William M. Campbell
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
William M. Campbell has a total of 28 co-authors. Here are the top co-authors by number of shared publications:
- Douglas A. Reynolds
MIT USA
7 shared publications - Lin Li
Massachusetts Institute of Technology Lincoln Laboratory, Lexington, USA
4 shared publications - Wade Shen
MIT, Lincoln Lab, Informat Syst Technol Grp, Lexington, MA 02420 USA
3 shared publications - Elliot Singer
Massachusetts Institute of Technology
2 shared publications A. Solomonoff
MIT Lincoln Lab., Lexington, MA
1 shared publication- C. Dagli
Department of Applied Math and Statistics Human Language Technology Group MIT Lincoln Laboratory Johns Hopkins University Baltimore MD Lexington MA
1 shared publication
Latest Publications
- Open Access
Graph Model Selection via Random Walks
In this paper, we present a novel approach based on the random walk process for finding meaningful representations of a graph model. Our approach leverages the transient behavior of many short random walks with novel initialization mechanisms to generate model discriminative features. These features are able to capture a more comprehensive structural signature of the underlying graph model. The resulting representation is invariant to both node permutation and the size of the graph, allowing direct comparison between large classes of graphs. We test our approach on two challenging model selection problems: the discrimination in the sparse regime of an Erd\"{o}s-Renyi model from a stochastic block model and the planted clique problem. Our representation approach achieves performance that closely matches known theoretical limits in addition to being computationally simple and scalable to large graphs.
Open AccessGraph Model Selection via Random Walks
In this paper, we present a novel approach based on the random walk process for finding meaningful representations of a graph model. Our approach leverages the transient behavior of many short random walks with novel initialization mechanisms to generate model discriminative features. These features are able to capture a more comprehensive structural signature of the underlying graph model. The resulting representation is invariant to both node permutation and the size of the graph, allowing direct comparison between large classes of graphs. We test our approach on two challenging model selection problems: the discrimination in the sparse regime of an Erd\"{o}s-Renyi model from a stochastic block model and the planted clique problem. Our representation approach achieves performance that closely matches known theoretical limits in addition to being computationally simple and scalable to large graphs.
- Open Access
Making Sense of Unstructured Text Data
Many network analysis tasks in social sciences rely on pre-existing data sources that were created with explicit relations or interactions between entities under consideration. Examples include email logs, friends and followers networks on social media, communication networks, etc. In these data, it is relatively easy to identify who is connected to whom and how they are connected. However, most of the data that we encounter on a daily basis are unstructured free-text data, e.g., forums, online marketplaces, etc. It is considerably more difficult to extract network data from unstructured text. In this work, we present an end-to-end system for analyzing unstructured text data and transforming the data into structured graphs that are directly applicable to a downstream application. Specifically, we look at social media data and attempt to predict the most indicative words from users' posts. The resulting keywords can be used to construct a context+content network for downstream processing such as graph-based analysis and learning. With that goal in mind, we apply our methods to the application of cross-domain entity resolution. The performance of the resulting system with automatic keywords shows improvement over the system with user-annotated hashtags.
Open AccessMaking Sense of Unstructured Text Data
Many network analysis tasks in social sciences rely on pre-existing data sources that were created with explicit relations or interactions between entities under consideration. Examples include email logs, friends and followers networks on social media, communication networks, etc. In these data, it is relatively easy to identify who is connected to whom and how they are connected. However, most of the data that we encounter on a daily basis are unstructured free-text data, e.g., forums, online marketplaces, etc. It is considerably more difficult to extract network data from unstructured text. In this work, we present an end-to-end system for analyzing unstructured text data and transforming the data into structured graphs that are directly applicable to a downstream application. Specifically, we look at social media data and attempt to predict the most indicative words from users' posts. The resulting keywords can be used to construct a context+content network for downstream processing such as graph-based analysis and learning. With that goal in mind, we apply our methods to the application of cross-domain entity resolution. The performance of the resulting system with automatic keywords shows improvement over the system with user-annotated hashtags.
- Open Access
Cross-Domain Entity Resolution in Social Media
The challenge of associating entities across multiple domains is a key problem in social media understanding. Successful cross-domain entity resolution provides integration of information from multiple sites to create a complete picture of user and community activities, characteristics, and trends. In this work, we examine the problem of entity resolution across Twitter and Instagram using general techniques. Our methods fall into three categories: profile, content, and graph based. For the profile-based methods, we consider techniques based on approximate string matching. For content-based methods, we perform author identification. Finally, for graph-based methods, we apply novel cross-domain community detection methods and generate neighborhood-based features. The three categories of methods are applied to a large graph of users in Twitter and Instagram to understand challenges, determine performance, and understand fusion of multiple methods. Final results demonstrate an equal error rate less than 1%.
Open AccessCross-Domain Entity Resolution in Social Media
The challenge of associating entities across multiple domains is a key problem in social media understanding. Successful cross-domain entity resolution provides integration of information from multiple sites to create a complete picture of user and community activities, characteristics, and trends. In this work, we examine the problem of entity resolution across Twitter and Instagram using general techniques. Our methods fall into three categories: profile, content, and graph based. For the profile-based methods, we consider techniques based on approximate string matching. For content-based methods, we perform author identification. Finally, for graph-based methods, we apply novel cross-domain community detection methods and generate neighborhood-based features. The three categories of methods are applied to a large graph of users in Twitter and Instagram to understand challenges, determine performance, and understand fusion of multiple methods. Final results demonstrate an equal error rate less than 1%.
- Open Access
Matching Community Structure Across Online Social Networks
The discovery of community structure in networks is a problem of considerable interest in recent years. In online social networks, often times, users are simultaneously involved in multiple social media sites, some of which share common social relationships. It is of great interest to uncover a shared community structure across these networks. However, in reality, users typically identify themselves with different usernames across social media sites. This creates a great difficulty in detecting the community structure. In this paper, we explore several approaches for community detection across online social networks with limited knowledge of username alignment across the networks. We refer to the known alignment of usernames as seeds. We investigate strategies for seed selection and its impact on networks with a different fraction of overlapping vertices. The goal is to study the interplay between network topologies and seed selection strategies, and to understand how it affects the detected community structure. We also propose several measures to assess the performance of community detection and use them to measure the quality of the detected communities in both Twitter-Twitter networks and Twitter-Instagram networks.
Open AccessMatching Community Structure Across Online Social Networks
The discovery of community structure in networks is a problem of considerable interest in recent years. In online social networks, often times, users are simultaneously involved in multiple social media sites, some of which share common social relationships. It is of great interest to uncover a shared community structure across these networks. However, in reality, users typically identify themselves with different usernames across social media sites. This creates a great difficulty in detecting the community structure. In this paper, we explore several approaches for community detection across online social networks with limited knowledge of username alignment across the networks. We refer to the known alignment of usernames as seeds. We investigate strategies for seed selection and its impact on networks with a different fraction of overlapping vertices. The goal is to study the interplay between network topologies and seed selection strategies, and to understand how it affects the detected community structure. We also propose several measures to assess the performance of community detection and use them to measure the quality of the detected communities in both Twitter-Twitter networks and Twitter-Instagram networks.
- 1 January 2008
- Published by Springer Nature
- p. 811-824
Automatic Language Recognition Via Spectral and Token Based Approaches
- 1 January 2008
- Published by Springer Nature
- p. 811-824
Automatic Language Recognition Via Spectral and Token Based Approaches
- 1 April 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, IV-989-IV-992
Language Recognition with Word Lattices and Support Vector Machines
Language recognition is typically performed with methods that exploit phonotactics - a phone recognition language modeling (PRLM) system. A PRLM system converts speech to a lattice of phones and then scores a language model. A standard extension to this scheme is to use multiple parallel phone recognizers (PPRLM). In this paper, we modify this approach in two distinct ways. First, we replace the phone tokenizer by a powerful speech-to-text system. Second, we use a discriminative support vector machine for language modeling. Our goals are twofold. First, we explore the ability of a single speech-to-text system to distinguish multiple languages. Second, we fuse the new system with an SVM PRLM system to see if it complements current approaches. Experiments on the 2005 NIST language recognition corpus show the new word system accomplishes these goals and has significant potential for language recognition.
- 1 April 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, IV-989-IV-992
Language Recognition with Word Lattices and Support Vector Machines
Language recognition is typically performed with methods that exploit phonotactics - a phone recognition language modeling (PRLM) system. A PRLM system converts speech to a lattice of phones and then scores a language model. A standard extension to this scheme is to use multiple parallel phone recognizers (PPRLM). In this paper, we modify this approach in two distinct ways. First, we replace the phone tokenizer by a powerful speech-to-text system. Second, we use a discriminative support vector machine for language modeling. Our goals are twofold. First, we explore the ability of a single speech-to-text system to distinguish multiple languages. Second, we fuse the new system with an SVM PRLM system to see if it complements current approaches. Experiments on the 2005 NIST language recognition corpus show the new word system accomplishes these goals and has significant potential for language recognition.
- 1 January 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, IV-49-IV-52
Robust Speaker Recognition with Cross-Channel Data: MIT-LL Results on the 2006 NIST SRE Auxiliary Microphone Task
One particularly difficult challenge for cross-channel speaker verification is the auxiliary microphone task introduced in the 2005 and 2006 NIST Speaker Recognition Evaluations, where training uses telephone speech and verification uses speech from multiple auxiliary microphones. This paper presents two approaches to compensate for the effects of auxiliary microphones on the speech signal. The first compensation method mitigates session effects through Latent Factor Analysis (LFA) and Nuisance Attribute Projection (NAP). The second approach operates directly on the recorded signal with noise reduction techniques. Results are presented that show a reduction in the performance gap between telephone and auxiliary microphone data.
- 1 January 2007
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 4, IV-49-IV-52
Robust Speaker Recognition with Cross-Channel Data: MIT-LL Results on the 2006 NIST SRE Auxiliary Microphone Task
One particularly difficult challenge for cross-channel speaker verification is the auxiliary microphone task introduced in the 2005 and 2006 NIST Speaker Recognition Evaluations, where training uses telephone speech and verification uses speech from multiple auxiliary microphones. This paper presents two approaches to compensate for the effects of auxiliary microphones on the speech signal. The first compensation method mitigates session effects through Latent Factor Analysis (LFA) and Nuisance Attribute Projection (NAP). The second approach operates directly on the recorded signal with noise reduction techniques. Results are presented that show a reduction in the performance gap between telephone and auxiliary microphone data.
- Top Cited Papers
- 2 August 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1 (15206149) , 97
SVM Based Speaker Verification using a GMM Supervector Kernel and NAP Variability Compensation
Gaussian mixture models with universal backgrounds (UBMs) have become the standard method for speaker recognition. Typically, a speaker model is constructed by MAP adaptation of the means of the UBM. A GMM supervector is constructed by stacking the means of the adapted mixture components. A recent discovery is that latent factor analysis of this GMM supervector is an effective method for variability compensation. We consider this GMM supervector in the context of support vector machines. We construct a support vector machine kernel using the GMM supervector. We show similarities based on this kernel between the method of SVM nuisance attribute projection (NAP) and the recent results in latent factor analysis. Experiments on a NIST SRE 2005 corpus demonstrate the effectiveness of the new technique
- 2 August 2006
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- Vol. 1 (15206149) , 97
Top Cited PapersSVM Based Speaker Verification using a GMM Supervector Kernel and NAP Variability Compensation
Gaussian mixture models with universal backgrounds (UBMs) have become the standard method for speaker recognition. Typically, a speaker model is constructed by MAP adaptation of the means of the UBM. A GMM supervector is constructed by stacking the means of the adapted mixture components. A recent discovery is that latent factor analysis of this GMM supervector is an effective method for variability compensation. We consider this GMM supervector in the context of support vector machines. We construct a support vector machine kernel using the GMM supervector. We show similarities based on this kernel between the method of SVM nuisance attribute projection (NAP) and the recent results in latent factor analysis. Experiments on a NIST SRE 2005 corpus demonstrate the effectiveness of the new technique