Product updates: GraphGrailAi Platform 1.o ver and CryptoPythia 1.1

GraphGrailAi LLC.
4 min readSep 10, 2018

--

Announcing two product updates: GraphGrailAi Platform 1.o version and CryptoPythia 1.1

GraphGrailAi Ai designer is almost ready — check it now and register https://platform.graphgrail.com

Ai designer is an essential component of the GraphGrailAi Platform (2 others — Data labeler available here: https://labelapp.graphgrail.com and Marketplace, that is under development)

With GraphGrailAi Ai designer you now are able to classify text using neural network (CNN).

How it works:

  1. Upload your dataset or choose from demo (csv file with text samples). Left menu — Datasets.
    You need to follow special format of dataset, but we provided users with well know example datasets — IMDB Movie reviews and Toxic comments. Simply choose one of them if you want to test it fast. We now support 2 languages: English and Russian, as you can see in brackets — [en] or [ru] near dataset’s name.
List of datasets
Sample dataset — movie review with pos or neg label in Category column

2. Create new Ai model.
Left menu — Ai Models.
Press +add button, and the press select button. You will see our Ai designer — an drag-n-drop interface.
a) Write Name
b) Choose dataset from dropdown list
c) Adjust settings. If you are not familiar with it — do nothing and press Save button at the bottom — launch a neural network training.
But, if you assume you are a hero — you can add your stop words in Stopwords remover block component. Don’t forget to move component to Customize solution palette on the right to apply settings.
Also, set Number of epochs to 3–5 to get fast-dirty result, if you set it to 25 you need to wait some time — neural network increase training time significantly.

Ai model creation
Ai designer drag-n-drop inerface
Training preparation
Training has started

3. Check training status — if it’s Ready you are done!
Link for status: https://platform.graphgrail.com/ai-models column Status

Training status

4. Test your neural network live!
Choose Ai model from list of Ready statuses and press on it’s name. You will be redirected to our test facility and API settings page.
In this tutorial i show how i tested it: i just go to IMDB website and choose my favorite Terminator 2: Judgment Day that is rated 8.5. Now i paste part of the review (some part, not full necessary) i found on it’s page in Text field and press Submit.
Result is:

Category: "positive"
Confidence: "0.72780907154083"

That means, that neural network now has a human brain ability — it understand (classify) movie review automatically — is it positive or negative.

Paste movie review in Text box to get category of this review
Terminator 2: Judgment Day page
Actual review to test neural network
GraphGrailAi neural network solve classification task successfully

That’s it!
Imagine now full power of GraphGrailAi solution. You can use it for chat-bots smart question answering, call center automatization, etc.

New users tutorial update.
For new registered users we have created a simple how-to tutorial with explanation of every section of the account.

How-to tutorial

CryptoPythia 1.1 version update

  1. New name — CryptoPythia (old is CryptoSherlock). Remember Pythia from Matrix movie and https://en.wikipedia.org/wiki/Pythia
  2. We have added data collection on every day basis — get valuable insights every day. Also you can compare sentiment trend — simply choose a period from filters dropdown.
https://cryptopythia.graphgrail.com

https://platform.graphgrail.com
https://cryptopythia.graphgrail.com
Register now.

Also, check latest updates and news in our media channels:
https://t.me/GraphGrailAi_en
https://t.me/GraphGrailAi
https://twitter.com/GraphGrail
https://www.facebook.com/graphgrailcom/

GraphGrailAi Team.

--

--

GraphGrailAi LLC.

http://t.co/EOzkcrk1cu - Artificial Intelligence platform for Blockchain built on top of Natural Language Understanding technology with the DApps marketplace