Flooding is one of the leading threats natural disasters to human life and property, especially in densely populated urban areas. Rapid precise extraction flooded areas key supporting emergency-response planning providing damage assessment both spatial temporal measurements. Unmanned Aerial Vehicles (UAV) technology has recently been recognized as an efficient photogrammetry data acquisition platform quickly deliver high-resolution imagery because its cost-effectiveness, ability fly at lower...

This article reports on the novel task of spatial role labeling in natural language text. It proposes machine learning methods to extract roles and their relations. work experiments with both a step-wise approach, where prepositions are found related trajectors, landmarks then extracted, joint relation its composing indicator, trajector, landmark classified collectively. Context-dependent techniques, such as skip-chain conditional random field, yield good results GUM-evaluation (Maptask)...

Compositional generalization is crucial for artificial intelligence agents to solve complex vision-language reasoning tasks. Neuro-symbolic approaches have demonstrated promise in capturing compositional structures, but they face critical challenges: (a) reliance on predefined predicates symbolic representations that limit adaptability, (b) difficulty extracting from raw data, and (c) using non-differentiable operations combining primitive concepts. To address these issues, we propose...

Human languages exhibit a variety of strategies for communicating spatial information, including toponyms, nominals, locations that are described in relation to other locations, and movements along paths. SpaceEval is combined information extraction classification task with the goal identifying categorizing such information. In this paper, we describe task, annotation schema, corpora, evaluate performance several supervised semi-supervised machine learning systems developed automating task.

Roshanak Mirzaee, Hossein Rajaby Faghihi, Qiang Ning, Parisa Kordjamshidi. Proceedings of the 2021 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies. 2021.

This work deals with the challenge of learning and reasoning over language vision data for related downstream tasks such as visual question answering (VQA) natural (NLVR). We design a novel cross-modality relevance module that is used in an end-to-end framework to learn representation between components various input modalities under supervision target task, which more generalizable unobserved compared merely reshaping original space. In addition modeling textual entities entities, we model...

Spatial Reasoning is an important component of human cognition and area in which the latest Vision-language models (VLMs) show signs difficulty. The current analysis works use image captioning tasks visual question answering. In this work, we propose using Referring Expression Comprehension task instead as a platform for evaluation spatial reasoning by VLMs. This provides opportunity deeper comprehension grounding abilities when there 1) ambiguity object detection, 2) complex expressions...

This paper considers the challenges Large Language Models (LLMs) face when reasoning over text that includes information involving uncertainty explicitly quantified via probability values. type of is relevant to a variety contexts ranging from everyday conversations medical decision-making. Despite improvements in mathematical capabilities LLMs, they still exhibit significant difficulties it comes probabilistic reasoning. To deal with this problem, we introduce Bayesian Linguistic Inference...

We aim to automatically extract species names of bacteria and their locations from webpages. This task is important for exploiting the vast amount biological knowledge which expressed in diverse natural language texts putting this databases easy access by biologists. The challenging previous results are far below an acceptable level performance, particularly extraction localization relationships. Therefore, we design a new system such extractions, using framework structured machine learning...

This work deals with the challenge of learning and reasoning over multi-hop question answering (QA). We propose a graph network based on semantic structure sentences to learn cross paragraph paths find supporting facts answer jointly. The proposed is heterogeneous document-level that contains nodes type sentence (question, title, other sentences), role labeling sub-graphs per contain arguments as predicates edges. Incorporating argument types, phrases, semantics edges originated from SRL...

Tracking entities throughout a procedure described in text is challenging due to the dynamic nature of world process. Firstly, we propose formulate this task as question answering problem. This enables us use pre-trained transformer-based language models on other QA benchmarks by adapting those procedural understanding. Secondly, since cannot encode flow events themselves, Time-Stamped Language Model (TSLM) event information LMs architecture introducing timestamp encoding. Our model...

This work is on a previously formalized semantic evaluation task of spatial role labeling (SpRL) that aims at extraction formal meaning from text. Here, we report the results initial efforts towards exploiting visual information in form images to help language understanding. We discuss way designing new models framework declarative learning-based programming (DeLBP). The DeLBP facilitates combining modalities and representing various data unified graph. learning inference exploit structure...