Abstract: This book presents the theory of output-driven maps and provides a fresh perspective on the extent to which phonologies can be characterized in terms of restrictions on outputs. Closely related to traditional conceptions of process opacity, but differing in notable ways, the theory of output-driven maps applies equally to SPE-style ordered rules, Optimality Theory, and other phonological theories. It permits a formally rigorous analysis of the issues in Optimality Theory that is not possible with traditional process opacity. Also presented is a theory of phonological learning. Building on prior work on learning in Optimality Theory, the learning theory exploits the formal structure of output-driven maps to achieve learning that is far more computationally efficient than comparable prior approaches. In this book Bruce Tesar, one of the founders of the study of learnability in Optimality Theory, presents fresh perspectives in an accessible way for graduate students and academic researchers.

Abstract: This paper explores the relative merits of constraint ranking vs. weighting in the context of a major outstanding learnability problem in phonology: learning in the face of hidden structure. Specifically, the paper examines a well-known approach to the structural ambiguity problem, Robust Interpretive Parsing (RIP; Tesar & Smolensky 1998), focusing on its stochastic extension first described by Boersma (2003). Two related problems with the stochastic formulation of RIP are revealed, rooted in a failure to take full advantage of probabilistic information available in the learner's grammar. To address these problems, two novel parsing strategies are introduced and applied to learning algorithms for both probabilistic ranking and weighting. The novel parsing strategies yield significant improvements in performance, asymmetrically improving performance of OT learners. Once RIP is replaced with the proposed modifications, the apparent advantage of HG over OT learners reported in previous work disappears (Boersma & Pater 2008).

Abstract: This study examines the intergenerational transfer of human communication systems. It tests if human communication systems evolve to be easy to learn or easy to use (or both), and how population size affects learnability and usability. Using an experimental-semiotic task, we find that human communication systems evolve to be easier to use (production efficiency and reproduction fidelity), but harder to learn (identification accuracy) for a second generation of naive participants. Thus, usability trumps learnability. In addition, the communication systems that evolve in larger populations exhibit distinct advantages over those that evolve in smaller populations: the learnability loss (from the Initial signs) is more muted and the usability benefits are more pronounced. The usability benefits for human communication systems that evolve in a small and large population is explained through guided variation reducing sign complexity. The enhanced performance of the communication systems that evolve in larger populations is explained by the operation of a content bias acting on the larger pool of competing signs. The content bias selects for information-efficient iconic signs that aid learnability and enhance usability.

Abstract: Various authors have recently endorsed Harmonic Grammar (HG) as a replacement for Optimality Theory (OT). One argument for this move is that OT seems not to have close correspondents within machine learning while HG allows methods and results from machine learning to be imported into computational phonology. Here, I prove that this argument in favor of HG and against OT is wrong. In fact, I show that any algorithm for HG can be turned into an algorithm for OT. Hence, HG has no computational advantages over OT. This result allows tools from machine learning to be systematically adapted to OT. As an illustration of this new toolkit for computational OT, I prove convergence for a slight variant of Boersma’s (1998) (nonstochastic) Gradual Learning Algorithm.

Abstract: A one-dimensional input system is provided. The input system comprises a plurality of characters arranged on a single dimension. User input along the dimension is continuously disambiguated. The characters may be arranged based on factors including motor efficiency, optimization of disambiguation and learnability. A touchscreen interface of the one-dimensional input system is provided. A gesture-based interface of the one-dimensional input system is also provided.

Abstract: A striking cross-linguistic generalisation about the semantics of determiners is that they never express non-conservative relations. To account for this one might hypothesise that the mechanisms underlying human language acquisition are unsuited to non-conservative determiner meanings. We present experimental evidence that 4- and 5-year-olds fail to learn a novel non-conservative determiner but succeed in learning a comparable conservative determiner, consistent with the learnability hypothesis.

Abstract: The present work determines the exact nature of {\em linear time computable} notions which characterise automatic functions (those whose graphs are recognised by a finite automaton). The paper also determines which type of linear time notions permit full learnability for learning in the limit of automatic classes (families of languages which are uniformly recognised by a finite automaton). In particular it is shown that a function is automatic iff there is a one-tape Turing machine with a left end which computes the function in linear time where the input before the computation and the output after the computation both start at the left end. It is known that learners realised as automatic update functions are restrictive for learning. In the present work it is shown that one can overcome the problem by providing work tapes additional to a resource-bounded base tape while keeping the update-time to be linear in the length of the largest datum seen so far. In this model, one additional such work tape provides additional learning power over the automatic learner model and two additional work tapes give full learning power. Furthermore, one can also consider additional queues or additional stacks in place of additional work tapes and for these devices, one queue or two stacks are sufficient for full learning power while one stack is insufficient.

Abstract: Mobile text entry methods traditionally have been designed with the assumption that users can devote full visual and mental attention on the device, though this is not always possible. The authors present their iterative design and evaluation of Escape-Keyboard, a sight-free text entry method for mobile touch-screen devices. Escape-Keyboard allows the user to type letters with one hand by pressing the thumb on different areas of the screen and performing a flick gesture. The authors then examine the performance of Escape-Keyboard in a study that included 16 sessions in which participants typed in sighted and sight-free conditions. Qualitative results from this study highlight the importance of reducing the mental load with using Escape-Keyboard to improve user performance over time. The authors thus also explore features to mitigate this learnability issue. Finally, the authors investigate the upper bound on the sight-free performance with Escape-Keyboard by performing theoretical analysis of the expert peak performance.

Abstract: Children learn their native language by exposure to their linguistic and communicative environment, but apparently without requiring that their mistakes be corrected. Such learning from "positive evidence" has been viewed as raising "logical" problems for language acquisition. In particular, without correction, how is the child to recover from conjecturing an over-general grammar, which will be consistent with any sentence that the child hears? There have been many proposals concerning how this "logical problem" can be dissolved. In this study, we review recent formal results showing that the learner has sufficient data to learn successfully from positive evidence, if it favors the simplest encoding of the linguistic input. Results include the learnability of linguistic prediction, grammaticality judgments, language production, and form-meaning mappings. The simplicity approach can also be "scaled down" to analyze the learnability of specific linguistic constructions, and it is amenable to empirical testing as a framework for describing human language acquisition.

Abstract: One universal feature of human languages is the division between grammatical functors and content words. From a learnability point of view, functors might provide entry points or anchors into the syntactic structure of utterances due to their high frequency. Despite its potentially universal scope, this hypothesis has not yet been tested on typologically different languages and on populations of different ages. Here we report a corpus study and an artificial grammar learning experiment testing the anchoring hypothesis in Basque, Japanese, French and Italian adults. We show that adults are sensitive to the distribution of functors in their native language and use them when learning new linguistic material. However, compared to infants’ performance on a similar task, adults exhibit a slightly different behavior, matching the frequency distributions of their native language more closely than infants do. This finding bears on the issue of the continuity of language learning mechanisms.

Abstract: Certain first languages (L1) seem to impede the acquisition of a specific L2 more than other L1s do This study investigates to what extent different L1s have an impact on the proficiency levelsattainedinL2Dutch(DutchL2learnability)Ourhypothesisisthatthevaryingeffects across the L1s are explainable by morphological similarity patterns between the L1s and L2 DutchCorrelationalanalysesontypologicallydefinedmorphologicaldifferencesbetween49 L1s and L2 Dutch show that L2 learnability co-varies systematically with similarities in morphologicalfeaturesWeinvestigateasetof28morphologicalfeatures,lookingbothatindividual features and the total set of features We then divide the differences in features into a class of increasing and a class of decreasing morphological complexity I t turns out that observed Dutch L2 proficiency correlates more strongly with features based on increasing morphological complexity (r= -67, p ⟨ 0001) than with features based on decreasing morphological complexity (r= -45, p ⟨ 005) Degree of similarity matters (r = -77, p ⟨ 0001), but increasing complexity seems to be the decisive property in establishing L2 learnability Our findings mayofferabetterunderstandingofL2learnabilityandofthedifferentproficiencylevelsofL2 speakers L2 learnability and L2 proficiency co-vary in terms of the morphological make-up of the mother tongue and the second language to be learned

Abstract: Driving can be dangerous, especially for young and inexperienced drivers. To help address the issue of inexperience a gamified logbook application was developed for Learner drivers. The application aims to encourage learners to undertake a wider range of practice, while also making it easier to record their mandatory practice sessions. This paper reports on the design of this application, focusing on the effect that adding gamification can have on the usability and user experience of the application and the importance of playability testing for gamified systems. Two versions of the application were developed, one with game elements and one without game elements. This paper presents findings from a study that compares the user experience of these two versions of the application with twelve recent Learner drivers. Overall, participants reported that the gamified version was more engaging and motivating than the non-gamified version, however neither versions were preferred over the other. We theorise that this may have occurred due to a number of usability issues that arose, including an increased difficulty in learnability due to the added game elements. These design issues are important to address in future gamified system designs.

Abstract: End-user programmers such as scientists and engineers often adopt a visual domain-specific language due to its easy learnability, but then they later encounter problems when trying to create high-performance programs. In response, they typically have had to resort to learning and using general textual languages such as C or Fortran as a supplement or replacement for the visual language. This paper proposes a technique, called Smell-driven performance analysis, for helping end-user programmers to overcome performance problems without leaving the visual dataflow paradigm. The technique involves statically analyzing programs to heuristically detect areas with potential performance problems (“bad smells”), alerting enduser programmers about problems, and advising on how to fix those problems. We have implemented a prototype for applying this technique and conducted a user study in which end-user programmers diagnosed performance problems. The experiment showed our technique increased participants' success rates at finding problems and decreased the time required for finding solutions. Qualitatively, 92% of participants said our technique was helpful, and they listed numerous specific benefits provided.

Abstract: We report the results of a study on the learnability of haptic icons used in a system for incoming-call identification in mobile phones. The aim was to explore the feasibility of using haptic icons to create new assistive technologies for people with visual impairments. We compared the performance and satisfaction of users with different visual capacities (visually impaired vs. sighted) and using different learning processes (unimodal vs. multimodal). A better recognition rate and user experience were observed for the visually impaired than for sighted users and for multimodal rather than unimodal learning processes.

Abstract: Learning theory has frequently been applied to language acquisition, but discussion has largely focused on information theoretic problems-in particular on the absence of direct negative evidence. Such arguments typically neglect the probabilistic nature of cognition and learning in general. We argue first that these arguments, and analyses based on them, suffer from a major flaw: they systematically conflate the hypothesis class and the learnable concept class. As a result, they do not allow one to draw significant conclusions about the learner. Second, we claim that the real problem for language learning is the computational complexity of constructing a hypothesis from input data. Studying this problem allows for a more direct approach to the object of study--the language acquisition device-rather than the learnable class of languages, which is epiphenomenal and possibly hard to characterize. The learnability results informed by complexity studies are much more insightful. They strongly suggest that target grammars need to be objective, in the sense that the primitive elements of these grammars are based on objectively definable properties of the language itself. These considerations support the view that language acquisition proceeds primarily through data-driven learning of some form.

Abstract: Anyone who has tried to learn a language with a very different sound system will understand the challenges faced by speakers of a language as different as Vietnamese who are attempting to learn to speak English in a way that is intelligible to non-speakers of Vietnamese. Many learners have very limited opportunity to hear model pronunciations other than their teacher’s, and no opportunity at all to speak in English outside the classroom. Vietnamese-accented English is characterised by a number of features which ride roughshod over English morphosyntax, resulting in speech that is extremely difficult to reconstruct for the non-Vietnamese-speaking listener. Some of these features appear to be more difficult to learn to avoid than others. Phonotactic constraints in L1 appear to be persistent even in L2, and L1 phonological rules will, apparently, often apply in L2 unless they are blocked in some way. Perception of salient (to native listeners) target pronunciations is often lacking, and learners may not be aware that their pronunciation is not intelligible. Despite years of language study, many learners are unable to produce some native speaker targets. Vietnamese learners typically exhibit a set of characteristic pronunciation features in English, and the aim of this study is to see which of these are susceptible to remediation through explicit teaching. This explicit teaching is compared with a less direct, less interactive kind of teaching, involving drawing native and native-like pronunciation of problematic features of English pronunciation to the learners’ attention. The results of this study can then be interpreted in terms of teachability and learnability, which do not always go hand in hand. If we understand what kinds of phonetic features are teachable and how learnability varies for different features, we can target those features where there is a good return for effort spent, resulting in efficient teaching.

Abstract: Eliciting user preferences constitutes a major step towards developing recommender systems and decision support tools. Assuming that preferences are ceteris paribus allows for their concise representation as Conditional Preference Networks (CP-nets). This work presents the first empirical investigation of an algorithm for reliably and efficiently learning CP-nets in a manner that is minimally intrusive. At the same time, it introduces a novel process for efficiently reasoning with (the learned) preferences.

Abstract: [Context] Numerous combinatorial testing techniques are available for generating test cases. However, many of them are never used in practice. [Objective] Considering that learn ability plays a vital role in initial adoption or rejection of a technology, in this paper we aim to investigate the learnability of a combinatorial testing tool in an industrial environment. [Method] A case study research method was designed and conducted, by including i) the definition of learnability measures for test cases models built using a combinatorial testing tool. ii) A training program was also implemented. iii) Qualitative and quantitative evaluation based on a three-level strategy was carried out (Reaction, Learning, and Performance). [Results] At the first level, the tool was perceived as easy to learn by the trainees (from a five-point ordinal scale). However, at the second level, during hands-on learning, it changed slightly: According to the working diaries, there were major difficulties. At third level, analyzing the learning curve of each trainee, we observe that semantic errors made per each subject were reduced slightly over the time.

Abstract: Most studies were devoted to the design of efficient algorithms and the evaluation and application on diverse ranking problems, whereas few work has been paid to the theoretical studies on ranking learnability. In this paper, we study the relation between uniform convergence, stability and learnability of ranking. In contrast to supervised learning where the learnability is equivalent to uniform convergence, we show that the ranking uniform convergence is sufficient but not necessary for ranking learnability with AERM, and we further present a sufficient condition for ranking uniform convergence with respect to bipartite ranking loss. Considering the ranking uniform convergence being unnecessary for ranking learnability, we prove that the ranking average stability is a necessary and sufficient condition for ranking learnability.

Abstract: This introduction presents an overview of key concepts discussed in the book. The book addresses central issues in the domain of language development: the range of evidence in favor of a language instinct, the existence of a critical period for language acquisition, the issue of maturation in the context of language acquisition, and the impact of language on other cognitive systems. It focuses on the interplay between mind, brain, and behavior. The book deals with the nature of theoretically informed experiments, working memory and language processing, modularity, language deficits, and pathologies. It focuses on a range of issues relevant to the study of language evolution: the cognitive capacities of non-human primates, the abilities of non-human vocal learners, the potential use of fossil records to shed light on the evolution of language, the possible role of natural selection, and the insights from computational modeling in the context of language evolution.

Abstract: Esperanto is a constructed or planned language which is actively used by a multifaceted speech community all over the world. For more than a century, the nature of the parts-of-speech (PoS) system of Esperanto has been an item of controversy among its grammarians. Based on a study of the language as it is actually used by its speech community and supported by historical and contemporary documents, the author argues that the language's current PoS system can be described adequately as a mixture of contentive lexemes and verbally specialized lexemes. A system of contentive lexemes, which allows the user to create four different words with predictable meanings out of each single root of any semantic category without having to use derivational tools, does justice to the learnability claim which underlies the structure of this language, which was made to be easy. The increasing number of verbally specialized lexemes is a development which is known to, but unsatisfactorily explained by many esperantologists. The paper develops the hypothesis that the diachrony of Esperanto (constructed in its origin, but freely developing) provides support to the claim that natural languages can be classified in terms of an implicational hierarchy of their PoS systems, in which verbal specialization precedes all other forms of syntactic specialization of lexemes. The analysis is done within the framework of functional discourse grammar.

Abstract: As visualization tools get more complicated, users often find it increasingly difficult to learn interaction sequences, recall past queries, and interpret visual states. We examine a query-to-question (Q2Q) supporting system that takes advantage of natural language generation (NLG) techniques to automatically translate and display query interactions as natural language questions. We focus on a symmetric pattern of multiple coordinated views, cross-filtered views, that involves only nominal/categorical data. We describe a study of the effects of pairing a visualization with a Q2Q interface on several aspects of usability. Q2Q produces considerable improvements in learnability, efficiency, and memorability of visualization in terms of speed and the length of interaction sequences that users follow, along with a modest decrease in error ratio. From a visual language perspective, we analyze how Q2Q speeds up users' comprehension of interaction, particularly when a visualization representation has deficiencies in illustrating hidden items or relationships.

Abstract: In a beautifully written, cogently argued paper MacDonald (2013), presents the theoretical framework that guides one of the most creative and influential research programs in the language sciences. The PDC began with empirical demonstrations that readers are remarkably sensitive to distributional patterns in the input. These empirical demonstrations were accompanied by theoretical arguments that ambiguity resolution can be modeled by constraint-based (probabilistic) systems that learn these patterns from experience (also see Tanenhaus and Trueswell, 1995; Tabor et al., 1997). This research was part of a wave of research in the 1990's that answered long-standing questions in real-time language comprehension and early language acquisition, with variations of “It's the input, stupid” (e.g., Saffran, Aslin, and Newport's seminal work on statistical learning, Saffran et al., 1996). The probabilistic constraints that comprehenders learn and use must arise from the output created by speakers and writers. But why does that output exhibit systematic patterns both within and across language? One answer is that languages maximize learnability for the child. A second is that speaking and more generally the structure of language is shaped by considerations of communicative efficiency (see Jaeger's commentary). MacDonald proposes a production-based answer. Grounding her arguments in insights from the motor planning and control literature, MacDonald makes a convincing case that that constraints on planning processes in language production play an important role in shaping the form of utterances. She argues that the demands of memory retrieval, planning, and linearization for sequential behavior that requires a hierarchical control structure play the central role both in determining the forms that speakers use to convey their intentions and, as a consequence, the patterns of linguistic forms that are observed both within and across languages. These arguments are supported by a clear exposition of principles and a summary of some elegant experiments focusing on the production of relative clauses. Less convincing is MacDonald's argument that these production constraints comprise most of the story. Here is the cartoon view of the assumptions that underlie this claim. Speaking is extremely hard while understanding is comparatively easy. It is costly for speakers to take into account the listener, especially at the temporal grain required for influencing the planning process in production. Listeners, however, are really good at learning probabilistic constraints, making use of context, and adapting to speakers. Given these considerations it makes sense for speakers (and languages) to promote forms that make speaking easier. In this commentary, I raise question about some of the assumption that underlie the claim that production demands—in particular planning and linearization—are most of the story. I begin by noting some parallels with earlier arguments that were based on assumptions about the difficulty of comprehension. I note that much of what we know about how naturally listeners use context, emerged only when psycholinguists began to examine language comprehension in richer contexts and more natural interactive tasks. I suggest that we actually don't know much about how speakers might adapt to addressees in real-time language production. There is a paucity of research that examines production in those interactive environments where addressees provide feedback. When production is examined in interactive settings, there is tantalizing evidence that speakers do, in fact, monitor addressees, and might adapt on the fly.

Abstract: The present work determines the exact nature of {\em linear time computable} notions which characterise automatic functions (those whose graphs are recognised by a finite automaton). The paper also determines which type of linear time notions permit full learnability for learning in the limit of automatic classes (families of languages which are uniformly recognised by a finite automaton). In particular it is shown that a function is automatic iff there is a one-tape Turing machine with a left end which computes the function in linear time where the input before the computation and the output after the computation both start at the left end. It is known that learners realised as automatic update functions are restrictive for learning. In the present work it is shown that one can overcome the problem by providing work tapes additional to a resource-bounded base tape while keeping the update-time to be linear in the length of the largest datum seen so far. In this model, one additional such work tape provides additional learning power over the automatic learner model and two additional work tapes give full learning power. Furthermore, one can also consider additional queues or additional stacks in place of additional work tapes and for these devices, one queue or two stacks are sufficient for full learning power while one stack is insufficient.

Abstract: Performance metrics are powerful tools used to evaluate the usability of any product. They are the cornerstone of usability and can inform key decisions, such as whether a new product is ready to launch. Performance metrics are always based on users’ behavior rather than what they say. There are five general types of performance metrics: task success, time on task, errors, efficiency, and learnability. Task success metrics are used when you are interested in whether users are able to complete tasks using the product. Sometimes you might only be interested in whether a user is successful or not based on a strict set of criteria (binary success). Other times you might be interested in defining different levels of success based on the degree of completion, the user’s experience in finding an answer, or the quality of the answer given. Time on task is helpful when you are concerned about how quickly users can perform tasks with the product. You might look at the time it takes to complete a task for all users, a subset of users, or the proportion of users who can complete a task within a desired time limit. Errors are based on the number of mistakes users make while attempting to complete a task. A task might have a single error opportunity or multiple error opportunities, and some types of errors may be more important than others. Efficiency is a way of evaluating the amount of effort (cognitive and physical) required to complete a task. Efficiency is often measured by the number of steps or actions required to complete a task or by the ratio of the task success rate to the average time per task. Learnability involves looking at how any efficiency metric changes over time. Learnability is useful if you want to examine how and when users reach proficiency in using a product.