Working Sentimental Analysis OF transcript Data

1. Performed extensive analysis in order to understand the district name present in full address and whichplay key role in district wise search in address.

2: Explore data visualization of actual district of all state and prepare dataset of each state separately.

3: Trained model using LSTM, W2VEC with Keras network and predict the district name in full address
with accuracy 98

Diva is equipped with the multiple features to empower agents with relevant information. Diva also enables
Process automation to minimize manual efforts.

•  Build Predictive Model to out the repeat call of customers.

1. Performed extensive analysis in order to understand the data which play key role in the decision
Of the repeat call.
2. Explore and visualize the data, build base model for benchmark.
3. Conceptualize feature engineering and leveraging the business insight from case review and set up
Validation framework consistent with the evaluation metrics.
4. Used various ML models like Liner, Decision Tree but Random Forest outperforms

• Build Predictive Model to find out the sentiment from transcript data.

1: Performed extensive analysis into the transcript data and find the some important key point on the
transcript data and find the category and subcategory.
2: Tagging all the transcript on the basis of category and subcategory.
3: Train and Test ML models like LSTM to predict sentiment of sentence
 

Automated quality system to check quality of question asking by agent or customer using NLP 

Verification of the proof of identity (POI) and proof of address (POA) is a key requirement for access to financial
products. In EKYC verification we get POI and POA data from UIDAI as per user’s provided data.
Authentication Mode: Demographics, Biometrics, Otp and Iris
1: Prepared technical documents and end to end implementation.
2: Client Support API.

Detected text from image using deep learning

1: Performed extensive analysis to prepare dataset for training.
2: Train model using CNN and KERAS and predict text present in image with 86?curacy