I worked as an REU in the Behavior Information Technology (BIT) lab at Michigan State University on the crowdfunding project supported by NSF Grant CCF-1101266. For more information about the BIT Lab click here
Donors Choose is a non-profit crowdfunding platform where teachers can request donations to facilitate classroom projects. If we can better understand what gets a project funded, we can better educate project creators on how to create a successful project. For this project, we were interest in what words, phrases, and topics are more likely to lead to a successful project
Analyzed projects using Linguistics Inquiry and Word Count (LIWC)
Used a generalized linear model to compare LIWC categories to whether the project was funded or not
The LIWC results were not comprehensive alone, but gave insight on how to approach further research. From the LIWC results, we discovered that:
Death words were frequently about the holocaust or war
Colon, SemiC, and Parenth provided more detail about the project
Ingest words were in projects about eating healthier
discrep words told donors what their donation could or would do
future words described what a project will enable students to do
We calculated the frequency of all of the unigrams, bigrams, and trigrams in each project description. For this, we ignored punctuation and capitlization, removed n-grams taht did not occur in all 7 focus areas, removed n-grams taht occured less than 50 total times, and used the sparse model matrix lasso regression to find the correlation between our n-grams and getting funded.
Our n-gram results were much more self-explanatory than the LIWC results. Projects that mentioned technology frequently seemed much less likely to get funded than other projects. This is not because technology projects are more expensive than other projects (we controlled for total price). If your school is from an urban setting, it is positive to mention that, and vice versa if the school is from a rural setting.