Search

Commonsense knowledge is essential for machines to reason about the world. Large language models (LLMs) have demonstrated their ability to perform almost human-like text generation. Despite this success, they fall short as trustworthy intelligent systems, due to the opacity of the basis for their answers and a tendency to confabulate facts when questioned about obscure entities or technical domains. We hypothesize, however, that their general knowledge about objects in the everyday world is largely sound. Based on that hypothesis, this paper investigates LLMs' ability to formulate explicit knowledge about common physical artifacts, focusing on their parts and materials. Our work distinguishes between the substances that comprise an entire object and those that constitute its parts$\unicode{x2014}$a previously underexplored distinction in knowledge base construction. Using few-shot with five in-context examples and zero-shot multi-step prompting, we produce a repository of data on the parts and materials of about 2,300 objects and their subtypes. Our evaluation demonstrates LLMs' coverage and soundness in extracting knowledge. This contribution to knowledge mining should prove useful to AI research on reasoning about object structure and composition and serve as an explicit knowledge source (analogous to knowledge graphs) for LLMs performing multi-hop question answering.

The zero-shot chain of thought (CoT) approach is often used in question answering (QA) by language models (LMs) for tasks that require multiple reasoning steps. However, some QA tasks hinge more on accessing relevant knowledge than on chaining reasoning steps. We introduce a simple prompting technique, called PREP, that involves using two instances of LMs: the first (LM1) generates relevant information, and the second (LM2) receives the information from the user and answers the question. This design is intended to make better use of the LM's instruction-following capability. PREP is applicable across various QA tasks without domain-specific prompt engineering. PREP is developed on a dataset of 100 QA questions, derived from an extensive schematic dataset specifying artifact parts and material composition. These questions ask which of two artifacts is less likely to share materials with another artifact. Such questions probe the LM's knowledge of shared materials in the part structure of different artifacts. We test our method on our parts-and-materials dataset and three published commonsense reasoning datasets. The average accuracy of our method is consistently higher than that of all the other tested methods across all the tested datasets.

Algorithmic theories of randomness can be related to theories of probabilistic sequence prediction through the notion of a predictor, defined as a function which supplies lower bounds on initial-segment probabilities of infinite sequences. An infinite binary sequence $z$ is called unpredictable iff its initial-segment "redundancy" $n+\log p(z(n))$ remains sufficiently low relative to every effective predictor $p$. A predictor which maximizes the initial-segment redundancy of a sequence is called optimal for that sequence. It turns out that a sequence is random iff it is unpredictable. More generally, a sequence is random relative to an arbitrary computable distribution iff the distribution is itself an optimal predictor for the sequence. Here "random" can be taken in the sense of Martin-L\"{o}f by using weak criteria of effectiveness, or in the sense of Schnorr by using stronger criteria of effectiveness. Under the weaker criteria of effectiveness it is possible to construct a universal predictor which is optimal for all infinite sequences. This predictor assigns nonvanishing limit probabilities precisely to the recursive sequences. Under the stronger criteria of effectiveness it is possible to establish a law of large numbers for sequences random relative to a computable distribution, which may be useful as a criterion of "rationality" for methods of probabilistic prediction. A remarkable feature of effective predictors is the fact that they are expressible in the special form first proposed by Solomonoff. In this form sequence prediction reduces to assigning high probabilities to initial segments with short and/or numerous encodings. This fact provides the link between theories of randomness and Solomonoff's theory of prediction., Comment: 30 pages + refs. A re-typeset University of Alberta Technical Report, no longer available as such

In many NLP applications that involve interpreting sentences within a rich context -- for instance, information retrieval systems or dialogue systems -- it is desirable to be able to preserve the sentence in a form that can be readily understood without context, for later reuse -- a process known as ``decontextualization''. While previous work demonstrated that generative Seq2Seq models could effectively perform decontextualization after being fine-tuned on a specific dataset, this approach requires expensive human annotations and may not transfer to other domains. We propose a few-shot method of decontextualization using a large language model, and present preliminary results showing that this method achieves viable performance on multiple domains using only a small set of examples.

Many practical applications of dialogue technology require the generation of responses according to a particular developer-specified persona. While a variety of personas can be elicited from recent large language models, the opaqueness and unpredictability of these models make it desirable to be able to specify personas in an explicit form. In previous work, personas have typically been represented as sets of one-off pieces of self-knowledge that are retrieved by the dialogue system for use in generation. However, in realistic human conversations, personas are often revealed through story-like narratives that involve rich habitual knowledge -- knowledge about kinds of events that an agent often participates in (e.g., work activities, hobbies, sporting activities, favorite entertainments, etc.), including typical goals, sub-events, preconditions, and postconditions of those events. We capture such habitual knowledge using an explicit schema representation, and propose an approach to dialogue generation that retrieves relevant schemas to condition a large language model to generate persona-based responses. Furthermore, we demonstrate a method for bootstrapping the creation of such schemas by first generating generic passages from a set of simple facts, and then inducing schemas from the generated passages.

Effective communication between a clinician and their patient is critical for delivering healthcare maximizing outcomes. Unfortunately, traditional communication training approaches that use human standardized patients and expert coaches are difficult to scale. Here, we present the development and validation of a scalable, easily accessible, digital tool known as the Standardized Online Patient for Health Interaction Education (SOPHIE) for practicing and receiving feedback on doctor-patient communication skills. SOPHIE was validated by conducting an experiment with 30 participants. We found that participants who underwent SOPHIE performed significantly better than the control in overall communication, aggregate scores, empowering the patient, and showing empathy ($p < 0.05$ in all cases). One day, we hope that SOPHIE will help make communication training resources more accessible by providing a scalable option to supplement existing resources., Comment: 10 pages, 5 figures, 2 tables

A schema-guided approach to dialogue management has been shown in recent work to be effective in creating robust customizable virtual agents capable of acting as friendly peers or task assistants. However, successful applications of these methods in open-ended, mixed-initiative domains remain elusive -- particularly within medical domains such as virtual standardized patients, where such complex interactions are commonplace -- and require more extensive and flexible dialogue management capabilities than previous systems provide. In this paper, we describe a general-purpose schema-guided dialogue management framework used to develop SOPHIE, a virtual standardized cancer patient that allows a doctor to conveniently practice for interactions with patients. We conduct a crowdsourced evaluation of conversations between medical students and SOPHIE. Our agent is judged to produce responses that are natural, emotionally appropriate, and consistent with her role as a cancer patient. Furthermore, it significantly outperforms an end-to-end neural model fine-tuned on a human standardized patient corpus, attesting to the advantages of a schema-guided approach.

We present NESL (the Neuro-Episodic Schema Learner), an event schema learning system that combines large language models, FrameNet parsing, a powerful logical representation of language, and a set of simple behavioral schemas meant to bootstrap the learning process. In lieu of a pre-made corpus of stories, our dataset is a continuous feed of "situation samples" from a pre-trained language model, which are then parsed into FrameNet frames, mapped into simple behavioral schemas, and combined and generalized into complex, hierarchical schemas for a variety of everyday scenarios. We show that careful sampling from the language model can help emphasize stereotypical properties of situations and de-emphasize irrelevant details, and that the resulting schemas specify situations more comprehensively than those learned by other systems., Comment: To appear at ACL SRW 2022

"Episodic Logic:Unscoped Logical Form" (EL-ULF) is a semantic representation capturing predicate-argument structure as well as more challenging aspects of language within the Episodic Logic formalism. We present the first learned approach for parsing sentences into ULFs, using a growing set of annotated examples. The results provide a strong baseline for future improvement. Our method learns a sequence-to-sequence model for predicting the transition action sequence within a modified cache transition system. We evaluate the efficacy of type grammar-based constraints, a word-to-symbol lexicon, and transition system state features in this task. Our system is available at https://github.com/genelkim/ulf-transition-parser We also present the first official annotated ULF dataset at https://www.cs.rochester.edu/u/gkim21/ulf/resources/.

In this paper, we describe the iterative participatory design of SOPHIE, an online virtual patient for feedback-based practice of sensitive patient-physician conversations, and discuss an initial qualitative evaluation of the system by professional end users. The design of SOPHIE was motivated from a computational linguistic analysis of the transcripts of 383 patient-physician conversations from an essential office visit of late stage cancer patients with their oncologists. We developed methods for the automatic detection of two behavioral paradigms, lecturing and positive language usage patterns (sentiment trajectory of conversation), that are shown to be significantly associated with patient prognosis understanding. These automated metrics associated with effective communication were incorporated into SOPHIE, and a pilot user study identified that SOPHIE was favorably reviewed by a user group of practicing physicians.

It is essential for dialogue-based spatial reasoning systems to maintain memory of historical states of the world. In addition to conveying that the dialogue agent is mentally present and engaged with the task, referring to historical states may be crucial for enabling collaborative planning (e.g., for planning to return to a previous state, or diagnosing a past misstep). In this paper, we approach the problem of spatial memory in a multi-modal spoken dialogue system capable of answering questions about interaction history in a physical blocks world setting. This work builds upon a full spatial question-answering pipeline consisting of a vision system, speech input and output mediated by an animated avatar, a dialogue system that robustly interprets spatial queries, and a constraint solver that derives answers based on 3-D spatial modelling. The contributions of this work include a symbolic dialogue context registering knowledge about discourse history and changes in the world, as well as a natural language understanding module capable of interpreting free-form historical questions and querying the dialogue context to form an answer., Comment: 16 pages, 4 figures

The blocks world is a classic toy domain that has long been used to build and test spatial reasoning systems. Despite its relative simplicity, tackling this domain in its full complexity requires the agent to exhibit a rich set of functional capabilities, ranging from vision to natural language understanding. There is currently a resurgence of interest in solving problems in such limited domains using modern techniques. In this work we tackle spatial question answering in a holistic way, using a vision system, speech input and output mediated by an animated avatar, a dialogue system that robustly interprets spatial queries, and a constraint solver that derives answers based on 3-D spatial modeling. The contributions of this work include a semantic parser that maps spatial questions into logical forms consistent with a general approach to meaning representation, a dialog manager based on a schema representation, and a constraint solver for spatial questions that provides answers in agreement with human perception. These and other components are integrated into a multi-modal human-computer interaction pipeline., Comment: 9 pages (with references), 2 figures

We present some results concerning the dialogue behavior and inferred sentiment of a group of older adults interacting with a computer-based avatar. Our avatar is unique in its ability to hold natural dialogues on a wide range of everyday topics---27 topics in three groups, developed with the help of gerontologists. The three groups vary in ``degrees of intimacy", and as such in degrees of difficulty for the user. Each participant interacted with the avatar for 7-9 sessions over a period of 3-4 weeks; analysis of the dialogues reveals correlations such as greater verbosity for more difficult topics, increasing verbosity with successive sessions, especially for more difficult topics, stronger sentiment on topics concerned with life goals rather than routine activities, and stronger self-disclosure for more intimate topics. In addition to their intrinsic interest, these results also reflect positively on the sophistication of our dialogue system.

A growing interest in tasks involving language understanding by the NLP community has led to the need for effective semantic parsing and inference. Modern NLP systems use semantic representations that do not quite fulfill the nuanced needs for language understanding: adequately modeling language semantics, enabling general inferences, and being accurately recoverable. This document describes underspecified logical forms (ULF) for Episodic Logic (EL), which is an initial form for a semantic representation that balances these needs. ULFs fully resolve the semantic type structure while leaving issues such as quantifier scope, word sense, and anaphora unresolved; they provide a starting point for further resolution into EL, and enable certain structural inferences without further resolution. This document also presents preliminary results of creating a hand-annotated corpus of ULFs for the purpose of training a precise ULF parser, showing a three-person pairwise interannotator agreement of 0.88 on confident annotations. We hypothesize that a divide-and-conquer approach to semantic parsing starting with derivation of ULFs will lead to semantic analyses that do justice to subtle aspects of linguistic meaning, and will enable construction of more accurate semantic parsers., Comment: Accepted for publication at The 13th International Conference on Computational Semantics (IWCS 2019)

We address the problem of designing a conversational avatar capable of a sequence of casual conversations with older adults. Users at risk of loneliness, social anxiety or a sense of ennui may benefit from practicing such conversations in private, at their convenience. We describe an automatic spoken dialogue manager for LISSA, an on-screen virtual agent that can keep older users involved in conversations over several sessions, each lasting 10-20 minutes. The idea behind LISSA is to improve users' communication skills by providing feedback on their non-verbal behavior at certain points in the course of the conversations. In this paper, we analyze the dialogues collected from the first session between LISSA and each of 8 participants. We examine the quality of the conversations by comparing the transcripts with those collected in a WOZ setting. LISSA's contributions to the conversations were judged by research assistants who rated the extent to which the contributions were "natural", "on track", "encouraging", "understanding", "relevant", and "polite". The results show that the automatic dialogue manager was able to handle conversation with the users smoothly and naturally.

We present the design of an online social skills development interface for teenagers with autism spectrum disorder (ASD). The interface is intended to enable private conversation practice anywhere, anytime using a web-browser. Users converse informally with a virtual agent, receiving feedback on nonverbal cues in real-time, and summary feedback. The prototype was developed in consultation with an expert UX designer, two psychologists, and a pediatrician. Using the data from 47 individuals, feedback and dialogue generation were automated using a hidden Markov model and a schema-driven dialogue manager capable of handling multi-topic conversations. We conducted a study with nine high-functioning ASD teenagers. Through a thematic analysis of post-experiment interviews, identified several key design considerations, notably: 1) Users should be fully briefed at the outset about the purpose and limitations of the system, to avoid unrealistic expectations. 2) An interface should incorporate positive acknowledgment of behavior change. 3) Realistic appearance of a virtual agent and responsiveness are important in engaging users. 4) Conversation personalization, for instance in prompting laconic users for more input and reciprocal questions, would help the teenagers engage for longer terms and increase the system's utility.

We characterize probabilities in Bayesian networks in terms of algebraic expressions called quasi-probabilities. These are arrived at by casting Bayesian networks as noisy AND-OR-NOT networks, and viewing the subnetworks that lead to a node as arguments for or against a node. Quasi-probabilities are in a sense the "natural" algebra of Bayesian networks: we can easily compute the marginal quasi-probability of any node recursively, in a compact form; and we can obtain the joint quasi-probability of any set of nodes by multiplying their marginals (using an idempotent product operator). Quasi-probabilities are easily manipulated to improve the efficiency of probabilistic inference. They also turn out to be representable as square-wave pulse trains, and joint and marginal distributions can be computed by multiplication and complementation of pulse trains., Comment: Appears in Proceedings of the Twentieth Conference on Uncertainty in Artificial Intelligence (UAI2004)

Natural language understanding applications such as interactive planning and face-to-face translation require extensive inferencing. Many of these inferences are based on the meaning of particular open class words. Providing a representation that can support such lexically-based inferences is a primary concern of lexical semantics. The representation language of first order logic has well-understood semantics and a multitude of inferencing systems have been implemented for it. Thus it is a prime candidate to serve as a lexical semantics representation. However, we argue that FOL, although a good starting point, needs to be extended before it can efficiently and concisely support all the lexically-based inferences needed., Comment: Presented at the "Future of the Dictionary" workshop, Grenoble, France (October, 1994), 12 pages PostScript