Partially observable Markov decision processes (POMDPs) have received significant interest in research on spoken dialogue systems, due to among many benefits its ability to naturally model the dialogue strategy selection problem under unreliable automated speech recognition. However, the POMDP approaches are essentially model-based, and as a result, the dialogue strategy computed from POMDP is still subject to the correctness of the model. In this paper, we extend some of the previous MDP user models to POMDPs, and evaluate the effects of user models on the dialogue strategy computed from POMDPs. We experimentally show that the strategies computed from POMDPs perform better than those from MDPs, and the strategies computed from poor user models fail severely when tested on different user models. This paper further investigates the evaluation methods for dialogue strategies, and proposes a method based on the bias-variance analysis for reliably estimating the dialogue performance.

Although HMM is widely used for online handwriting recognition, there is no simple and well-established method of designing the HMM topology. We propose a data-driven systematic method to design HMM topology. Data samples in a single pattern class are structurally simplified into a sequence of straight-line segments, and then these simplified representations of the samples are clustered. An HMM is constructed for each of these clusters, by assigning a state to each straight-line segments. Then the resulting multiple models of the class are combined to form an architecture of a multiple parallel-path HMM, which behaves as a single HMM. To avoid excessive growing of the number of the states, parameter tying is applied such that structural similarity among patterns is reflected. Experiments on online Hangul recognition showed about 19% of error reductions, compared to the intuitive design method.

We propose a new parameter estimation method called minimum entropy estimation (MEE), which tries to minimize the conditional entropy of the models given the input data. Since there is no assumption in MEE for the correctness of the parameter space of models, MEE will perform not less than the other estimation methods such as maximum likelihood estimation and maximum mutual information estimation, under the condition that the training data size is large enough. In the experiments, the three estimation methods are applied to the parameter estimation of hierarchical random graphs so that their estimation performance can be compared with each other.

A new approach for on-line cursive script recognition that combines a letter spotting technique with an island-driven lattice search algorithm is presented. Initially, all plausible letter components within an input pattern are detected, using a letter spotting technique based on hidden Markov models. A word hypothesis lattice is generated as a result of the letter spotting. Then an island-driven search algorithm is performed to find the optimal path on the word hypothesis lattice, which corresponds to the most probable word among the dictionary words. The results of this experiment suggest that the proposed method works effectively in recognizing English cursive words. In a word recognition test, the average 85.4% word accuracy was obtained.

This paper presents a hybrid architecture of hidden Markov models (HMMs) and a multilayer perceptron (MLP). This exploits the discriminative capability of a neural network classifier while using HMM formalism to capture the dynamics of input patterns. The main purpose is to improve the discriminative power of the HMM-based recognizer by additionally classifying the likelihood values inside them with an MLP classifier. To appreciate the performance of the presented method, we apply it to the recognition problem of on-line handwritten characters. Simulations show that the proposed architecture leads to a significant improvement in generalization performance over conventional approaches to sequential pattern recognition.

Multiplayer feedforward networks trained by minimizing the mean squared error and by using a one of c teaching function yield network outputs that estimate posterior class probabilities. This provides a sound basis for combining the results from multiple networks to get more accurate classification. This paper presents a method for combining multiple networks based on fuzzy logic, especially the fuzzy integral. This method non-linearly combines objective evidence, in the form of a network output, with subjective evaluation of the importance of the individual neural networks. The experimental results with the recognition problem of on-line handwriting characters show that the performance of individual networks could be improved significantly.

In the area of artificial neural networks, the concept of combining multiple networks has been proposed as a new direction for the development of highly reliable neural network systems. The authors propose a method for multinetwork combination based on the fuzzy integral. This technique nonlinearly combines objective evidence, in the form of a fuzzy membership function, with subjective evaluation of the worth of the individual neural networks with respect to the decision. The experimental results with the recognition problem of on-line handwriting characters confirm the superiority of the presented method to the other voting techniques.>

Our goal is to perform handwritten character recognition using a bank of multilayer feedforward neural networks. This paper presents both the front-end and the back-end of such a recognition system. The front-end relies on a data pre-classification scheme based on the concept of segment. A segment can be viewed as a representative building block of handwriting. The back-end hinges on a connectionist approach. Instead of a single large network, a bank of parallel networks is developed to overcome commonly encountered difficulties such as slow training process and requirement for a large amount of training data. The recognition system has been evaluated, on tasks involving (i) discrimination between similarly shaped characters and (ii) recognition of discretely written upper-case characters.>

An operational domain-independent decision-aiding system for situation assessment tasks is described. The system elicits the user's perception of a given situation through a stylized English dialogue and focuses the user's attention on the issue of highest relevancy. Elicited problems are structured as networks where nodes represent variables and directed links represent causal relationships. The system uses a Bayesian inference procedure which combines causal and diagnostic reasoning using a bidirectional propagation of evidence in the form of belief parameters. Upon completion of the dialogue the system provides a formal structure representing relevant propositions, their interrelations, and their updated belief distributions.

The paper introduces three extensions of the A* search algorithm which improve the search efficiency by relaxing the admissibility condition. 1) A* employs an admissible heuristic function but invokes quicker termination conditions while still guaranteeing that the cost of the solution found will not exceed the optimal cost by a factor greater than 1 + . 2) R¿* may employ heuristic functions which occasionally violate the admissibility condition, but guarantees that at termination the risk of missing the opportunity for further cost reduction is at most ¿. 3) R¿*,* is a speedup version of R¿*, combining the termination condition of A* with the risk-admissibility condition of R¿*. The Traveling Salesman problem was used as a test vehicle to examine the performances of the algorithms A* and R¿*. The advantages of A* are shown to be significant in difficult problems, i.e., problems requiring a large number of expansions due to the presence of many subtours of roughly equal costs. The use of R¿* is shown to produce a 4:1 reduction in search time with only a minor increase in final solution cost.

With the exponential explosion in chip complexity there is a growing need for high level design aids. A preliminary experiment was conducted in mating a hierarchical, top-down DA system for data paths with an existing IC placement and router. Nine designs ranging in complexity from 7 to 150 register transfers were synthesized. Strong correlations were observed between high level, abstract measures and final placed and routed chip area. It was observed that use of logic primitives of a moderate level abstraction yielded a 50% savings in placed and routed chip area.