In lifelong learning, data are used to improve performance not only on the present task, but also on past and future (unencountered) tasks. While typical transfer learning algorithms can improve performance on future tasks, their performance on prior tasks degrades upon learning new tasks (called forgetting). Many recent approaches for continual or lifelong learning have attempted to maintain performance on old tasks given new tasks. But striving to avoid forgetting sets the goal unnecessarily low. The goal of lifelong learning should be to use data to improve performance on both future tasks (forward transfer) and past tasks (backward transfer). In this paper, we show that a simple approach-representation ensembling-demonstrates both forward and backward transfer in a variety of simulated and benchmark data scenarios, including tabular, vision (CIFAR-100, 5-dataset, Split Mini-Imagenet, Food1k, and CORe50), and speech (spoken digit), in contrast to various reference algorithms, which typically failed to transfer either forward or backward, or both. Moreover, our proposed approach can flexibly operate with or without a computational budget.

Learning to rank – producing a ranked list of items specific to a query and with respect to a set of supervisory items – is a problem of general interest. The setting we consider is one in which no analytic description of what constitutes a good ranking is available. Instead, we have a collection of representations and supervisory information consisting of a (target item, interesting items set) pair. We demonstrate analytically, in simulation, and in real data examples that learning to rank via combining representations using an integer linear program is effective when the supervision is as light as “these few items are similar to your item of interest.” While this nomination task is quite general, for specificity we present our methodology from the perspective of vertex nomination in graphs. The methodology described herein is model agnostic.

Abstract People are becoming increasingly sophisticated in their ability to navigate information spaces using search, hyperlinks, and visualization. But, mobile phones preclude the use of multiple coordinated views that have proven effective in the desktop environment (e.g., for business intelligence or visual analytics). In this work, we propose to model information as multivariate heterogeneous networks to enable greater analytic expression for a range of sensemaking tasks while suggesting a new, list‐based paradigm with gestural navigation of structured information spaces on mobile phones. We also present a mobile application, called Orchard, which combines ideas from both faceted search and interactive network exploration in a visual query language to allow users to collect facets of interest during exploratory navigation. Our study showed that users could collect and combine these facets with Orchard, specifying network queries and projections that would only have been possible previously using complex data tools or custom data science.

The application of modern NLP and ML techniques to large-scale datasets can generate implicit graphs that are so densely connected as to be unusable when rendered as node-link diagrams. We present a two-stage approach to extracting usable, map-like layouts from large, dense input graphs. This approach uses edge-cutting strategies based on node and edge metrics to reduce a graph to a skeletal structure showing only essential relationships, before filling in the resulting communities to create dense but usable layouts. Through a case study on a 145k-document adversarial health communication dataset, we show that each edge-cutting strategy has advantages and disadvantages, and that the appropriate choice of strategy depends on the data, user, and task.

Technical Support Scams delivered through malicious webpages and linked advertisements are an endemic problem on the web. To avoid detection, scammers systematically vary the URLs, web page designs, and phone numbers of scam delivery. We describe a web-scale pipeline powered by Cloud AI services that continuously detects and links evidence of such scams. By integrating this evidence in a graph structure and exposing it for forensic analysis in a user interface, we enable investigators and law enforcement partners to track the evolving scope and sophistication of scam operations. This approach automates and scales investigative tradecraft that contributed to major enforcement actions by the FTC in 2016. From 2016-2018, it helped reduce consumer exposure to such scams by 5 percent.

As Visual Analytics (VA) research grows and diversifies to encompass new systems, techniques, and use contexts, gaining a holistic view of analytic practices is becoming ever more challenging. However, such a view is essential for researchers and practitioners seeking to develop systems for broad audiences that span multiple domains. In this paper, we interpret VA research through the lens of Activity Theory (AT)-a framework for modelling human activities that has been influential in the field of Human-Computer Interaction. We first provide an overview of Activity Theory, showing its potential for thinking beyond tasks, representations, and interactions to the broader systems of activity in which interactive tools are embedded and used. Next, we describe how Activity Theory can be used as an organizing framework in the construction of activity typologies, building and expanding upon the tradition of abstract task taxonomies in the field of Information Visualization. We then apply the resulting process to create an activity typology for Visual Analytics, synthesizing a wide range of systems and activity concepts from the literature. Finally, we use this typology as the foundation of an activity-centered design process, highlighting both tensions and opportunities in the design space of VA systems.

DARPA has been investing in data science and building open source tools for applications ranging from counter threat finance, through radar operations and cancer research, to anti-human trafficking. This presentation will cover previous work at DARPA, experience building real-world applications for defense and law enforcement to analyze data, and the future of computer science as an enabler for content discovery, information extraction, relevance determination, and information visualization. The talk will be a mix of background, detailed examples, and software demonstration. It will cover the importance of anchoring in applications, minimization of design-to-testing time, development with users-in-the-loop, error tolerance of machine learning, design for diverse user populations, and the necessity of open source software integration. It will conclude by covering a few next directions for special projects at Microsoft.

This paper examines a query-by-example approach to spoken term detection in audio files. The approach is designed for low-resource situations in which limited or no in-domain training material is available and accurate word-based speech recognition capability is unavailable. Instead of using word or phone strings as search terms, the user presents the system with audio snippets of desired search terms to act as the queries. Query and test materials are represented using phonetic posteriorgrams obtained from a phonetic recognition system. Query matches in the test data are located using a modified dynamic time warping search between query templates and test utterances. Experiments using this approach are presented using data from the Fisher corpus.

The spoken term detection (STD) task aims to return relevant segments from a spoken archive that contain the query terms whether or not they are in the system vocabulary. This paper focuses on pronunciation modeling for out-of-vocabulary (OOV) terms which frequently occur in STD queries. The STD system described in this paper indexes word-level and sub-word level lattices or confusion networks produced by an LVCSR system using weighted finite state transducers (WFST).We investigate the inclusion of n-best pronunciation variants for OOV terms (obtained from letter-to-sound rules) into the search and present the results obtained by indexing confusion networks as well as lattices. The following observations are worth mentioning: phone indexes generated from sub-words represent OOVs well and too many variants for the OOV terms degrade performance if pronunciations are not weighted.

This paper addresses selecting between candidate pronunciations for out-of-vocabulary words in speech processing tasks. We introduce a simple, unsupervised method that outperforms the conventional supervised method of forced alignment with a reference. The success of this method is independently demonstrated using three metrics from large-scale speech tasks: word error rates for large vocabulary continuous speech recognition, decision error tradeoff curves for spoken term detection, and phone error rates compared to a handcrafted pronunciation lexicon. The experiments were conducted using state-of-the-art recognition, indexing, and retrieval systems. The results were compared across many terms, hundreds of hours of speech, and well known data sets.

In this paper we examine an alternative interface for phonetic search, namely query-by-example, that avoids OOV issues associated with both standard word-based and phonetic search methods. We develop three methods that compare query lattices derived from example audio against a standard ngrambased phonetic index and we analyze factors affecting the performance of these systems. We show that the best systems under this paradigm are able to achieve 77 % precision when retrieving utterances from conversational telephone speech and returning 10 results from a single query (performance that is better than a similar dictionary-based approach) suggesting significant utility for applications requiring high precision. We also show that these systems can be further improved using relevance feedback: By incorporating four additional queries the precision of the best system can be improved by 13.7 % relative. Our systems perform well despite high phone recognition error rates (> 40%) and make use of no pronunciation or letter-to-sound resources.

This paper addresses the adjustment of the language model (LM) scaling factor of an automatic speech recognition (ASR) system for a new domain using only un-transcribed speech. The main idea is to replace the (unavailable) reference transcript with an automatic transcript generated by an independent ASR system, and adjust parameters using this sloppy reference. It is shown that despite its fairly high error rate (ca. 35%), choosing the scaling factor to minimize disagreement with the erroneous transcripts is still an effective recipe for model selection. This effectiveness is demonstrated by adjusting an ASR system trained on Broadcast News to transcribe the MIT Lectures corpus. An ASR system for telephone speech produces the sloppy reference, and optimizing towards it yields a nearly optimal LM scaling factor for the MIT Lectures corpus. Index Terms: LVCSR, language modeling, domain adaptation, semi-supervised learning, unsupervised learning

Indexing and retrieval of speech content in various forms such as broadcast news, customer care data and on-line media has gained a lot of interest for a wide range of applications, from customer analytics to on-line media search. For most retrieval applications, the speech content is typically first converted to a lexical or phonetic representation using automatic speech recognition (ASR). The first step in searching through indexes built on these representations is the generation of pronunciations for named entities and foreign language query terms. This paper summarizes the results of the work conducted during the 2008 JHU Summer Workshop by the Multilingual Spoken Term Detection team, on mining the web for pronunciations and analyzing their impact on spoken term detection. We will first present methods to use the vast amount of pronunciation information available on the Web, in the form of IPA and ad-hoc transcriptions. We describe techniques for extracting candidate pronunciations from Web pages and associating them with orthographic words, filtering out poorly extracted pronunciations, normalizing IPA pronunciations to better conform to a common transcription standard, and generating phonemic representations from ad-hoc transcriptions. We then present an analysis of the effectiveness of using these pronunciations to represent Out-Of-Vocabulary (OOV) query terms on the performance of a spoken term detection (STD) system. We will provide comparisons of Web pronunciations against automated techniques for pronunciation generation as well as pronunciations generated by human experts. Our results cover a range of speech indexes based on lattices, confusion networks and one-best transcriptions at both word and word fragments levels.

This paper addresses the detection of OOV segments in the output of a large vocabulary continuous speech recognition (LVCSR) system. First, standard confidence measures from frame-based wordand phone- posteriors are investigated. Substantial improvement is obtained when posteriors from two systems — strongly constrained (LVCSR) and weakly constrained (phone posterior estimator) are combined. We show that this approach is also suitable for detection of general recognition errors. All results are presented on WSJ task with reduced recognition vocabulary.

Current approaches to automatic spoken language identification (LID) assume the availability of a large corpus of manually language-labeled speech samples for training statistical classifiers. We investigate two methods of active learning to significantly reduce the amount of labeled speech needed for training LID systems. Starting with a small training set, an automated method is used to select samples from a corpus of unlabeled speech, which are then labeled and added to the training pool - one selection method is based on a previously known entropy criterion, and another on a novel likelihood-ratio criterion. We demonstrate LID performance comparable to a large training corpus using only a tenth of the training data. A further 40% improvement in LID performance is obtained using a third of the training data. Finally, we show that our novel selection method is more robust to variance in the unlabeled pool than the entropy based method.

Automatic speech recognition (ASR) systems continue to make errors during search when handling various phenomena including noise, pronunciation variation, and out of vocabulary (OOV) words. Predicting the probability that a word is incorrect can prevent the error from propagating and perhaps allow the system to recover. This paper addresses the problem of detecting errors and OOVs for read Wall Street Journal speech when the word error rate (WER) is very low. It augments a traditional confidence estimate by introducing two novel methods: phone-level comparison using multi-string alignment (MSA) and word-level comparison using phone-to-word transduction. We show that features from phone and word string comparisons can be added to a standard maximum entropy framework thereby substantially improving performance in detecting both errors and OOVs. Additionally we show an extension to detecting English and accented English for the language identification (LID) task.

Multilingual speech processing continues to develop as speech technology spreads to heterogeneous clients and applications. We address a distinct problem of code-switching — the spontaneous but occasional use, within speech in one language (referred to as L1), of words, phrases, expressions or idioms from a second language (L2). We examine two alternatives for modeling the acoustics of such words: creation of L1 pronunciations for the out-of-language (OOL) words for use with L1 acoustic models, and retention of their L2 pronunciations for use with multilingual acoustic models. We test the hypothesis that the latter is a better acoustic model for OOL words. We develop a set of lexica in IPA form, a global phoneme inventory, and handle the problem of L2 word pronunciation by creating linguistically motivated pairwise mappings. We show that retention of L2 pronunciations with multilingual acoustic models better explains the observations when restricted to a forced alignment. Index Terms: Cross-lingual and multi-lingual processing, Automatic speech recognition, Accent and language identification, Spoken language resources and annotation 1.

This work details the development of a hybrid evolutionary algorithm for solving the traveling salesman problem (TSP). The strategy of the algorithm is to complement and extend the successful results of a genetic algorithm (GA) using a distance preserving crossover (DPX) by incorporating memory in the form of ant pheromone during the city selection process. The synergistic combination of the DPX-GA with city selection based on probability determined by both distance and previous success incorporates additional information into the search mechanism. This combination into a hybrid GA facilitates finding quality solutions for TSP problems with lower computation complexity. This study represents a preliminary investigation with direct comparison to show the feasibility and promise of this hybrid approach.

We describe a prototype spoken language system that loosely integrates a speech recognition component based on hidden Markov models with a constraint dependency grammar (CDG) parser using a word graph to pass sentence candidates between the two modules. This loosely coupled system was able to improve the sentence selection accuracy and concept accuracy over the level achieved by the acoustic module with a stochastic grammar. Timing profiles suggest that a tighter coupling of the modules could reduce parsing times of the system, as could the development of better acoustic models and tighter parsing constraints for conjunctions.