I have this lovely ‘Keep Pursuing’ backpack with the new EV logo (retail $295). I wanted to give it to the person in our team that has a proven track record of (1) collaborating with others, (2) being helpful, & (3) adding true technical value to our team. So maybe we start by looking at the tacos people on our team have recieved.
But wait!...
As we are all aware, tacos are inflated within the EV institution. People seem to get and give tacos for just showing up to work. So let’s make this more rigorous.
To create a level-playing field across our team, I used the taco stats starting Jan. 20, 2019 to March 1, 2019. This ensures that our recent new hires are on the taco leaderboard and not starting with a handicap. I am really only interested in who on our team has been getting tacos. (This is a one-sided perspective, but so be it.) I immediately removed anyone from the dataset who received tacos but whom I do not consider core to the ML team (at this time). Team members were allowed to receive tacos by people outside of the ML team. The ML team is defined as:
team =['tony.wang','alex.westover','allison.lechner','anton.sakharov','dan.chen','daniel.keogh','david.murr','haylie.murray','indra','shay.strong','jieun.lee','josh.leaverton','kai.johnson','lilly.thomas','paul.park','oriol.torras','robert.miller',’sng’,’abbey.moore']
I had to normalize some of the names of people, such as ‘robert.miller’ and ‘robert_miller', since there had been the recent slack account updates that caused changes to some original usernames.
I then used an NLP (natural language processing) model in python called VADER (Valence Aware Dictionary & sEntiment Reasoner) to rank the messages sent with embedded tacos as positive, neutral, or negative. As you might surmise, a lot of the language used in the taco messages are truncated, tweet-like statements that could skew the algorithm’s judgment.
If VADER deemed the output neutral, the taco received was weighted by 0.5x. Positive and negative assessments were weighted 1x and -1x, respectively.
From this analysis, here are the top 5 recipients of the most meaningful tacos:
| team member | weighted tacos |
|---|---|
| jieun.lee | 62.0 |
| kai.johnson | 39.5 |
| tony.wang | 20.0 |
| paul.park | 15.0 |
| josh.leaverton | 13.0 |
Jieun is the winner of the backpack!! If she does not want the backpack, it will go to the next in line, and so on.
Here is our Taco GraphTM:
You'll notice lines & nodes (circles). Lines represent the exchange of tacos. The color of the line shows the sentiment, red is negative, orange/yellow is neutral, and green is positive. The line thickness is the number of tacos given. Arrowheads show the from:to relationship (but may be hard to see on this visualization). The node size scales with the number of outgoing tacos given.
I think the Taco GraphTM will be updated to either include a better language model, or perhaps more weighting factors to understand the significane of tacos given (aka the Golden TacosTM). TBD.