Search Results
Why reinvent the wheel: Let's build question answering systems together
Modern question answering (QA) systems need to flexibly integrate a number of components specialised to fulfil specific tasks in a QA pipeline. Key QA tasks include Named Entity Recognition and Disambiguation, Relation Extraction, and Query Building. Since a number of different software components exist that implement different strategies for each of these tasks, it is a major challenge to select and combine the most suitable components into a QA system, given the characteristics of a question. We study this optimisation problem and train classifiers, which take features of a question as input and have the goal of optimising the selection of QA components based on those features. We then devise a greedy algorithm to identify the pipelines that include the suitable components and can effectively answer the given question. We implement this model within Frankenstein, a QA framework able to select QA components and compose QA pipelines. We evaluate the effectiveness of the pipelines generated by Frankenstein using the QALD and LC-QuAD benchmarks. These results not only suggest that Frankenstein precisely solves the QA optimisation problem but also enables the automatic composition of optimised QA pipelines, which outperform the static Baseline QA pipeline. Thanks to this flexible and fully automated pipeline generation process, new QA components can be easily included in Frankenstein, thus improving the performance of the generated pipelines.
EVENTSKG: A 5-Star Dataset of Top-Ranked Events in Eight Computer Science Communities
Metadata of scientific events has become increasingly available on the Web, albeit often as raw data in various formats, disregarding its semantics and interlinking relations. This leads to restricting the usability of this data for, e.g., subsequent analyses and reasoning. Therefore, there is a pressing need to represent this data in a semantic representation, i.e., Linked Data. We present the new release of the EVENTSKG dataset, comprising comprehensive semantic descriptions of scientific events of eight computer science communities. Currently, EVENTSKG is a 5-star dataset containing metadata of 73 top-ranked event series (almost 2,000 events) established over the last five decades. The new release is a Linked Open Dataset adhering to an updated version of the Scientific Events Ontology, a reference ontology for event metadata representation, leading to richer and cleaner data. To facilitate the maintenance of EVENTSKG and to ensure its sustainability, EVENTSKG is coupled with a Java API that enables users to add/update events metadata without going into the details of the representation of the dataset. We shed light on events characteristics by analyzing EVENTSKG data, which provides a flexible means for customization in order to better understand the characteristics of renowned CS events.
Semantic segmentation of non-linear multimodal images for disease grading of inflammatory bowel disease: A segnet-based application
Non-linear multimodal imaging, the combination of coherent anti-stokes Raman scattering (CARS), two-photon excited fluorescence (TPEF) and second harmonic generation (SHG), has shown its potential to assist the diagnosis of different inflammatory bowel diseases (IBDs). This label-free imaging technique can support the ‘gold-standard’ techniques such as colonoscopy and histopathology to ensure an IBD diagnosis in clinical environment. Moreover, non-linear multimodal imaging can measure biomolecular changes in different tissue regions such as crypt and mucosa region, which serve as a predictive marker for IBD severity. To achieve a real-time assessment of IBD severity, an automatic segmentation of the crypt and mucosa regions is needed. In this paper, we semantically segment the crypt and mucosa region using a deep neural network. We utilized the SegNet architecture (Badrinarayanan et al., 2015) and compared its results with a classical machine learning approach. Our trained SegNet mod el achieved an overall F1 score of 0.75. This model outperformed the classical machine learning approach for the segmentation of the crypt and mucosa region in our study.
A Case for Integrated Data Processing in Large-Scale Cyber-Physical Systems
Large-scale cyber-physical systems such as manufacturing lines generate vast amounts of data to guarantee precise control of their machinery. Visions such as the Industrial Internet of Things aim at making this data available also to computation systems outside the lines to increase productivity and product quality. However, rising amounts and complexities of data and control decisions push existing infrastructure for data transmission, storage, and processing to its limits. In this paper, we exemplarily study a fine blanking line which can produce up to 6.2 Gbit/s worth of data to showcase the extreme requirements found in modern manufacturing. We consequently propose integrated data processing which keeps inherently local and small-scale tasks close to the processes while at the same time centralizing tasks relying on more complex decision procedures and remote data sources. Our approach thus allows for both maintaining control of field-level processes and leveraging the benefits of “big data” applications.
DoMoRe – A recommender system for domain modeling
Domain modeling is an important activity in early phases of software projects to achieve a shared understanding of the problem field among project participants. Domain models describe concepts and relations of respective application fields using a modeling language and domain-specific terms. Detailed knowledge of the domain as well as expertise in model-driven development is required for software engineers to create these models. This paper describes DoMoRe, a system for automated modeling recommendations to support the domain modeling process. We describe an approach in which modeling benefits from formalized knowledge sources and information extraction from text. The system incorporates a large network of semantically related terms built from natural language data sets integrated with mediator-based knowledge base querying in a single recommender system to provide context-sensitive suggestions of model elements.
Temporal Role Annotation for Named Entities
Natural language understanding tasks are key to extracting structured and semantic information from text. One of the most challenging problems in natural language is ambiguity and resolving such ambiguity based on context including temporal information. This paper, focuses on the task of extracting temporal roles from text, e.g. CEO of an organization or head of a state. A temporal role has a domain, which may resolve to different entities depending on the context and especially on temporal information, e.g. CEO of Microsoft in 2000. We focus on the temporal role extraction, as a precursor for temporal role disambiguation. We propose a structured prediction approach based on Conditional Random Fields (CRF) to annotate temporal roles in text and rely on a rich feature set, which extracts syntactic and semantic information from text. We perform an extensive evaluation of our approach based on two datasets. In the first dataset, we extract nearly 400k instances from Wikipedia through distant supervision, whereas in the second dataset, a manually curated ground-truth consisting of 200 instances is extracted from a sample of The New York Times (NYT) articles. Last, the proposed approach is compared against baselines where significant improvements are shown for both datasets.
SemSur: A Core Ontology for the Semantic Representation of Research Findings
The way how research is communicated using text publications has not changed much over the past decades. We have the vision that ultimately researchers will work on a common structured knowledge base comprising comprehensive semantic and machine-comprehensible descriptions of their research, thus making research contributions more transparent and comparable. We present the SemSur ontology for semantically capturing the information commonly found in survey and review articles. SemSur is able to represent scientific results and to publish them in a comprehensive knowledge graph, which provides an efficient overview of a research field, and to compare research findings with related works in a structured way, thus saving researchers a significant amount of time and effort. The new release of SemSur covers more domains, defines better alignment with external ontologies and rules for eliciting implicit knowledge. We discuss possible applications and present an evaluation of our approach with the retrospective, exemplary semantification of a survey. We demonstrate the utility of the SemSur ontology to answer queries about the different research contributions covered by the survey. SemSur is currently used and maintained at OpenResearch.org.
When humans and machines collaborate: Cross-lingual Label Editing in Wikidata
The quality and maintainability of a knowledge graph are determined by the process in which it is created. There are different approaches to such processes; extraction or conversion of available data in the web (automated extraction of knowledge such as DBpedia from Wikipedia), community-created knowledge graphs, often by a group of experts, and hybrid approaches where humans maintain the knowledge graph alongside bots. We focus in this work on the hybrid approach of human edited knowledge graphs supported by automated tools. In particular, we analyse the editing of natural language data, i.e. labels. Labels are the entry point for humans to understand the information, and therefore need to be carefully maintained. We take a step toward the understanding of collaborative editing of humans and automated tools across languages in a knowledge graph. We use Wikidata as it has a large and active community of humans and bots working together covering over 300 languages. In this work, we analyse the different editor groups and how they interact with the different language data to understand the provenance of the current label data.
The Research Core Dataset (KDSF) in the Linked Data context
This paper describes our efforts to implement the Research Core Dataset (“Kerndatensatz Forschung”; KDSF) as an ontology in VIVO. KDSF is used in VIVO to record the required metadata on incoming data and to produce reports as an output. While both processes need an elaborate adaptation of the KDSF specification, this paper focusses on the adaptation of the KDSF basic data model for recording data in VIVO. In this context, the VIVO and KDSF ontologies were compared with respect to domain, syntax, structure, and granularity in order to identify correspondences and mismatches. To produce an alignment, different matching approaches have been applied. Furthermore, we made necessary modifications and extensions on KDSF classes and properties.
Interaction Network Analysis Using Semantic Similarity Based on Translation Embeddings
Biomedical knowledge graphs such as STITCH, SIDER, and Drugbank provide the basis for the discovery of associations between biomedical entities, e.g., interactions between drugs and targets. Link prediction is a paramount task and represents a building block for supporting knowledge discovery. Although several approaches have been proposed for effectively predicting links, the role of semantics has not been studied in depth. In this work, we tackle the problem of discovering interactions between drugs and targets, and propose SimTransE, a machine learning-based approach that solves this problem effectively. SimTransE relies on translating embeddings to model drug-target interactions and values of similarity across them. Grounded on the vectorial representation of drug-target interactions, SimTransE is able to discover novel drug-target interactions. We empirically study SimTransE using state-of-the-art benchmarks and approaches. Experimental results suggest that SimTransE is competitive with the state of the art, representing, thus, an effective alternative for knowledge discovery in the biomedical domain.