Ekaterina Ovchinnikova

Abductive reasoning with a large knowledge base for discourse processing

This paper presents a discourse processing framework based on weighted abduction. We elaborate on ideas described in Hobbs et al. (1993) and implement the abductive inference procedure in a system called Mini-TACITUS. Particular attention is paid to constructing a large and reliable knowledge base for supporting inferences. For this purpose we exploit such lexical-semantic resources as WordNet and FrameNet. We test the proposed procedure and the obtained knowledge base on the Recognizing Textual Entailment task using the data sets from the RTE-2 challenge for evaluation. In addition, we provide an evaluation of the semantic role labeling produced by the system taking the Frame-Annotated Corpus for Textual Entailment as a gold standard.

Using collective expert judgements to evaluate quality measures of mass spectrometry images

Motivation: Imaging mass spectrometry (IMS) is a maturating technique of molecular imaging. Confidence in the reproducible quality of IMS data is essential for its integration into routine use. However, the predominant method for assessing quality is visual examination, a time consuming, unstandardized and non-scalable approach. So far, the problem of assessing the quality has only been marginally addressed and existing measures do not account for the spatial information of IMS data. Importantly, no approach exists for unbiased evaluation of potential quality measures. Results: We propose a novel approach for evaluating potential measures by creating a gold-standard set using collective expert judgements upon which we evaluated image-based measures. To produce a gold standard, we engaged 80 IMS experts, each to rate the relative quality between 52 pairs of ion images from MALDI-TOF IMS datasets of rat brain coronal sections. Experts’ optional feedback on their expertise, the task and the survey showed that (i) they had diverse backgrounds and sufficient expertise, (ii) the task was properly understood, and (iii) the survey was comprehensible. A moderate inter-rater agreement was achieved with Krippendorff’s alpha of 0.5. A gold-standard set of 634 pairs of images with accompanying ratings was constructed and showed a high agreement of 0.85. Eight families of potential measures with a range of parameters and statistical descriptors, giving 143 in total, were evaluated. Both signal-to-noise and spatial chaos-based measures performed highly with a correlation of 0.7 to 0.9 with the gold standard ratings. Moreover, we showed that a composite measure with the linear coefficients (trained on the gold standard with regularized least squares optimization and lasso) showed a strong linear correlation of 0.94 and an accuracy of 0.98 in predicting which image in a pair was of higher quality. Availability and implementation: The anonymized data collected from the survey and the Matlab source code for data processing can be found at: https://github.com/alexandrovteam/IMS_quality. Contact: theodore.alexandrov@embl.de

Natural Language Understanding and World Knowledge

In artificial intelligence and computational linguistics, natural language understanding (NLU) is a subfield of natural language processing that deals with machine reading comprehension. The goal of an NLU system is to interpret an input text fragment. The process of interpretation can be viewed as a translation of the text from a natural language to a representation in an unambiguous formal language. This representation, supposed to expressthe text’s content, is further used for performing concrete tasks implied by a user request

I-Cog: a computational framework for integrated cognition of higher cognitive abilities

There are several challenges for AI models of higher cognitive abilities like the profusion of knowledge, different forms of reasoning, the gap between neuro-inspired approaches and conceptual representations, the problem of inconsistent data, and the manifold of computational paradigms. The I-Cog architecture - proposed as a step towards a solution for these problems - consists of a reasoning device based on analogical reasoning, a rewriting mechanism operating on the knowledge base, and a neuro-symbolic interface for robust learning from noisy data. I-Cog is intended as a framework for human-level intelligence (HLI).

Morph Moulder: Teaching Software for HPSG and Description Logics

The graphical software Morph Moulder (MoMo) presented here was originally created for teaching the logical foundations of Head-Driven Phrase Structure Grammar (HPSG) in an e-Learning environment. It has then been extended to a treatment of description logics (DL), which are at present the standard formalism for building ontologies. With MoMo, students can construct interpretations of sets of formulae and check whether their interpretations model these formulae (called theory in HPSG and axioms in DL). MoMo also supports reasoning such as the construction of well-formed interpretations in the feature logic of HPSG and the automatic extraction of subsumption hierarchies in DL. It has been used successfully in several courses on HPSG linguistics, on computational grammar implementation and on the logical foundations of constraint-based grammar frameworks.

Knowledge Base Construction

In order to understand a natural language expression it is usually not enough to know the literal (“dictionary”) meaning of the words used in this expression and compositional rules of the corresponding language. Much more knowledge is actually involved in discourse processing; knowledge, which may have nothing to do with the linguistic competence but is rather related to our general conception of the world. Suppose we are reading the following text fragment.

Adaptation of Ontological Knowledge from Structured Textual Data

This paper provides a general framework for the extraction and adaptation of ontological knowledge from new structured information. The cycle of this process is described starting with the extraction of semantic knowledge from syntactically given information, the transformation of this information into an appropriate format of description logic, and the dynamic update of a given ontology with this new information where certain types of potentially occurring inconsistencies are automatically resolved. The framework uses crucially certain tools for this incremental update. In addition to WordNet, the usage of FrameNet plays an important role, in order to provide a consistent basis for reasoning applications. The cycle of rewriting textual definitions into description logic axioms is prototypically implemented as well as the resolution of certain types of inconsistencies in the dynamic update of ontologies.

Abductive Reasoning with the Integrative Knowledge Base

This chapter is concerned with extensions of the abductive inference procedure implemented in the reasoning system Mini-TACITUS (Mulkar et al., 2007). The extensions are intended to make the system able to reason with the developed integrative knowledge base.

Sources of World Knowledge

In the area of artificial intelligence, interest to model world knowledge computationally arose in the late 1960s. The first proposals in this direction were Quillian’s (1968) semantic networks and Minsky’s (1975) frame-based representations. These proposals quickly attracted particular attention in the NLP community, because they seemed to provide a solution to the problems in natural language semantics, which required world knowledge.The two examples of the early classical approaches to natural language understanding employing semantic networks and frame representations are presented in (Woods et al., 1980) and (Bobrow et al., 1977).

Reasoning for Natural Language Understanding

In order to understand a natural language expression it is usually not enough to know the literal (“dictionary”) meaning of the words used in this expression and compositional rules of the corresponding language. Much more knowledge is actually involved in discourse processing; knowledge, which may have nothing to do with the linguistic competence but is rather related to our general conception of the world. Suppose we are reading the following text fragment.