Multilingual Knowledge Base Research Articles

Improving Open-Domain Dialogue Response Generation with Multi-Source Multilingual Commonsense Knowledge

Knowledge-grounded Dialogue Response Generation (KRG) can facilitate informative and fidelity dialogues using external knowledge. Prior monolingual works can only use the knowledge of the corresponding native language. Thus, due to the prohibitive costs of collecting and constructing external knowledge bases, the limited scale of accessible external knowledge always constrains the ability of KRG, especially in low-resource language scenarios. To this end, we propose a new task, Multi-Source Multilingual Knowledge-Grounded Response Generation (MMKRG), which simultaneously uses multiple knowledge sources of different languages. We notice that simply combining knowledge of different languages is inefficient due to the Cross-Conflict issue and Cross-Repetition issue. Thus, we propose a novel approach MMK-BART, which uses a simple but elegant Estimate-Cluster-Penalize mechanism to overcome the mentioned issues and adopts the multilingual language model mBART as the backbone. Meanwhile, based on the recent multilingual corpus XDailyDialog, we propose an MMKRG dataset MMK-DailyDialog, which has been aligned to the large-scale multilingual commonsense knowledge base ConceptNet and supports four languages (English, Chinese, German, and Italian). Extensive experiments have verified the effectiveness of our dataset and approach in monolingual, cross-lingual, and multilingual scenarios.

Modelling Research Expeditions in Wikidata: Best Practice for Standardisation and Contextualisation

Expeditions and other collecting events are a major source of objects in natural history museums (e.g., Mesibov 2021). Historically, these trips were often transdisciplinary: biological and Earth science specimens were collected at the same time as ethnological or anthropological objects. As a result, specimens and other material gathered during the same expedition, as well as the related data and metadata, are often distributed across multiple institutions. Many expeditions were driven by colonial agendas, aiming to discover new resources to exploit, and their findings were seldom shared with the source countries and local people. Understanding these expeditions illuminates the colonial origins of museum collections, and contributes to recognizing and addressing their impacts (e.g., Das and Lowe 2018, Ashby and Machin 2021). Research expeditions continue to contribute to natural history collections. There is a need to link historical or contemporary research expeditions to other entities, requiring the unambiguous labelling (and persistent identifiers) of such events. Stable identifiers for expeditions plus the sharing of metadata and descriptions in a wide range of languages will facilitate access to scattered information about the event, the institutions housing specimens and objects, the participants and the locations visited, and assist with the linking of distributed material and related research data. However, structured data for scientific expeditions are currently lacking. While identifier systems have been created for many entities over the last few decades, there is no dedicated identifier for research expeditions and similar events. Several studies have shown the importance of people identifiers for linking collection data (e.g., Groom et al. 2022), and we argue the same is true for expeditions. Wikidata is a multilingual community-curated knowledge base containing data structured in a human- and machine-readable format. It allows easy creation, updating and enriching of items on expeditions, and provides stable identifiers for them that can be used in collection management systems. Expeditions can be linked to participants and other agents, regions, localities, objects, archival material, maps, publications, field notebooks, documentary footage and art works resulting from the expeditions, thus making historical information more easily accessible and assisting with the acknowledgment of any imperial or colonial impact that may have resulted from the expedition. Expeditions in Wikidata can be hierarchical, e.g., linking a series of related events or under an umbrella project together providing a machine-readable way to harvest all project data. Wikidata also can provide links between present day countries and historical names for locations (e.g., former colonial names). Expeditions published as Linked Open Data make datasets more FAIR (Findable, Accessible, Interoperable, Reusable), and are also useful in data transcription and validation processes. Visualisation of itinerary data and travel routes also facilitate data quality checks. An informal working group of people interested in the topic was formed to discuss standards and share best practices and recommendations regarding terminology, data modelling and contextualisation. Building upon previous work (e.g., Bauer et al. 2022, von Mering et al. 2022, Leachman 2023), we aim to work towards the enrichment, linking and standardisation of data about research expeditions. If the Wikidata identifiers of these expeditions and participants are added to the records of the corresponding entities in the collection management system, institutions can link from their own collection metadata to the relations made in Wikidata, including to collections in other institutions. The participants of the expedition can be further linked to specimens gathered during the expedition with the use of tools, such as Bionomia, which can facilitate data round-tripping between these collections and specimen records, the Global Biodiversity Information Facility (GBIF) and Wikidata (Shorthouse 2020). Other initiatives such as the Distributed System of Scientific Collections (DiSSCo) are also interested in incorporating these identifiers as links and annotations.

Multilingual Autoregressive Entity Linking

Abstract We present mGENRE, a sequence-to- sequence system for the Multilingual Entity Linking (MEL) problem—the task of resolving language-specific mentions to a multilingual Knowledge Base (KB). For a mention in a given language, mGENRE predicts the name of the target entity left-to-right, token-by-token in an autoregressive fashion. The autoregressive formulation allows us to effectively cross-encode mention string and entity names to capture more interactions than the standard dot product between mention and entity vectors. It also enables fast search within a large KB even for mentions that do not appear in mention tables and with no need for large-scale vector indices. While prior MEL works use a single representation for each entity, we match against entity names of as many languages as possible, which allows exploiting language connections between source input and target name. Moreover, in a zero-shot setting on languages with no training data at all, mGENRE treats the target language as a latent variable that is marginalized at prediction time. This leads to over 50% improvements in average accuracy. We show the efficacy of our approach through extensive evaluation including experiments on three popular MEL benchmarks where we establish new state-of-the-art results. Source code available at https://github.com/facebookresearch/GENRE.

BabelRelate! A Joint Multilingual Approach to Computing Semantic Relatedness

We present a knowledge-rich approach to computing semantic relatedness which exploits the joint contribution of different languages. Our approach is based on the lexicon and semantic knowledge of a wide-coverage multilingual knowledge base, which is used to compute semantic graphs in a variety of languages. Complementary information from these graphs is then combined to produce a 'core' graph where disambiguated translations are connected by means of strong semantic relations. We evaluate our approach on standard monolingual and bilingual datasets, and show that: i) we outperform a graph-based approach which does not use multilinguality in a joint way; ii) we achieve uniformly competitive results for both resource-rich and resource-poor languages.

EcoLexicon and by-products

Abstract Reutilization and interoperability are major issues in the fields of knowledge representation and extraction, as reflected in initiatives such as the Semantic Web and the Linked Open Data Cloud. This paper shows how terminological resources can be integrated and reused within different types of application. EcoLexicon is a multilingual terminological knowledge base (TKB) on environmental science that integrates conceptual, linguistic and visual information. It has led to the following by-products: (i) the EcoLexicon English Corpus; (ii) EcoLexiCAT, a terminology-enhanced translation tool; and (iii) Manzanilla, an image annotation tool. This paper explains EcoLexicon and its by-products, and shows how the latter exploit and enhance the data in the TKB.

Bilingual embeddings with random walks over multilingual wordnets

Bilingual word embeddings represent words of two languages in the same space, and allow to transfer knowledge from one language to the other without machine translation. The main approach is to train monolingual embeddings first and then map them using bilingual dictionaries. In this work, we present a novel method to learn bilingual embeddings based on multilingual knowledge bases (KB) such as WordNet. Our method extracts bilingual information from multilingual wordnets via random walks and learns a joint embedding space in one go. We further reinforce cross-lingual equivalence adding bilingual constraints in the loss function of the popular Skip-gram model. Our experiments on twelve cross-lingual word similarity and relatedness datasets in six language pairs covering four languages show that: 1) our method outperforms the state-of-the-art mapping method using dictionaries; 2) multilingual wordnets on their own improve over text-based systems in similarity datasets; 3) the combination of wordnet-generated information and text is key for good results. Our method can be applied to richer KBs like DBpedia or BabelNet, and can be easily extended to multilingual embeddings. All our software and resources are open source.

Recent Advances in EcoLexicon

EcoLexicon is a multilingual terminological knowledge base concerning the environment. It is the practical application of Frame-based Terminology, a cognitive approach to the representation of specialized knowledge. Recent enhancements include the EcoLexicon English corpus, a phraseological module, and a flexible approach to terminological definitions.

DBpedia – A large-scale, multilingual knowledge base extracted from Wikipedia

The DBpedia community project extracts structured, multilingual knowledge from Wikipedia and makes it freely available on the Web using Semantic Web and Linked Data technologies. The project extracts knowledge from 111 different language editions of Wikipedia. The largest DBpedia knowledge base which is extracted from the English edition of Wikipedia consists of over 400 million facts that describe 3.7 million things. The DBpedia knowledge bases that are extracted from the other 110 Wikipedia editions together consist of 1.46 billion facts and describe 10 million additional things. The DBpedia project maps Wikipedia infoboxes from 27 different language editions to a single shared ontology consisting of 320 classes and 1,650 properties. The mappings are created via a world-wide crowd-sourcing effort and enable knowledge from the different Wikipedia editions to be combined. The project publishes releases of all DBpedia knowledge bases for download and provides SPARQL query access to 14 out of the 111 language editions via a global network of local DBpedia chapters. In addition to the regular releases, the project maintains a live knowledge base which is updated whenever a page in Wikipedia changes. DBpedia sets 27 million RDF links pointing into over 30 external data sources and thus enables data from these sources to be used together with DBpedia data. Several hundred data sets on the Web publish RDF links pointing to DBpedia themselves and make DBpedia one of the central interlinking hubs in the Linked Open Data (LOD) cloud. In this system report, we give an overview of the DBpedia community project, including its architecture, technical implementation, maintenance, internationalisation, usage statistics and applications.

An evaluation framework for cross-lingual link discovery

Cross-Lingual Link Discovery (CLLD) is a new problem in Information Retrieval. The aim is to automatically identify meaningful and relevant hypertext links between documents in different languages. This is particularly helpful in knowledge discovery if a multi-lingual knowledge base is sparse in one language or another, or the topical coverage in each language is different; such is the case with Wikipedia. Techniques for identifying new and topically relevant cross-lingual links are a current topic of interest at NTCIR where the CrossLink task has been running since the 2011 NTCIR-9. This paper presents the evaluation framework for benchmarking algorithms for cross-lingual link discovery evaluated in the context of NTCIR-9.This framework includes topics, document collections, assessments, metrics, and a toolkit for pooling, assessment, and evaluation. The assessments are further divided into two separate sets: manual assessments performed by human assessors; and automatic assessments based on links extracted from Wikipedia itself. Using this framework we show that manual assessment is more robust than automatic assessment in the context of cross-lingual link discovery.

Implementing an open source knowledge base

The concept of knowledge bases originated in artificial intelligence as one side of expert systems-namely, the fundamental body of knowledge available to a domain. KBs are particularly appropriate in knowledge-intensive activities like software development. They offer context-based access to complex information, including informal documents and multimedia, as well as a centralized means of storing and preserving digital assets. KBs can help software engineers with many tasks-from project management and design rationale to version control, defect tracking, code reuse, and staff training and development. We recently implemented an open source KB to support the Consortium for Studying Open Source in Public Administrations. COSPA originated in a EU initiative to study the use of open source software to reduce public administrative software and system support costs. The KB project aimed to build a multilingual knowledge base for comparing and pooling knowledge and experience.

Access to lexical knowledge in modular interpersonal communication aids

This paper discusses the role of lexical linguistic information in the framework of modular architectures for interpersonal communication aids and describes the User Vocabulary Definition and Meaning Mapping Module (UVDMM), developed in the context of the TIDE-ACCESS TP1001 project. The UVDMM Module is a multilingual lexical knowledge base that contains linguistic information and lexical translation relations for orthographic languages and symbolic systems. UVDMM has been designed to operate in the framework of component- or module-based communication aid architectures such as the ATIC architecture, which was also developed in the context of the TIDE-ACCESS project.