Mayuri Duggirala

Speech-interface prompt design: lessons from the field

Designers of IVR systems often shy away from using speech prompts; preferring, where they can, to use keypad input. Part of the reason is that speech processing is expensive and often error prone. This work attempts to address this problem by offering guidelines for prompt design based on field experiments. It is shown, specifically, that accuracy can be influenced by prompt examples, depending on the nature of the information requested.

Empirical evaluation of emotion classification accuracy for non-acted speech

Emotion recognition is important at the workplace because it impacts a multitude of outcomes, such as performance, engagement and well-being. Emotion recognition from audio is an attractive option due to its non-obtrusive nature and availability of microphones in devices at the workplace. We describe building a classifier that analyzes the para-linguistic features of audio streams to classify them into positive, neutral and negative affect. Since speech at the workplace is different from acted speech, and because it is important that the training data be situated in the right context, we designed and executed an emotion induction procedure to generate a corpus of non-acted speech data of 33 speakers. The corpus was used to train a set of classification models and a comparative analysis of these models was used to choose the feature parameters. Bootstrap aggregation (bagging) was then used on the best combination of algorithm (Random Forest) and features (60 millisecond window size). The resulting classification accuracy of 73% is on par with, or exceeds, accuracies reported in the current literature for non-acted speech for a speaker-dependent test set. For reference, we also report the speaker-dependent recognition accuracy (95%) of the same classifier trained and tested on acted speech for three emotions in the Emo-DB database.

Towards fine grained human behaviour simulation models

Agent based simulation modelers have found it difficult to build grounded fine grained simulation models of human behavior. By grounded we mean that the model elements must rest on valid observations of the real world, by fine grained we mean the ability to factor in multiple dimensions of behavior such as personality, affect and stress. In this paper, we present a set of guidelines to build such models that use fragments of behavior mined from past literature in the social sciences as well as behavioral studies conducted in the field. The behavior fragments serve as the building blocks to compose grounded fine grained behavior models. The models can be used in simulations for studying the dynamics of any set of behavioral dimensions in some situation of interest. These guidelines are a result of our experience with creating a fine grained simulation model of a support services organization.

Towards Better Crisis Management in Support Services Organizations Using Fine Grained Agent Based Simulation

Critical support service operations have to run 24 × 7 and 365 days a year. Support operations therefore do contingency planning to continue operations during a crisis. In this paper we explore the use of fine-grained agent-based simulation models, which factor in human-behavioral dimensions such as stress, as a means to do better people planning for such situations. We believe the use of this approach may allow support operations managers to do more nuanced planning leading to higher resilience, and quicker return to normalcy. We model a prototypical support operation, which runs into different crisis severity levels, and show for each case, a reasonable size of the crisis team that would be required. We identify two contributions in this paper: First, emergency planning using agent based simulations have mostly focused, naturally, on societal communities such as urban populations. There has not been much attention paid to study crisis responses within support services organizations and our work is an attempt to address this deficit. Second, our use of grounded behavioral elements in our agent models allows us to build complex human behavior into the agents without sacrificing validity.

Can a Buffering Strategy Reduce Workload Related Stress

The support services industry remains a competitive business operating on very stringent budgets, metrics, and milestones. Given the heavy and varying workloads that are characteristic in this business, stress is a common all-too-common phenomenon and emerges from different sources. In this context, we use an agent based approach to examine whether using a workload buffering strategy based loosely on the Leaky Bucket Algorithm can help to manage work-load-related stress. We use our workplace stress model to see the implications of such a buffering approach on workload related stress at an individual level and on macro indicators such as turn-around time (TAT). The experiments show that the strategy can help not only reduce the stress but also provide knobs to the operations managers to manage workload and to ensure compliance. We conclude with implications for future research and practice.