Sources of commonsense knowledge aim to support applications in natural language understanding, computer vision, and knowledge graphs. These sources contain complementary knowledge to each other, which makes their integration desired. Yet, such integration is not trivial because of their different foci, modeling approaches, and sparse overlap. In this paper, we propose to consolidate commonsense knowledge by following five principles. We apply these principles to combine seven key sources into a first integrated CommonSense Knowledge Graph (CSKG). We perform analysis of CSKG and its various text and graph embeddings, showing that CSKG is a well-connected graph and that its embeddings provide a useful entry point to the graph. Moreover, we show the impact of CSKG as a source for reasoning evidence retrieval, and for pre-training language models for generalizable downstream reasoning. CSKG and all its embeddings are made publicly available to support further research on commonsense knowledge integration and reasoning.

CSKG: The CommonSense Knowledge Graph

In the SPARQL query processing community, as well as in the wider databases community, benchmark reproducibility is based on releasing datasets and query workloads. However, this paradigm breaks down for federated query processors, as these systems do not manage the data they serve to their clients but provide a data-integration abstraction over the actual query processors that are in direct contact with the data. As a consequence, benchmark results can be greatly affected by the performance and characteristics of the underlying data services. This is further aggravated when one considers benchmarking in more realistic conditions, where internet latency and throughput between the federator and the federated data sources is also a key factor.
In this paper we present KOBE, a benchmarking system that leverages modern containerization and Cloud computing technologies in order to reproduce collections of data sources. In KOBE, data sources are formally described in more detail than what is conventionally provided, covering not only the data served but also the specific software that serves it and its configuration as well as the characteristics of the network that connects them. KOBE provides a specification formalism and a command-line interface that completely hides from the user the mechanics of provisioning and orchestrating the benchmarking process on Kubernetes based infrastructures; and of simulating network latency. Finally, KOBE automates the process of collecting and comprehending logs, and extracting and visualizing evaluation metrics from these logs.

KOBE: Cloud-native Open Benchmarking Engine for Federated Query Processors

Small and medium-sized enterprises increasingly adopt electronic invoices and digitized purchase-to-pay processes. A purchase-to-pay process begins with making a purchase order and ends with completing the payment process. Even when organizations adopt electronic invoices, knowledge work in such processes is characterized by assimilating information distributed over heterogeneous sources among different stages in the process. By integrating such information and enabling a shared understanding of stakeholders in such processes, ontologies and knowledge graphs can serve as an appropriate infrastructure for enabling knowledge services. However, no suitable ontology is available for current electronic invoices and digitized purchase-to-pay processes. Therefore, this paper presents P2P-O, a dedicated purchase-to-pay ontology developed in cooperation with industry domain experts. P2P-O enables organizations to create semantic invoices, which are invoices following linked data principles. The European Standard EN 16931-1:2017 for electronic invoices was the main non-ontological resource for developing P2P-O. The evaluation approach is threefold: (1) to follow ontology engineering best practices, we applied OOPS! (OntOlogy Pitfall Scanner!) and OntoDebug; (2) to evaluate competency questions, we constructed a purchase-to-pay knowledge graph with RML technologies and executed corresponding SPARQL queries; (3) to illustrate a P2P-O-based knowledge service and use case, we implemented an invoicing dashboard within a corporate memory system and thus enabled an entity-centric view on invoice data. Organizations can immediately start experimenting with P2P-O by generating semantic invoices with provided RML mappings.

P2P-O: A Purchase-To-Pay Ontology for Enabling Semantic Invoices

Log files are a vital source of information for keeping systems running and healthy. However, analyzing raw log data, i.e., textual records of system events, typically involves tedious searching for and inspecting clues, as well as tracing and correlating them across log sources. Existing log management solutions ease this process with efficient data collection, storage and normalization mechanisms, but identifying and linking entities across log sources and enriching them with background knowledge is largely an unresolved challenge. To facilitate a knowledge-based approach to log analysis, this paper introduces SLOGERT, a flexible framework and workflow for automated construction of knowledge graphs from arbitrary raw log messages. At its core, it automatically identifies rich RDF graph modelling patterns to represent types of events and extracted parameters that appear in a log stream. We present the workflow, the developed vocabularies for log integration, and our prototypical implementation. To demonstrate the viability of this approach, we conduct a performance analysis and illustrate its application on a large public log dataset in the security domain.

The SLOGERT Framework for Automated Log Knowledge Graph Construction

Graph neural networks and other machine learning models offer a promising direction for interpretable machine learning on relational and multimodal data. Until now, however, progress in this area is difficult to gauge. This is primarily due to a limited number of datasets with (a) a high enough number of labeled nodes in the test set for precise measurement of performance, and (b) a rich enough variety of multimodal information to learn from. We introduce a set of new benchmark tasks for node classification on RDF-encoded knowledge graphs. We focus primarily on node classification, since this setting cannot be solved purely by node embedding models. For each dataset, we provide test and validation sets of at least 1 000 instances, with some over 10 000. Each task can be performed in a purely relational manner, or with multimodal information. All datasets are packaged in a CSV format that is easily consumable in any machine learning environment, together with the original source data in RDF and pre-processing code for full provenance. We provide code for loading the data into numpy and pytorch. We compute performance for several baseline models.

kgbench: A Collection of Knowledge Graph Datasets for Evaluating Relational and Multimodal Machine Learning

This paper presents ParaQA, a question answering (QA) dataset with multiple paraphrased responses for single-turn conversation over knowledge graphs (KG). The dataset was created using a semi-automated framework for generating diverse paraphrasing of the answers using techniques such as back-translation. The existing datasets for conversational question answering over KGs (single-turn/multi-turn) focus on question paraphrasing and provide only up to one answer verbalization. However, ParaQA contains 5000 question-answer pairs with a minimum of two and a maximum of eight unique paraphrased responses for each question. We complement the dataset with baseline models and illustrate the advantage of having multiple paraphrased answers through commonly used metrics such as BLEU and METEOR. The ParaQA dataset is publicly available on a persistent URI for broader usage and adaptation in the research community.

ParaQA: A Question Answering Dataset with Paraphrase Responses for Single-Turn Conversation

Today, Javascript runtimes intend to process data both at server and client levels. In this paper, we study how Rust and Web Assembly can contribute to implement efficient Semantic Web libraries for the Javascript ecosystem. We propose WasmTree, a new implementation of the RDF.JS Store and Dataset interfaces in which the processing is carefully split between the Web Assembly and Javascript layers. Experiments show that our best setup outperforms state-of-the-art implementations for fine-grained data access and SPARQL queries.

WasmTree: Web Assembly for the Semantic Web

The RDF Stream Processing (RSP) community has proposed several models and languages for continuously querying and reasoning over RDF streams over the last decade. They each have their semantics, making them hard to compare. The variety of approaches has fostered both empirical and theoretical research and led to the design of RSPQL, i.e., a unifying model for RSP. However, an RSP API  for the development under RSPQL semantics was still missing. RSP community would benefit from an RSP API because it can foster comparable and reproducible research by providing programming abstractions based on RSPQL semantics. Moreover, it can encourage further development and in-use research. Finally, it can stimulate practical activities such as tutorials, lectures, and challenges, e.g., during the Stream Reasoning Workshop. 

In this paper, we present RSP4J, a flexible API for the development of RSP engines and applications under RSPQL semantics. RSP4J offers all the necessary abstractions required for fast-prototyping of RSP engines under the proposed RSPQL semantics. Users can configure it to reproduce the variety of RSP engine behaviours in a comparable software environment. To promote systematic and comparative research, RSP4J is open-source, provides canonical citation, permanent web identifiers, and a comprehensive user guide for developers.

RSP4J: An API for RDF Stream Processing

This paper presents a formal knowledge organization system (KOS) to represent the United Nations Sustainable Development Goals (SDGs). The SDGs are a set of objectives adopted by all United Nations member states in 2015 to achieve a better and sustainable future. The developed KOS consists of an ontology that models the core elements of the Global SDG indicator framework, which currently includes 17 Goals, 169 Targets and 231 unique indicators, as well as more than 450 related statistical data series maintained by the global statistical community to monitor progress towards the SDGs, and of a dataset containing these elements. In addition to formalizing and establishing unique identifiers for the components of the SDGs and their indicator framework, the ontology includes mappings of each goal, target, indicator and data series to relevant terms and subjects in the United Nations Bibliographic Information System (UNBIS) and the EuroVoc vocabularies, thus facilitating multilingual semantic search and content linking.

A Knowledge Organization System for the United Nations Sustainable Development Goals

We describe the second version of RuBQ, a Russian dataset for knowledge base question answering (KBQA) over Wikidata. Whereas the first version builds on Q&A pairs harvested online, the extension is based on questions obtained through search engine query suggestion services. The questions underwent crowdsourced and in-house annotation in a quite different fashion compared to the first edition. The freely available RuBQ 2.0 contains 2,910 questions along with answers and SPARQL queries, i.e. the dataset doubled in size. The dataset also incorporates answer-bearing paragraphs from Wikipedia for majority of questions. RuBQ 2.0 is suitable for evaluation of KBQA, machine reading comprehension (MRC), and hybrid questions answering models, as well as for semantic parsing research. We provide analysis of the dataset and report several KBQA and MRC baseline results. 

RuBQ 2.0: An Innovated Russian Question Answering Dataset

Since 2017, the goal of the two-million song WASABI database has been to build a knowledge graph linking collected metadata (artists,discography, producers, dates, etc.) with metadata generated by the analysis of both the songs’ lyrics (topics, places, emotions, structure, etc.) and audio signal (chords, sound, etc.). It relies on natural language processing and machine learning methods for extraction, and semantic Web frameworks  for  representation  and  integration.  It  describes  more  than 2  millions  commercial  songs,  200K  albums  and  77K  artists.  It  can  be exploited by music search engines, music professionals (e.g. journalists,radio presenters, music teachers) or scientists willing to analyze popular music published since 1950. It is available under an open license, in multiple formats and with online and open source services including an interactive navigator, a REST API and a SPARQL endpoint.

ESWC 2021 Conference Resources Track Submissions