Despite improvements in performances on different natural language generation tasks, deep neural models are prone to hallucinating facts that are incorrect or nonexistent.Different hypotheses are proposed and examined separately for different tasks, but no systematic explanations are available across these tasks.In this study, we draw connections between hallucinations and predictive uncertainty in conditional language generation.We investigate their relationship in both image captioning and data-to-text generation and propose a simple extension to beam search to reduce hallucination.Our analysis shows that higher predictive uncertainty corresponds to a higher chance of hallucination.Epistemic uncertainty is more indicative of hallucination than aleatoric or total uncertainties.It helps to achieve better results of trading performance in standard metric for less hallucination with the proposed beam search variant.

Reliable uncertainty quantification is a first step towards building explainable, transparent, and accountable artificial intelligent systems. Recent progress in Bayesian deep learning has made such quantification realizable. In this paper, we propose novel methods to study the benefits of characterizing model and data uncertainties for natural language processing (NLP) tasks. With empirical experiments on sentiment analysis, named entity recognition, and language modeling using convolutional and recurrent neural network models, we show that explicitly modeling uncertainties is not only necessary to measure output confidence levels, but also useful at enhancing model performances in various NLP tasks.

Different from other sequential data, sentences in natural language are structured by linguistic grammars. Previous generative conversational models with chain-structured decoder ignore this structure in human language and might generate plausible responses with less satisfactory relevance and fluency. In this study, we aim to incorporate the results from linguistic analysis into the process of sentence generation for high-quality conversation generation. Specifically, we use a dependency parser to transform each response sentence into a dependency tree and construct a training corpus of sentence-tree pairs. A tree-structured decoder is developed to learn the mapping from a sentence to its tree, where different types of hidden states are used to depict the local dependencies from an internal tree node to its children. For training acceleration, we propose a tree canonicalization method, which transforms trees into equivalent ternary trees. Then, with a proposed tree-structured search method, the model is able to generate the most probable responses in the form of dependency trees, which are finally flattened into sequences as the system output. Experimental results demonstrate that the proposed X2Tree framework outperforms baseline methods over 11.15% increase of acceptance ratio.

Neural models aiming at generating meaningful and diverse response is attracting increasing attention over recent years. For a given post, the conventional encoder-decoder models tend to learn high-frequency but trivial responses, or are difficult to determine which speaking styles are suitable to generate responses. To address this issue, we propose the elastic responding machine (ERM), which is based on a proposed encoder-diverter-filter-decoder framework. ERM models the multiple responding mechanisms to not only generate acceptable responses for a given post but also improve the diversity of responses. Here, the mechanisms could be regraded as some latent variables, and for a given post different responses may be generated by different mechanisms. The experiments demonstrate the quality and diversity of the generated responses, intuitively show how the learned model controls response mechanism when responding, and reveal some underlying relationship between mechanism and language style.

This paper presents a general framework that aims to address the task of segmenting three-dimensional (3-D) scan data representing the human form into subsets which correspond to functional human body parts. Such a task is challenging due to the articulated and deformable nature of the human body. A salient feature of this framework is that it is able to cope with various body postures and is in addition robust to noise, holes, irregular sampling and rigid transformations. Although whole human body scanners are now capable of routinely capturing the shape of the whole body in machine readable format, they have not yet realized their potential to provide automatic extraction of key body measurements. Automated production of anthropometric databases is a prerequisite to satisfying the needs of certain industrial sectors (e.g., the clothing industry). This implies that in order to extract specific measurements of interest, whole body 3-D scan data must be segmented by machine into subsets corresponding to functional human body parts. However, previously reported attempts at automating the segmentation process suffer from various limitations, such as being restricted to a standard specific posture and being vulnerable to scan data artifacts. Our human body segmentation algorithm advances the state of the art to overcome the above limitations and we present experimental results obtained using both real and synthetic data that confirm the validity, effectiveness, and robustness of our approach.

This paper presents a novel curve based algorithm of stereo vision to reconstruct 3D line-like objects. B-spline approximations of 2D edge curves are selected as primitives for the reconstruction of their corresponding space curves so that, under the assumption of affine camera model, a 3D curve can be derived from reconstructing its control points according to the affine invariant property of B-Spline curves. The superiority of B-spline model in representing free-form curves gives good geometric properties of reconstruction results. Both theoretical analysis and experimental results demonstrate the validity of our approach.