Matt Marx Profile picture
@Cornell entrepreneurship/innovation professor & NBER research associate. Open datasets: https://t.co/IN9SXP0bXn. Former startups: Speechworks, Tellme, Vlingo.
May 24, 2023 6 tweets 3 min read
Thank you to the organizers of #oisconf23, who let me zoom in when teaching ruled out Vienna😢

At the Open Innovation in Science Research Conference I announced preview release of a new open dataset: Patent Paper Pairs (PPPs)

Available now at

🧵1/ https://t.co/8X5v0OrIkprelianceonscience.org
twitter.com/i/web/status/1… Where does scientific discovery and commercial technology intersect?

Scholars incl. @Fiona_MIT @sstern_mit @fab_montobbio @leflix311 @HSauermann @dfehder @ProfNeilT @ArhoSuominen @SamiraRanaei @SamiraRanaei @ProfAroraAshish @SharonBelenzon have built PPPs to study this

2/
Mar 18, 2020 10 tweets 4 min read
I'm relieved that my first online class session was not a disaster. so anxious I could not sleep the night before. some learnings from first time through:

1/ logged in 15m early, half the students were already there! great conversations, got to know them much better. 2/ did a pre-class Time Zone poll to figure out who would be groggy. shout-outs to those logging in after midnight!

3/ instead of sitting at desk w/bulky headset mic, stood by whiteboard with camera @ eye level + wireless mic. students said it felt more "classroomy"
Sep 24, 2019 4 tweets 1 min read
Stata hack for fuzzy-matching large datasets when waiting for Levenshtein or bigram matching seems interminable:

1/ generate the soundex for the matching key in both datasets. this is a quasi-phonetic representation of the first several syllables of the phrase 2/ joinby the two datasets on the soundex match. note THIS WILL FAIL if the beginning of the strings is crufty. (in my experience suffixes tend to be a bigger problem, at least for company names: Inc Corp LLC etc.)

3/ run expensive fuzzy matching on the subset that share soundex
Aug 28, 2019 13 tweets 3 min read
nobody asked but here are some teaching tips:
1/ first day, show up extra early. wander around the room introducing yourself to students before class.
2/ send around a (voluntary) survey asking about their backgrounds/interests/etc. & then work in a couple of those per class, e.g it's a case on XYZ Pharmaceuticals, so "Pat, you're premed / worked in pharma - what are we missing here?" give them a chance to shine whenever possible.
3/ if you are teaching case method, make everyone aware that class participation is an *inherently unfair* grading method b/c