Milos Gligoric (PhD candidate in Computer Science at the University of Illinois at Urbana-Champaign) will be presenting Regression Testing: Theory and Practice in the Computer Science Seminar series in CS150 at 4 p.m. Thursday, April 2. Everyone is welcome – the abstract is below.
Abstract: Developers often build regression test suites that are automatically run for each code revision to check that code changes did not break any functionality. While regression testing is important, it is also expensive due to both the number of revisions and the number of tests. For example, Google recently reported that they observed a quadratic increase in daily test-suite run time (a linear increase in the number of revisions per day and a linear increase in the number of tests per revision).
In this talk, I present a technique, called Ekstazi, to substantially reduce test-suite run time. Ekstazi introduces a novel approach to regression test selection, which runs only a subset of tests whose dependencies may be affected by the latest changes; Ekstazi keeps file dependencies for each test. Ekstazi also speeds up test-suite runs for software that uses modern distributed version-control systems; by modeling different branch and merge commands directly, Ekstazi computes test sets that can be significantly smaller than the entire test suite. I developed Ekstazi for JVM languages and evaluated it on several hundred revisions of 32 open-source projects (totaling 5M lines of code). Ekstazi can reduce test-suite run time an order of magnitude, including runs for merge revisions. Finally, only a few months after the initial release, Ekstazi was adopted and used daily by many developers from several open-source projects, including Apache Camel, Commons Math, and CXF.
Miriam Munoz (postbac student in Lisa Sander’s NeuroCognition and Perception Lab) and others (including students in Alexandra Jesse’s Language, Intersensory Perception, and Speech Lab, Rebecca Spencer’s Cognition and Action Lab, and David Moorman’s neuroscience lab; click here for a full list) will be presenting their research in UMass’s 2015 Annual Postbaccalaureate Research Education Program (PREP) Poster Session in the Marriott Center, 11th floor of the Campus Center from 5-7 p.m. Friday, April 3. Everyone is welcome.
Joshua Levy (graduate student in PBS) will be presenting Category Learning of Musical Chords in the Cognitive Brown Bag series in Tobin 521B at noon Wednesday, April 1. Everyone is welcome – the abstract is below.
Abstract: Musical chords can vary with respect to several music-theoretic attributes, such as root, quality, and inversion. In ear training classes, musicians are expected to learn to identify these attributes of a chord upon as little as a single presentation. We present data from a supervised category learning task (Shepard et al., 1961) examining the rate of learning of these attributes. In the task, subjects are asked to induce a rule that groups eight chords into two categories. While some subjects learn a simple music-theoretic rule (e.g. root position vs. first inversion), other subjects learn a complex music-atheoretic rule (e.g. major root position or minor first inversion vs. minor root position or major first inversion). The results show unequal learning rates of each attribute when learning music-theoretic rules. Morever, learning rates are in some cases even higher when learning music-atheoretic rules that partition the eight chords according to the average frequency of the three notes in a chord. Average frequency is a perceptually salient attribute that subjects rely upon when first learning to categorize chords.
Jason Weston of Facebook AI Research will be giving a talk titled “Memory Networks” at the Machine Learning and Friends Lunch this Thursday, 2 April at 12:00 pm in CS150. Everyone is welcome.
Larry Hyman will be giving a colloquium talk titled “Multiple argument marking in Bantoid” in the Linguistics department this Friday, 3 April at 3:30pm in ILC N400 (abstract follows). All are welcome.
“Multiple argument marking in Bantoid”
Given the typological similarities between the geographically and genetically distant Atlantic, Kordofanian and (Narrow) Bantu languages, it is generally assumed that early Niger-Congo had a synthetic structure with extensive noun class marking and derivational verb extensions (causative, applicative, etc.). However, other Niger-Congo languages show varying degrees of analyticity, some coming close to the endpoint of one morpheme per word. Much of the variation is quite clearly areal. In this paper I am concerned both with the mechanisms of change that lead from syntheticity to analyticity in the Nigeria-Cameroon borderland area, as well as the different strategies that are adopted as these languages lose their verbal extensions. The Bantoid languages I report on are typologically quite diverse. I am particularly concerned with what has replaced dative (recipient, benefactive) and instrumental applicative marking on verbs which historically allowed multiple object noun phrases (send-APPL chief letter, cook-APPL child rice, cut-APPL knife meat). The languages of this area show a remarkable variation: (i) Some languages allow multiple objects (typically with restrictions), while others do not (including some which have relic verb extensions); (ii) some have replaced verb extensions with serial verbs (take knife cut meat give child), others use adpositions (cut meat with knife for child). Still others have adopted multiple strategies for marking such arguments. In this study I sort out these strategies and attempt a micro-mapping of who has developed what where—and grammaticalizing from what (nouns? verbs?)? I will show that even adpositional languages have extensive verb serialization which they exploit for other functions (aspectual, directional, comparatives etc.), thereby raising the question of why only some Bantoid languages use serial verbs for argument marking. Although information is lacking for many languages, there does appear to be a southerly band of languages which mark datives and instruments with serial verbs. Information from Nigeria suggests that a similar distinction separates much of Lower Cross-River from Upper Cross River languages as well.
Francesca Rossi of the University of Padova, Italy will be giving a talk at the Machine Learning and Friends lunch titled “Probabilistic Conditional Preferences” (Joint work with C. Cornelio, U. Grandi, J. Goldsmith, N. Mattei, and K. B. Venable) this Thursday, 26 March at 12pm in CS150. The abstract follows.
“Probabilistic Conditional Preferences”
Preferences are ubiquitous in our everyday life. We use them to take all our decisions, either in isolation or together with others. They change over time and they get affected by our relationship with friends. We sometimes describe them explicitly but most often we show them implicitly via our actions (like clicks, follows, tweets, blogs, choices). Being able to model them faithfully and reason with them efficiently is essential in every intelligent environment.
Several formalisms exist to handle preferences of various kinds: qualitative, quantitative, conditional, etc. However, they usually assume preferences to be certain and do not provide help for describing and reasoning with uncertain preferences. Uncertainty may be present in a single agent setting, where an individual is not sure about its preferences or some noise is present, or also in a multi-agent setting, where several individuals may have conflicting preferences and uncertainty may be used to reconcile them. Probabilistic CP-nets (PCP-nets) are a formalism to model conditional qualitative preferences with probabilistic uncertainty. Under reasonable restrictions on the topology of their dependency graph, it is computationally easy to perform various tasks in PCP-nets. Thus they provide an efficient tool to model and reason with preferences. In this talk, I will describe PCP-nets and how to respond to optimality and dominance queries over them. I will then advocate for the need of a logical language to model them, and hint at their possible use in various scenarios. I will also compare them to other frameworks and will discuss preference elicitation/learning issues.
Tim Hunter of the University of Minnesota will be giving a talk titled “On Finding New Ways to Test Old Theories” this Thursday, 26 March at 9:30 am in ILC N400. The abstract follows.
“On Finding New Ways to Test Old Theories”
This talk will consist of two loosely-connected halves.
The point of departure for the first part is the observation of a surprising exception to the phenomenon known as “vehicle change”. These observations have direct consequences for our understanding of the relationship between ellipsis and movement. The generalization that emerges can also be used to construct novel tests that bear on the island-insensitivity of constructions like sluicing, and on whether the EPP is active inside ellipsis sites.
In the second part, I will present an approach to integrating minimalist syntax with information-theoretic sentence complexity metrics, from which it follows that two grammars that are extensionally equivalent — two grammars which produce the same structures, and differ only in what primitive derivational operations they use to build those structures — can nonetheless give rise to distinct predictions concerning sentence comprehension difficulty. This provides a linking hypothesis that connects sentence processing observations to subtle questions about the derivational operations that comprise human grammars (merge, move, re-merge, agree, etc.).
Jill Thorston of Northeastern University will be giving a talk titled “The development of intonation and information structure in early speech perception and production” this Tuesday, 24 March at 5:30 pm in Herter Hall room 217. The abstract follows.
“The development of intonation and information structure in early speech perception and production”
Infants are born with sensitivities to their native language s melody and rhythm. This attunement to prosody affects language development over the first years of life, and impacts early attentional processing, word learning, and speech production. The motivation for the first line of research is to investigate how American English-acquiring toddlers are guided by the mapping between intonation and information structure during on-line reference resolution and novel word learning. Specifically, I ask how specific pitch movements (deaccented/H*/L+H*) systematically predict patterns of attention and subsequent novel word learning abilities depending on the referring or learning condition (new/given/contrastive). Results show that the presence of either newness or a pitch accent facilitates attention, and that toddlers learn better from more prominent learning conditions. A second line of research examines the phonological and phonetic realizations of information categories as produced by toddler and adult speakers of English. During a spontaneous speech task designed as an interactive game, a set of target nouns are labeled and analyzed as new, given, or contrastive. Results reveal that toddlers reflect adult phonological patterns for new and contrastive information, as well as demonstrate a sophisticated usage of the acoustic correlates of intonation. Together, this set of studies demonstrates how higher-level components combine to direct attention to a referent in discourse and how this process helps explain mechanisms that are important for novel word learning and early speech production.
Vincent Homer and Rajesh Bhatt of the UMass Linguistics department will be giving a talk titled “Move Something” this Friday, 27 March at 3:30pm in ILC N400 (abstract below). All are welcome.
Typically, PPIs cannot be interpreted in the scope of a clausemate negation (barring shielding and rescuing). This means that when a given PPI is such that its scope is uniquely determined by its surface position, as is the case with e.g.would rather, the effect of putting it under a clause-mate negation is plain ungrammaticality. With indefinites, such as some, things are different: they can appear in that same configuration, provided that they are interpreted with wide scope over negation, which, in their case, is an available option.
In fact, indefinites are independently known to be able to take free wide scope: it is thus a priori possible that a mechanism whereby indefinite PPIs escape out of anti-licensing environments is the same that gives them wide scope out of syntactic islands, i.e. they can be interpreted by choice functions. In this talk, we address the question of the nature of the mechanism at play when, for polarity purposes, elements take wider scope than where they appear on the surface. We present arguments from Hindi-Urdu that, when a PPI surfaces in an anti-licensing environment, the wide scope mechanism that salvages it is movement (overt in Hindi-Urdu), not existential closure of a function-variable.
Gaja Jarosz of Yale University will be giving a job talk in the Linguistics department this Friday, 13 March at 3:30pm in ILC N400. Her talk is titled “Sonority Sequencing Effects in Polish: Defying the Stimulus?” (abstract below). All are welcome.
Sonority Sequencing Effects in Polish: Defying the Stimulus?
The Sonority Sequencing Principle (SSP: Steriade 1982; Selkirk 1984; Clements 1988, 1992) states that syllables with a sonority rise in the transition from the onset to the nucleus are preferred cross-linguistically. Experimental evidence indicates that English speakers exhibit gradient sensitivity to the SSP for onset clusters that are not attested in English (Davidson 2006, 2007; Berent et al. 2007, 2009; Daland et al. 2011). Berent et al. (2007, 2009) show that several lexical statistics of English fail to predict these preferences and suggest that the principle may therefore be innate. However, Daland et al. (2011) show that computational models with the ability to form abstract generalizations on the basis of phonological features and phonological context can detect SSP preferences on the basis of English lexical statistics. In this talk, I explore this controversy using computational and developmental approaches in a language (Polish) with very different sonority sequencing patterns from English. Using computational modeling, I show that a) the lexical statistics of Polish contradict the SSP, b) computational models applied to input estimated from Polish child-directed speech predict reverse-SSP preferences, and c) computational models that encode the SSP straightforwardly predict earlier acquisition of clusters with higher sonority rises. Thus, Polish provides a rare example where predictions of input-based models, even phonologically sophisticated ones, diverge dramatically from predictions expected on the basis of universal principles. I test these predictions by examining the acquisition of onset clusters in Polish. The data come from the spontaneous speech of four typically-developing, monolingual, Polish children aged 1;7-2;6 in the Weist-Jarosz Corpus (Weist and Witkowska-Stadnik 1986; Weist et al. 1984; Jarosz 2010; Jarosz et al. submitted). In conflict with the input-based predictions, the acquisition analyses indicate that development is significantly and gradiently sensitive to the SSP. I discuss the implications for phonological theory.