OCTOPUS

OCTOPUS is an AI services project, that provide AI use cases as REST API
the services we have now:

Language analysis: An AI use case that takes text as input and predict sentiment, speech act and sarcasm.

Why OCTOPUS

Most of AI companies and startups provides AI services in a form of (get-data > train-data > tuning > deploy) but in octopus we provide AI sevices as REST API in form of (subscribe > use AI use case) no need to get customer data and make the customer waiting for training and tuning.

Language analysis project

project overview

The project consist of two phases:

Simi supervised phase : In this phase we will create the data by making 2 models, one for sarcasm and one for sppech act.
Final model phase : In this phase we will train the model on the data that we produce in the first phase.

Data set

First dataset link: HERE
Second dataset link: HERE

AI Model

The algorithem we choose is MARBERT which is a BERT model trained on huge arabic corpus.
The github link for MARBERT : HERE
The model architecture:

The model architecture with more details:

Model performance

Sentiment classifier

Class	Precision	Recall	F1-score
Positive	0.83	0.71	0.77
Neutral	0.78	0.77	0.78
Negative	0.78	0.90	0.83
AVG	0.80	0.79	0.79

Speech act classifier

Class	Precision	Recall	F1-score
Expression	0.95	0.91	0.93
Assertion	0.88	0.93	0.90
AVG	0.92	0.92	0.92

Sarcasm classifier

Class	Precision	Recall	F1-score
Not sarcastic	0.94	0.86	0.90
Sarcastic	0.72	0.86	0.79
AVG	0.87	0.86	0.86

Team members

Name	Role
Faris adel alahmadi	ML developer
Saed alshahrani	Data analysis
Moath almoghirah	UX/UI developer

Faris-ML/KhalayaProject