I came across Peggys Twitter request:
I'm looking for some badass ladies who @GraphQL to follow on Twitter. ν ½ν²ν ΌνΏ½ν ½ν²ͺ
— Peggy Rayzis ν ½ν±©ν ΌνΏΌβν ½ν²» (@peggyrayzis) August 29, 2017
Who are your favs? Comment below & RT please!
Which got a really response from Bonnie that warmed my heart:
Hi Peggy! ν ½ν± I'm talking about @GraphQL @apollographql @neo4j at @AngularMix in Orlando in October, super excited! β¨ https://t.co/n8GuRlJbZk
— Bonnie Brennan (@bonnster75) August 29, 2017
And triggered an idea. As you might know, we’re having a lot of fun showing the impressive engagement of developers in their communities (e.g. GraphQL, Neo4j, …) in a single place by importing them into a "Community Graph". Usually it is really hard to follow the flurry of activity across Twitter, Slack, StackOverflow, GitHub, etc. to keep on top of whats happening. Especially if your community is growing quickly.
So we scratched an itch and imported the activity for the Neo4j community into a single graph, which can then be queried and visualized and which is accessible here.
We did the same for the GraphQL community, their data is also accessible via GraphiQL and documented here.
So as we had all the activities of GraphQL of the last few months in our graph database, I thought it would be cool to use it to answer Peggys request.
I googled for "gender api" and found http://gender-api.com which looked really nice and with 500 free monthly requests and a simple HTTP API. Perfect for my goal.
I tested a few of the names with that came back to Peggys request, unfortunately only a few got recommended: Peggy, Bonnie, BelΓ©n, Robin and Morgan. So I hoped that I could do better.
I used the interactive check at the gender-api homepage, which resulted in these. I had to change the country to "US" as my default ("DE") didn’t have the correct mapping for Robin and Morgan.
| name | response
| Peggy | {"name":"peggy","country":"US","gender":"female","samples":3015,"accuracy":99,"duration":"51ms"}
| Bonnie | {"name":"bonnie","country":"US","gender":"female","samples":3984,"accuracy":98,"duration":"25ms"}
| Morgan | {"name":"morgan","country":"US","gender":"female","samples":5956,"accuracy":76,"duration":"33ms"}
| BelΓ©n | {"name":"belΓ©n","country":"US","gender":"female","samples":35,"accuracy":97,"duration":"64ms"}
| Danielle | {"name":"danielle","country":"US","gender":"female","samples":12284,"accuracy":99,"duration":"47ms"}
| Robin | {"name":"robin","country":"US","gender":"female","samples":8088,"accuracy":83,"duration":"31ms"}
Cool so I had some baseline data, I only gonna look at results with an accuracy of > 75 and at least 10 samples.
I can execute the HTTP API like this: https://gender-api.com/get?key=<key>&country=US&name=peggy
Back to our community graph, which you can access here: http://107.170.69.23:7474/browser/ using "graphql" as username and password.
Let’s just see if one of the women tweeting about GraphQL is in here, and show her tweets and their tags.
MATCH (user:Twitter:User)-[:POSTED]->(t:Tweet)-[:TAGGED]->(tag:Tag)
WHERE user.screen_name = 'bonnster75'
RETURN *
Now I wanted to test the gender api call with our known names.
So we find twitter users by a list of screen-names, and split their name by space and take the first word. Which we then send to the "gender-api" API and get the result back as an map-value. We only want to return a few attributes from our user node.
MATCH (user:Twitter:User) WHERE user.screen_name IN ['bonnster75','peggyrayzis','okbel','morgancodes','robin_heinze','danimman']
WITH user, head(split(user.name," ")) as firstname
CALL apoc.load.json("https://gender-api.com/get?key=<key>&country=US&name="+firstname) YIELD value
RETURN user { .screen_name, .name, .followers, .statuses} as user_data, firstname, value;This worked, only Morgan would probably not show up, b/c she has not tweeted.
| {"name":"Bonnie Brennan","screen_name":"bonnster75","followers":"467","statuses":"2831"}
| {"name":"bonnie","accuracy":"98","samples":"3984","country":"US","gender":"female"}
| {"name":"BelΓ©n Curcio","screen_name":"okbel","followers":"3821","statuses":"35721"}
| {"name":"belΓ©n","accuracy":"97","samples":"35","country":"US","gender":"female"}
| {"name":"Morgan Laco","screen_name":"morgancodes","followers":null,"statuses":null}
| {"name":"morgan","accuracy":"76","samples":"5956","country":"US","gender":"female"}
Now we want to find the "most active" women who tweet about GraphQL. A "score" could contain the number of tweets, and how often those have been favorited, retweeted or replied to. This is what we do, we find twitter accounts with tweets, compute our score per user and return the top 500 sorted by score.
MATCH (u:Twitter:User)-[:POSTED]->(t:Tweet)
WITH u, count(*) as tweets, sum(t.favorites+size((t)<-[:RETWEETED|REPLIED_TO]-())) as score
WHERE tweets > 5 AND tweets * score > 100
RETURN u.name, u.screen_name, tweets, score
ORDER BY tweets * score DESC LIMIT 500Looking at the results it also makes sense:
ββββββββββββββββββββββββ€ββββββββββββββββββ€ββββββββ€ββββββββ β"u.name" β"u.screen_name" β"tweets"β"score"β ββββββββββββββββββββββββͺββββββββββββββββββͺββββββββͺββββββββ‘ β"Sashko Stubailo" β"stubailo" β"538" β"1567" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"Apollo" β"apollographql" β"150" β"1389" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"ReactDOM" β"ReactDOM" β"221" β"596" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"KOYCHEV.DE" β"K0YCHEV" β"309" β"341" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"Graphcool" β"graphcool" β"84" β"859" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"adeeb" β"_adeeb" β"179" β"328" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"ReactJS News" β"ReactJS_News" β"93" β"517" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"Max Stoiber" β"mxstbr" β"102" β"450" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"Caleb Meredith" β"calebmer" β"135" β"273" β ββββββββββββββββββββββββΌββββββββββββββββββΌββββββββΌββββββββ€ β"Lee Byron" β"leeb" β"53" β"652" β
Cool, now we can combine our two statements. To save some repeated API calls, I just store the gender information on the account (also the accuracy and samples) so that we can reuse it later.
MATCH (u:Twitter:User)-[:POSTED]->(t:Tweet)
// name has have at least 2 parts, and gender not yet retrieved
WHERE u.name contains " " AND NOT exists(u.gender)
WITH u, count(*) AS tweets, sum(t.favorites+size((t)<-[:RETWEETED|REPLIED_TO]-())) AS score
WHERE tweets > 5 AND tweets * score > 100
WITH u, tweets, score, head(split(u.name," ")) as firstname
ORDER BY tweets * score DESC LIMIT 500
// call gender api
CALL apoc.load.json("https://gender-api.com/get?key=<key>&name="+firstname) YIELD value
// set result values as properties
SET u.gender = value.gender, u.gender_meta = [value.accuracy,value.samples]
RETURN count(*)So we get our the 500 top accounts with a space in their name resolved via the API.
Now we can look at our data, and hopefully find some accounts that we can recommend to Peggy.
MATCH (u:Twitter:User)-[:POSTED]->(t:Tweet)
WHERE u.gender = "female" and u.gender_meta[0] > 75 and u.gender_meta[1] > 10
WITH u, count(*) AS tweets, sum(t.favorites+size((t)<-[:RETWEETED|REPLIED_TO]-())) AS score
ORDER BY tweets * score DESC LIMIT 50
RETURN u { .screen_name, .name, .followers, .following, .statuses} as user, tweets, score;user |
tweets |
score |
ok |
{name: "Danielle Man", statuses: 155, followers: 581, screen_name: "danimman"} |
38 |
76 |
β |
{name: "Peggy Rayzis", statuses: 802, followers: 1870, screen_name: "peggyrayzis"} |
28 |
86 |
β |
{name: "Jess Telford", statuses: 5022, followers: 1241, screen_name: "jesstelford"} |
37 |
42 |
x |
{name: "Tracy Lee ladyleet", statuses: 53446, followers: 19030, screen_name: "ladyleet"} |
13 |
74 |
β! |
{name: "Bonnie Brennan", statuses: 2831, followers: 467, screen_name: "bonnster75"} |
24 |
39 |
β |
{name: "Ruby Inside", statuses: 4706, followers: 41121, screen_name: "RubyInside"} |
6 |
78 |
x |
{name: "Brittney Bond", statuses: 20753, followers: 310, screen_name: "_KarimaTounsya"} |
31 |
15 |
β! |
{name: "Jira Vinyoopongphan", statuses: 1558, followers: 243, screen_name: "thekamahele"} |
21 |
22 |
β! |
{name: "Laura Linda Laugwitz", statuses: 10007, followers: 1421, screen_name: "lauralindal"} |
9 |
29 |
β! |
{name: "Thea Lamkin", statuses: 882, followers: 580, screen_name: "thelamkin"} |
9 |
12 |
β! |
{name: "Eve Porcello", statuses: 1334, followers: 626, screen_name: "eveporcello"} |
8 |
13 |
β! |
{name: "Nitya Narasimhan", statuses: 15764, followers: 3568, screen_name: "nitya"} |
6 |
9 |
β! |
{name: "Nastia Bratochkina", statuses: 327, followers: 88, screen_name: "Sestri4kina"} |
7 |
6 |
β! |
{name: "BergΓ© Greg", statuses: 3229, followers: 631, screen_name: "neoziro"} |
8 |
5 |
x |
{name: "Samantha Demi", statuses: 93214, followers: 2311, screen_name: "queersorceress"} |
6 |
5 |
β! |
{name: "Jen Luker", statuses: 1731, followers: 468, screen_name: "knitcodemonkey"} |
6 |
4 |
β! |
{name: "FΓΌsun Wehrmann", statuses: 924, followers: 570, screen_name: "fuesunw"} |
7 |
3 |
β! |
{name: "BelΓ©n Curcio", statuses: 35721, followers: 3821, screen_name: "okbel"} |
18 |
1 |
β |
{name: "Krisztina Hirth", statuses: 7605, followers: 394, screen_name: "YellowBrickC"} |
6 |
3 |
β! |
{name: "Berrak", statuses: 46891, followers: 1525, screen_name: "toggleModal"} |
6 |
2 |
β! |
{name: "Hera Mudho", statuses: 1333, followers: 160, screen_name: "wuodaboge"} |
10 |
1 |
x |
{name: "Brooke Smith", statuses: 460, followers: 93, screen_name: "BossOSmith"} |
7 |
1 |
x |
{name: "Else if", statuses: 39079, followers: 1693, screen_name: "getelseif"} |
7 |
1 |
x |
{name: "Ruby on Rails News", statuses: 3563, followers: 139, screen_name: "RubyonRailsNew1"} |
11 |
0 |
x |
{name: "Amelia Warner", statuses: 8641, followers: 2689, screen_name: "facetimeJS"} |
11 |
0 |
β! |
{name: "Lori Wilkins", statuses: 6278, followers: 565, screen_name: "webnerd_lw"} |
14 |
0 |
β! |
{name: "Reme J", statuses: 38284, followers: 828, screen_name: "remelehane"} |
41 |
0 |
x |
{name: "Hanène Maghrebi", statuses: 762, followers: 4300, screen_name: "SmartSemantics"} |
6 |
0 |
β! |
{name: "Nati Namvong", statuses: 7233, followers: 464, screen_name: "_nati"} |
12 |
0 |
? |
{name: "Kelly Goetsch", statuses: 878, followers: 417, screen_name: "KellyGoetsch"} |
6 |
0 |
x |
{name: "Piyali gradgospel", statuses: 2142, followers: 458, screen_name: "gradgospel"} |
6 |
0 |
β! |
{name: "Camilla AEscΓ³rcio", statuses: 14899, followers: 939, screen_name: "benedictusSit"} |
9 |
0 |
β! |
Besides the funny (Ruby Inside, Else if) and the incorrectly classified (Jess,Brooke), we get a number of active women in the GraphQL community that were not recommended before: ladyleet, _KarimaTounsya, thekamahele, lauralindal, thelamkin, eveporcello and more.
I manually went over the screen-names, looked at these twitter profiles and set a check-mark β for female accounts and an ! for new names.
We found 22 women in total which is of course not a lot if you look at the totals, but hopefully a growing number.
And I hope that Peggy, Bonnie and the others will engage with each other (and everyone else) and can encourage more women to become active in the GraphQL community.
PS: We heard from Nikolas, the curator of @graphqlweekly, that our "this week in GraphQL" overview page helped them a lot, compiling the weekly newsletter. It also features an "Twitter Active" tab which should help you too to find people to follow.
We’re happy to offer the community graph service also to other communities, feel free to reach out to us via devrel@neo4j.com, if you’re interested.