Monthly Archives: February 2019

Zobel in Cognitive bag lunch Weds. Feb. 20 at noon

The next cognitive brown bag is on 2/20 (12:00, Tobin 521B). The speaker is Ben Zobel (UMass PBS); title and abstract are below. Please note that there is no cognitive brown bag on 2/27.

Effects of Age on Spatial Release from Informational Masking

Spatial release from informational masking (SRIM) describes the reduction in perceptual/cognitive confusion between relevant speech (target) and irrelevant speech (masker) when target and masker are perceived as spatially separated compared to spatially co-located. Under complex listening conditions in which peripheral (head shadow) and low-level binaural (interaural time differences) cues are washed out by multiple noise sources and reverberation, SRIM is a crucial mechanism for successful speech processing. It follows that any age-related declines in SRIM would contribute to the speech-processing difficulties older adults often report within noisy environments. Some research indicates age-related declines in SRIM (e.g., Gallun et al., 2013) while other research does not (e.g., Li et al., 2004). This talk will describe research designed to add clarity to two fundamental questions: 1) Does SRIM decline with age and, if so, 2) does age predict this decline independent of hearing loss? To answer these questions, younger and older participants listened to low-pass-filtered noise-vocoded speech and were asked to detect whether a target talker was presented along with two-talker masking babble. Spatial separation was perceptually manipulated without changing peripheral and low-level cues (Freyman et al., 1999). Results showed that detection thresholds were nearly identical across age groups in the co-located condition but markedly higher for older adults compared to younger adults when target and masker were spatially separated. Multiple regression analysis showed that age predicted a decline in SRIM controlling for hearing loss (based on pure-tone audiometry), while there was no indication that hearing loss predicted a decline in SRIM controlling for age. These results provide strong evidence that SRIM declines with age, and that the source of this decline may begin at higher perceptual/cognitive levels of auditory processing. Such declines are likely to contribute to the greater speech-processing difficulties older adults often experience in complex, noisy environments.

Syrett in Linguistics Fri. Feb. 22 at 3:30

Kristen Syrett of Rutgers University (https://sites.rutgers.edu/kristen-syrett/) will present “Playing with semantic building blocks: Acquiring the lexical representations of verbs and adjectives” in ILC N400 Friday Feb. 22 at 3:30. All are welcome!

ABSTRACT: Early lexicons and initial child productions reflect a preponderance of object-denoting lexical items (nouns), while those that denote properties of objects or events (adjectives and verbs) lag behind. If nouns are the “Marsha” of the Brady Bunch, adjectives and verbs compete for the role of “Jan.” In many ways, this asymmetry privileging nouns makes sense: it’s much easier to track event participants than to track ephemeral events and the properties of those participants, which are much less stable, and both verbs and adjectives require nominal elements both syntactically and semantically. But the process of language acquisition is rapid, and within a matter of a few years, the child fairly quickly achieves proficiency, enough so to appreciate polysemy or word play. Given this state of affairs, we might ask two questions about the acquisition of these predicates: (1) What strategies or information sources do children recruit to pin down the lexical meaning of verbs and adjectives?, and (2) When they enter into the lexicon, how rich is children’s semantic knowledge of these words? In this talk, I provide one answer to (1), showcasing the role of the linguistic context. I then highlight a set of examples in response to (2), illustrating children’s early command of selectional restrictions for both categories. In doing so, I also demonstrate that once these words are established as part of the children’s receptive and productive vocabulary, there are certain advantages afforded to the language learner—although here, we uncover an asymmetry between verbs and adjectives implicating other aspects of the grammar and the context. Together, what this body of work reveals is the complex, interrelated process of acquiring and assembling the semantic building blocks of language.

Verga in Machine Learning and Friends Thurs. Feb. 14 at 11:45

who: Patrick Verga
when: Feb 14 11:45am
where: Computer Science Building Rm 150
food: Athena’s Pizza

“Neural Knowledge Representation And Reasoning”

Abstract: Making complex decisions in science, government policy, finance, and clinical treatments all require integrating and reasoning over disparate sources of information. While some decisions can be made from a single piece of evidence, others require considering information that exists outside of the current context. A long-term store of abstracted knowledge over related concepts can facilitate this type of reasoning, while also influencing interpretation as well as enhancing the acquisition of new knowledge. A symbolic graph over a fixed, human-defined schema encoding facts about entities and their relations is the predominant method of representing knowledge, but this method is brittle, lacks specificity, and is inevitably highly incomplete. On the other extreme, recent work on purely text based knowledge models lack abstractions necessary for complex reasoning.

In this talk I will present a middle ground incorporating powerful neural network models with rich structured ontologies and unstructured raw text to improve the representations of entities and their relations. We first discuss our work on universal schema, a method for learning a latent schema over both existing structured resources and unstructured free text data, embedding them jointly within a shared semantic space. Next we inject additional hierarchical structure into the embedding space of concepts, resulting in more efficient statistical sharing amongst related concepts and improving accuracy in both fine-grained entity typing and linking. We then present initial work to represent knowledge in context, including a single model for extracting all entities and long-range relations simultaneously over full paragraphs while jointly linking these entities to a knowledge graph. Lastly, we propose future directions for representing knowledge in context by incorporating cognitive theories of human memory systems and discuss how these models can address longstanding shortcomings in knowledge representation.

Bio: Patrick Verga is a final year PhD candidate in the College of Information and Computer Sciences at UMass Amherst, advised by Andrew McCallum. His research contributes to knowledge representation and reasoning, with a focus on large knowledge base construction from unstructured text, with applications to general domain, commonsense, and biomedicine. Pat previously interned at Google and the Chan Zuckerberg Initiative and received the best long paper award at EMNLP 2018. Over the past several years he has advised multiple M.S. and junior PhD students, resulting in published research in fine-grained entity typing, unsupervised parsing, and partially labeled named entity extraction. He holds M.S. and B.A degrees in computer science as well as a B.S. in neuroscience.

Zhang in Computer Science Weds. Feb. 13 at 4 pm

Systems to Improve Online Discussion

Amy X. Zhang
Computer Science and Artificial Intelligence Laboratory (CSAIL)
MIT

Wednesday, February 13, 2019
4:00 – 5:00 p.m.
Computer Science 151

Abstract — Discussions online are integral to everyday life, affecting how we learn, work, socialize, and participate in public society. Yet the systems that we use to conduct online discourse, whether they be email, chat, or forums, have changed little since their inception many decades ago. As more people participate and more venues for discourse migrate online, new problems have arisen, and old problems have intensified. People are still drowning in information and must now juggle dozens of disparate discussion silos in addition. Finally, an unfortunately significant proportion of this online interaction is unwanted or unpleasant, with clashing norms leading to people bickering or getting harassed into silence. My research in human-computer interaction is on reimagining outdated designs towards designing novel online discussion systems that fix what’s broken about online discussion. To solve these problems, I develop tools that empower users and communities to have direct control over their experiences and information. These include: 1) summarization tools to make sense of large discussions, 2) annotation tools to situate conversations in the context of what is being discussed, as well as 3) moderation tools to give users more fine-grained control over content delivery.

Bio — Amy X. Zhang is a fifth-year PhD student at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), focusing on human-computer interaction and social computing. She is also a 2018-19 Fellow at the Berkman Klein Center at Harvard. She has interned at Microsoft Research and Google Research and was a software engineer at a news startup before her PhD. Her work has received awards at CHI and CSCW and has been featured by ABC News, BBC, CBC, The Verge, and New Scientist. She has an M.Phil. in advanced computer science at the University of Cambridge on a Gates Fellowship and a B.S. in computer science at Rutgers, where she was captain of the Division I Women’s tennis team. Her research is supported by a Google PhD Fellowship and an NSF Graduate Research Fellowship.

A reception for attendees will be held at 3:30 p.m. in CS 150

Serre in Cognitive Brown Bag Weds. Feb. 13 at noon

The next Cognitive brown bag speaker will be Thomas Serre of Brown University (http://serre-lab.clps.brown.edu/). The talk is on Wednesday 2/13, 12:00, Tobin 521B; title and abstract are below.

What are the computations underlying primate versus machine vision?

Primates excel at object recognition: For decades, the speed and accuracy of their visual system have remained unmatched by computer algorithms. But recent advances in Deep Convolutional Networks (DCNs) have led to vision systems that are starting to rival human decisions. A growing body of work also suggests that this recent surge in accuracy is accompanied by a concomitant improvement in our ability to account for neural data in higher areas of the primate visual cortex. Overall, DCNs have become de facto computational models of visual recognition.

In this talk, I will review recent work by our group which brings into relief limitations of modern DCNs as computational models of primate vision. I will show that DCNs are limited in their ability to solve seemingly simple visual reasoning problems involving incremental grouping, similarity and spatial relation judgments suggesting the need for additional neural computations beyond those implemented in current architectures. I will further demonstrate how neuroscience principles may help guide the future design for more robust computer vision architectures.

Perkins in Linguistics Fri. Feb. 15 at 3:30

Laurel Perkins of the University of Maryland (http://ling.umd.edu/~perkinsl/) will present “How to Grow a Grammar: Syntactic Development in 1-Year-Olds” on Friday Feb. 15th at 3:30 PM in N400. All are welcome – an abstract follows.

ABSTRACT: What we can learn depends on what we already know; a child who can’t count cannot learn arithmetic, and a child who can’t segment words cannot identify properties of verbs in her language. Language acquisition, like learning in general, is incremental. How do children draw the right generalizations about their language using incomplete and noisy representations of their linguistic input?

In this talk, I’ll examine some of the first steps of syntax acquisition in 1-year-old infants, using behavioral methods to probe their linguistic representations, and computational methods to ask how they learn from those representations. Taking argument structure as my case study, I will show: (1) that infants represent core clause arguments like “subject” and “object” when learning verbs, (2) that infants can cope with “non-basic” clause types, where those arguments have been displaced, by ignoring some of their input, and (3) that it is possible for infants to learn what kind of data to ignore, even before they can parse it. I will argue that the approach I take for studying this particular learning problem will generalize widely, allowing us to build new models for understanding the role of development in grammar learning.

Suhr in Machine Learning and Friends Thurs. Feb. 7 at 11:45

who: Alane Suhr
when: Feb 7 11:45am
where: Computer Science Building Rm 150
food: Athena’s Pizza

“Modeling and Learning Agents that Understand Language in Context”

Abstract: The meaning of a natural language utterance is influenced by the context in which it occurs, including interaction history and situated context. I will discuss two recent projects in context-dependent natural language understanding for building natural language interfaces to databases and following sequences of instructions. In the first part, I will introduce a model for mapping from natural language to executable SQL queries in an interaction. To resolve the meaning of later utterances, the system must consider the interaction history, including previous user utterances and previously-generated queries. We show how using both implicit and explicit mechanisms for making use of interaction history allows the system to effectively generate context-dependent representations. In the second part, I will describe an approach to map sequences of natural language instructions to system actions that modify an environment, focusing on learning without direct supervision on action sequences. We introduce an exploration-based learning approach that effectively learns to compose system actions to carry out user instructions in context of the environment and interaction.

Bio: Alane Suhr is a PhD student in the Computer Science department at Cornell University, focusing on building agents that understand natural language grounded in complex interactions. She is the recipient of an AI2 Key Scientific Challenges Award and a Microsoft Research Women’s Fellowship, and is a National Science Foundation Graduate Research Fellow. She has received paper awards at ACL 2017 and NAACL 2018. Alane received a Bachelor’s degree in Computer Science and Engineering from Ohio State University in 2016.

Momma in Linguistics Fri. Feb. 8 at 3:30

Shota Momma of UC San Diego (https://shotam.github.io) will present “Unifying parsing and generation” at 3:30, Friday February 8th in ILC N400. All are welcome!

Abstract: We use our grammatical knowledge in at least two ways. On one hand, we use our grammatical knowledge to say what we want to convey to others. On the other hand, we use our grammatical knowledge to understand what others say. In either case, we need to assemble sentence structures in a systematic fashion, in accordance with the grammar of our language. In this talk, I will advance the view that the same syntactic structure building mechanism is shared between comprehension and production, specifically focusing on sentences involving long-distance dependencies. I will argue that both comprehenders and speakers anticipatorily build (i.e., predict and plan) the gap structure, soon after they represent the filler and before representing the words and structures that intervene the filler and the gap. I will discuss the basic properties of the algorithm for establishing long-distance dependencies that I hypothesize to be shared between comprehension and production, and suggest that it resembles the derivational steps for establishing long-distance dependencies in an independently motivated grammatical formalism, known as Tree Adjoining Grammar.

Hopper in Cognitive bag lunch Weds. Feb. 6 at noon

Will Hopper (https://people.umass.edu/whopper/) will present “Comparing discrete and continuous evidence models of recognition memory response times” on 2/6 at 12:00 in Tobin 521B (abstract below). All are welcome.

Memory theorists have long debated whether recognition decisions are mediated by considering the strength of a continuous memory strength signal, or by entering discrete evidence states. Historically, only models which utilized a continuous memory strength signal were able to account for both the distribution of response times and choice probabilities of recognition decisions. Recently, discrete state models have been extended to account for response times distributions, assuming the observed response times arise as a mixture of latent response time distributions associated with each discrete evidence state (Heck & Erdfelder, 2016, Starns, 2018). Here, we compare models from each class (the discrete-race model and the Ratcliff diffusion model), testing their ability to account for both speeded and unspeeded recognition decisions for items tested multiple times within a session. We conclude that the Ratcliff diffusion model provides a better account of the data, as the discrete-race model overestimates memory strength on unspeeded tests in order to describe the response times on speeded tests.