DOCSLIB.ORG

Farsbase: the Persian Knowledge Graph

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text

Load more

Semantic Web 1 (0) 1–5 1 IOS Press 1 1 2 2 3 3 4 FarsBase: The Persian Knowledge Graph 4 5 5 a a a,* 6 Majid Asgari , Ali Hadian and Behrouz Minaei-Bidgoli 6 a 7 Department of Computer Engineering, Iran University of Science and Technology, Tehran, Iran 7 8 8 9 9 10 10 11 11 Abstract. Over the last decade, extensive research has been done on automatically constructing knowledge graphs from web 12 12 resources, resulting in a number of large-scale knowledge graphs such as YAGO, DBpedia, BabelNet, and Wikidata. Despite 13 13 that some of these knowledge graphs are multilingual, they contain few or no linked data in Persian, and do not support tools 14 for extracting Persian information sources. FarsBase is a multi-source knowledge graph especially designed for semantic search 14 15 engines in Persian language. FarsBase uses some hybrid and flexible techniques to extract and integrate knowledge from various 15 16 sources, such as Wikipedia, web tables and unstructured texts. It also supports an entity linking that allow integrating with 16 17 other knowledge bases. In order to maintain a high accuracy for triples, we adopt a low-cost mechanism for verifying candidate 17 18 knowledge by human experts, which are assisted by automated heuristics. FarsBase is being used as the semantic-search system 18 19 of a Persian search engine and efficiently answers hundreds of semantic queries per day. 19 20 20 Keywords: Semantic Web, Linked Date, Persian, Knowledge Graph 21 21 22 22 23 23 24 24 1. Introduction In this paper we present FarsBase, a Persian knowl- 25 25 edge graph constructed from various information 26 26 Extracting a knowledge graph (KG) from open- sources, including Wikipedia, web tables and raw text. 27 27 FarsBase is specifically designed to fit the require- 28 access structural data such as Wikipedia has paved 28 ments of structural query answering in Persian search 29 the path for a revolution in information retrieval sys- 29 engines. Our contribution are as follows: 30 tems, including search engines and personal assistants 30 like Siri, Google Assistant, Alexa, and Cortana. In a 31 – We provide a hybrid architecture for knowl- 31 search query, users prefer to find exact answer instead 32 edge graph construction from multiple sources 32 of showing a list of web pages. For example, the de- 33 that leverages both top-down and bottom-up ap- 33 sired response for “How many children does the Queen 34 proaches: A preliminary version of the knowl- 34 have?” is simply "Four". This requires an credible and 35 edge graph is constructed from Wikipedia info- 35 up-to-date knowledge graph with comprehensive in- 36 boxes, which is consequently used to extract more 36 formation for responding to most user queries. In fact, 37 knowledge from other knowledge graphs, raw 37 most of the efforts for information acquisition that 38 text, and tables. 38 was traditionally done by the user —including fact- 39 – Contrary to other knowledge graphs, FarsBase is 39 checking, conflict resolution, and credibility analysis 40 tailor-made for [Persian] search engines. There- 40 of the information sources— are to be done when the 41 fore, the entire process of data collection, data 41 42 knowledge graph is being constructed. filtering, and query processing is specifically de- 42 43 In the last decade, extensive research have been signed to boost the user experience. In that re- 43 44 done on constructing knowledge graphs. This includes spect, the query log has a key role in prioritizing 44 45 knowledge graphs constructed from Wikipedia such as data sources, entities, classes, infoboxes, proper- 45 46 DBPedia [1]; systems that extract knowledge raw text, ties, and images, in various stages of the system. 46 47 e.g. NELL [2]; as well as the hybrid systems that ex- The workload-driven design of FarsBase requires 47 48 ploit multiple types of information sources, including fewer experts for building a knowledge graph, be- 48 49 Yago [3]. cause the human resources and system-tuning ef- 49 50 forts focus on data records that are more impor- 50 51 *Corresponding author. E-mail: [email protected] tant to the user, e.g. frequently searched. 51 1570-0844/0-1900/$35.00 c 0 – IOS Press and the authors. All rights reserved 2 M. Asgari et al. / FarsBase: The Persian Knowledge Graph 1 – FarsBase supports rule-based methods that enable 2.1. Knowledge Base and Knowledge Graph 1 2 flexibility for data extraction and manipulation in 2 3 several components of our architecture, including A knowledge base contains a set of facts, assump- 3 4 infobox extraction, raw text extraction, data trans- tions and rules that allows storing knowledge in a com- 4 5 form, and data cleansing. puter system. Knowledge bases can be specific to cer- 5 6 – FarsBase supports efficient human labeling for tain domains, e.g. a medical knowledge base contain- 6 7 managing and cleansing data from various sources ing facts about medical drugs (such as their proper- 7 8 and in multiple versions. It benefits from various ties and interactions). Also, knowledge from multiple 8 9 types of metadata provided by the different ex- domains can be integrated to build a general-domain 9 10 tractors, e.g. the time-flags and the accuracy/con- knowledge base. For example, DBPedia [1] is a multi- 10 11 fidence of different extraction modules for each domain knowledge base that is semi-automatically 11 12 constructed based on the entire Wikipedia articles. 12 triple. Such features can be used for prioritizing 13 Knowledge bases require a data model to organize 13 and grouping the entities for cost-effective batch 14 the facts. A typical approach is to define an ontology, 14 verification of triples by human experts. 15 where data instances (a.k.a. entities) are assigned to 15 – We provide a mechanism for integrating data 16 classes. Each class can be a subclass of another class, 16 from different knowledge extractors. Our mecha- 17 which results in a hierarchy known as ontology tree. 17 18 nism handles different versions from data sources The facts of a knowledge base are commonly repre- 18 19 with minimum expert intervention. This requires sented using a knowledge representation format. Mod- 19 20 extracting temporal facts and triple versioning for ern multi-domain knowledge bases use the Resource 20 21 handling further conflicts between the new and Description Framework (RDF) for knowledge repre- 21 22 current information. To the best of our knowl- sentation. RDF is primarily designed to represent re- 22 23 edge, FarsBase is the only multi-source knowl- sources on the web, but it can also be used for knowl- 23 24 edge base that supports timeliness[4] by handling edge management and supports essential features for 24 25 different versions of data from multiple sources. constructing a knowledge base, such as Is-A relations 25 26 and object properties. 26 The remainder of this paper is organized as follows. 27 In Semantic Web and linked data, there are differ- 27 The preliminaries and motivation is briefly introduced 28 ent definition of knowledge graph (KG); Ehrlinger et 28 in section 2. Section 3 describes a cost-based solu- 29 al tried to clarify the term in [5]. They mentioned 5 se- 29 30 tion to select knowledge sources for FarsBase. We give lected definitions of knowledge graph and presented an 30 31 an overview about FarsBase architecture in section 4. architecture for it. They assumed a knowledge graph 31 32 Section 5 explicates knowledge extraction from differ- is somehow superior and more complex than a knowl- 32 33 ent sources, including Wikipedia, web table and raw edge base because it contains a reasoning engine and 33 34 text. In section 6, we describe how extracted triple are also integrates knowledge from one or more sources. 34 35 mapped and integrated into a unified knowledge graph. 35 36 Evaluation and statistics about FarsBase has been re- 2.2. RDF Knowledge Representation Format 36 37 ported in Section 7. Section 8 describes related work 37 38 in knowledge graph construction, quality assessment, RDF is a standard for conceptualizing structural 38 39 mapping, relation extraction from raw texts, never end- data. In this model, data is represented as a set of triples 39 40 ing paradigms and knowledge augmentation. Finally, consisting of a subject, a predicate, and an object.A 40 41 section 9 concludes the paper with directions for future set of triples forms an RDF graph. For example, the 41 42 work. phrase "Einstein was born in Ulm" can be represented 42 43 as (subj:Albert_Einstein, pred:birth_place, obj:Ulm). 43 44 Triples are the atomic component of the RDF data 44 45 model. 45 46 2. Preliminaries and Motivation The RDF format enables knowledge representa- 46 47 tion using web resources, where each resource has a 47 48 In this section, we briefly introduce the basics Unique Resource Identifier (URI). For instance, the 48 49 of knowledge graph construction and representation. URI corresponding to “Albert Einstein” can be defined 49 50 Also, we explain challenges for constructing a multi- as http://example.name/Albert_Einstein 50 51 domain Persian knowledge graph. where http://example.name is the prefix address of 51 M. Asgari et al. / FarsBase: The Persian Knowledge Graph 3 1 each entity in the knowledge base. In RDF, subjects ing language[9], such as Persian. This is mostly due 1 2 and predicates are URIs, and objects can be either to the fact that tools required for automatic knowledge 2 3 URIs or literal values.
Recommended publications
  • One Knowledge Graph to Rule Them All? Analyzing the Differences Between Dbpedia, YAGO, Wikidata & Co

    One Knowledge Graph to Rule Them All? Analyzing the Differences Between Dbpedia, YAGO, Wikidata & Co

    One Knowledge Graph to Rule them All? Analyzing the Differences between DBpedia, YAGO, Wikidata & co. Daniel Ringler and Heiko Paulheim University of Mannheim, Data and Web Science Group Abstract. Public Knowledge Graphs (KGs) on the Web are consid- ered a valuable asset for developing intelligent applications. They contain general knowledge which can be used, e.g., for improving data analyt- ics tools, text processing pipelines, or recommender systems. While the large players, e.g., DBpedia, YAGO, or Wikidata, are often considered similar in nature and coverage, there are, in fact, quite a few differences. In this paper, we quantify those differences, and identify the overlapping and the complementary parts of public KGs. From those considerations, we can conclude that the KGs are hardly interchangeable, and that each of them has its strenghts and weaknesses when it comes to applications in different domains. 1 Knowledge Graphs on the Web The term \Knowledge Graph" was coined by Google when they introduced their knowledge graph as a backbone of a new Web search strategy in 2012, i.e., moving from pure text processing to a more symbolic representation of knowledge, using the slogan \things, not strings"1. Various public knowledge graphs are available on the Web, including DB- pedia [3] and YAGO [9], both of which are created by extracting information from Wikipedia (the latter exploiting WordNet on top), the community edited Wikidata [10], which imports other datasets, e.g., from national libraries2, as well as from the discontinued Freebase [7], the expert curated OpenCyc [4], and NELL [1], which exploits pattern-based knowledge extraction from a large Web corpus.
  • Wikipedia Knowledge Graph with Deepdive

    Wikipedia Knowledge Graph with Deepdive

    The Workshops of the Tenth International AAAI Conference on Web and Social Media Wiki: Technical Report WS-16-17 Wikipedia Knowledge Graph with DeepDive Thomas Palomares Youssef Ahres [email protected] [email protected] Juhana Kangaspunta Christopher Re´ [email protected] [email protected] Abstract This paper is organized as follows: first, we review the related work and give a general overview of DeepDive. Sec- Despite the tremendous amount of information on Wikipedia, ond, starting from the data preprocessing, we detail the gen- only a very small amount is structured. Most of the informa- eral methodology used. Then, we detail two applications tion is embedded in unstructured text and extracting it is a non trivial challenge. In this paper, we propose a full pipeline that follow this pipeline along with their specific challenges built on top of DeepDive to successfully extract meaningful and solutions. Finally, we report the results of these applica- relations from the Wikipedia text corpus. We evaluated the tions and discuss the next steps to continue populating Wiki- system by extracting company-founders and family relations data and improve the current system to extract more relations from the text. As a result, we extracted more than 140,000 with a high precision. distinct relations with an average precision above 90%. Background & Related Work Introduction Until recently, populating the large knowledge bases relied on direct contributions from human volunteers as well With the perpetual growth of web usage, the amount as integration of existing repositories such as Wikipedia of unstructured data grows exponentially. Extract- info boxes. These methods are limited by the available ing facts and assertions to store them in a struc- structured data and by human power.
  • Knowledge Graphs on the Web – an Overview Arxiv:2003.00719V3 [Cs

    Knowledge Graphs on the Web – an Overview Arxiv:2003.00719V3 [Cs

    January 2020 Knowledge Graphs on the Web – an Overview Nicolas HEIST, Sven HERTLING, Daniel RINGLER, and Heiko PAULHEIM Data and Web Science Group, University of Mannheim, Germany Abstract. Knowledge Graphs are an emerging form of knowledge representation. While Google coined the term Knowledge Graph first and promoted it as a means to improve their search results, they are used in many applications today. In a knowl- edge graph, entities in the real world and/or a business domain (e.g., people, places, or events) are represented as nodes, which are connected by edges representing the relations between those entities. While companies such as Google, Microsoft, and Facebook have their own, non-public knowledge graphs, there is also a larger body of publicly available knowledge graphs, such as DBpedia or Wikidata. In this chap- ter, we provide an overview and comparison of those publicly available knowledge graphs, and give insights into their contents, size, coverage, and overlap. Keywords. Knowledge Graph, Linked Data, Semantic Web, Profiling 1. Introduction Knowledge Graphs are increasingly used as means to represent knowledge. Due to their versatile means of representation, they can be used to integrate different heterogeneous data sources, both within as well as across organizations. [8,9] Besides such domain-specific knowledge graphs which are typically developed for specific domains and/or use cases, there are also public, cross-domain knowledge graphs encoding common knowledge, such as DBpedia, Wikidata, or YAGO. [33] Such knowl- edge graphs may be used, e.g., for automatically enriching data with background knowl- arXiv:2003.00719v3 [cs.AI] 12 Mar 2020 edge to be used in knowledge-intensive downstream applications.
  • Extracting Common Sense Knowledge from Text for Robot Planning

    Extracting Common Sense Knowledge from Text for Robot Planning

    Extracting Common Sense Knowledge from Text for Robot Planning Peter Kaiser1 Mike Lewis2 Ronald P. A. Petrick2 Tamim Asfour1 Mark Steedman2 Abstract— Autonomous robots often require domain knowl- edge to act intelligently in their environment. This is particu- larly true for robots that use automated planning techniques, which require symbolic representations of the operating en- vironment and the robot’s capabilities. However, the task of specifying domain knowledge by hand is tedious and prone to error. As a result, we aim to automate the process of acquiring general common sense knowledge of objects, relations, and actions, by extracting such information from large amounts of natural language text, written by humans for human readers. We present two methods for knowledge acquisition, requiring Fig. 1: The humanoid robots ARMAR-IIIa (left) and only limited human input, which focus on the inference of ARMAR-IIIb working in a kitchen environment ([5], [6]). spatial relations from text. Although our approach is applicable to a range of domains and information, we only consider one type of knowledge here, namely object locations in a kitchen environment. As a proof of concept, we test our approach using domain knowledge based on information gathered from an automated planner and show how the addition of common natural language texts. These methods will provide the set sense knowledge can improve the quality of the generated plans. of object and action types for the domain, as well as certain I. INTRODUCTION AND RELATED WORK relations between entities of these types, of the kind that are commonly used in planning. As an evaluation, we build Autonomous robots that use automated planning to make a domain for a robot working in a kitchen environment decisions about how to act in the world require symbolic (see Fig.
  • Open Mind Common Sense: Knowledge Acquisition from the General Public

    Open Mind Common Sense: Knowledge Acquisition from the General Public

    Open Mind Common Sense: Knowledge Acquisition from the General Public Push Singh, Grace Lim, Thomas Lin, Erik T. Mueller Travell Perkins, Mark Tompkins, Wan Li Zhu MIT Media Laboratory 20 Ames Street Cambridge, MA 02139 USA {push, glim, tlin, markt, wlz}@mit.edu, [email protected], [email protected] Abstract underpinnings for commonsense reasoning (Shanahan Open Mind Common Sense is a knowledge acquisition 1997), there has been far less work on finding ways to system designed to acquire commonsense knowledge from accumulate the knowledge to do so in practice. The most the general public over the web. We describe and evaluate well-known attempt has been the Cyc project (Lenat 1995) our first fielded system, which enabled the construction of which contains 1.5 million assertions built over 15 years a 400,000 assertion commonsense knowledge base. We at the cost of several tens of millions of dollars. then discuss how our second-generation system addresses Knowledge bases this large require a tremendous effort to weaknesses discovered in the first. The new system engineer. With the exception of Cyc, this problem of scale acquires facts, descriptions, and stories by allowing has made efforts to study and build commonsense participants to construct and fill in natural language knowledge bases nearly non-existent within the artificial templates. It employs word-sense disambiguation and intelligence community. methods of clarifying entered knowledge, analogical inference to provide feedback, and allows participants to validate knowledge and in turn each other. Turning to the general public 1 In this paper we explore a possible solution to this Introduction problem of scale, based on one critical observation: Every We would like to build software agents that can engage in ordinary person has common sense of the kind we want to commonsense reasoning about ordinary human affairs.
  • Knowledge Graph Identification

    Knowledge Graph Identification

    Knowledge Graph Identification Jay Pujara1, Hui Miao1, Lise Getoor1, and William Cohen2 1 Dept of Computer Science, University of Maryland, College Park, MD 20742 fjay,hui,[email protected] 2 Machine Learning Dept, Carnegie Mellon University, Pittsburgh, PA 15213 [email protected] Abstract. Large-scale information processing systems are able to ex- tract massive collections of interrelated facts, but unfortunately trans- forming these candidate facts into useful knowledge is a formidable chal- lenge. In this paper, we show how uncertain extractions about entities and their relations can be transformed into a knowledge graph. The ex- tractions form an extraction graph and we refer to the task of removing noise, inferring missing information, and determining which candidate facts should be included into a knowledge graph as knowledge graph identification. In order to perform this task, we must reason jointly about candidate facts and their associated extraction confidences, identify co- referent entities, and incorporate ontological constraints. Our proposed approach uses probabilistic soft logic (PSL), a recently introduced prob- abilistic modeling framework which easily scales to millions of facts. We demonstrate the power of our method on a synthetic Linked Data corpus derived from the MusicBrainz music community and a real-world set of extractions from the NELL project containing over 1M extractions and 70K ontological relations. We show that compared to existing methods, our approach is able to achieve improved AUC and F1 with significantly lower running time. 1 Introduction The web is a vast repository of knowledge, but automatically extracting that knowledge at scale has proven to be a formidable challenge.
  • Google Knowledge Graph, Bing Satori and Wolfram Alpha * Farouk Musa Aliyu and Yusuf Isah Yahaya

    Google Knowledge Graph, Bing Satori and Wolfram Alpha * Farouk Musa Aliyu and Yusuf Isah Yahaya

    International Journal of Scientific & Engineering Research Volume 12, Issue 1, January-2021 11 ISSN 2229-5518 An Investigation of the Accuracy of Knowledge Graph-base Search Engines: Google knowledge Graph, Bing Satori and Wolfram Alpha * Farouk Musa Aliyu and Yusuf Isah Yahaya Abstract— In this paper, we carried out an investigation on the accuracy of two knowledge graph driven search engines (Google knowledge Graph and Bing’s Satori) and a computational knowledge system (Wolfram Alpha). We used a dataset consisting of list of books and their correct authors and constructed queries that will retrieve the author(s) of a book given the book’s name. We evaluate the result from each search engine and measure their precision, recall and F1 score. We also compared the result of these two search engines to the result from the computation knowledge engine (Wolfram Alpha). Our result shows that Google performs better than Bing. While both Google and Bing performs better than Wolfram Alpha. Keywords — Knowledge Graphs, Evaluation, Information Retrieval, Semantic Search Engines.. —————————— —————————— 1 INTRODUCTION earch engines have played a significant role in helping leveraging knowledge graphs or how reliable are the result S web users find their search needs. Most traditional outputted by the semantic search engines? search engines answer their users by presenting a list 2. How are the accuracies of semantic search engines as of ranked documents which they believe are the most relevant compare with computational knowledge engines? to their
  • Commonsense Knowledge Base Completion with Structural and Semantic Context

    Commonsense Knowledge Base Completion with Structural and Semantic Context

    Commonsense Knowledge Base Completion with Structural and Semantic Context Chaitanya Malaviya}, Chandra Bhagavatula}, Antoine Bosselut}|, Yejin Choi}| }Allen Institute for Artificial Intelligence |University of Washington fchaitanyam,[email protected], fantoineb,[email protected] Abstract MotivatedByGoal prevent go to tooth Automatic KB completion for commonsense knowledge dentist decay eat graphs (e.g., ATOMIC and ConceptNet) poses unique chal- candy lenges compared to the much studied conventional knowl- HasPrerequisite edge bases (e.g., Freebase). Commonsense knowledge graphs Causes brush use free-form text to represent nodes, resulting in orders of your tooth HasFirstSubevent Causes magnitude more nodes compared to conventional KBs ( ∼18x tooth bacteria decay more nodes in ATOMIC compared to Freebase (FB15K- pick 237)). Importantly, this implies significantly sparser graph up your toothbrush structures — a major challenge for existing KB completion ReceivesAction methods that assume densely connected graphs over a rela- HasPrerequisite tively smaller set of nodes. good Causes breath NotDesires In this paper, we present novel KB completion models that treat by can address these challenges by exploiting the structural and dentist infection semantic context of nodes. Specifically, we investigate two person in cut key ideas: (1) learning from local graph structure, using graph convolutional networks and automatic graph densifi- cation and (2) transfer learning from pre-trained language Figure 1: Subgraph from ConceptNet illustrating semantic models to knowledge graphs for enhanced contextual rep- diversity of nodes. Dashed blue lines represent potential resentation of knowledge. We describe our method to in- edges to be added to the graph. corporate information from both these sources in a joint model and provide the first empirical results for KB com- pletion on ATOMIC and evaluation with ranking metrics on ConceptNet.
  • Towards a Knowledge Graph for Science

    Towards a Knowledge Graph for Science

    Towards a Knowledge Graph for Science Invited Article∗ Sören Auer Viktor Kovtun Manuel Prinz TIB Leibniz Information Centre for L3S Research Centre, Leibniz TIB Leibniz Information Centre for Science and Technology and L3S University of Hannover Science and Technology Research Centre at University of Hannover, Germany Hannover, Germany Hannover [email protected] [email protected] Hannover, Germany [email protected] Anna Kasprzik Markus Stocker TIB Leibniz Information Centre for TIB Leibniz Information Centre for Science and Technology Science and Technology Hannover, Germany Hannover, Germany [email protected] [email protected] ABSTRACT KEYWORDS The document-centric workflows in science have reached (or al- Knowledge Graph, Science and Technology, Research Infrastructure, ready exceeded) the limits of adequacy. This is emphasized by Libraries, Information Science recent discussions on the increasing proliferation of scientific lit- ACM Reference Format: erature and the reproducibility crisis. This presents an opportu- Sören Auer, Viktor Kovtun, Manuel Prinz, Anna Kasprzik, and Markus nity to rethink the dominant paradigm of document-centric schol- Stocker. 2018. Towards a Knowledge Graph for Science: Invited Article. arly information communication and transform it into knowledge- In WIMS ’18: 8th International Conference on Web Intelligence, Mining and based information flows by representing and expressing informa- Semantics, June 25–27, 2018, Novi Sad, Serbia. ACM, New York, NY, USA, tion through semantically rich, interlinked knowledge graphs. At 6 pages. https://doi.org/10.1145/3227609.3227689 the core of knowledge-based information flows is the creation and evolution of information models that establish a common under- 1 INTRODUCTION standing of information communicated between stakeholders as The communication of scholarly information is document-centric.

  • Common Sense Reasoning with the Semantic Web

    Common Sense Reasoning with the Semantic Web Christopher C. Johnson and Push Singh MIT Summer Research Program Massachusetts Institute of Technology, Cambridge, MA 02139 [email protected], [email protected] http://groups.csail.mit.edu/dig/2005/08/Johnson-CommonSense.pdf Abstract Current HTML content on the World Wide Web has no real meaning to the computers that display the content. Rather, the content is just fodder for the human eye. This is unfortunate as in fact Web documents describe real objects and concepts, and give particular relationships between them. The goal of the World Wide Web Consortium’s (W3C) Semantic Web initiative is to formalize web content into Resource Description Framework (RDF) ontologies so that computers may reason and make decisions about content across the Web. Current W3C work has so far been concerned with creating languages in which to express formal Web ontologies and tools, but has overlooked the value and importance of implementing common sense reasoning within the Semantic Web. As Web blogging and news postings become more prominent across the Web, there will be a vast source of natural language text not represented as RDF metadata. Common sense reasoning will be needed to take full advantage of this content. In this paper we will first describe our work in converting the common sense knowledge base, ConceptNet, to RDF format and running N3 rules through the forward chaining reasoner, CWM, to further produce new concepts in ConceptNet. We will then describe an example in using ConceptNet to recommend gift ideas by analyzing the contents of a weblog.
  • Conceptnet 5.5: an Open Multilingual Graph of General Knowledge

    Conceptnet 5.5: an Open Multilingual Graph of General Knowledge

    Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence (AAAI-17) ConceptNet 5.5: An Open Multilingual Graph of General Knowledge Robyn Speer Joshua Chin Catherine Havasi Luminoso Technologies, Inc. Union College Luminoso Technologies, Inc. 675 Massachusetts Avenue 807 Union St. 675 Massachusetts Avenue Cambridge, MA 02139 Schenectady, NY 12308 Cambridge, MA 02139 Abstract In this paper, we will concisely represent assertions such Machine learning about language can be improved by sup- as the above as triples of their start node, relation label, and plying it with specific knowledge and sources of external in- end node: the assertion that “a dog has a tail” can be repre- formation. We present here a new version of the linked open sented as (dog, HasA, tail). data resource ConceptNet that is particularly well suited to ConceptNet also represents links between knowledge re- be used with modern NLP techniques such as word embed- sources. In addition to its own knowledge about the English dings. term astronomy, for example, ConceptNet contains links to ConceptNet is a knowledge graph that connects words and URLs that define astronomy in WordNet, Wiktionary, Open- phrases of natural language with labeled edges. Its knowl- Cyc, and DBPedia. edge is collected from many sources that include expert- The graph-structured knowledge in ConceptNet can be created resources, crowd-sourcing, and games with a pur- particularly useful to NLP learning algorithms, particularly pose. It is designed to represent the general knowledge in- those based on word embeddings, such as (Mikolov et al. volved in understanding language, improving natural lan- 2013). We can use ConceptNet to build semantic spaces that guage applications by allowing the application to better un- are more effective than distributional semantics alone.
  • Exploiting Semantic Web Knowledge Graphs in Data Mining

    Exploiting Semantic Web Knowledge Graphs in Data Mining

    Exploiting Semantic Web Knowledge Graphs in Data Mining Inauguraldissertation zur Erlangung des akademischen Grades eines Doktors der Naturwissenschaften der Universit¨atMannheim presented by Petar Ristoski Mannheim, 2017 ii Dekan: Dr. Bernd Lübcke, Universität Mannheim Referent: Professor Dr. Heiko Paulheim, Universität Mannheim Korreferent: Professor Dr. Simone Paolo Ponzetto, Universität Mannheim Tag der mündlichen Prüfung: 15 Januar 2018 Abstract Data Mining and Knowledge Discovery in Databases (KDD) is a research field concerned with deriving higher-level insights from data. The tasks performed in that field are knowledge intensive and can often benefit from using additional knowledge from various sources. Therefore, many approaches have been proposed in this area that combine Semantic Web data with the data mining and knowledge discovery process. Semantic Web knowledge graphs are a backbone of many in- formation systems that require access to structured knowledge. Such knowledge graphs contain factual knowledge about real word entities and the relations be- tween them, which can be utilized in various natural language processing, infor- mation retrieval, and any data mining applications. Following the principles of the Semantic Web, Semantic Web knowledge graphs are publicly available as Linked Open Data. Linked Open Data is an open, interlinked collection of datasets in machine-interpretable form, covering most of the real world domains. In this thesis, we investigate the hypothesis if Semantic Web knowledge graphs can be exploited as background knowledge in different steps of the knowledge discovery process, and different data mining tasks. More precisely, we aim to show that Semantic Web knowledge graphs can be utilized for generating valuable data mining features that can be used in various data mining tasks.