Conversation Sentiment Analysis (CSAT)

Documentation

Inputs and Outputs

The toolkit receives and processes a sequence of spoken terms (i.e., a user’s utterance) up to the length of 150 terms per utterance. It outputs a classification score for each of the three sentiment classes, the positive, the neutral, and the negative class in a form suitable for the ALAMEDA project.

Background Info

HolC (Agathangelou et al, 2022) is a deep learning model that employs and advances state of the art modules for text and sentiment analysis. It is a feed forward model that consists of three blocks. The first, is a Convolutional network that receives word embeddings and extracts sentence embeddings. The second, is a Bi-directional Long Short Memory network (Bi-LSTM) that receives sentences embeddings and outputs unregularized predictions. Then an attention layer augments, filters and transfers neural features into a single layer perceptron which gathers the unregularized predictions and outputs the final prediction. The described HolC’s architecture is illustrated in Figure 1. HolC provides accurate results outperforming other known architectures of Neural Networks like Convolutional Neural Networks and LSTM networks.

Installation Ιnstructions

The Conversation Sentiment Analysis Toolkit has been developed following industrial standards. In this sense, the toolkit has been packaged and uploaded to the AI Toolkit as a dockerized container. It requires the following commands to open ports and run the model:

IMPORTANT NOTE

• All requests to the VarCls service require an ALAMEDA IAM authenticated client. This translates into a requirement for all requests to include an Authorization Header in the HTTP call with the format: Bearer , where the access_token has been obtained after authenticating with the ALAMEDA IAM backend endpoint.

• Open port:

  sudo ufw allow 5050

  ---

• Login to container registry using guest account:

  docker login gitlab.telecom.ntua.gr:5050 -u alameda_ai_toolkit_registry_guest -p ByeYyNesUxqQGs91FzzW

  ---

• Pull and run the docker image:

  docker run -p 5050:5052 gitlab.telecom.ntua.gr:5050/alameda/alameda-source-code/ai-toolkit/ai-toolkit-registry/csat

  ---

• Logout from container registry:

  docker logout gitlab.telecom.ntua.gr:5050

  ---

Datasets & Samples

The dataset that was used for training and fine tuning the Conversation Sentiment Analysis Toolkit included samples from several domains from Amazon’s review data. Below are some samples:

OpenAPI Documentation