Large Graphs such as YAGO, DBPedia, and Wikidata, form the backbone for many applications including chatbots, personal assistants and question answering systems. Graph database (GDB) users typically use structured queries to precisely express their information needs. However, given the exact match semantics of these languages, a common challenge that they face is that of getting empty or too few results. In this talk, I will discuss some solutions that enable GDB users to explore unfamiliar graphs. First, we integrate connecting tree patterns (CTPs) with existing graph query languages like GPML, SPARQL or Cypher, leading to Extended Queries (EQs). This allows finding general connections between two or more groups of nodes in a graph, without the need to specify exactly how. I will then present an efficient algorithm to evaluate the CTPs. Further, I will introduce Spec-QP, a system that performs automatic reformulations on user queries and efficiently evaluates them. I will conclude my talk with some ongoing works and future research directions.
Bio: Madhulika Mohanty is a postdoc in the CEDAR Team at Inria, Saclay. Previously, she was an Assistant Professor at IIT Dhanbad, India. She obtained her PhD from IIT Delhi in 2020 on the topic of “Techniques for effective search and exploration over Knowledge Graphs”. Her research interests are in Semantic Web Data Management, Querying Graphs and Information Retrieval.
