Santanu Chaudhury

As enterprises increasingly integrate cloud-based large language models (LLMs) such as ChatGPT and Gemini into their legal document workflows, protecting sensitive contractual information - including Personally Identifiable Information (PII) and commercially sensitive clauses - has emerged as a critical challenge. In this work, we propose CON-QA, a hybrid privacy-preserving framework designed specifically for secure question answering over enterprise contracts, effectively combining local and cloud-hosted LLMs. The CON-QA framework operates through three stages: (i) semantic query decomposition and query-aware document chunk retrieval using a locally deployed LLM analysis, (ii) anonymization of detected sensitive entities via a structured one-to-many mapping scheme, ensuring semantic coherence while preventing cross-session entity inference attacks, and (iii) anonymized response generation by a cloud-based LLM, with accurate reconstruction of the original answer locally using a session-consistent many-to-one reverse mapping. To rigorously evaluate CON-QA, we introduce CUAD-QA, a corpus of 85k question-answer pairs generated over 510 real-world CUAD contract documents, encompassing simple, complex, and summarization-style queries. Empirical evaluations, complemented by detailed human assessments, confirm that CON-QA effectively maintains both privacy and utility, preserves answer quality, maintains fidelity to legal clause semantics, and significantly mitigates privacy risks, demonstrating its practical suitability for secure, enterprise-level contract documents.

Ensemble learning is one of the most successful approaches for concept-drift adaptation due to its versatility and high predictive performance. However, a practical challenge in using ensembles for high-speed data stream mining is the associated large computational cost. In this paper, we introduce a computationally efficient heterogeneous ensemble classifier named SUHEN (Successive Halving Ensemble) which adapts to concept-drift using online ensemble selection. We model ensemble selection as a fixed budget best arm identification bandit problem and solve it using Successive Halving Algorithm (SHA). SUHEN identifies a single best performing member for a stream segment and utilizes it for training and prediction until a drift is detected. Upon detecting drift, SHA identifies the new best performer for the segment. As stream characteristics evolve, manually choosing a fixed SHA budget would be challenging. To this end, we extend SUHEN by posing budget selection as a hyperparameter tuning problem and solve it using meta-learning. Our evaluation on 20 benchmark datasets reveal that SUHEN provides accuracy statistically at par with state-of-the-art ensemble algorithms, while providing significant computational resource savings. This makes our proposal attractive for high-speed stream mining problems in resource-constrained settings.