Clinical trials are important tools to improve knowledge about the effectiveness of new treatments for all diseases, including cancers. However, studies show that fewer than 5% of cancer patients are enrolled in any type of research study or clinical trial. Although there is a wide variety of reasons for the low participation rate, we address this issue by designing a chatbot to help users determine their eligibility via interactive, two-way communication. The chatbot is supported by a user-centered classifier that uses an active deep learning approach to separate complex eligibility criteria into questions that can be easily answered by users and information that requires verification by their doctors. We collected all the available clinical trial eligibility criteria from the National Cancer Institute's website to evaluate the chatbot and the classifier. Experimental results show that the active deep learning classifier outperforms the baseline k-nearest neighbor method. In addition, an in-person experiment was conducted to evaluate the effectiveness of the chatbot. The results indicate that the participants who used the chatbot achieved better understanding about eligibility than those who used only the website. Furthermore, interfaces with chatbots were rated significantly better in terms of perceived usability, interactivity, and dialogue.

Publics' perceptions of new scientific advances such as AI are often informed and influenced by news coverage. To understand how artificial intelligence (AI) was framed in U.S. newspapers, a content analysis based on framing theory in journalism and science communication was conducted. This study identified the dominant topics and frames, as well as the risks and benefits of AI covered in five major American newspapers from 2009 to 2018. Results indicated that business and technology were the primary topics in news coverage of AI. The benefits of AI were discussed more frequently than its risks, but risks of AI were generally discussed with greater specificity. Additionally, episodic issue framing and societal impact framing were more frequently used.

We propose an end-to-end approach for modeling polyphonic music with a novel graphical representation, based on music theory, in a deep neural network. Despite the success of deep learning in various applications, it remains a challenge to incorporate existing domain knowledge in a network without affecting its training routines. In this paper we present a novel approach for predictive music modeling and music generation that incorporates domain knowledge in its representation. In this work, music is transformed into a 2D representation, inspired by tonnetz from music theory, which graphically encodes musical relationships between pitches. This representation is incorporated in a deep network structure consisting of multilayered convolutional neural networks (CNN, for learning an efficient abstract encoding of the representation) and recurrent neural networks with long short-term memory cells (LSTM, for capturing temporal dependencies in music sequences). We empirically evaluate the nature and the effectiveness of the network by using a dataset of classical music from various composers. We investigate the effect of parameters including the number of convolution feature maps, pooling strategies, and three configurations of the network: LSTM without CNN, LSTM with CNN (pre-trained vs. not pre-trained). Visualizations of the feature maps and filters in the CNN are explored, and a comparison is made between the proposed tonnetz-inspired representation and pianoroll, a commonly used representation of music in computational systems. Experimental results show that the tonnetz representation produces musical sequences that are more tonally stable and contain more repeated patterns than sequences generated by pianoroll-based models, a finding that is directly useful for tackling current challenges in music and AI such as smart music generation.

In this paper we propose an active deep learning approach to automatically classify eligibility criteria of clinical trials, an application that has not been explored in machine learning. We collected all clinical trial data from the National Cancer Institute website, and applied word2vec to learn word embeddings for eligibility criteria. Criteria encoded with word embeddings were then fed into a multi-layer convolution neural network (CNN) for classification. To overcome the challenge of non-existing class labels, we designed an active learning algorithm that uses uncertainty cluster sampling to navigate the dataset and strategically propagate obtained labels to expand the training set for CNN. Experimental results show that word2vec successfully learns meaningful embeddings in criteria data, and the active deep learning approach reports a significant lower error rate in classification than the baseline k-nearest neighbor method.

We propose an integrated system using active learning for audio-to-score alignment. Audio-to-score alignment is a fundamental task in music information retrieval. Although various machine learning techniques have been applied to this task, it is not the case for active learning. To show how beneficial active learning is in audio-to-score alignment, we demonstrate a system that integrates it with dynamic time warping, a commonly used algorithm for time series alignment. We propose a simple parametric model for selecting queries—a crucial step in active learning. We evaluate the system using synthesized audio as well as real performances. The alignment accuracy is improved with a range from 20% to 50% using only less than 10% query instances, a promising result that hopefully can inspire the creation of a collaborative framework between human and machine for audio-to-score alignment in the future.

Keys provide musical context and key modulation (changes) forms a crucial feature of music. In the age of big music data collections, automatic key segmentation is an important step towards music indexing and structure analysis. When using template-based key-finding methods, the best segmentation must minimize intra-segment distance to keys while maximizing inter-segment distance for neighboring keys. We present a general dynamic programming (DP) solution to this segmentation problem that is applicable to all distance-based key-finding methods and that does not require the number of segments to be pre-defined. This meta-algorithm is applied to the Kostka-Payne and Beatles datasets with three widely used distance-based key-finding methods. The key-finding results are evaluated using a compound score, and precision and recall. Statistical analysis of the results show that a precision value of 0.9 can be achieved with both datasets, for excerpts in one key, an average compound score above 0.8 is reported.

This paper presents a pilot study on an intelligent tutoring system for domain-independent argument making. Students' responses to an open-ended question were collected as the instances for supervised text classification based on the grade given by the instructor using structured outcome of the learning observation taxonomy. The responses were processed using Cohmetrix as well as n-gram models to generate attributes for the classification task. The best result of 81.74% in classification correct rate was obtained when all grade classes were used.

In this paper, we present an American Sign Language recognition system using a compact and affordable 3D motion sensor. The palm-sized Leap Motion sensor provides a much more portable and economical solution than Cyblerglove or Microsoft kinect used in existing studies. We apply k-nearest neighbor and support vector machine to classify the 26 letters of the English alphabet in American Sign Language using the derived features from the sensory data. The experiment result shows that the highest average classification rate of 72.78% and 79.83% was achieved by k-nearest neighbor and support vector machine respectively. We also provide detailed discussions on the parameter setting in machine learning methods and accuracy of specific alphabet letters in this paper.

This paper proposed and tested a model that provides quick search and retrieval of continuations for time series, particularly musical data, using relational databases. The model extends an existing interactive music-generation system by focusing on large input sequences. Experiments using textural and musical data provided satisfactory performance results for the model.

Data mining tasks such as music indexing, information retrieval, and similarity search, require an understanding of how listeners process music internally. Many algorithms for automatically analyzing the structure of recorded music assume that a large change in one or another musical feature suggests a section boundary. However, this assumption has not been tested: while our understanding of how listeners segment melodies has advanced greatly in the past decades, little is known about how this process works with more complex, full-textured pieces of music, or how stable this process is across genres. Knowing how these factors affect how boundaries are perceived will help researchers to judge the viability of algorithmic approaches with different corpora of music. We present a statistical analysis of a large corpus of recordings whose formal structure was annotated by expert listeners. We find that the acoustic properties of boundaries in these recordings corroborate findings of previous perceptual experiments. Nearly all boundaries correspond to peaks in novelty functions, which measure the rate of change of a musical feature at a particular time scale. Moreover, most of these boundaries match peaks in novelty for several features at several time scales. We observe that the boundary-novelty relationship can vary with listener, time scale, genre, and musical feature. Finally, we show that a boundary profile derived from a collection of novelty functions correlates with the estimated salience of boundaries indicated by listeners.

This paper presents a multimodal approach to style identification in pop/rock music. Considering the intuitive feelings of similarity from the listener's perspective, this study focuses on features that are computed using similarity metrics for melodies, harmonies, and audio signals for style identification. Support vector machine is used as a binary classifier to determine if two songs are created by the same artist given their similarity distances in the three aspects. Experiments are conducted using songs of four well-known pop/rock bands from 6 albums. The preliminary result shows that the approach achieves the best result in correct rate of 85% using only seven similarity metrics.

In this paper, we present statistical models to predict the difficulty of recognizing musical keys from polyphonic audio signals. Automatic audio key finding has been studied for many years, and various approaches have been proposed and reported. Reports of these methods' performance are usually based on the proposers' own data sets. Without details on the data set, i.e., how challenging the data set is, directly comparing the effectiveness of these methods is not meaningful or even possible. Thus, in this study we focus on predicting the difficulty level of key recognition as perceived by human experts. Given an audio recording, represented as the extracted acoustic features, we apply multiple linear regression and proportional odds model to predict the difficulty level of the recording, annotated by experts as an integer on a 5-point Likert scale. We use four metrics to evaluate our prediction results: root mean square error, Pearson correlation coefficient, exact accuracy, and adjacent accuracy. We also examine the difference between experts' annotations and discuss their consistency.

In this paper, we present an audio classification system using wavelets for extracting low-level acoustic features. We perform multiple-level decomposition using Discrete Wavelet Transform to extract acoustic features at different scales and time from audio recordings. The extracted features are then translated into a compact vector representation. Gaussian Mixture Models with Expectation Maximization algorithm are then used to build models for sound classes. Specifically, three types of audio classification tasks are designed to evaluate the system, including speech/music classification, male/female speech classification, and music genre (classical, pop, jazz, and electronic) classification. By evaluating the system through 5-fold cross validation, the experimental result shows the promising capability of wavelets for speech and music analyses.

N-gram models have been successfully applied to harmonic analysis for differentiating a composer's style based on all the pieces in a large corpus of the composer. In this paper, we focus on each individual song and explore the effectiveness of the n-gram model when it is applied to a different but equally important musical task: harmonic style-based song retrieval. A chord profile is generated for a song by using the n-gram model as the descriptor of the song's harmonic features. The system retrieves songs based on the similarity between the chord profile in the query and that of the songs in the database. The retrieval result is evaluated in terms of a style-based retrieval score as well as traditional information retrieval metrics such as precision and recall. Finally, we list the most common pop-rock chord patterns from the most frequently retrieved songs, and compare the patterns with those described in previous works.

I describe a method for analyzing pop-rock musical styles by extracting chord patterns based on their structural roles in relation to melody and lyrics. Previous work applied models such as n-grams, treating chords as equally chopped fragments without considering their functions with respect to musical structures like phrases. In this paper I focus on chord patterns at the end of segments as suggested in lyrics and melody. I study punctuation marks in lyrics and apply established melodic segmentation methods to retrieve segmental structure of songs, and extract two-chord patterns at the end of retrieved segments. I also examine the manner in which lyric punctuation coincides with melodic boundaries. In the experiment consisting of 101 pop-rock songs, it is observed that 68% of extracted patterns differed from those defined as standard cadences in Western classical music. Focusing on those non-standard patterns seems beneficial for differentiating composition styles among pop-rock bands.

In this paper, we explore the effect of musical context on audio onset detection using machine learning techniques. We extract the signal intensity and frequency energy of audio as the attributes of input instances for the machine learning techniques. The audio is synthesized from MIDI files, providing exact information of onset events. We test three state-of-the-art machine learning algorithms, support vector machines (SVM), neural networks (NN), and Naive Bayes (NB) with Ada boosting, for learning and classifying audio onsets. We found that SVMs perform best in general, based on the average of training and 10-fold cross validation errors as the evaluation criterion. We then test the SVM and NN, the two best performing methods, on Bach's Prelude in C major BWV 943, transposed to different keys and time-stretched to various tempi. The error rates ranged from 23.91% (when training set key and tempo equals those of the test set) to 37.22% (when the key is off by four accidentals) and 37.91% (when tempo is 20 beats per minute faster). The results show that audio onset detection performs significantly better when the key and tempo attributes of the test and training sets concur, than when they are different, thus supporting the utility of tempo and key knowledge in designing onset detection systems, or in prescribing confidence statistics to onset detection outcomes

Key finding is an integral step in content-based music indexing and retrieval. In this paper, we present an O(n) real-time algorithm for determining key from polyphonic audio. We use the standard Fast Fourier Transform with a local maximum detection scheme to extract pitches and pitch strengths from polyphonic audio. Next, we use Chew's Spiral Array Center of Effect Generator (CEG) algorithm to determine the key from pitch strength information. We test the proposed system using Mozart's Symphonies. The test data is audio generated from MIDI source. The algorithm achieves a maximum correct key recognition rate of 96% within the first fifteen seconds, and exceeds 90% within the first three seconds. Starting from the extracted pitch strength information, we compare the CEG algorithm's performance to the classic Krumhansl-Schmuckler (K-S) probe tone profile method and Temperley's modified version of the K-S method. Correct key recognition rates for the K-S and modified K-S methods remain under 50% in the first three seconds, with maximum values of 80% and 87% respectively within the first fifteen seconds for the same test set. The CEG method consistently scores higher throughout the fifteen-second selections.

Mobile ad hoc networks have gained more and more research attention. They provide wireless communications without location limitations and pre-built fixed infrastructures. Because of the absence of any static support structure, ad hoc networks are prone to link failure. This has become the most serious cause of throughput degradation when using TCP over ad hoc networks. Some researchers chose dynamic source routing (DSR) as the routing protocol and showed that disabling the assignment of a route directly from cache gives better performance. We introduce an efficient cache management mechanism to increase the TCP throughput by replying with a route directly from the cache of DSR and perform cache recovery when a host failure has occurred. We use simulations to compare the performance of our algorithm with the original DSR under the link failure prone environment due to mobility. We also provide the simulation results when host failures are considered in the ad hoc networks.