Aurélie Herbelot

Natalia Flechas Manrique, Wanqian Bao, Aurelie Herbelot et al.

Interpretability methods in NLP aim to provide insights into the semantics underlying specific system architectures. Focusing on word embeddings, we present a supervised-learning method that, for a given domain (e.g., sports, professions), identifies a subset of model features that strongly improve prediction of human similarity judgments. We show this method keeps only 20-40% of the original embeddings, for 8 independent semantic domains, and that it retains different feature sets across domains. We then present two approaches for interpreting the semantics of the retained features. The first obtains the scores of the domain words (co-hyponyms) on the first principal component of the retained embeddings, and extracts terms whose co-occurrence with the co-hyponyms tracks these scores' profile. This analysis reveals that humans differentiate e.g. sports based on how gender-inclusive and international they are. The second approach uses the retained sets as variables in a probing task that predicts values along 65 semantically annotated dimensions for a dataset of 535 words. The features retained for professions are best at predicting cognitive, emotional and social dimensions, whereas features retained for fruits or vegetables best predict the gustation (taste) dimension. We discuss implications for alignment between AI systems and human knowledge.

Ludovica Pannitto, Aurélie Herbelot

This paper presents a novel framework for evaluating Neural Language Models' linguistic abilities using a constructionist approach. Not only is the usage-based model in line with the underlying stochastic philosophy of neural architectures, but it also allows the linguist to keep meaning as a determinant factor in the analysis. We outline the framework and present two possible scenarios for its application.

Andrea Bruera, Aurélie Herbelot

In human semantic cognition, proper names (names which refer to individual entities) are harder to learn and retrieve than common nouns. This seems to be the case for machine learning algorithms too, but the linguistic and distributional reasons for this behaviour have not been investigated in depth so far. To tackle this issue, we show that the semantic distinction between proper names and common nouns is reflected in their linguistic distributions by employing an original task for distributional semantics, the Doppelgänger test, an extensive set of models, and a new dataset, the Novel Aficionados dataset. The results indicate that the distributional representations of different individual entities are less clearly distinguishable from each other than those of common nouns, an outcome which intriguingly mirrors human cognition.

Ludovica Pannitto, Aurélie Herbelot

Recurrent Neural Networks (RNNs) have been shown to capture various aspects of syntax from raw linguistic input. In most previous experiments, however, learning happens over unrealistic corpora, which do not reflect the type and amount of data a child would be exposed to. This paper remedies this state of affairs by training a Long Short-Term Memory network (LSTM) over a realistically sized subset of child-directed input. The behaviour of the network is analysed over time using a novel methodology which consists in quantifying the level of grammatical abstraction in the model's generated output (its "babbling"), compared to the language it has been exposed to. We show that the LSTM indeed abstracts new structuresas learning proceeds.

Katrin Erk, Aurelie Herbelot

In this paper, we derive a notion of 'word meaning in context' that characterizes meaning as both intensional and conceptual. We introduce a framework for specifying local as well as global constraints on word meaning in context, together with their interactions, thus modelling the wide range of lexical shifts and ambiguities observed in utterance interpretation. We represent sentence meaning as a 'situation description system', a probabilistic model which takes utterance understanding to be the mental process of describing to oneself one or more situations that would account for an observed utterance. We show how the system can be implemented in practice, and apply it to examples containing various contextualisation phenomena.

Aurelie Herbelot, Marco Baroni

Distributional semantics models are known to struggle with small data. It is generally accepted that in order to learn 'a good vector' for a word, a model must have sufficient examples of its usage. This contradicts the fact that humans can guess the meaning of a word from a few occurrences only. In this paper, we show that a neural language model such as Word2Vec only necessitates minor modifications to its standard architecture to learn new terms from tiny data, using background knowledge from a previously learnt semantic space. We test our model on word definitions and on a nonce task involving 2-6 sentences' worth of context, showing a large increase in performance over state-of-the-art models on the definitional task.

Ionut Sorodoc, Sandro Pezzelle, Aurélie Herbelot et al.

Major advances have recently been made in merging language and vision representations. But most tasks considered so far have confined themselves to the processing of objects and lexicalised relations amongst objects (content words). We know, however, that humans (even pre-school children) can abstract over raw data to perform certain types of higher-level reasoning, expressed in natural language by function words. A case in point is given by their ability to learn quantifiers, i.e. expressions like 'few', 'some' and 'all'. From formal semantics and cognitive linguistics, we know that quantifiers are relations over sets which, as a simplification, we can see as proportions. For instance, in 'most fish are red', most encodes the proportion of fish which are red fish. In this paper, we study how well current language and vision strategies model such relations. We show that state-of-the-art attention mechanisms coupled with a traditional linguistic formalisation of quantifiers gives best performance on the task. Additionally, we provide insights on the role of 'gist' representations in quantification. A 'logical' strategy to tackle the task would be to first obtain a numerosity estimation for the two involved sets and then compare their cardinalities. We however argue that precisely identifying the composition of the sets is not only beyond current state-of-the-art models but perhaps even detrimental to a task that is most efficiently performed by refining the approximate numerosity estimator of the system.