ABSTRACT. Knowledge acquisition from text is process of automatically acquiring, organizing and structuring knowledge from text which can be used to perform question answering or complex reasoning. However, current state-of-the-art systems are limited by the fact that they are not able to construct the knowledge base with high quality as knowledge representation and reasoning (KRR) has a keen requirement for the accuracy of data. Controlled Natural Languages (CNLs) emerged as a way to improve the quality problem by restricting the flexibility of the language. However, they still fail to do so as sentences that express the same information may be represented by different forms. Current CNL systems have limited power to standardize the sentences into the same logical form. We solved this problem by building the Knowledge Acquisition Logic Machine (KALM), which performs semantic analysis of English sentences and achieves superior accuracy of standardizing sentences that express the same meaning to the same logical representation. Besides, we developed the query part of KALM to perform question answering, which also achieves very high accuracy in query understanding.

ABSTRACT. The paper addresses the problem of automatic generation of natural language descriptions for ontology-described artifacts. The motivation for the work is the challenge of providing textual descriptions of automatically generated scientific workflows (e.g., paragraphs that scientists can include in their publications). The extended abstract presents a system which generates descriptions of sets of atoms derived from a collection of ontologies. The system, called \nlgP, demonstrates the feasibility of the task in the \emph{Phylotastic} project, that aims at providing evolutionary biologists with a platform for automatic generation of phylogenetic trees given some suitable inputs. \nlgP\ utilizes the fact that the Grammatical Framework (GF) is suitable for the natural language generation (NLG) task; the abstract shows how elements of the ontologies in Phylotastic, such as web services, inputs and outputs of web services, can be encoded in GF for the NLG task.

ABSTRACT. We present an ongoing research on a probabilistic extension of action language BC+. Just like BC+ is defined as a high-level notation of answer set programs for describing transition systems, the proposed language, which we call pBC+, is defined as a high-level notation of lpmln programs - a probabilistic extension of answer set programs.

As preliminary results accomplished, we illustrate how probabilistic reasoning about transition systems, such as prediction, postdiction, and planning problems, as well as probabilistic diagnosis for dynamic domains, can be modeled in pBC+ and computed using an implementation of LP^MLN.

For future work, we plan to develop a compiler that automatically translates pBC+ description into LP^MLN programs, as well as parameter learning in probabilistic action domains through lpmln weight learning. We will work on defining useful extensions of pBC+ to facilitate hypothetical/counterfactual reasoning. We will also find real-world applications, possibly in robotic domains, to empirically study the performance of this approach to probabilistic reasoning in action domains.

ABSTRACT. The grounding bottleneck is an important open issue in Answer Set Programming. Lazy grounding addresses it by interleaving grounding and search. The performance of current lazy-grounding solvers is not yet comparable to that of ground-and-solve systems, however. The aim of this thesis is to extend prior work on lazy grounding by novel heuristics and other techniques like non-ground conflict learning in order to speed up solving. Parts of expected results will be beneficial for ground-and-solve systems as well.

ABSTRACT. Recently, biological data has been increasingly produced calling for the existence of computational models able to organize and computationally reproduce existing observations. In particular, biological regulatory networks have been modeled relying on the Sign Consistency Model or the logical formalism. However, their construction still completely relies on a domain expert to choose the best functions for every network component. Due to the number of possible functions for k arguments, this is typically a process prone to error. Here, we propose to assist the modeler using logic-based tools to verify the model, identifying crucial network components responsible for model inconsistency. We intend to obtain a model building procedure capable of providing the modeler with repaired models satisfying a set of pre-defined criteria, therefore minimizing possible modeling errors.

ABSTRACT. In causality, an actual cause is often defined as an event responsible for bringing about a given outcome in a scenario. In practice, however, identifying this event alone is not always sufficient to provide a satisfactory explanation of how the outcome came to be. In this paper, we motivate this claim using well-known examples and present a novel framework for reasoning more deeply about actual causation. The framework reasons over a scenario and domain knowledge to identify additional events that helped to ``set the stage’’ for the outcome. By leveraging techniques from Reasoning about Actions and Change, the proposed approach supports reasoning over domains in which the evolution of the state of the world over time plays a critical role and enables one to identify and explain the circumstances that led to an outcome of interest. We utilize action language $\mathcal{AL}$ for defining the constructs of the framework. This language lends itself quite naturally to an automated translation to Answer Set Programming, using which, reasoning tasks of considerable complexity can be specified and executed. We speculate that a similar approach can also lead to the development of algorithms for our framework.

ABSTRACT. Answer set programming (ASP) is a particularly useful approach for nonmonotonic reasoning in knowledge representation. In order to handle quantitative and qualitative reasoning, a number of different extensions of ASP have been invented, such as quantitative extensions LP^{MLN} and P-log, and qualitative extensions LPOD, and CR-Prolog2.

Although each of these formalisms introduced some new and unique concepts, we present reductions of each of these languages into the standard ASP language, which not only gives us an alternative insight into the semantics of these extensions in terms of the standard ASP language, but also shows that the standard ASP is capable of representing quantitative uncertainty and qualitative uncertainty. What's more, our translations yield a way to tune the semantics of LPOD and CR-Prolog2. Since the semantics of each formalism is represented in ASP rules, we can modify their semantics by modifying the corresponding ASP rules.

For future work, we plan to create a new formalism that is capable of representing quantitative and qualitative uncertainty at the same time. Since LPOD rules are simple and informative, we will first try to include quantitative preference into LPOD by adding the concept of weights, and tune the semantics of LPOD by modifying the translated standard ASP rules.

ABSTRACT. Over time, Answer Set Programming (ASP) has gained traction as a versatile logic programming semantics with performant processing systems, used by a growing number of significant applications in academia and industry. However, this development is threatened by a lack of commonly accepted design patterns and techniques for ASP to address dynamic application on a real-world scale. To this end, we identified robotic intra-logistics as representative scenario, a major domain of interest in the context of the fourth industrial revolution. For this setting, we aim to provide a scalable and efficient ASP-based solutions by (1) stipulating a standardized test and benchmark framework; (2) leveraging existing ASP techniques through new design patterns; and (3) extending ASP with new functionalities. In this paper we will expand on the subject matter as well as detail our current progress and future plans.

ABSTRACT. Proof-relevant resolution is a new variant of resolution in Horn-clause logic and its extensions. We propose proof-relevant resolution as a systematic approach to elaboration in programming languages that is close to formal specification and hence allows for analysis of semantics of the language. We demonstrate the approach on two case studies; we describe a novel, proof-relevant approach to type inference and term synthesis in dependently types languages and we show how proof-relevant resolution allows for analysis of inductive and coinductive soundness of type class resolution. We conclude by a discussion of overall contributions of our current work.

ABSTRACT. Given a text, several questions can be asked. For some of these questions, the answer can be directly looked up from the text. However for several other questions, one might need to use additional knowledge and sophisticated reasoning to find the answer. Developing AI agents that can answer this kinds of questions and can also justify their answer is the focus of this research. Towards this goal, we use the language of Answer Set Programming as the knowledge representation and reasoning language for the agent. The question then arises, is how to obtain the additional knowledge? In this work we show that using existing Natural Language Processing parsers and a scalable Inductive Logic Programming algorithm it is possible to learn this additional knowledge (containing mostly commonsense knowledge) from question-answering datasets which then can be used for inference.