Offering:
AI Jumpstart

Business Category:
Dental Services aggregator with Providers and Insurance/Wellness Companies

Project Years:
2024 – 2025

Problem/Opportunity

The client wishes to add value added features to the provider management system. The dental provider could ask a chatbot for a range of questions on dental health, and the system would provide satisfactory answers to the questions. The model should be built that is specific to the client and its domain.

Outcome

When user enters a query, the system processes the data using various techniques and provides reasonably accurate answers.

Execution

This project integrates BERT embeddings (paraphrase-mpnet-base-v2) for semantic similarity and TF-IDF for lexical matching, ensuring precise FAQ retrieval. A hybrid search mechanism combines semantic (70%) and lexical (30%) scores, with a threshold-based filtering (0.6) for accurate responses. Query expansion handles synonyms and variations, improving partial query recognition. All queries and matches are logged for tracking and enhancement.  Model performance is evaluated using Precision, Recall, and F1-score metrics to evaluate the model accuracy. The dataset, enhanced_dentist_questions_dataset.csv contains two well-structured columns: ‘Prompt’ (user questions) and ‘Expected Response’ (ideal answers),  ensuring efficient training and response accuracy.