Justine T. Kao

Nonliteral understanding of number words

Significance Human communication is rife with nonliteral language, ranging from metaphor to irony to hyperbole. How do people go so far beyond the literal meaning of an utterance to infer the speaker’s intended meaning? We present a computational model that understands hyperbolic and other nonliteral uses of number words (e.g., “That watch costs 10,000 dollars”). Our model integrates empirically measured background knowledge, principles of communication, and reasoning about communicative goals to explain the computational basis of nonliteral language understanding. This framework sheds light on the nature of communication, marking a significant advancement in the flexibility and richness of formal models of language understanding. One of the most puzzling and important facts about communication is that people do not always mean what they say; speakers often use imprecise, exaggerated, or otherwise literally false descriptions to communicate experiences and attitudes. Here, we focus on the nonliteral interpretation of number words, in particular hyperbole (interpreting unlikely numbers as exaggerated and conveying affect) and pragmatic halo (interpreting round numbers imprecisely). We provide a computational model of number interpretation as social inference regarding the communicative goal, meaning, and affective subtext of an utterance. We show that our model predicts humans’ interpretation of number words with high accuracy. Our model is the first to our knowledge to incorporate principles of communication and empirically measured background knowledge to quantitatively predict hyperbolic and pragmatic halo effects in number interpretation. This modeling framework provides a unified approach to nonliteral language understanding more generally.

Speech recognitionwith segmental conditional random fields: A summary of the JHU CLSP 2010 Summer Workshop

This paper summarizes the 2010 CLSP Summer Workshop on speech recognition at Johns Hopkins University. The key theme of the workshop was to improve on state-of-the-art speech recognition systems by using Segmental Conditional Random Fields (SCRFs) to integrate multiple types of information. This approach uses a state-of-the-art baseline as a springboard from which to add a suite of novel features including ones derived from acoustic templates, deep neural net phoneme detections, duration models, modulation features, and whole word point-process models. The SCRF framework is able to appropriately weight these different information sources to produce significant gains on both the Broadcast News and Wall Street Journal tasks.

A Computational Model of Linguistic Humor in Puns

Abstract Humor plays an essential role in human interactions. Precisely what makes something funny, however, remains elusive. While research on natural language understanding has made significant advancements in recent years, there has been little direct integration of humor research with computational models of language understanding. In this paper, we propose two information‐theoretic measures—ambiguity and distinctiveness—derived from a simple model of sentence processing. We test these measures on a set of puns and regular sentences and show that they correlate significantly with human judgments of funniness. Moreover, within a set of puns, the distinctiveness measure distinguishes exceptionally funny puns from mediocre ones. Our work is the first, to our knowledge, to integrate a computational model of general language understanding and humor theory to quantitatively predict humor at a fine‐grained level. We present it as an example of a framework for applying models of language processing to understand higher level linguistic and cognitive phenomena.

Discriminative duration modeling for speech recognition with segmental conditional random fields

This paper describes a new approach to modeling duration for LVCSR using SCARF, a toolkit for speech recognition with segmental conditional random fields. We utilize SCARFs ability to integrate long-span, segment-level features to design and test duration models that help discriminate between correct and incorrect word hypotheses. We show that the duration distributions of correct and incorrect word hypotheses differ. Given a word hypothesis in the lattice and its duration, conditional length probabilities are integrated to the SCARF system as duration features. We evaluate three kinds of duration features on Broadcast News: word, pre- and post-pausal durations, and word span confusions. Adding the duration features to SCARF results in an up to 0.3% improvement over a state-of-the-art discriminatively trained baseline of 15.3% WER on a Broadcast News task.