Workshop on Multiword Expressions (2019)
Hear about Verbal Multiword Expressions in the Bulgarian and the Romanian Wordnets Straight from the Horse’s Mouth
Multiword expressions (MWEs) feature prominently in the mental lexicon of native speakers (Jackendoff, 1997) in all languages and domains, from informal to technical contexts (Biber et al., 1999) with about four MWEs being produced per minute of discourse (Glucksberg, 1989). MWEs come in all shapes and forms, including idioms like rock the boat (as cause problems or disturb a situation) and compound nouns like monkey business (as dishonest behaviour). Their accurate detection and understanding may often require more than knowledge about individual words and how they can be combined (Fillmore, 1979), as they may display various degrees of idiosyncrasy, including lexical, syntactic, semantic and statistical (Sag et al., 2002; Baldwin and Kim, 2010), which provide new challenges and opportunities for language processing (Constant et al., 2017). For instance, while for some combinations the meaning can be inferred from their parts like olive oil (oil made of olives) this is not always the case, as in dark horse (meaning an unknown candidate who unexpectedly succeeds), and when processing a sentence some of the challenges are to identify which words form an expression (Ramisch, 2015), and whether the expression is idiomatic (Cordeiro et al., 2019). In this talk I will give an overview of advances on the identification and treatment of multiword expressions, in particular concentrating on techniques for identifying their degree of idiomaticity.
Hear about Verbal Multiword Expressions in the Bulgarian and the Romanian Wordnets Straight from the Horse’s Mouth
Verginica Barbu Mititelu | Ivelina Stoyanova | Svetlozara Leseva | Maria Mitrofan | Tsvetana Dimitrova | Maria Todorova
In this paper we focus on verbal multiword expressions (VMWEs) in Bulgarian and Romanian as reflected in the wordnets of the two languages. The annotation of VMWEs relies on the classification defined within the PARSEME Cost Action. After outlining the properties of various types of VMWEs, a cross-language comparison is drawn, aimed to highlight the similarities and the differences between Bulgarian and Romanian with respect to the lexicalization and distribution of VMWEs. The contribution of this work is in outlining essential features of the description and classification of VMWEs and the cross-language comparison at the lexical level, which is essential for the understanding of the need for uniform annotation guidelines and a viable procedure for validation of the annotation.
This paper reports on the Romanian journalistic corpus annotated with verbal multiword expressions following the PARSEME guidelines. The corpus is sentence split, tokenized, part-of-speech tagged, lemmatized, syntactically annotated and verbal multiword expressions are identified and classified. It offers insights into the frequency of such Romanian word combinations and allows for their characterization. We offer data about the types of verbal multiword expressions in the corpus and some of their characteristics, such as internal structure, diversity in the corpus, average length, productivity of the verbs. This is a language resource that is important per se, as well as for the task of automatic multiword expressions identification, which can be further used in other systems. It was already used as training and test material in the shared tasks for the automatic identification of verbal multiword expressions organized by PARSEME.
We describe work consisting in porting two large German lexical resources into the OntoLex-Lemon model in order to establish complementary interlinkings between them. One resource is OdeNet (Open German WordNet) and the other is a further development of the German version of the MMORPH morphological analyzer. We show how the Multiword Expressions (MWEs) contained in OdeNet can be morphologically specified by the use of the lexical representation and linking features of OntoLex-Lemon, which also support the formulation of restrictions in the usage of such expressions.
We introduce temporally and contextually-aware models for the novel task of predicting unseen but plausible concepts, as conveyed by noun-noun compounds in a time-stamped corpus. We train compositional models on observed compounds, more specifically the composed distributed representations of their constituents across a time-stamped corpus, while giving it corrupted instances (where head or modifier are replaced by a random constituent) as negative evidence. The model captures generalisations over this data and learns what combinations give rise to plausible compounds and which ones do not. After training, we query the model for the plausibility of automatically generated novel combinations and verify whether the classifications are accurate. For our best model, we find that in around 85% of the cases, the novel compounds generated are attested in previously unseen data. An additional estimated 5% are plausible despite not being attested in the recent corpus, based on judgments from independent human raters.
This article describes a dependency-based strategy that uses compositional distributional semantics and cross-lingual word embeddings to translate multiword expressions (MWEs). Our unsupervised approach performs translation as a process of word contextualization by taking into account lexico-syntactic contexts and selectional preferences. This strategy is suited to translate phraseological combinations and phrases whose constituent words are lexically restricted by each other. Several experiments in adjective-noun and verb-object compounds show that mutual contextualization (co-compositionality) clearly outperforms other compositional methods. The paper also contributes with a new freely available dataset of English-Spanish MWEs used to validate the proposed compositional strategy.
This paper presents an exploration of different statistical association measures to automatically identify collocations from corpora in English, Portuguese, and Spanish. To evaluate the impact of the association metrics we manually annotated corpora with three different syntactic patterns of collocations (adjective-noun, verb-object and nominal compounds). We took advantage of the PARSEME 1.1 Shared Task corpora by selecting a subset of 155k tokens in the three referred languages, in which we annotated 1,526 collocations with the corresponding Lexical Functions according to the Meaning-Text Theory. Using the resulting gold-standard, we have carried out a comparison between frequency data and several well-known association measures, both symmetric and asymmetric. The results show that the combination of dependency triples with raw frequency information is as powerful as the best association measures in most syntactic patterns and languages. Furthermore, and despite the asymmetric behaviour of collocations, directional approaches perform worse than the symmetric ones in the extraction of these phraseological combinations.
A substantial body of research has demonstrated that native speakers are sensitive to the frequencies of multiword sequences (MWS). Here, we ask whether and to what extent intermediate-advanced L2 speakers of English can also develop the sensitivity to the statistics of MWS. To this end, we aimed to replicate the MWS frequency effects found for adult native language speakers based on evidence from self-paced reading and sentence recall tasks in an ecologically more valid eye-tracking study. L2 speakers’ sensitivity to MWS frequency was evaluated using generalized linear mixed-effects regression with separate models fitted for each of the four dependent measures. Mixed-effects modeling revealed significantly faster processing of sentences containing MWS compared to sentences containing equivalent control items across all eyetracking measures. Taken together, these findings suggest that, in line with emergentist approaches, MWS are important building blocks of language and that similar mechanisms underlie both native and non-native language processing.
The paper presents the characteristics of the predominant types of MultiWord expressions (MWEs) in the BulTreeBank WordNet – BTB-WN. Their distribution in BTB-WN is discussed with respect to the overall hierarchical organization of the lexical resource. Also, a catena-based modeling is proposed for handling the issues of lexical semantics of MWEs.
Because most multiword expressions (MWEs), especially verbal ones, are semantically non-compositional, their automatic identification in running text is a prerequisite for semantically-oriented downstream applications. However, recent developments, driven notably by the PARSEME shared task on automatic identification of verbal MWEs, show that this task is harder than related tasks, despite recent contributions both in multilingual corpus annotation and in computational models. In this paper, we analyse possible reasons for this state of affairs. They lie in the nature of the MWE phenomenon, as well as in its distributional properties. We also offer a comparative analysis of the state-of-the-art systems, which exhibit particularly strong sensitivity to unseen data. On this basis, we claim that, in order to make strong headway in MWE identification, the community should bend its mind into coupling identification of MWEs with their discovery, via syntactic MWE lexicons. Such lexicons need not necessarily achieve a linguistically complete modelling of MWEs’ behavior, but they should provide minimal morphosyntactic information to cover some potential uses, so as to complement existing MWE-annotated corpora. We define requirements for such minimal NLP-oriented lexicon, and we propose a roadmap for the MWE community driven by these requirements.
Building meaningful representations of noun compounds is not trivial since many of them scarcely appear in the corpus. To that end, composition functions approximate the distributional representation of a noun compound by combining its constituent distributional vectors. In the more general case, phrase embeddings have been trained by minimizing the distance between the vectors representing paraphrases. We compare various types of noun compound representations, including distributional, compositional, and paraphrase-based representations, through a series of tasks and analyses, and with an extensive number of underlying word embeddings. We find that indeed, in most cases, composition functions produce higher quality representations than distributional ones, and they improve with computational power. No single function performs best in all scenarios, suggesting that a joint training objective may produce improved representations.
In this paper we argue that Frame Semantics (Fillmore, 1982) provides a good framework for semantic modelling of adjective-noun collocations. More specifically, the notion of a frame is rich enough to account for nouns from different semantic classes and to model semantic relations that hold between an adjective and a noun in terms of Frame Elements. We have substantiated these findings by considering a sample of adjective-noun collocations from German such as “enger Freund” ‘close friend’ and “starker Regen” ‘heavy rain’. The data sample is taken from different semantic fields identified in the German wordnet GermaNet (Hamp and Feldweg, 1997; Henrich and Hinrichs, 2010). The study is based on the electronic dictionary DWDS (Klein and Geyken, 2010) and uses the collocation extraction tool Wortprofil (Geyken et al., 2009). The FrameNet modelling is based on the online resource available at http://framenet.icsi.berkeley.edu. Since FrameNets are available for a range of typologically different languages, it is feasible to extend the current case study to other languages.
We propose to tackle the problem of verbal multiword expression (VMWE) identification using a neural graph parsing-based approach. Our solution involves encoding VMWE annotations as labellings of dependency trees and, subsequently, applying a neural network to model the probabilities of different labellings. This strategy can be particularly effective when applied to discontinuous VMWEs and, thanks to dense, pre-trained word vector representations, VMWEs unseen during training. Evaluation of our approach on three PARSEME datasets (German, French, and Polish) shows that it allows to achieve performance on par with the previous state-of-the-art (Al Saied et al., 2018).
An idiom is defined as a non-compositional multiword expression, one whose meaning cannot be deduced from the definitions of the component words. This definition does not explicitly define the compositionality of an idiom’s sentiment; this paper aims to determine whether the sentiment of the component words of an idiom is related to the sentiment of that idiom. We use the Dictionary of Affect in Language augmented by WordNet to give each idiom in the Sentiment Lexicon of IDiomatic Expressions (SLIDE) a component-wise sentiment score and compare it to the phrase-level sentiment label crowdsourced by the creators of SLIDE. We find that there is no discernible relation between these two measures of idiom sentiment. This supports the hypothesis that idioms are not compositional for sentiment along with semantics and motivates further work in handling idioms for sentiment analysis.
We report on the ongoing development of IDION, a web resource of richly documented multiword expressions (MWEs) of Modern Greek addressed to the human user and to NLP. IDION contains about 2000 verb MWEs (VMWEs) of which about 850 are fully documented as regards their syntactic flexibility, their semantics and the semantic relations with other VMWEs. Sets of synonymous MWEs are defined in a bottom-up manner revealing the conceptual organization of the MG VMWE domain.
Neologism detection is a key task in the constructing of lexical resources and has wider implications for NLP, however the identification of multiword neologisms has received little attention. In this paper, we show that we can effectively identify the distinction between compositional and non-compositional adjective-noun pairs by using pretrained language models and comparing this with individual word embeddings. Our results show that the use of these models significantly improves over baseline linguistic features, however the combination with linguistic features still further improves the results, suggesting the strength of a hybrid approach.
This article focuses on the lemmatization of multiword expressions (MWEs). We propose a deep encoder-decoder architecture generating for every MWE word its corresponding part in the lemma, based on the internal context of the MWE. The encoder relies on recurrent networks based on (1) the character sequence of the individual words to capture their morphological properties, and (2) the word sequence of the MWE to capture lexical and syntactic properties. The decoder in charge of generating the corresponding part of the lemma for each word of the MWE is based on a classical character-level attention-based recurrent model. Our model is evaluated for Italian, French, Polish and Portuguese and shows good performances except for Polish.
Automatic Term Extraction (ATE) extracts terminology from domain-specific corpora. ATE is used in many NLP tasks, including Computer Assisted Translation, where it is typically applied to individual documents rather than the entire corpus. While corpus-level ATE has been extensively evaluated, it is not obvious how the results transfer to document-level ATE. To fill this gap, we evaluate 16 state-of-the-art ATE methods on full-length documents from three different domains, on both corpus and document levels. Unlike existing studies, our evaluation is more realistic as we take into account all gold terms. We show that no single method is best in corpus-level ATE, but C-Value and KeyConceptRelatendess surpass others in document-level ATE.
Recent developments in deep learning have prompted a surge of interest in the application of multitask and transfer learning to NLP problems. In this study, we explore for the first time, the application of transfer learning (TRL) and multitask learning (MTL) to the identification of Multiword Expressions (MWEs). For MTL, we exploit the shared syntactic information between MWE and dependency parsing models to jointly train a single model on both tasks. We specifically predict two types of labels: MWE and dependency parse. Our neural MTL architecture utilises the supervision of dependency parsing in lower layers and predicts MWE tags in upper layers. In the TRL scenario, we overcome the scarcity of data by learning a model on a larger MWE dataset and transferring the knowledge to a resource-poor setting in another language. In both scenarios, the resulting models achieved higher performance compared to standard neural approaches.
This paper describes the categorisation of Irish MWEs, and the construction of the first version of a lexicon of Irish MWEs for NLP purposes (Ilfhocail, meaning ‘Multiwords’), collected from a number of resources. For the purposes of quality assurance, 530 entries of this lexicon were examined and manually annotated for POS information and MWE category.
Recent initiatives such as the PARSEME shared task allowed the rapid development of MWE identification systems. Many of those are based on recent NLP advances, using neural sequence models that take continuous word representations as input. We study two related questions in neural MWE identification: (a) the use of lemmas and/or surface forms as input features, and (b) the use of word-based or character-based embeddings to represent them. Our experiments on Basque, French, and Polish show that character-based representations yield systematically better results than word-based ones. In some cases, character-based representations of surface forms can be used as a proxy for lemmas, depending on the morphological complexity of the language.