PhDStudent

 Segue o link para posterior leitura: artigos sobre ontologia https://jessicatalisman.substack.com/p/from-metadata-to-meaning-the-knowledge?utm_source=substack&utm_medium=email  

 Esse vídeo mostra o poder da autoestima. Eu até comentei em um post do linkedin sobre esse vídeo: " d á pra filosofar sobre isso de diversas formas. Vou colocar em uma perspectiva acadêmica e se for pensar no reconhecimento por pares ele foi aceito pela galera que até replicou o que ele estava fazendo mas cada um ao seu modo. E deu certo! É preciso muita coragem para sustentar suas ideias e atitudes mesmo com julgamento alheio. Mas é importante acreditar no que se faz independente de tudo. Isso é ter muita autoestima! É o que precisamos hoje em dia!"   https://www.youtube.com/watch?v=nU7dxkIz1Vs&list=RDnU7dxkIz1Vs&start_radio=1  

  𝗪𝗵𝘆 𝗶𝘀 𝗕𝗙𝗢 (Basic Formal Ontology) so important in ontology engineering? Basic Formal Ontology is one of the most widely adopted upper ontologies in scientific domains because it helps organize reality at a very fundamental level. Its central idea is simple: Before defining domain concepts, we must first understand what kind of entity each concept is . BFO starts by separating reality into two major categories: 1️⃣ Continuants Things that persist through time while maintaining identity. Examples: • a patient • a mosquito • a hospital • a virus sample These entities continue to exist even while undergoing change. 2️⃣ Occurrents Things that unfold in time. Examples: • an infection • a diagnosis • a notification event • a hospitalization process These are temporal phenomena — they happen. This distinction is extremely powerful because many modeling problems happen when we mix objects and processes. Example: A common mistake is to model diagnosis as i...

 𝗪𝗵𝗮𝘁 𝗶𝘀 𝗮𝗻 𝘂𝗽𝗽𝗲𝗿 𝗼𝗻𝘁𝗼𝗹𝗼𝗴𝘆? An upper ontology defines the most general concepts that can be reused across many domains. It does not describe dengue, hospitals, cities, books, or sensors directly. Instead, it defines concepts such as: • entity • object • event • process • role • quality • relation These concepts work as a semantic foundation for more specific ontologies. Example: A domain ontology may define: • Patient • Mosquito • Notification • Fever But an upper ontology helps explain: • Patient is an object • Fever is a quality • Notification is an event • Diagnosis is a process This creates conceptual consistency. Why does this matter? Because when different ontologies share the same upper conceptual layer, semantic integration becomes more reliable. That is why upper ontologies are often used in: • knowledge graphs • semantic interoperability • ontology integration • explainable AI Some well-known examples: • Basic Formal O...

 Tabela → agregação → JSONL → embedding → FAISS → retrieval   1. Dados em tabela = estrutura para máquinas tradicionais Dados tabulares são organizados em linhas e colunas . Cada linha representa um registro. Cada coluna representa um atributo. Exemplo simples: municipio semana febre mialgia casos Campinas 12 51 47 80 Aqui: linha = um registro coluna = significado fixo Isso é excelente para: ✅ SQL ✅ filtros ✅ agregações ✅ estatística ✅ joins ✅ processamento analítico Ou seja: tabelas são ótimas para cálculo.   2. Problema: LLM não "pensa" naturalmente em colunas Para um modelo de linguagem, isso: Campinas | 12 | 51 | 47 | 80 não é naturalmente interpretável. Porque o embedding funciona melhor quando existe contexto linguístico . A LLM entende melhor frases como: "No município de Campinas, na semana epidemiológica 12, foram observados 80 casos, com febre em 51 registros e mialgia em 47." ✅ aqui existe semântica explícita   3. JSONL = cada...

  𝗡𝗼𝘁 𝗮𝗹𝗹 𝗼𝗻𝘁𝗼𝗹𝗼𝗴𝗶𝗲𝘀 𝗽𝗹𝗮𝘆 𝘁𝗵𝗲 𝘀𝗮𝗺𝗲 𝗿𝗼𝗹𝗲. When people hear the word ontology, they often imagine a single structure of classes and relationships. But in practice, ontologies can be organized into different types, depending on their level of abstraction and purpose. 𝗧𝗵𝗲 𝗺𝗮𝗶𝗻 𝘁𝘆𝗽𝗲𝘀 𝗮𝗿𝗲: 𝗢𝗻𝗲 𝗼𝗳 𝘁𝗵𝗲𝗺 𝗶𝘀 𝘂𝗽𝗽𝗲𝗿 𝗼𝗻𝘁𝗼𝗹𝗼𝗴𝗶𝗲𝘀. They describe very general concepts such as: • entity • process • event • object • relation • time Examples include Basic Formal Ontology (BFO) and Unified Foundational Ontology (UFO). Their role is to provide a shared conceptual foundation across domains. For example, concepts such as disease, city, patient, or sensor can all be aligned under broader categories. 𝗔𝗻𝗼𝘁𝗵𝗲𝗿 𝘁𝘆𝗽𝗲 𝗶𝘀 𝗱𝗼𝗺𝗮𝗶𝗻 𝗼𝗻𝘁𝗼𝗹𝗼𝗴𝗶𝗲𝘀. These represent knowledge specific to a field. Examples: • healthcare • education • agriculture • transportation • epidemiological surveillance A health ontology may d...

  Conversa interessante sobre ia que passa por temas como RAG, Yann LeCun (cientista chefe da Meta), Fake news com IA, política. https://open.spotify.com/episode/3naFVUgMKaFmIfXlOzX04o?si=X9pw3VKsTyy41Oypp0GbNA   Interessante pois fui olhar quem foi Yann LeCun e descobri que Mark Zukerberg contratou Alexandr Wang no lugar dele devido sua visão de mundo mas focada em LLM e rotulagem de dados.  https://www.youtube.com/watch?v=y12yZ7bQizk&t=1151s     

  M𝗼𝗻𝗼𝘁𝗼𝗻𝗶𝗰 𝗿𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 does allow revision — but not by automatically withdrawing previous conclusions. Monotonicity means that once a conclusion is logically inferred, adding new knowledge does not remove that conclusion. Instead, if revision is needed, the knowledge model itself must be explicitly reformulated. Example: • every bird flies ∀x (Bird(x) → Flies(x)) • penguin is a bird Bird(Penguin) From these statements, the inference is: • penguin flies Flies(Penguin) If we later add: • penguin does not fly ¬Flies(Penguin) The previous conclusion remains logically available: • Flies(Penguin) • ¬Flies(Penguin) This is not a loss of monotonicity. It is logical inconsistency. So how is revision achieved in monotonic systems? By refining the ontology itself. Instead of: • every bird flies ∀x (Bird(x) → Flies(x)) it is usually better to define: • flying bird is a subclass of bird ∀x (FlyingBird(x) → Bird(x)) • penguin is a bird Bird(Penguin) • penguin does not fly ¬Flies...

  Why inference changes what you can retrieve from an ontology One of the most interesting aspects of ontologies is that the same query can produce different results depending on whether reasoning has been applied or not . Without inference, a query only retrieves what is explicitly asserted. With inference, the ontology can reveal knowledge that was never directly written — but logically follows from the axioms. Example Suppose we define: SevereDengue ⊑ DengueCase Patient123 rdf:type SevereDengue Now consider the query: SELECT ?x WHERE {   ?x a :DengueCase . } Without inference Result: no result (if Patient123 was only declared as SevereDengue ) With inference enabled Result: Patient123 Why? Because the reasoner derives: Patient123 rdf:type DengueCase . even though this triple was never explicitly stored. This means: Inference transforms ontologies from static structures into knowledge generators. The query remains the same. The ontology becomes ric...

 Ao entrar no wordpress acabei me deparando com um blog antigo que construí quando estava cursando  ciências e humanidades na UFABC. Segue o link do blog: www.misturabrasileira.wordpress.com

Esse é meu primeiro post e irei resgatar alguns assuntos que compartilhei no linkedin. Resolvi fazer publicações (posts) no linkedin mas pretendo fazer um backup nesse blog para facilitar minhas buscas. No meu primeiro post escrevi com auxílio do chatGPT sobre semantic web e o conceito de anyone can say anything about any topic. Aqui está o link: https://www.linkedin.com/posts/renato-bueno-domingos-de-oliveira-14a4491b_semanticweb-ontology-linkeddata-share-7439662439807246336-mmXl?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAQooYoBiO6MjINiIumiwZCA_yPABO6gTWE    Meu  segundo post foi sobre metadados e sua importância para os dados. Segue o link do post https://www.linkedin.com/posts/renato-bueno-domingos-de-oliveira-14a4491b_semanticweb-ontology-linkeddata-share-7439662439807246336-mmXl?utm_source=share&utm_medium=member_desktop&rcm=ACoAAAQooYoBiO6MjINiIumiwZCA_yPABO6gTWE     O último post foi sobre o modelo de 5 estrelas proposto por Tim ...