Prompt engineering is an increasingly important skill set needed to converse effectively with large language models (LLMs), such as ChatGPT. Prompts are instructions given LLM enforce rules, automate processes, and ensure specific qualities (and quantities) of generated output. also a form programming that can customize the outputs interactions LLM. This paper describes catalog prompt techniques presented in pattern have been applied solve common problems when conversing LLMs. patterns...

This paper presents prompt design techniques for software engineering, in the form of patterns, to solve common problems when using large language models (LLMs), such as ChatGPT automate engineering activities, ensuring code is decoupled from third-party libraries and simulating a web application API before it implemented. provides two contributions research on LLMs engineering. First, catalog patterns that classifies according types they solve. Second, explores several have been applied...

The rapid advent of Large Language Models (LLMs), such as ChatGPT and Claude, is revolutionizing various fields, from education healthcare to the engineering reliable software systems. These LLMs operate through "prompts," which are natural language inputs that users employ query leverage models' capabilities. Given novelty LLMs, understanding how effectively use prompts remains largely anecdotal, based on isolated cases. This fragmented approach limits reliability utility especially when...

Spoof speech can be used to try and fool speaker verification systems that determine the identity of based on voice characteristics. This paper compares popular learnable front-ends this task. We categorize by defining two generic architectures then analyze filtering stages both types in terms learning constraints. pro-pose replacing fixed filterbanks with a layer better adapt anti-spoofing tasks. The proposed FastAudio front-end is tested back-ends measure performance Logical Access track...

Data visualization has become a vital tool to help people understand the driving forces behind real-world phenomena. Although learning curve of tools have been reduced, domain experts still often require significant amounts training use them effectively. To reduce this even further, paper proposes Sketch2Vis, novel solution using deep techniques and generate source code for multi-platform data visualizations automatically from hand-drawn sketches provided by experts, which is similar how an...

Research in the past several years has boosted performance of automatic speaker verification systems and countermeasure to deliver low Equal Error Rates (EERs) on each system. However, research joint optimization both is still limited. The Spoofing-Aware Speaker Verification (SASV) 2022 challenge was proposed encourage development integrated SASV with new metrics evaluate model performance. This paper proposes an ensemble-free end-to-end solution, known as Spoof-Aggregated-SASV (SA-SASV)...

The NLC2CMD Competition hosted at NeurIPS 2020 aimed to bring the power of natural language processing command line. Participants were tasked with building models that can transform descriptions line tasks in English their Bash syntax. This is a report on competition details task, metrics, data, attempted solutions, and lessons learned.

Translating natural language into Bash Commands is an emerging research field that has gained attention in recent years. Most efforts have focused on producing more accurate translation models. To the best of our knowledge, only two datasets are available, with one based other. Both involve scraping through known data sources (through platforms like stack overflow, crowdsourcing, etc.) and hiring experts to validate correct either English text or Commands. This paper provides contributions...

An emerging trend in audio processing is capturing low-level speech representations from raw waveforms. These have shown promising results on a variety of tasks, such as recognition and separation. Compared to handcrafted features, learning features via backpropagation provides the model greater flexibility how it represents data for different tasks theoretically. However, empirical study shows that, some voice spoof detection, are more competitive than learned features. Instead evaluating...

This paper explores the translation of natural language into Bash Commands, which developers commonly use to accomplish command-line tasks in a terminal. In our approach terminal takes command as sentence plain English and translates it corresponding string Commands. The analyzes performance several architectures on this problem using data from NLC2CMD competition at NeurIPS 2020 conference. presented is best performing architecture date improves current state-of-the-art accuracy task 13.8% 53.2%.

Voice assistants, such as smart speakers, have exploded in popularity. It is currently estimated that the speaker adoption rate has exceeded 35% US adult population. Manufacturers integrated identification technology, which attempts to determine identity of person speaking, provide personalized services different members same family. Speaker can also play an important role controlling how used. For example, it not critical correctly identify user when playing music. However, reading user's...

Turbine blade is one of the most important structural components aeroengine, and it may be broken in process using. So meaningful to inspect blades accurately by optical method. In this paper, a laser scanning sensor designed for inspecting aeronautic casting turbine blade. First, some useful equations parameters performance are obtained through establishment line projection scheme. Second, which affect analyzed. Third, requirements that must obeyed given when designed. Furthermore,...

An emerging trend in audio processing is capturing low-level speech representations from raw waveforms. These have shown promising results on a variety of tasks, such as recognition and separation. Compared to handcrafted features, learning features via backpropagation can potentially provide the model greater flexibility how it represents data for different tasks. However, empirical studies show that, some spoof detection, still currently outperform learned features. Instead evaluating...

Translating natural language into Bash Commands is an emerging research field that has gained attention in recent years. Most efforts have focused on producing more accurate translation models. To the best of our knowledge, only two datasets are available, with one based other. Both involve scraping through known data sources (through platforms like stack overflow, crowdsourcing, etc.) and hiring experts to validate correct either English text or Commands. This paper provides contributions...

Meetings are an essential form of communication for all types organizations, and remote collaboration systems have been much more widely used since the COVID-19 pandemic. One major issue with meetings is that it challenging participants to interrupt speak. We recently developed first speech interruption analysis model WavLM_SI, which detects failed interruptions, shows very promising performance, being deployed in cloud. To deliver this feature a cost-efficient environment-friendly way, we...

GPUs have been favored for training deep learning models due to their highly parallelized architecture. As a result, most studies on optimization focus GPUs. There is often trade-off, however, between cost and efficiency when deciding how choose the proper hardware training. In particular, CPU servers can be beneficial if CPUs was more efficient, as they incur fewer update costs better utilizing existing infrastructure. This paper makes several contributions research using CPUs. First, it...

Research in the past several years has boosted performance of automatic speaker verification systems and countermeasure to deliver low Equal Error Rates (EERs) on each system. However, research joint optimization both is still limited. The Spoofing-Aware Speaker Verification (SASV) 2022 challenge was proposed encourage development integrated SASV with new metrics evaluate model performance. This paper proposes an ensemble-free end-to-end solution, known as Spoof-Aggregated-SASV (SA-SASV)...

Meetings are an essential form of communication for all types organizations, and remote collaboration systems have been much more widely used since the COVID-19 pandemic. One major issue with meetings is that it challenging participants to interrupt speak. We recently developed first speech interruption analysis model, which detects failed interruptions, shows very promising performance, being deployed in cloud. To deliver this feature a cost-efficient environment-friendly way, we reduced...

GPUs have been favored for training deep learning models due to their highly parallelized architecture. As a result, most studies on optimization focus GPUs. There is often trade-off, however, between cost and efficiency when deciding how choose the proper hardware training. In particular, CPU servers can be beneficial if CPUs was more efficient, as they incur fewer update costs better utilizing existing infrastructure. This paper makes several contributions research using CPUs. First, it...