Ontology Learning and Reasoning
Create an ontology from structured and semi-structured data sources
Identify mapping from the data ontology to the standard ontology
Extract the relevant sub-graph from the standard ontology to enhance the data ontology
Use the enhanced data ontology to bootstrap conversational space for ontology enhanced query answering
Focused on natural language querying over knowledge bases, as well as platforms and infrastructure for large-scale data analysis, storage and querying of knowledge bases.
Built cognitive querying systems that utilize ontologies, integrate both semi-structured and unstructured content sources, and expose natural language interfaces.
Designed an intelligence video analytics system that blends, in real-time, the intelligence in high-velocity streaming data sources with contextual information from many other data sources to generate complex, data-driven insights.
Built a complex situation detection solution based on spatio-temporal relationships among objects extracted from video streams.
Designed event trend optimization techniques that trade off between CPU and memory costs to execute event queries with Kleene patterns over high-rate event streams.
Proposed an online aggregation approach to dynamically compute event trend aggregation without ever constructing the actual trends.
Scalable Complex Event Analytics
Designed a share-aware optimizer that identifies opportunities for effective shared processing among CEP queries by leveraging time-based event correlations.
Developed a scalable framework that offers stream transactions to assure concurrent shared maintenance and reuse of sub-patterns across queries.
Recurring Query Processing on Big Data
Built a scalable data infrastructure that treats recurring query over big evolving data as first class citizens during query processing.
Tackled issues from massive query workloads to approximate processing to achivew low-latency execution with limited resources.
Robust Distributed Stream Processing
- Designed a new query optimization paradigm capable of coping with data fluctuations in distributed data streams arriving in large volumes, and with near-real time response requirement.