Marie and BERT─A Knowledge Graph Embedding Based Question Answering System for Chemistry

Abstract

This paper presents a novel knowledge graph question answering (KGQA) system for chemistry, which is implemented on hybrid knowledge graph embeddings, aiming to provide fact-oriented information retrieval for chemistry-related research and industrial applications. Unlike other existing designs, the system operates on multiple embedding spaces, which use various embedding methods and queries the embedding spaces in parallel. With the answers returned from multiple embedding spaces, the system leverages a score alignment model to adjust the answer scores and rerank the answers. Further, the system implements an algorithm to derive implicit multihop relations to handle the complexities of deep ontologies and improve multihop question answering. The system also implements a BERT-based bidirectional entity-linking model to enhance the robustness and accuracy of the entity-linking module. The system uses a joint numerical embedding model to efficiently handle numerical filtering questions. Further, it can invoke semantic agents to perform dynamic calculations autonomously. Finally, the KGQA system handles numerous chemical reaction mechanisms using semantic parsing supported by a Linked Data Fragment server. This paper evaluates the accuracy of each module within the KGQA system with a chemistry question data set.

Mixtures Recomposition by Neural Nets: A Multidisciplinary Overview

“…As suggested by systems theory, ANN-based recomposition complements traditional computational decomposition and involves organizing mixture components to understand how they dynamically work together to produce specific behaviors, especially through orchestrated mechanisms comprising feedback loops and including the role of noise and interactions with the environment. In fact, ANN graph representations of chemical mixtures demonstrate exceptional flexibility (up to hypergraphs) and expressivity for complex system analysis, visualization, interpretation, and transdisciplinary communication. − Combining them with probabilistic models and natural language processing may enable multifaceted deciphering of intricate relationships within a knowledge graph. , The concept of “fluid spatiality” enhances our understanding of networks by introducing dynamics into the connections, transitioning from “clocks” to “clouds” . Probabilistic Graph Models provide such a framework, elements within categories being treated as random variables with conditional dependencies …”

Leveraging Text-to-Text Pretrained Language Models for Question Answering in Chemistry

“…Doing away with the need for explicit logical forms, information retrieval approaches rely on signals obtained from input questions to enumerate candidate answers, rank them, and return the top candidate. Inspired by growing research in information retrieval-based QA systems − and transfer learning methods using pretrained language models (PLMs), the Marie and BERT system leverages the PLM BERT and KG embedding techniques to map KG constituents and natural language questions into low-dimensional vector spaces where ranking of candidate answers can be efficiently performed as vector operations. However, KG embeddings are generally expensive to train and have to be retrained whenever new factual assertions are added to the knowledge store.…”

“…The developed KGQA system is lightweight and able to not just handle multihop questions and but also balance between correctness and speed in a CPU-only setting. This new approach offers significant advantages over the prior implementation that relies on KG embeddings . Specifically, the updated system boasts higher accuracy and greater flexibility in accommodating changes and evolution of the data stored in the KG without necessitating retraining.…”

“…Marie − is a KGQA system for chemistry. It is part of a wider effort to develop a user-friendly interface to interact with The World Avatar (TWA)a world model built upon a dynamic KG with the aim to interoperate across heterogeneous ontologies and disparate domains, including not just chemistry but also power systems, 3D models of cities and landscapes, among others .…”

“…Additionally, it is not a trivial task to find a KG embedding model suitable for a given ontology. In fact, a poor modeling strategy can inhibit not only the end performance of KGQA systems but also the extent of complexity that reasoning can be done in the embedding space; as a case in point, the Marie and BERT system 12 is unable to handle questions with more than one relation. Additionally, as noted by Yu et al, 27 information retrieval-based systems tend to be outperformed by their semantic parsing-based counterparts.…”