Setup.txt

(Michael 3-15-20)
How to setup and run:

1. The app is based off this tutorial: https://x-team.com/blog/slack-app-step-by-step/
Parts 1 and 2.

2. Install Node.js, the Yarn package manager, Express.js, and ngrok.

3. Git clone. Type
npm install
to install dependencies. Login to Slack SwiftDemo and navigate to "My Apps".

4. Create a .env file. This will hold the verification token under Basic Info->App Credentials
and the access token under Install App.

slack_app_token=...
verification_token=...

Git will ignore the .env file when you push later.

5. From the swift git repo:
Navigate to /src/ner-server/stanford-ner-2018-10-16.
Type the shell script .\ner-server.sh to start the NER server.
You can also just type the java command in the stanford folder.

6. From the top-level git repo:
yarn run dev
And on a separate terminal:
ngrok http 3000
Copy the https address. 

Log into swiftdemo on Slack, and go to My Apps page at: https://api.slack.com/apps. Select Swift app.
Paste the https address onto the Event Subscriptions section's Request URL:
[url]/api/events
Save changes.

7. Go to Swift workspace. Now when you post a new message in the support channel, the 
message is read, the tagged entities are logged, and the message is reposted in the general channel.

8. To close, individually CTRL-C all terminals. Otherwise the app or ner server may be listening at ports
even after you've closed your text editor. If that happens, use

netstat -ano | findstr :yourPortNumber
taskkill /PID typeyourPIDhere /F

where yourPortNumber is either the localhost port 3000 or the ner port 9191,
and PID is the Process Identifier found by the first command.
See this link: https://stackoverflow.com/questions/39632667/how-do-i-kill-the-process-currently-using-a-port-on-localhost-in-windows


Training Instructions:
1. Get a train and test CSV file from Mockaroo website. Specify a category per CSV - itsupport, health, or finance.

2. Run csv_to_tsv.py, replacing CSV name with your CSV file name, and the string after \t with your category,
ex. '\tHEALTH\n'

3. Run extra_space_generator.py to separate words as documents and set them to lowercase. You have
to replace all labels with all-caps again, ex. health to HEALTH

4. Repeat for your test CSV file. You should have two TSV files now, one for train and test.

5. Replace train4.tsv in train.prop with your new train TSV. Now you can train and check accuracy. Use:
java -cp stanford-ner.jar edu.stanford.nlp.ie.crf.CRFClassifier -prop train.prop
java -cp stanford-ner.jar edu.stanford.nlp.ie.crf.CRFClassifier -loadClassifier ner-model.ser.gz -testFile test2.tsv