Using Analogy to Acquire Commonsense Knowledge from Human Contributors

The goal of the work reported here is to capture the commonsense knowledge of non-expert human contributors. Achieving this goal will enable more intelligent human-computer interfaces and pave the way for computers to reason about our world. In the domain of natural language processing, it will provide the world knowledge much needed for semantic processing of natural language. To acquire knowledge from contributors not trained in knowledge engineering, I take the following four steps: (i) develop a knowledge representation (KR) model for simple assertions in natural language, (ii) introduce cumulative analogy, a class of nearest-neighbor based analogical reasoning algorithms over this representation, (iii) argue that cumulative analogy is well suited for knowledge acquisition (KA) based on a theoretical analysis of effectiveness of KA with this approach, and (iv) test the KR model and the effectiveness of the cumulative analogy algorithms empirically. To investigate effectiveness of cumulative analogy for KA empirically, Learner, an open source system for KA by cumulative analogy has been implemented, deployed, and evaluated. (The site "1001 Questions," is available at http://teach-computers.org/learner.html). Learner acquires assertion-level knowledge by constructing shallow semantic analogies between a KA topic and its nearest neighbors and posing these analogies as natural language questions to human contributors. Suppose, for example, that based on the knowledge about "newspapers" already present in the knowledge base, Learner judges "newspaper" to be similar to "book" and "magazine." Further suppose that assertions "books contain information" and "magazines contain information" are also already in the knowledge base. Then Learner will use cumulative analogy from the similar topics to ask humans whether "newspapers contain information." Because similarity between topics is computed based on what is already known about them, Learner exhibits bootstrapping behavior --- the quality of its questions improves as it gathers more knowledge. By summing evidence for and against posing any given question, Learner also exhibits noise tolerance, limiting the effect of incorrect similarities. The KA power of shallow semantic analogy from nearest neighbors is one of the main findings of this thesis. I perform an analysis of commonsense knowledge collected by another research effort that did not rely on analogical reasoning and demonstrate that indeed there is sufficient amount of correlation in the knowledge base to motivate using cumulative analogy from nearest neighbors as a KA method. Empirically, evaluating the percentages of questions answered affirmatively, negatively and judged to be nonsensical in the cumulative analogy case compares favorably with the baseline, no-similarity case that relies on random objects rather than nearest neighbors. Of the questions generated by cumulative analogy, contributors answered 45% affirmatively, 28% negatively and marked 13% as nonsensical; in the control, no-similarity case 8% of questions were answered affirmatively, 60% negatively and 26% were marked as nonsensical.

A Statistical Image-Based Shape Model for Visual Hull Reconstruction and 3D Structure Inference

We present a statistical image-based shape + structure model for Bayesian visual hull reconstruction and 3D structure inference. The 3D shape of a class of objects is represented by sets of contours from silhouette views simultaneously observed from multiple calibrated cameras. Bayesian reconstructions of new shapes are then estimated using a prior density constructed with a mixture model and probabilistic principal components analysis. We show how the use of a class-specific prior in a visual hull reconstruction can reduce the effect of segmentation errors from the silhouette extraction process. The proposed method is applied to a data set of pedestrian images, and improvements in the approximate 3D models under various noise conditions are shown. We further augment the shape model to incorporate structural features of interest; unknown structural parameters for a novel set of contours are then inferred via the Bayesian reconstruction process. Model matching and parameter inference are done entirely in the image domain and require no explicit 3D construction. Our shape model enables accurate estimation of structure despite segmentation errors or missing views in the input silhouettes, and works even with only a single input view. Using a data set of thousands of pedestrian images generated from a synthetic model, we can accurately infer the 3D locations of 19 joints on the body based on observed silhouette contours from real images.

Fast Perceptual Learning in Visual Hyperacuity

In many different spatial discrimination tasks, such as in determining the sign of the offset in a vernier stimulus, the human visual system exhibits hyperacuity-level performance by evaluating spatial relations with the precision of a fraction of a photoreceptor"s diameter. We propose that this impressive performance depends in part on a fast learning process that uses relatively few examples and occurs at an early processing stage in the visual pathway. We show that this hypothesis is plausible by demonstrating that it is possible to synthesize, from a small number of examples of a given task, a simple (HyperBF) network that attains the required performance level. We then verify with psychophysical experiments some of the key predictions of our conjecture. In particular, we show that fast timulus-specific learning indeed takes place in the human visual system and that this learning does not transfer between two slightly different hyperacuity tasks.

Sharing visual features for multiclass and multiview object detection

We consider the problem of detecting a large number of different classes of objects in cluttered scenes. Traditional approaches require applying a battery of different classifiers to the image, at multiple locations and scales. This can be slow and can require a lot of training data, since each classifier requires the computation of many different image features. In particular, for independently trained detectors, the (run-time) computational complexity, and the (training-time) sample complexity, scales linearly with the number of classes to be detected. It seems unlikely that such an approach will scale up to allow recognition of hundreds or thousands of objects. We present a multi-class boosting procedure (joint boosting) that reduces the computational and sample complexity, by finding common features that can be shared across the classes (and/or views). The detectors for each class are trained jointly, rather than independently. For a given performance level, the total number of features required, and therefore the computational cost, is observed to scale approximately logarithmically with the number of classes. The features selected jointly are closer to edges and generic features typical of many natural structures instead of finding specific object parts. Those generic features generalize better and reduce considerably the computational cost of an algorithm for multi-class object detection.

The Role of Knowledge in Visual Shape Representation

This report shows how knowledge about the visual world can be built into a shape representation in the form of a descriptive vocabulary making explicit the important geometrical relationships comprising objects' shapes. Two computational tools are offered: (1) Shapestokens are placed on a Scale-Space Blackboard, (2) Dimensionality-reduction captures deformation classes in configurations of tokens. Knowledge lies in the token types and deformation classes tailored to the constraints and regularities ofparticular shape worlds. A hierarchical shape vocabulary has been implemented supporting several later visual tasks in the two-dimensional shape domain of the dorsal fins of fishes.

Learning Shape Descriptions: Generating and Generalizing Models of Visual Objects

We present the results of an implemented system for learning structural prototypes from grey-scale images. We show how to divide an object into subparts and how to encode the properties of these subparts and the relations between them. We discuss the importance of hierarchy and grouping in representing objects and show how a notion of visual similarities can be embedded in the description language. Finally we exhibit a learning algorithm that forms class models from the descriptions produced and uses these models to recognize new members of the class.

Natural Language Input for a Computer Problem Solving System

The STUDENT problem solving system, programmed in LISP, accepts as input a comfortable but restricted subset of English which can express a wide variety of algebra story problems. STUDENT finds the solution to a large class of these problems. STUDENT can utilize a store of global information not specific to any one problem, and may make assumptions about the interpretation of ambiguities in the wording of the problem being solved. If it uses such information or makes any assumptions, STUDENT communicates this fact to the user. The thesis includes a summary of other English language questions-answering systems. All these systems, and STUDENT, are evaluated according to four standard criteria. The linguistic analysis in STUDENT is a first approximation to the analytic portion of a semantic theory of discourse outlined in the thesis. STUDENT finds the set of kernel sentences which are the base of the input discourse, and transforms this sequence of kernel sentences into a set of simultaneous equations which form the semantic base of the STUDENT system. STUDENT then tries to solve this set of equations for the values of requested unknowns. If it is successful it gives the answers in English. If not, STUDENT asks the user for more information, and indicates the nature of the desired information. The STUDENT system is a first step toward natural language communication with computers. Further work on the semantic theory proposed should result in much more sophisticated systems.

The Logical Problem of Language Change

This paper considers the problem of language change. Linguists must explain not only how languages are learned but also how and why they have evolved along certain trajectories and not others. While the language learning problem has focused on the behavior of individuals and how they acquire a particular grammar from a class of grammars ${cal G}$, here we consider a population of such learners and investigate the emergent, global population characteristics of linguistic communities over several generations. We argue that language change follows logically from specific assumptions about grammatical theories and learning paradigms. In particular, we are able to transform parameterized theories and memoryless acquisition algorithms into grammatical dynamical systems, whose evolution depicts a population's evolving linguistic composition. We investigate the linguistic and computational consequences of this model, showing that the formalization allows one to ask questions about diachronic that one otherwise could not ask, such as the effect of varying initial conditions on the resulting diachronic trajectories. From a more programmatic perspective, we give an example of how the dynamical system model for language change can serve as a way to distinguish among alternative grammatical theories, introducing a formal diachronic adequacy criterion for linguistic theories.

Perceptually-based Comparison of Image Similarity Metrics

The image comparison operation ??sessing how well one image matches another ??rms a critical component of many image analysis systems and models of human visual processing. Two norms used commonly for this purpose are L1 and L2, which are specific instances of the Minkowski metric. However, there is often not a principled reason for selecting one norm over the other. One way to address this problem is by examining whether one metric better captures the perceptual notion of image similarity than the other. With this goal, we examined perceptual preferences for images retrieved on the basis of the L1 versus the L2 norm. These images were either small fragments without recognizable content, or larger patterns with recognizable content created via vector quantization. In both conditions the subjects showed a consistent preference for images matched using the L1 metric. These results suggest that, in the domain of natural images of the kind we have used, the L1 metric may better capture human notions of image similarity.