can be used at each step and conduct a
systematic multifactor evaluation experiment using human subjects to
identify the methods that dominate. We also show that our framework can be
used to produce matches with higher subjective match quality than current
state-of-the-art techniques. We then apply our chosen method to a
substantive debate in the study of media bias using a novel data set of
front page news articles from thirteen news sources. Media bias is composed
of topic selection bias and presentation bias; using our matching method to
control for topic selection, we find that both components contribute
significantly to media bias, though some news sources rely on one component
more than the other.
Hi everyone!
This week at the Applied Statistics Workshop we will be welcoming *Luke
Miratrix*, Professor of Education at Harvard University. He will be
presenting work entitled *Matching with Text Data: An Experimental
Evaluation of Methods for Matching Documents and of Measuring Match Quality*.
Please find the abstract below and on the Applied Stats website here
<https://projects.iq.harvard.edu/applied.stats.workshop-gov3009>.
As usual, we will meet at noon in CGIS Knafel Room 354 and lunch will be
provided. See you all there!
-- Dana Higgins
*Title:* *Matching with Text Data: An Experimental Evaluation of Methods
for Matching Documents and of Measuring Match Quality*
*Abstract:* How should one perform matching in observational studies when
the units are text documents? The lack of randomized assignment of
documents into treatment and control groups may lead to systematic
differences between groups on high-dimensional and latent features of text
such as topical content and sentiment. Standard balance metrics, used to
measure the quality of a matching method, fail in this setting. We present
a framework for matching documents that decomposes matching methods into
two parts: (1) a text representation, and (2) a distance metric. We
consider various methods that can be used at each step and conduct a
systematic multifactor evaluation experiment using human subjects to
identify the methods that dominate. We also show that our framework can be
used to produce matches with higher subjective match quality than current
state-of-the-art techniques. We then apply our chosen method to a
substantive debate in the study of media bias using a novel data set of
front page news articles from thirteen news sources. Media bias is composed
of topic selection bias and presentation bias; using our matching method to
control for topic selection, we find that both components contribute
significantly to media bias, though some news sources rely on one component
more than the other.
Hi everyone!
This week at the Applied Statistics Workshop we will be welcoming *Tianxiao
Shen*, a graduate student at MIT. She will be presenting work entitled*
Language Style Transfer*. Please find the abstract below and on the
Applied Stats website here
<https://projects.iq.harvard.edu/applied.stats.workshop-gov3009>.
As usual, we will meet at noon in CGIS Knafel Room 354 and lunch will be
provided. See you all there!
-- Dana Higgins
*Title:* *Language Style Transfer*
*Abstract:* Recent advances in text generation tasks such as machine
translation and summarization rely on the use of massive amounts of
parallel data, which is costly to collect or nonexistent in many scenarios.
In this talk, I will present a novel model to perform style transfer on the
basis of non-parallel text. This is an instance of a broad family of
problems including machine translation, decipherment, and sentiment
modification. I will talk about how we deal with the challenge to
disentangle content from style, as well as the techniques we use for
adversarial training over discrete samples. I will conclude with the
experiments we design which allow qualitative and quantitative evaluation
of the effectiveness of our method.