Abstract Analysts like to pose a variety of questions over large relational databases containing data on the domain that they are analyzing. Enabling natural language question answering over such data for analysts requires mechanisms to extract exceptions in data, find steps to transform data, detect implications in the data, and apply classifications on the data. Motivated by this problem, we propose... Analysts like to pose a variety of questions over large relational databases containing data on the domain that they are analyzing. Enabling natural language question answering over such data for analysts requires mechanisms to extract exceptions in data, find steps to transform data, detect implications in the data, and apply classifications on the data. Motivated by this problem, we propose a semantically enriched deep learning pipeline that supports natural language question answering over relational databases and uses Formal Concept Analysis to find exceptions, classification and transformation steps. Our frame- work is based on a set of deep learning sequence tagging networks which extracts information from the NL sentence and constructs an equivalent intermediate sketch, and then maps it into the actual tables and columns of the database. The output data of the query is converted into a lattice structure which results into the (extent, intent) tuples. These tuples are then analyzed to find the exceptions, classification and transformation steps. Read More Read Less

Abstract We convert existing training datasets into the ones closed under linguistic generalization operations to expand infrequent cases. We transfer the definition of the least general generalization from logical formulas to linguistic structures, from words to phrases, sentences, speech acts and discourse trees. The main advantage of the resultant frameworks is explainability and learnability from a small set of samples. Learning... We convert existing training datasets into the ones closed under linguistic generalization operations to expand infrequent cases. We transfer the definition of the least general generalization from logical formulas to linguistic structures, from words to phrases, sentences, speech acts and discourse trees. The main advantage of the resultant frameworks is explainability and learnability from a small set of samples. Learning via generalization of linguistic structures turned out to be well suited for industrial linguistic applications with limited training datasets. Read More Read Less

Abstract While the traditional philosophical epistemology stresses the importance of distinguishing knowledge from true beliefs, the formalisation of this distinction with standard logical means turns out to be problematic. In Knowledge Representation (KR) as a Computer Science discipline this crucial distinction is largely neglected. A practical consequence of this neglect is that the existing KR systems store and communicate knowledge that... While the traditional philosophical epistemology stresses the importance of distinguishing knowledge from true beliefs, the formalisation of this distinction with standard logical means turns out to be problematic. In Knowledge Representation (KR) as a Computer Science discipline this crucial distinction is largely neglected. A practical consequence of this neglect is that the existing KR systems store and communicate knowledge that cannot be verified and justified by users of these systems without external means. Information obtained from such systems does not qualify as knowledge in the sense of philosophical epistemology. Recent advances in the research area at the crossroad of the computational mathematical logic, formal epistemology and computer science open new perspectives for an effective computational realisation of justificatory procedures in KR. After exposing the problem of justification in logic, epistemology and KR, we sketch a novel framework for representing knowledge along with relevant justificatory procedures, which is based on the Homotopy Type theory (HoTT). This formal framework supports representation of both propositional knowledge, aka knowledge that, and non-propositional knowledge, aka knowledge-how or procedural knowledge. The default proof-theoretic semantics of HoTT allows for combining the two sorts of represented knowledge at the formal level by interpreting all permissible constructions as justification terms (witnesses) of associated propositions. Read More Read Less