show and tell for hackers in DC.

Round 38: Return of the Function

15 Nov 2016

Marie Whittaker - @mariecwhittaker

I made a thing! The thing is the Techlady Hackathon Visual Collateral for this year. The way I made the thing is with Adobe Illustrator and my creative juices. I’ll talk about those things.

4th TechLady Hackathon – day of learning and hacking in late October

Marie created visual materials to help brand the hackathon as its own standalone event

Logo was used for stickers and selfie machine

Identifont: which font did someone use in their materials? This helps designers narrow down which font was used when you don’t know or can’t ask the designer who made it.

Used identifont and Adobe Creative Cloud.

Adobe Illustrator: traced over an similar tattoed-knuckles photo, then made it cartoony by tracing over it with the pen tool; adding shading

Result was an SVG – specifically so it would be vector (scalable up or down), which is better when working with text and when working with lines. The edges are crisper, and it’s scalable without losing any image quality.

Inspiration came from a Data Viz leather jacket and the prompt to make something “badass”

Good logo design conveys emotion

Kate Rabinowitz - @datalensdc

Gender diversity of major tech + data meetups plus a way to help!;

How many speakers of DC Tech Meetups were female?

Used Python & the Meetup API to collect the data; used pronouns to code the gender.

Focused on Meetup with 1000+ members, and narrowed it to Meetups that focused on

In every case, there were zero female speakers at single-speaker events, across all DC Tech Meetups.

Multi-speaker events weren’t much better

Diversity in meetups is a good thing!

So Kate built to build a bridge: this is a directory of talented, creative women who should be speaking at your Meetups!

Having an enforceable Code of Conduct also helps to make it so that Meetups can be a safer, better space for all attendees

Psst: DC Hack And Tell’s Code of Conduct is here:

Daniela Noir

Metro pricing comparison map

This compares how long it will take and how much it will cost to go from one Metro Station to another.

Select one station and hover over the second station in order to get the off-peak and peak fares.

Javascript, D3, Ajax, Node, SQL, Microsoft Azure

Pulled all of the (very messy) data initially from the WMATA API; added unique IDs to each metro station; added columns to create every possible combination of stations and then added in the fares

The map is a static background; each station point is an SVG; the coordinates of each station are calculated with Javascript

Structure is all there so it wouldn’t be too difficult to repurpose for another city.

Matthew Coates - @PeaceTechLab

OSRx is designed as a portal for peace builders and other actors to monitor and visualize various conflict data. Currently, two large event databases are represented. Various social media monitors powered by Crimson Hexagon are also shown. In the future, specific topic areas will be featured (for now, see the South Sudan page).

Open Situation Room Exchange – Mapping and Visualizing Conflict Data

Producing data for peacebuilders

Real-time data (social media, news reports)

Periodic indices (World bank, economic stability, etc)

Local data & insights (crowdsourced data sets, etc)

Overview by country; two different databases.

Social media analysis (Crimson Hexagon) monitors social media and visualizes conversations by topic

GDELT Instability comparison: normalizes conflict news vs. all other news to give a % of how much news about conflict there is to give a rough metric of how much conflict there is

Chris Nguyen - @uncompiled

Alexa + Metro = :emoji_1f525:

Flash briefings!

Most Alexa flash briefings are Customized weather reports or News alerts

Expected response format is XML or JSON

Chris set up a flash briefing to pull from the WMATA API and uses Alexa to speak the – Fetch rail incidents

Uses Scala microservice to process the data (roughly 5 lines)

WMATA uses shorthand notation in their incident API, which doesn’t sound natural if sent through text to speech so it needs to be normalized.

When latency is a constraint, avoid Java (the JVM is super slow to start).

Output the results of a lambda function to an S3 bucket

Jessica Garson - @jessicagarson

Tech Lady Hackathon

Started in 2013

Full of training, workshops, and projects

Second Tech Lady Hackathon in 2014 was where Buscando was created

About 160 women at Techlady Hackathon #4

Featured guest: DC CTO Archana Vemulapalli

Workshops on everything from intersectional feminism and user experience to front-end development and security

Trended on Twitter!

Usually happens once a year; might happen more frequently as well!

What if it happened in multiple cities?

People of all skill levels coming together – for most attendees, this was their first hackathon!

Travis Hoppe - @metasemantic

Trained tiny neural nets to emulate authors, and then have them reproduce text from another genre. Quantifiable author-to-author measures!

The data we feed a model shapes its personality

If each bot (a deep learning model) is a book: Wizard of Oz bot, Wuthering Heights bot, etc.

But what if the Edgar Allen Poe bot reads Wuthering Heights?

Top 100 books pulled from project Gutenbergs. Manually cleaned headers and footers

Each book is used to train one bot Long short term memory Recurrent neural network. Each book took about 1.5 hours to train. Did 80 books

Bots get confused when they see new text (a different book); we can measure that confusion to see author similarity

The Pride and Prejudice bot is least confused by other Jane Austen books. The King James Bible bot is not similar to anything else

Created a similarity map. Comparing two bots takes several seconds, and there are N^2 comparisons

Could a lot of similarity be attributable to say books written in the same tense/ perspective?

…Maybe you do that hack!