19 research outputs found
MonaLIA 1.0 : étude préliminaire sur le couplage apprentissage-raisonnement pour la reconnaissance d'images et l’enrichissement de notices de la base Joconde
No full textInternational audienceThe MonaLIA 1.0 project is a preliminary study on the coupling of learning methods (Deep Neural Networks) and knowledge-based methods (Semantic Web) for image recognition and the enhancement of descriptive documentary records. The approach is applied and evaluated on the collection and data in the Joconde database in order to identify the possibilities and challenges offered by this coupling in assisting in the creation and maintenance of such an annotated collection.Le projet MonaLIA 1.0 est une étude préliminaire sur le couplage de méthodes d’apprentissage (Réseaux de Neurones Profonds) et de méthodes à base de connaissances (Web Sémantique) pour la reconnaissance d'images et l’enrichissement de notices descriptives documentaires. L’approche est appliquée et évaluée sur la collection et les données de la base Joconde afin d’identifier les possibilités et les verrous offerte par ce couplage dans l’assistance à la création et la maintenance d’une telle collection annotée
Learning and Reasoning for Cultural Metadata Quality
No full textInternational audienceThis work combines semantic reasoning and machine learning to create tools that allow curators of the visual art collections to identify and correct the annotations of the artwork as well as to improve the relevance of the content-based search results in these collections. The research is based on the Joconde database maintained by French Ministry of Culture that contains illustrated artwork records from main French public and private museums representing archeological objects, decorative arts, fine arts, historical and scientific documents, etc. The Joconde database includes semantic metadata that describes properties of the artworks and their content. The developed methods create a data pipeline that processes metadata, trains a Convolutional Neural Network image classification model, makes prediction for the entire collection and expands the metadata to be the base for the SPARQL search queries. We developed a set of such queries to identify noise and silence in the human annotations and to search image content with results ranked according to the relevance of the objects quantified by the prediction score provided by the deep learning model. We also developed methods to discover new contextual relationships between the concepts in the metadata by analyzing the contrast between the concepts similarities in the Joconde's semantic model and other vocabularies and we tried to improve the model prediction scores based on the semantic relations. Our results show that cross-fertilization between symbolic AI and machine learning can indeed provide the tools to address the challenges of the museum curators work describing the artwork pieces and searching for the relevant images
Stunning Doodle: un outil pour la visualisation et l'analyse conjointe de graphes de connaissances et leurs plongements
No full textInternational audienceCes dernières années, l'utilisation croissante des graphes de connaissances dans des domaines nouveaux et variés nécessite de rendre ces ressources accessibles et compréhensibles par des utilisateurs aux profils de plus en plus divers. De plus, l'application de méthodes d'apprentissage automatique sur des plongements de graphes de connaissances renforce encore la visibilité de ce type de représentation, mais soulève un nouveau problème de compréhension et d'interprétabilité de ces plongements. Dans ce travail, nous montrons comment des techniques de visualisation peuvent être utilisées pour explorer et interpréter conjointement les graphes de connaissances et les plongements de graphes. Mots-clés Plongements de graphes de connaissances, Visualisation
issa-pipeline
No full text<p>Enhanced and refactored ISSA pipeline into a new version that supports easily configurable multi-instance pipelines.</p>
<ul>
<li>Upgraded ISSA pipeline environment<ul>
<li>upgraded the execution environment with available upgrades of the tools: Grobid, Entity-Fishing, Pyclinrec</li>
<li>added Morph-xR2RML Dockers instead of MongoDB and host-based morph-xr2rml java application</li>
</ul>
</li>
<li>Extended Agritrop KG<ul>
<li>annotate documents beyond articles, e.g. theses, conference papers, etc. without full-text extraction</li>
<li>added document domain descriptors</li>
</ul>
</li>
<li>Refactored the code to support multi-instance deployment<ul>
<li>with a minimal effort to launch a new ISSA pipeline for another document corpus</li>
<li>and the support for running multiple pipelines and SPARQL endpoints on the same machine</li>
</ul>
</li>
<li>Added the configuration for the HAL EuroMov ISSA pipeline</li>
</ul>If you use this software, please cite it as below
issa-pipeline
No full text<p>Minor updates after branching to ISSA-2.</p>If you use this software, please cite it as below
ISSA Pipeline
No full textThe ISSA pipeline was developed by the ISSA project (https://issa.cirad.fr/) . It orchestrates the automatic indexing of a scientific archive by extracting from the articles full-text thematic descriptors and named entities, and linking them with terminological resources in the Semantic Web format.The repository consists of various tools, scripts and configuration files involved in each step of the pipeline:- retrieve the articles metadata from the archive's API;- download and pre-process the PDF files of the articles;- process the output to extract thematic descriptors and named entities;- translate the output of each processing step into a unified, consistent RDF dataset;- retrieve additional metadata from OpenAlex: topics, Sustainable Devlopment Goals (SDG), authorship with institutions- upload the resulting dataset to a triple store equipped with a SPARQL endpoint
ISSA Agritrop Dataset
No full textThis dataset, produced in the context of the ISSA project, provides a semantic index of the articles of the Agritrop scientific archive. It is built by extracting thematic descriptors and named entities from the articles full-text, and linking them with resources from DBpedia, Wikidata and AGROVOC.
Licensing:
Different licenses apply to the different subsets of the ISSA Agritrop dataset:
articles metadata is provided under the Agritrop open licence;
full text content extracted from the articles retains the same licence as the original article;
additional data produced by mining the articles (thematic descriptors, named entities) is published under the Open Data Commons Attribution License 1.0 (ODC-By
Stunning Doodle: a Tool for Joint Visualization and Analysis of Knowledge Graphs and Graph Embeddings
No full textInternational audienceIn recent years, the growing application of Knowledge Graphs to new and diverse domains has created the need to make these resources accessible and understandable by users with increasingly diverse backgrounds. Visualization techniques have been widely employed as means to facilitate the exploration and comprehension of such data sources. Moreover, the emerging use of Knowledge Graph Embeddings as input features of Machine Learning methods has given even more visibility to this kind of representation, but raising the new issue of understandability and interpretability of such embeddings. In this paper, we show how visualization techniques can be used to jointly explore and interpret both Knowledge Graphs and Graph Embeddings. We present Stunning Doodle, a tool that enriches the classical visualization of Knowledge Graphs with additional information meant to enable the visual analysis and comprehension of Graph Embeddings. The idea is to help the user figure out the logical connection between (1) the information captured by the Graph Embeddings and (2) the structure and semantics of the Knowledge Graph from which they are generated. We detail the use of Stunning Doodle in a real-world scenario and we show how it has been helpful to interpret different Graph Embeddings and to choose the most suitable with respect to a specific final goal
A systematic approach to identify the information captured by Knowledge Graph Embeddings
No full textInternational audienceIn the last decade Knowledge Graphs have undergone an impressive expansion, mainly due to their extensive use in AI-related applications, such as query answering or recommender systems. This growth has been powered by the expanding landscape of Graph Embedding techniques, which facilitate the manipulation of the vast and sparse information described by Knowledge Graphs. Graph Embedding algorithms create a low-dimensional vector representation of the elements in the graph, i.e. nodes and edges, suitable as input for Machine Learning tasks. Although their effectiveness has been proved on many occasions and for many contexts, the interpretability of such vector representations remains an open issue. In this work, we aim to tackle this issue by providing a systematic approach to decode and make sense of the knowledge captured by Graph Embeddings. We propose a technique for verifying whether Graph Embeddings are able to encode certain properties of the graph elements and we present a categorization for such properties. We test our approach by evaluating the embeddings computed from the same Knowledge Graph through several embedding techniques. We analyze the results on the level of encoding of each property by all the benchmarked algorithms with the final goal of providing insights into the choice of the most suitable technique for each context and encouraging a more conscious use of such approaches
ISSA Agritrop Dataset
No full text<p>This dataset, produced in the context of the <a href="https://issa.cirad.fr/">ISSA</a> project, provides a semantic index of the articles of the <a href="https://agritrop.cirad.fr/">Agritrop scientific archive</a>. It is built by extracting thematic descriptors and named entities from the articles full-text, and linking them with resources from DBpedia, Wikidata and AGROVOC.</p>
<p><strong>Licensing:</strong></p>
<p>Different licenses apply to the different subsets of the ISSA Agritrop dataset:</p>
<ul>
<li>articles metadata is provided under the <a href="https://agritrop.cirad.fr/mention_legale.html">Agritrop open licence</a>;</li>
<li>full text content extracted from the articles retains the same licence as the original article;</li>
<li>additional data produced by mining the articles (thematic descriptors, named entities) is published under the Open <a href="http://opendatacommons.org/licenses/by/1.0/">Data Commons Attribution License 1.0</a> (ODC-By)</li>
</ul>
