Hovy, Eduard


Controllable Invariance through Adversarial Feature Learning

arXiv.org Artificial Intelligence

Learning meaningful representations that maintain the content necessary for a particular task while filtering away detrimental variations is a problem of great interest in machine learning. In this paper, we tackle the problem of learning representations invariant to a specific factor or trait of data. The representation learning process is formulated as an adversarial minimax game. We analyze the optimal equilibrium of such a game and find that it amounts to maximizing the uncertainty of inferring the detrimental factor given the representation while maximizing the certainty of making task-specific predictions. On three benchmark tasks, namely fair and bias-free classification, language-independent generation, and lighting-independent image classification, we show that the proposed framework induces an invariant representation, and leads to better generalization evidenced by the improved performance.


RACE: Large-scale ReAding Comprehension Dataset From Examinations

arXiv.org Artificial Intelligence

We present RACE, a new dataset for benchmark evaluation of methods in the reading comprehension task. Collected from the English exams for middle and high school Chinese students in the age range between 12 to 18, RACE consists of near 28,000 passages and near 100,000 questions generated by human experts (English instructors), and covers a variety of topics which are carefully designed for evaluating the students' ability in understanding and reasoning. In particular, the proportion of questions that requires reasoning is much larger in RACE than that in other benchmark datasets for reading comprehension, and there is a significant gap between the performance of the state-of-the-art models (43%) and the ceiling human performance (95%). We hope this new dataset can serve as a valuable resource for research and evaluation in machine comprehension. The dataset is freely available at http://www.cs.cmu.edu/~glai1/data/race/ and the code is available at https://github.com/qizhex/RACE_AR_baselines.


Softmax Q-Distribution Estimation for Structured Prediction: A Theoretical Interpretation for RAML

arXiv.org Machine Learning

Reward augmented maximum likelihood (RAML), a simple and effective learning framework to directly optimize towards the reward function in structured prediction tasks, has led to a number of impressive empirical successes. RAML incorporates task-specific reward by performing maximum-likelihood updates on candidate outputs sampled according to an exponentiated payoff distribution, which gives higher probabilities to candidates that are close to the reference output. While RAML is notable for its simplicity, efficiency, and its impressive empirical successes, the theoretical properties of RAML, especially the behavior of the exponentiated payoff distribution, has not been examined thoroughly. In this work, we introduce softmax Q-distribution estimation, a novel theoretical interpretation of RAML, which reveals the relation between RAML and Bayesian decision theory. The softmax Q-distribution can be regarded as a smooth approximation of the Bayes decision boundary, and the Bayes decision rule is achieved by decoding with this Q-distribution. We further show that RAML is equivalent to approximately estimating the softmax Q-distribution, with the temperature $\tau$ controlling approximation error. We perform two experiments, one on synthetic data of multi-class classification and one on real data of image captioning, to demonstrate the relationship between RAML and the proposed softmax Q-distribution estimation method, verifying our theoretical analysis. Additional experiments on three structured prediction tasks with rewards defined on sequential (named entity recognition), tree-based (dependency parsing) and irregular (machine translation) structures show notable improvements over maximum likelihood baselines.


Neural Probabilistic Model for Non-projective MST Parsing

arXiv.org Machine Learning

In this paper, we propose a probabilistic parsing model, which defines a proper conditional probability distribution over non-projective dependency trees for a given sentence, using neural representations as inputs. The neural network architecture is based on bi-directional LSTM-CNNs which benefits from both word- and character-level representations automatically, by using combination of bidirectional LSTM and CNN. On top of the neural network, we introduce a probabilistic structured layer, defining a conditional log-linear model over non-projective trees. We evaluate our model on 17 different datasets, across 14 different languages. By exploiting Kirchhoff's Matrix-Tree Theorem (Tutte, 1984), the partition functions and marginals can be computed efficiently, leading to a straight-forward end-to-end model training procedure via back-propagation. Our parser achieves state-of-the-art parsing performance on nine datasets.


An Active Learning Approach to Coreference Resolution

AAAI Conferences

In this paper, we define the problem of coreference resolution in text as one of clustering with pairwise constraints where human experts are asked to provide pairwise constraints (pairwise judgments of coreferentiality) to guide the clustering process. Positing that these pairwise judgments are easy to obtain from humans given the right context, we show that with significantly lower number of pairwise judgments and feature-engineering effort, we can achieve competitive coreference performance. Further, we describe an active learning strategy that minimizes the overall number of such pairwise judgments needed by asking the most informative questions to human experts at each step of coreference resolution. We evaluate this hypothesis and our algorithms on both entity and event coreference tasks and on two languages.


Using Part-Of Relations for Discovering Causality

AAAI Conferences

Historically, causal markers, syntactic structures and connectives have been the sole identifying features for automatically extracting causal relations in natural language discourse. However various connectives such as “and,” prepositions such as “as” and other syntactic structures are highly ambiguous in nature, and it is clear that one cannot solely rely on lexico-syntactic markers for detection of causal phenomenon in discourse. This paper introduces the theory of granularity and describes different approaches to identify granularity in natural language. As causality is often granular in nature, we use granularity relations to discover and infer the presence of causal relations in text. We compare this with causal relations identified using just causal markers. We achieve a precision of 0.91 and a recall of 0.79 using granularity for causal relation detection, as compared to a precision of 0.79 and a recall of 0.44 using pure causal markers for causality detection.


Toward Virtual Humans

AI Magazine

This article describes the virtual humans developed as part of the Mission Rehearsal Exercise project, a virtual reality-based training system. This project is an ambitious exercise in integration, both in the sense of integrating technology with entertainment industry content, but also in that we have joined a number of component technologies that have not been integrated before. This integration has not only raised new research issues, but it has also suggested some new approaches to difficult problems. We describe the key capabilities of the virtual humans, including task representation and reasoning, natural language dialogue, and emotion reasoning, and show how these capabilities are integrated to provide more human-level intelligence than would otherwise be possible.


Toward Virtual Humans

AI Magazine

This article describes the virtual humans developed as part of the Mission Rehearsal Exercise project, a virtual reality-based training system. This project is an ambitious exercise in integration, both in the sense of integrating technology with entertainment industry content, but also in that we have joined a number of component technologies that have not been integrated before. This integration has not only raised new research issues, but it has also suggested some new approaches to difficult problems. We describe the key capabilities of the virtual humans, including task representation and reasoning, natural language dialogue, and emotion reasoning, and show how these capabilities are integrated to provide more human-level intelligence than would otherwise be possible.


The AAAI Spring Symposia

AI Magazine

The Association for the Advancement of Artificial Intelligence, in cooperation with Stanford University's Department of Computer Science, held the 1998 Spring Symposium Series on 23 to 25 March at Stanford University. The topics of the eight symposia were (1) Applying Machine Learning to Discourse Processing, (2) Integrating Robotic Research: Taking the Next Leap, (3) Intelligent Environments, (4) Intelligent Text Summarization, (5) Interactive and Mixed-Initiative Decision-Theoretic Systems, (6) Multimodal Reasoning, (7) Prospects for a Common-Sense Theory of Causation, and (8) Satisficing Models.


The AAAI Spring Symposia

AI Magazine

The Association for the Advancement of Artificial Intelligence, in cooperation with Stanford University's Department of Computer Science, held the 1998 Spring Symposium Series on 23 to 25 March at Stanford University. The topics of the eight symposia were (1) Applying Machine Learning to Discourse Processing, (2) Integrating Robotic Research: Taking the Next Leap, (3) Intelligent Environments, (4) Intelligent Text Summarization, (5) Interactive and Mixed-Initiative Decision-Theoretic Systems, (6) Multimodal Reasoning, (7) Prospects for a Common-Sense Theory of Causation, and (8) Satisficing Models.