580 California St., Suite 400
San Francisco, CA, 94104
Key research themes
1. How can semantic understanding and query expansion improve document retrieval precision in knowledge retrieval systems?
This research area focuses on leveraging semantic similarity, query expansion, and knowledge integration techniques to enhance the precision of document retrieval by better capturing the user's intent and contextual meaning. It addresses challenges related to lexical ambiguity, vocabulary mismatch, and the limited expressiveness of traditional keyword-based systems. By incorporating semantic models such as WordNet, latent semantic analysis, and fuzzy ontologies, systems aim to return more relevant documents that accurately reflect the user's knowledge need.
Key finding: This paper presents a method that leverages alternate query formulations and Latent Semantic Analysis (LSA) to improve the ranking of documents in open-domain question answering systems. The approach creates semantically... Read more
Key finding: By utilizing an information content-based measure to compute semantic similarity between concepts within WordNet taxonomy, this research addresses the issue of vocabulary mismatch in keyword matching. The approach supplements... Read more
Key finding: The proposed system integrates fuzzy logic with ontology-based query expansion to handle polysemy and synonymy in queries and documents. By mapping query terms to their semantic equivalents and relating synonyms across... Read more
Key finding: Through empirical studies on user behavior and knowledge components in online catalog searches, this work highlights the importance of incorporating subject area knowledge, classification scheme knowledge, and system... Read more
2. What role do formal concept analysis and description logics play in structuring and enhancing knowledge-based retrieval systems?
This theme encompasses methods utilizing formal mathematical and logical frameworks, particularly Formal Concept Analysis (FCA) and Description Logics (DL), to organize, represent, and infer knowledge structures for information retrieval. It investigates how these frameworks facilitate semantic indexing, efficient querying, and reasoning over knowledge bases and ontologies, enabling more precise matching of user requests to knowledge resources beyond keyword matching.
Key finding: This paper introduces IRAFCA, an efficient information retrieval algorithm that leverages Formal Concept Analysis to represent document-term relationships as a lattice structure. The algorithm supports both conjunctive and... Read more
Key finding: The study presents a retrieval approach grounded in Description Logic semantics that models requests and resources within a shared ontology framework. By applying reasoning services such as subsumption and satisfiability, the... Read more
Key finding: This work proposes an inductive k-Nearest Neighbor based method extending standard nearest neighbor algorithms to OWL ontologies using an entropy-weighted semantic distance metric. The method enables efficient classification... Read more
Key finding: The paper introduces ASP-PROLOG, an integrated system combining Answer Set Programming and Prolog to enable modular, interactive reasoning over heterogeneous knowledge representations. This integration facilitates the dynamic... Read more
3. How can retrieval systems be optimized to support human learning and knowledge acquisition through personalized and cognitive-aware information retrieval?
This research investigates the design and implementation of retrieval algorithms and systems that actively enhance human learning by tailoring retrieved information to individual knowledge states, learning goals, and cognitive effort. Moving beyond generic relevance, it integrates cognitive models with retrieval objectives to select content that optimally progresses the user's understanding. This approach combines information retrieval, educational psychology, and machine teaching principles to provide retrieval results that serve as effective educational resources.
Key finding: Introducing a novel retrieval framework that integrates a cognitive learning model with traditional retrieval objectives, this paper formulates and approximates an optimization problem to select document sets that maximize... Read more
Key finding: This article discusses the RAG (Retrieve, Augment, Generate) model architecture that combines dynamic external knowledge retrieval with generative language models to produce more relevant, coherent, and reliable content. By... Read more
Key finding: This work details a hybrid retrieval system combining traditional inverted indices with embedding-based neural retrieval models to enhance e-commerce search relevance, particularly on tail queries with complex intent. The... Read more