How LSTM Encodes Syntax: Exploring Context Vectors and Semi-Quantization on Natural Text

TL;DR

This paper investigates how LSTM encodes syntactic information in natural language, revealing that internal vectors are quantized and correlated with phrase structure depth, aiding in understanding LSTM's syntactic representations.

Contribution

It demonstrates that LSTM context vectors encode syntactic depth and phrase structure information, with quantization and correlation analyses providing new insights into internal representations.

Findings

Context vectors are approximately binary or ternary, aiding nesting depth counting.

Activations correlate with phrase structure depth such as VP and NP.

Small components of context vectors can predict phrase membership with regularization.

Abstract

Long Short-Term Memory recurrent neural network (LSTM) is widely used and known to capture informative long-term syntactic dependencies. However, how such information are reflected in its internal vectors for natural text has not yet been sufficiently investigated. We analyze them by learning a language model where syntactic structures are implicitly given. We empirically show that the context update vectors, i.e. outputs of internal gates, are approximately quantized to binary or ternary values to help the language model to count the depth of nesting accurately, as Suzgun et al. (2019) recently show for synthetic Dyck languages. For some dimensions in the context vector, we show that their activations are highly correlated with the depth of phrase structures, such as VP and NP. Moreover, with an $L_1$ regularization, we also found that it can accurately predict whether a word is inside a phrase structure or not from a small number of components of the context vector. Even for the case of learning from raw text, context vectors are shown to still correlate well with the phrase structures. Finally, we show that natural clusters of the functional words and the part of speeches that trigger phrases are represented in a small but principal subspace of the context-update vector of LSTM.