-
Encoding Performance Data in MEI with the Automatic Music Performance Analysis and Comparison Toolkit (AMPACT)
Authors:
Johanna Devaney,
Cecilia Beauchamp
Abstract:
This paper presents a new method of encoding performance data in MEI using the recently added \texttt{<extData>} element. Performance data was extracted using the Automatic Music Performance Analysis and Comparison Toolkit (AMPACT) and encoded as a JSON object within an \texttt{<extData>} element linked to a specific musical note. A set of pop music vocals has was encoded to demonstrate both the r…
▽ More
This paper presents a new method of encoding performance data in MEI using the recently added \texttt{<extData>} element. Performance data was extracted using the Automatic Music Performance Analysis and Comparison Toolkit (AMPACT) and encoded as a JSON object within an \texttt{<extData>} element linked to a specific musical note. A set of pop music vocals has was encoded to demonstrate both the range of descriptors that can be encoded in <extData> and how AMPACT can be used for extracting performance data in the absence of a fully specified musical score.
△ Less
Submitted 19 November, 2023;
originally announced November 2023.
-
Rewarding Chatbots for Real-World Engagement with Millions of Users
Authors:
Robert Irvine,
Douglas Boubert,
Vyas Raina,
Adian Liusie,
Ziyi Zhu,
Vineet Mudupalli,
Aliaksei Korshuk,
Zongyi Liu,
Fritz Cremer,
Valentin Assassi,
Christie-Carol Beauchamp,
Xiaoding Lu,
Thomas Rialan,
William Beauchamp
Abstract:
The emergence of pretrained large language models has led to the deployment of a range of social chatbots for chitchat. Although these chatbots demonstrate language ability and fluency, they are not guaranteed to be engaging and can struggle to retain users. This work investigates the development of social chatbots that prioritize user engagement to enhance retention, specifically examining the us…
▽ More
The emergence of pretrained large language models has led to the deployment of a range of social chatbots for chitchat. Although these chatbots demonstrate language ability and fluency, they are not guaranteed to be engaging and can struggle to retain users. This work investigates the development of social chatbots that prioritize user engagement to enhance retention, specifically examining the use of human feedback to efficiently develop highly engaging chatbots. The proposed approach uses automatic pseudo-labels collected from user interactions to train a reward model that can be used to reject low-scoring sample responses generated by the chatbot model at inference time. Intuitive evaluation metrics, such as mean conversation length (MCL), are introduced as proxies to measure the level of engagement of deployed chatbots. A/B testing on groups of 10,000 new daily chatbot users on the Chai Research platform shows that this approach increases the MCL by up to 70%, which translates to a more than 30% increase in user retention for a GPT-J 6B model. Future work aims to use the reward model to realise a data fly-wheel, where the latest user conversations can be used to alternately fine-tune the language model and the reward model.
△ Less
Submitted 30 March, 2023; v1 submitted 10 March, 2023;
originally announced March 2023.
-
Demonstration of a laser powder bed fusion combinatorial sample for high-throughput microstructure and indentation characterization
Authors:
Jordan S. Weaver,
Adam L. Pintar,
Carlos Beauchamp,
Howie Joress,
Kil-Won Moon,
Thien Q. Phan
Abstract:
High-throughput experiments that use combinatorial samples with rapid measurements can be used to provide process-structure-property information at reduced time, cost, and effort. Develo** these tools and methods is essential in additive manufacturing where new process-structure-property information is required on a frequent basis as advances are made in feedstock materials, additive machines, a…
▽ More
High-throughput experiments that use combinatorial samples with rapid measurements can be used to provide process-structure-property information at reduced time, cost, and effort. Develo** these tools and methods is essential in additive manufacturing where new process-structure-property information is required on a frequent basis as advances are made in feedstock materials, additive machines, and post-processing. Here we demonstrate the design and use of combinatorial samples produced on a commercial laser powder bed fusion system to study 60 distinct process conditions of nickel superalloy 625: five laser powers and four laser scan speeds in three different conditions. Combinatorial samples were characterized using optical and electron microscopy, x-ray diffraction, and indentation to estimate the porosity, grain size, crystallographic texture, secondary phase precipitation, and hardness. Indentation and porosity results were compared against a regular sample. The smaller-sized regions (3 mm x 4 mm) in the combinatorial sample have a lower hardness compared to a larger regular sample (20 mm x 20 mm) with similar porosity (< 0.03 %). Despite this difference, meaningful trends were identified with the combinatorial sample for grain size, crystallographic texture, and porosity versus laser power and scan speed as well as trends with hardness versus stress-relief condition.
△ Less
Submitted 3 August, 2021; v1 submitted 2 March, 2021;
originally announced March 2021.
