404 research outputs found
LM-KBC 2023: 2<sup>nd</sup> Challenge on Knowledge Base Construction from Pre-trained Language Models
No full textLarge language models (LLMs) like chatGPT [1] have advanced a range of semantic tasks and are being ubiquitously used for knowledge extraction. Although several works have explored this ability by crafting prompts with in-context or instruction learning, the viability of complete and precise knowledge base construction from LMs is still in its nascent form. In the 2nd edition of this challenge, we invited participants to extract disambiguated knowledge triples from LMs for a given set of subjects and relations. In crucial difference to existing probing benchmarks like LAMA [2], we made no simplifying assumptions on relation cardinalities, i.e., a subject-entity can stand in relation with zero, one, or many object-entities. Furthermore, submissions needed to go beyond just ranking predicted surface strings, and materialize disambiguated entities in the output, which were evaluated using established KB metrics of precision, recall, and F1-score. The challenge had two tracks: (1) a small model track, where models with < 1 billion parameters could be probed, and (2) an open track, where participants could use any LM of their choice. We received seven submissions, two for track 1 and five for track 2. We present the contributions and insights of the submitted peer-reviewed submissions and lay out the possible paths for future work. All the details related to the challenge can be found on our website at https://lm-kbc.github.io/challenge2023/.</p
LM-KBC 2023: 2<sup>nd</sup> Challenge on Knowledge Base Construction from Pre-trained Language Models
No full textLarge language models (LLMs) like chatGPT [1] have advanced a range of semantic tasks and are being ubiquitously used for knowledge extraction. Although several works have explored this ability by crafting prompts with in-context or instruction learning, the viability of complete and precise knowledge base construction from LMs is still in its nascent form. In the 2nd edition of this challenge, we invited participants to extract disambiguated knowledge triples from LMs for a given set of subjects and relations. In crucial difference to existing probing benchmarks like LAMA [2], we made no simplifying assumptions on relation cardinalities, i.e., a subject-entity can stand in relation with zero, one, or many object-entities. Furthermore, submissions needed to go beyond just ranking predicted surface strings, and materialize disambiguated entities in the output, which were evaluated using established KB metrics of precision, recall, and F1-score. The challenge had two tracks: (1) a small model track, where models with < 1 billion parameters could be probed, and (2) an open track, where participants could use any LM of their choice. We received seven submissions, two for track 1 and five for track 2. We present the contributions and insights of the submitted peer-reviewed submissions and lay out the possible paths for future work. All the details related to the challenge can be found on our website at https://lm-kbc.github.io/challenge2023/.</p
LM-KBC 2023: 2<sup>nd</sup> Challenge on Knowledge Base Construction from Pre-trained Language Models
No full textLarge language models (LLMs) like chatGPT [1] have advanced a range of semantic tasks and are being ubiquitously used for knowledge extraction. Although several works have explored this ability by crafting prompts with in-context or instruction learning, the viability of complete and precise knowledge base construction from LMs is still in its nascent form. In the 2nd edition of this challenge, we invited participants to extract disambiguated knowledge triples from LMs for a given set of subjects and relations. In crucial difference to existing probing benchmarks like LAMA [2], we made no simplifying assumptions on relation cardinalities, i.e., a subject-entity can stand in relation with zero, one, or many object-entities. Furthermore, submissions needed to go beyond just ranking predicted surface strings, and materialize disambiguated entities in the output, which were evaluated using established KB metrics of precision, recall, and F1-score. The challenge had two tracks: (1) a small model track, where models with < 1 billion parameters could be probed, and (2) an open track, where participants could use any LM of their choice. We received seven submissions, two for track 1 and five for track 2. We present the contributions and insights of the submitted peer-reviewed submissions and lay out the possible paths for future work. All the details related to the challenge can be found on our website at https://lm-kbc.github.io/challenge2023/.</p
LM-KBC 2023: 2<sup>nd</sup> Challenge on Knowledge Base Construction from Pre-trained Language Models
No full textLarge language models (LLMs) like chatGPT [1] have advanced a range of semantic tasks and are being ubiquitously used for knowledge extraction. Although several works have explored this ability by crafting prompts with in-context or instruction learning, the viability of complete and precise knowledge base construction from LMs is still in its nascent form. In the 2nd edition of this challenge, we invited participants to extract disambiguated knowledge triples from LMs for a given set of subjects and relations. In crucial difference to existing probing benchmarks like LAMA [2], we made no simplifying assumptions on relation cardinalities, i.e., a subject-entity can stand in relation with zero, one, or many object-entities. Furthermore, submissions needed to go beyond just ranking predicted surface strings, and materialize disambiguated entities in the output, which were evaluated using established KB metrics of precision, recall, and F1-score. The challenge had two tracks: (1) a small model track, where models with < 1 billion parameters could be probed, and (2) an open track, where participants could use any LM of their choice. We received seven submissions, two for track 1 and five for track 2. We present the contributions and insights of the submitted peer-reviewed submissions and lay out the possible paths for future work. All the details related to the challenge can be found on our website at https://lm-kbc.github.io/challenge2023/.</p
A tücsök meg a hangyák
No full textThis children's book is composed of seven thick boards bound together. On the cover a grasshopper with moustache sits on a mushroom playing his fiddle as a row of ants marches by carrying or rolling food and an ant-baby. The next pages expand on their labors. They include a cut-out portion that looks past their hill to the flowers. On the following pages ants continue their workline while, I believe, young grasshoppers dance about and the older grasshopper continues to fiddle. Succeeding pages show more ant work, including carrying off a dead or exhausted ant on a stretcher. And we see lots of grasshopper fiddlers while other ants push carts full of food, both by day and by night. Soon there are rains and snows, and an ant finds the grasshopper lying next to his fiddle on the ground. The ants take him in, feed him, and dance to his music. I believe it is typical of the East Block countries that a Hungarian book was executed in Czechoslovakia. Might there have been a Czech original?Language note: Hungaria
It is not magic, it is smith: Comparison in a study of Jewish theology
No full textIn a search for a theoretical framework that would structure and orient a comparative analysis of diverse Jewish theological responses to the Holocaust, the author reached for J.Z. Smith’s discussions of comparative enterprise. The questions of similarity, difference and of the putative goal of comparison loomed large over her project. In J.Z. Smith’s work, the author found helpful clues, illuminating insights as well as somewhat confusing and counterintuitive examples
It is not magic, it is smith: Comparison in a study of Jewish theology
No full textIn a search for a theoretical framework that would structure and orient a comparative analysis of diverse Jewish theological responses to the Holocaust, the author reached for J.Z. Smith’s discussions of comparative enterprise. The questions of similarity, difference and of the putative goal of comparison loomed large over her project. In J.Z. Smith’s work, the author found helpful clues, illuminating insights as well as somewhat confusing and counterintuitive examples.acceptedVersion© 2019. This is the authors' accepted and refereed manuscript. Locked until 12 February 2021 due to copyright restrictions. The final authenticated version is available online at: https://doi.org/10.1163/15700682-12341460
J2RM: An ontology-based JSON-to-RDF mapping tool
No full textThis manuscript introduces J2RM: a tool to process mappings from JSON data to RDF triples guided by an OWL2 ontology structure. The mappings are defined as annotation properties associated with each ontology entity of interest. They are embedded in an ontology file so that they can be readily deployed and shared to automate RDF-graph creation. In this paper, we motivate the need for such mappings, describe some of their definitions on a use case example, present the formal grammar of the mapping language, and explain how these mappings work. We conclude with a discussion of the key aspects, main contributions, and future improvements
Active knowledge graph completion
No full textKnowledge graphs (KGs) proliferating on theWeb are known to be incomplete. Much research has been proposed for automatic com- pletion, sometimes by rule learning, that scales well. All existing methods learn closed rules. Here we introduce open path (OP) rules and present a novel algorithm, oprl, for learning them. While closed rules are used to complete a KG by answering given queries, OP rules identify the incom- pleteness of a KG by inducing such queries to ask. We use adaptations of Freebase, YAGO2, and a synthetic but complete Poker KG to evaluate oprl. We find that oprl mines hundreds of accurate rules from massive KGs with up to 1M facts. The learnt OP rules induce queries with preci- sion up to 98% and recall of 62% on a complete KG, demonstrating the first solution for active knowledge graph completion
- …
