Invited Speakers

Allyson Ettinger
(University of Chicago)

“Understanding” and prediction: Controlled examinations of meaning sensitivity in pre-trained models

In recent years, NLP has made what appears to be incredible progress, with performance even surpassing human performance on some benchmarks. How should we interpret these advances? Have these models achieved language “understanding”? Operating on the premise that “understanding” will necessarily involve the capacity to extract and deploy meaning information, in this talk I will discuss a series of projects leveraging targeted tests to examine NLP models’ ability to capture meaning in a systematic fashion. I will first discuss work probing model representations for compositional meaning, with a particular focus on disentangling compositional information from encoding of lexical properties. I’ll then explore models’ ability to extract and use meaning information when executing the basic pre-training task of word prediction in context. In all cases, these investigations apply tests that prioritize control of unwanted cues, so as to target the desired model capabilities with greater precision. The results of these studies suggest that although models show a good deal of sensitivity to word-level information, and to certain semantic and syntactic distinctions, when subjected to controlled tests they show little sign of representing higher-level compositional meaning, or of being able to retain and deploy such information robustly during word prediction. Instead, models show signs of heuristic predictive strategies that are unsurprising given their training, but that differ critically from systematic understanding of meaning. I will discuss potential implications of these findings with respect to the goals of achieving “understanding” with currently dominant pre-training paradigms.

Jacob Andreas
(Massachusetts Institute of Technology)

Models of meaning?

The extent to which language modeling induces representations of meaning—and the broader question of whether it is even in principle possible to learn about meaning from text alone—have remained a subject of ongoing debate across the language sciences. I’ll present some evidence that transformer language models build (rudimentary) structured representations of the meaning of input sentences; that these representations support LMs’ ability to reason about the entities and events described in a discourse; and that they can be modified with predictable effects on downstream language generation. Despite all this, even the largest LMs are prone to glaring semantic errors: they refer to entities that have not yet been mentioned, present contradictory facts, or describe impossible events. By understanding how (and where) LMs build models of meaning, we identify the causes of these errors, and in some cases correct them with extremely small amounts of targeted supervision.