Verint Da Vinci Research

An Open Intent Discovery Evaluation Framework

Discovering target intents is crucial but often overlooked in dialog systems. Creating labelled datasets is complex and manual. Open Intent Discovery automates grouping utterances and identifying intents. This framework offers various techniques for each discovery step, including human-readable label generation. It also analyzes dataset features to recommend optimal technique combinations, aiding users without exhaustive exploration.

Authors: Ian Beaver, Grant Anderson

Automated Human-Readable Label Generation in Open Intent Discovery

Determining user intent in dialog systems requires large, labelled datasets, which are complex to create. Many works focus on discovering intent clusters but not on generating human-readable labels. This study introduces a new candidate label extraction method, evaluating six combinations of extraction and selection methods on three datasets. Results show detailed labels can be generated from unlabelled data without costly pre-trained models.

Authors: Ian Beaver, Grant Anderson

FCBench: Cross-Domain Benchmarking of Lossless Compression for Floating-Point Data

The database and HPC communities use lossless compression for floating-point data but lack cross-domain evaluation. With HPC’s shift to in-situ analysis, more data is stored in databases, highlighting the need for a unified study. This research evaluates 13 compression methods on 33 datasets, using the roofline model to profile runtime bottlenecks, aiming to guide method selection and development.

Authors: Ian Beaver, Cynthia Freeman

Responsible and Ethical AI

Towards Awareness of Human Relational Strategies in Virtual Agents

This paper investigates how humans apply relational strategies to virtual agents and chatbots compared to human agents in a customer service environment.

Authors: Ian Beaver, Cynthia Freeman, Abdullah Mueen

An Adaptive Deep Clustering Pipeline to Inform Text Labeling at Scale

Learn how Verint researchers created a flexible and scalable clustering pipeline that integrates the fine-tuning of language models, a high performing k-NN library, and community detection techniques to help analysts quickly surface and organize relevant user intentions from conversational texts.

Authors: Xinyu Chen, Ian Beaver

Is AI at Human Parity Yet? A Case Study on Speech Recognition

What does it even mean for an AI system to reach human parity? How is progress towards that goal being measured? This article focuses on the current state of speech recognition and the recent developments in benchmarking and measuring performance of AI models built for speech processing.

Author: Ian Beaver

The Success of Conversational AI and the AI Challenge it Reveals

A massive increase in conversational artificial intelligence research and advancement in recent years has enabled systems to produce rich and varied turns in human-like conversations that often rely on crowd worker opinions as the primary measurement of success. The challenge, though, is that evaluation strategies need to mature alongside AI systems that are mature in more “human” tasks that involve creativity and variation,

Author: Ian Beaver

Timevae: A Variational Auto-Encoder for Multivariate Time Series Generation

Recent work in synthetic data generation in the time-series domain has focused on the use of Generative Adversarial Networks. In this paper, we propose a novel architecture for synthetically generating time-series data with the use of Variational Auto-Encoders (VAEs).

Authors: Abhyuday Desai, Cynthia Freeman, Zuhui Wang, Ian Beaver

Automated Conversation Review to Surface Virtual Assistant Misunderstandings: Reducing Cost and Increasing Privacy

With the rise of Intelligent Virtual Assistants (IVAs), there is a necessary rise in human effort to identify conversations containing misunderstood user inputs. Here, we present a scalable system for automated conversation review that can identify potential miscommunications and provide IVA designers with suggested actions to fix errors in IVA understanding, prioritizes areas of language model repair, and automates the review of conversations where desired.

Authors: Ian Beaver, Abdullah Mueen

Experimental Comparison and Survey of 12-Time Series Anomaly Detection Algorithms

There’s a plethora of AI anomaly detection methods across domains, which can be a burden for AI developers. To reduce this evaluation burden, we present guidelines to intelligently choose the optimal anomaly detection methods based on the characteristics the time series displays such as seasonality, trend, level change concept drift, and missing time steps.

Authors: Cynthia Freeman, Jonathan Merriman, Ian Beaver, Abdullah Mueen

Fine-Tuning Language Models for Semi-Supervised Text Mining

In this paper, we present an empirical study of a pipeline for semi-supervised text clustering tasks. Our proposed method utilizes a small number of labeled samples to fine-tune pre-trained language models. This fine-tuning step adapts the language models to produce task-specific contextualized representations, improving the performance of downstream text clustering tasks.

Authors: Xinyu Chen, Ian Beaver, Cynthia Freeman