One of the major challenges of machine translation (MT) is ambiguity, which can in some cases be resolved by accompanying context such as images. However, recent work in multimodal MT (MMT) has shown that obtaining improvements from images is challenging, limited not only by the difficulty of building effective cross-modal representations, but also by the lack of specific evaluation and training data. We present a new MMT approach based on a strong text-only MT model, which uses neural adapters, a novel guided self-attention mechanism and which is jointly trained on both visually-conditioned masking and MMT. We also introduce CoMMuTE, a Contrastive Multilingual Multimodal Translation Evaluation set of ambiguous sentences and their possible translations, accompanied by disambiguating images corresponding to each translation. Our approach obtains competitive results compared to strong text-only models on standard English→French, English→German and English→Czech benchmarks and outperforms baselines and state-of-the-art MMT systems by a large margin on our contrastive test set. Our code and CoMMuTE are freely available.

We present SpeechMatrix, a large-scale multilingual corpus of speech-to-speech translations mined from real speech of European Parliament recordings. It contains speech alignments in 136 language pairs with a total of 418 thousand hours of speech. To evaluate the quality of this parallel speech, we train bilingual speech-to-speech translation models on mined data only and establish extensive baseline results on EuroParl-ST, VoxPopuli and FLEURS test sets. Enabled by the multilinguality of SpeechMatrix, we also explore multilingual speech-to-speech translation, a topic which was addressed by few other works. We also demonstrate that model pre-training and sparse scaling using Mixture-of-Experts bring large gains to translation performance. The mined data and models will be publicly released

We introduce dGSLM, the first “textless” model able to generate audio samples of naturalistic spoken dialogues. It uses recent work on unsupervised spoken unit discovery coupled with a dual-tower transformer architecture with cross-attention trained on 2000 hours of two-channel raw conversational audio (Fisher dataset) without any text or labels. We show that our model is able to generate speech, laughter, and other paralinguistic signals in the two channels simultaneously and reproduces more naturalistic and fluid turn taking compared to a text-based cascaded model.1,2

Recent advances in NLP have significantly improved the performance of language models on a variety of tasks. While these advances are largely driven by the availability of large amounts of data and computational power, they also benefit from the development of better training methods and architectures. In this paper, we introduce CamemBERTa, a French DeBERTa model that builds upon the DeBERTaV3 architecture and training objective. We evaluate our model’s performance on a variety of French downstream tasks and datasets, including question answering, part-of-speech tagging, dependency parsing, named entity recognition, and the FLUE benchmark, and compare against CamemBERT, the state-of-the-art monolingual model for French. Our results show that, given the same amount of training tokens, our model outperforms BERT-based models trained with MLM on most tasks. Furthermore, our new model reaches similar or superior performance on downstream tasks compared to CamemBERT, despite being trained on only 30% of its total number of input tokens. In addition to our experimental results, we also publicly release the weights and code implementation of CamemBERTa, making it the first publicly available DeBERTaV3 model outside of the original paper and the first openly available implementation of a DeBERTaV3 training objective.

This paper describes the Inria ALMAnaCH team submission to the WMT 2022 general translation shared task. Participating in the language directions cs,ru,uk→en and cs↔uk, we experiment with the use of a dedicated Latin-script transcription convention aimed at representing all Slavic languages involved in a way that maximises character- and word-level correspondences between them as well as with the English language. Our hypothesis was that bringing the source and target language closer could have a positive impact on machine translation results. We provide multiple comparisons, including bilingual and multilingual baselines, with and without transcription. Initial results indicate that the transcription strategy was not successful, resulting in lower results than baselines. We nevertheless submitted our multilingual, transcribed models as our primary systems, and in this paper provide some indications as to why we got these negative results.

En dépit de leur qualité certaine, les ressources et outils disponibles pour l’analyse du français d’Ancien Régime ne sont plus à même de répondre aux enjeux de la recherche en linguistique et en littérature pour cette période. Après avoir précisément défini le cadre chronologique retenu, nous présentons les corpus mis à disposition et les résultats obtenus avec eux pour plusieurs tâches de TAL fondamentales à l’étude de la langue et de la littérature.

L’apprentissage par transfert basé sur le pré-entraînement de modèles de langue sur une grande quantité de données brutes est devenu la norme pour obtenir des performances état de l’art en TAL. Cependant, la façon dont cette approche devrait être appliquée pour des langues inconnues, qui ne sont couvertes par aucun modèle de langue multilingue à grande échelle et pour lesquelles seule une petite quantité de données brutes est le plus souvent disponible, n’est pas claire. Dans ce travail, en comparant des modèles multilingues et monolingues, nous montrons que de tels modèles se comportent de multiples façons sur des langues inconnues. Certaines langues bénéficient grandement de l’apprentissage par transfert et se comportent de manière similaire à des langues proches riches en ressource, alors que ce n’est manifestement pas le cas pour d’autres. En nous concentrant sur ces dernières, nous montrons dans ce travail que cet échec du transfert est largement lié à l’impact du script que ces langues utilisent. Nous montrons que la translittération de ces langues améliore considérablement le potentiel des larges modèles de langue neuronaux multilingues pour des tâches en aval. Ce résultat indique une piste prometteuse pour rendre ces modèles massivement multilingues utiles pour de nouveaux ensembles de langues absentes des données d’entraînement.

Character-based neural machine translation models have become the reference models for cognate prediction, a historical linguistics task. So far, all linguistic interpretations about latent information captured by such models have been based on external analysis (accuracy, raw results, errors). In this paper, we investigate what probing can tell us about both models and previous interpretations, and learn that though our models store linguistic and diachronic information, they do not achieve it in previously assumed ways.

Static subword tokenization algorithms have been an essential component of recent works on language modeling. However, their static nature results in important flaws that degrade the models’ downstream performance and robustness. In this work, we propose MANTa, a Module for Adaptive Neural TokenizAtion. MANTa is a differentiable tokenizer trained end-to-end with the language model. The resulting system offers a trade-off between the expressiveness of byte-level models and the speed of models trained using subword tokenization. In addition, our tokenizer is highly explainable since it produces an explicit segmentation of sequences into blocks. We evaluate our pre-trained model on several English datasets from different domains as well as on synthetic noise. We find that MANTa improves robustness to character perturbations and out-of-domain data. We then show that MANTa performs comparably to other models on the general-domain GLUE benchmark. Finally, we show that it is considerably faster than strictly byte-level models.

The 2022 Multilingual Representation Learning (MRL) Shared Task was dedicated to clause-level morphology. As the first ever benchmark that defines and evaluates morphology outside its traditional lexical boundaries, the shared task on multilingual clause-level morphology sets the scene for competition across different approaches to morphological modeling, with 3 clause-level sub-tasks: morphological inflection, reinflection and analysis, where systems are required to generate, manipulate or analyze simple sentences centered around a single content lexeme and a set of morphological features characterizing its syntactic clause. This year’s tasks covered eight typologically distinct languages: English, French, German, Hebrew, Russian, Spanish, Swahili and Turkish. The tasks has received submissions of four systems from three teams which were compared to two baselines implementing prominent multilingual learning methods. The results show that modern NLP models are effective in solving morphological tasks even at the clause level. However, there is still room for improvement, especially in the task of morphological analysis.

With the success of large-scale pre-training and multilingual modeling in Natural Language Processing (NLP), recent years have seen a proliferation of large, Web-mined text datasets covering hundreds of languages. We manually audit the quality of 205 language-specific corpora released with five major public datasets (CCAligned, ParaCrawl, WikiMatrix, OSCAR, mC4). Lower-resource corpora have systematic issues: At least 15 corpora have no usable text, and a significant fraction contains less than 50% sentences of acceptable quality. In addition, many are mislabeled or use nonstandard/ambiguous language codes. We demonstrate that these issues are easy to detect even for non-proficient speakers, and supplement the human audit with automatic analyses. Finally, we recommend techniques to evaluate and improve multilingual corpora and discuss potential risks that come with low-quality data releases.

Finding word boundaries in continuous speech is challenging as there is little or no equivalent of a ‘space’ delimiter between words. Popular Bayesian non-parametric models for text segmentation (Goldwater et al., 2006, 2009) use a Dirichlet process to jointly segment sentences and build a lexicon of word types. We introduce DP-Parse, which uses similar principles but only relies on an instance lexicon of word tokens, avoiding the clustering errors that arise with a lexicon of word types. On the Zero Resource Speech Benchmark 2017, our model sets a new speech segmentation state-of-the-art in 5 languages. The algorithm monotonically improves with better input representations, achieving yet higher scores when fed with weakly supervised inputs. Despite lacking a type lexicon, DP-Parse can be pipelined to a language model and learn semantic and syntactic representations as assessed by a new spoken word embedding benchmark. 1

The successes of contextual word embeddings learned by training large-scale language models, while remarkable, have mostly occurred for languages where significant amounts of raw texts are available and where annotated data in downstream tasks have a relatively regular spelling. Conversely, it is not yet completely clear if these models are also well suited for lesser-resourced and more irregular languages. We study the case of Old French, which is in the interesting position of having relatively limited amount of available raw text, but enough annotated resources to assess the relevance of contextual word embedding models for downstream NLP tasks. In particular, we use POS-tagging and dependency parsing to evaluate the quality of such models in a large array of configurations, including models trained from scratch from small amounts of raw text and models pre-trained on other languages but fine-tuned on Medieval French data.

Progress in sentence simplification has been hindered by a lack of labeled parallel simplification data, particularly in languages other than English. We introduce MUSS, a Multilingual Unsupervised Sentence Simplification system that does not require labeled simplification data. MUSS uses a novel approach to sentence simplification that trains strong models using sentence-level paraphrase data instead of proper simplification data. These models leverage unsupervised pretraining and controllable generation mechanisms to flexibly adjust attributes such as length and lexical complexity at inference time. We further present a method to mine such paraphrase data in any language from Common Crawl using semantic sentence embeddings, thus removing the need for labeled data. We evaluate our approach on English, French, and Spanish simplification benchmarks and closely match or outperform the previous best supervised results, despite not using any labeled simplification data. We push the state of the art further by incorporating labeled simplification data.

Spelling normalisation is a useful step in the study and analysis of historical language texts, whether it is manual analysis by experts or automatic analysis using downstream natural language processing (NLP) tools. Not only does it help to homogenise the variable spelling that often exists in historical texts, but it also facilitates the use of off-the-shelf contemporary NLP tools, if contemporary spelling conventions are used for normalisation. We present FREEMnorm, a new benchmark for the normalisation of Early Modern French (from the 17th century) into contemporary French and provide a thorough comparison of three different normalisation methods: ABA, an alignment-based approach and MT-approaches, (both statistical and neural), including extensive parameter searching, which is often missing in the normalisation literature.

anguage models for historical states of language are becoming increasingly important to allow the optimal digitisation and analysis of old textual sources. Because these historical states are at the same time more complex to process and more scarce in the corpora available, this paper presents recent efforts to overcome this difficult situation. These efforts include producing a corpus, creating the model, and evaluating it with an NLP task currently used by scholars in other ongoing projects.

The need for large corpora raw corpora has dramatically increased in recent years with the introduction of transfer learning and semi-supervised learning methods to Natural Language Processing. And while there have been some recent attempts to manually curate the amount of data necessary to train large language models, the main way to obtain this data is still through automatic web crawling. In this paper we take the existing multilingual web corpus OSCAR and its pipeline Ungoliant that extracts and classifies data from Common Crawl at the line level, and propose a set of improvements and automatic annotations in order to produce a new document-oriented version of OSCAR that could prove more suitable to pre-train large generative language models as well as hopefully other applications in Natural Language Processing and Digital Humanities.

Detecting divergences in the applications of the law (where the same legal text is applied differently by two rulings) is an important task. It is the mission of the French Cour de Cassation. The first step in the detection of divergences is to detect similar cases, which is currently done manually by experts. They rely on summarised versions of the rulings (syntheses and keyword sequences), which are currently produced manually and are not available for all rulings. There is also a high degree of variability in the keyword choices and the level of granularity used. In this article, we therefore aim to provide automatic tools to facilitate the search for similar rulings. We do this by (i) providing automatic keyword sequence generation models, which can be used to improve the coverage of the analysis, and (ii) providing measures of similarity based on the available texts and augmented with predicted keyword sequences. Our experiments show that the predictions improve correlations of automatically obtained similarities against our specially colelcted human judgments of similarity.

We present a new approach to perform zero-shot cross-modal transfer between speech and text for translation tasks. Multilingual speech and text are encoded in a joint fixed-size representation space. Then, we compare different approaches to decode these multimodal and multilingual fixed-size representations, enabling zero-shot translation between languages and modalities. All our models are trained without the need of cross-modal labeled translation data.Despite a fixed-size representation, we achieve very competitive results on several text and speech translation tasks. In particular, we significantly improve the state-of-the-art for zero-shot speech translation on Must-C. Incorporating a speech decoder in our framework, we introduce the first results for zero-shot direct speech-to-speech and text-to-speech translation.

Multilingual pretrained language models have demonstrated remarkable zero-shot cross-lingual transfer capabilities. Such transfer emerges by fine-tuning on a task of interest in one language and evaluating on a distinct language, not seen during the fine-tuning. Despite promising results, we still lack a proper understanding of the source of this transfer. Using a novel layer ablation technique and analyses of the model’s internal representations, we show that multilingual BERT, a popular multilingual language model, can be viewed as the stacking of two sub-networks: a multilingual encoder followed by a task-specific language-agnostic predictor. While the encoder is crucial for cross-lingual transfer and remains mostly unchanged during fine-tuning, the task predictor has little importance on the transfer and can be reinitialized during fine-tuning. We present extensive experiments with three distinct tasks, seventeen typologically diverse languages and multiple domains to support our hypothesis.

Recent impressive improvements in NLP, largely based on the success of contextual neural language models, have been mostly demonstrated on at most a couple dozen high- resource languages. Building language mod- els and, more generally, NLP systems for non- standardized and low-resource languages remains a challenging task. In this work, we fo- cus on North-African colloquial dialectal Arabic written using an extension of the Latin script, called NArabizi, found mostly on social media and messaging communication. In this low-resource scenario with data display- ing a high level of variability, we compare the downstream performance of a character-based language model on part-of-speech tagging and dependency parsing to that of monolingual and multilingual models. We show that a character-based model trained on only 99k sentences of NArabizi and fined-tuned on a small treebank of this language leads to performance close to those obtained with the same architecture pre- trained on large multilingual and monolingual models. Confirming these results a on much larger data set of noisy French user-generated content, we argue that such character-based language models can be an asset for NLP in low-resource and high language variability set- tings.

Coupled with the availability of large scale datasets, deep learning architectures have enabled rapid progress on the Question Answering task. However, most of those datasets are in English, and the performances of state-of-the-art multilingual models are significantly lower when evaluated on non-English data. Due to high data collection costs, it is not realistic to obtain annotated data for each language one desires to support. We propose a method to improve the Cross-lingual Question Answering performance without requiring additional annotated data, leveraging Question Generation models to produce synthetic samples in a cross-lingual fashion. We show that the proposed method allows to significantly outperform the baselines trained on English data only. We report a new state-of-the-art on four datasets: MLQA, XQuAD, SQuAD-it and PIAF (fr).

Transfer learning based on pretraining language models on a large amount of raw data has become a new norm to reach state-of-the-art performance in NLP. Still, it remains unclear how this approach should be applied for unseen languages that are not covered by any available large-scale multilingual language model and for which only a small amount of raw data is generally available. In this work, by comparing multilingual and monolingual models, we show that such models behave in multiple ways on unseen languages. Some languages greatly benefit from transfer learning and behave similarly to closely related high resource languages whereas others apparently do not. Focusing on the latter, we show that this failure to transfer is largely related to the impact of the script used to write such languages. We show that transliterating those languages significantly improves the potential of large-scale multilingual language models on downstream tasks. This result provides a promising direction towards making these massively multilingual models useful for a new set of unseen languages.

Cognate prediction and proto-form reconstruction are key tasks in computational historical linguistics that rely on the study of sound change regularity. Solving these tasks appears to be very similar to machine translation, though methods from that field have barely been applied to historical linguistics. Therefore, in this paper, we investigate the learnability of sound correspondences between a proto-language and daughter languages for two machine-translation-inspired models, one statistical, the other neural. We first carry out our experiments on plausible artificial languages, without noise, in order to study the role of each parameter on the algorithms respective performance under almost perfect conditions. We then study real languages, namely Latin, Italian and Spanish, to see if those performances generalise well. We show that both model types manage to learn sound changes despite data scarcity, although the best performing model type depends on several parameters such as the size of the training data, the ambiguity, and the prediction direction.

Les modèles de langue neuronaux contextuels sont désormais omniprésents en traitement automatique des langues. Jusqu’à récemment, la plupart des modèles disponibles ont été entraînés soit sur des données en anglais, soit sur la concaténation de données dans plusieurs langues. L’utilisation pratique de ces modèles — dans toutes les langues sauf l’anglais — était donc limitée. La sortie récente de plusieurs modèles monolingues fondés sur BERT (Devlin et al., 2019), notamment pour le français, a démontré l’intérêt de ces modèles en améliorant l’état de l’art pour toutes les tâches évaluées. Dans cet article, à partir d’expériences menées sur CamemBERT (Martin et al., 2019), nous montrons que l’utilisation de données à haute variabilité est préférable à des données plus uniformes. De façon plus surprenante, nous montrons que l’utilisation d’un ensemble relativement petit de données issues du web (4Go) donne des résultats aussi bons que ceux obtenus à partir d’ensembles de données plus grands de deux ordres de grandeurs (138Go).

We introduce the first treebank for a romanized user-generated content variety of Algerian, a North-African Arabic dialect known for its frequent usage of code-switching. Made of 1500 sentences, fully annotated in morpho-syntax and Universal Dependency syntax, with full translation at both the word and the sentence levels, this treebank is made freely available. It is supplemented with 50k unlabeled sentences collected from Common Crawl and web-crawled data using intensive data-mining techniques. Preliminary experiments demonstrate its usefulness for POS tagging and dependency parsing. We believe that what we present in this paper is useful beyond the low-resource language community. This is the first time that enough unlabeled and annotated data is provided for an emerging user-generated content dialectal language with rich morphology and code switching, making it an challenging test-bed for most recent NLP approaches.

We use the multilingual OSCAR corpus, extracted from Common Crawl via language classification, filtering and cleaning, to train monolingual contextualized word embeddings (ELMo) for five mid-resource languages. We then compare the performance of OSCAR-based and Wikipedia-based ELMo embeddings for these languages on the part-of-speech tagging and parsing tasks. We show that, despite the noise in the Common-Crawl-based OSCAR data, embeddings trained on OSCAR perform much better than monolingual embeddings trained on Wikipedia. They actually equal or improve the current state of the art in tagging and parsing for all five languages. In particular, they also improve over multilingual Wikipedia-based contextual embeddings (multilingual BERT), which almost always constitutes the previous state of the art, thereby showing that the benefit of a larger, more diverse corpus surpasses the cross-lingual benefit of multilingual embedding architectures.

In order to simplify a sentence, human editors perform multiple rewriting transformations: they split it into several shorter sentences, paraphrase words (i.e. replacing complex words or phrases by simpler synonyms), reorder components, and/or delete information deemed unnecessary. Despite these varied range of possible text alterations, current models for automatic sentence simplification are evaluated using datasets that are focused on a single transformation, such as lexical paraphrasing or splitting. This makes it impossible to understand the ability of simplification models in more realistic settings. To alleviate this limitation, this paper introduces ASSET, a new dataset for assessing sentence simplification in English. ASSET is a crowdsourced multi-reference corpus where each simplification was produced by executing several rewriting transformations. Through quantitative and qualitative experiments, we show that simplifications in ASSET are better at capturing characteristics of simplicity when compared to other standard evaluation datasets for the task. Furthermore, we motivate the need for developing better methods for automatic evaluation using ASSET, since we show that current popular metrics may not be suitable when multiple simplification transformations are performed.

Pretrained language models are now ubiquitous in Natural Language Processing. Despite their success, most available models have either been trained on English data or on the concatenation of data in multiple languages. This makes practical use of such models –in all languages except English– very limited. In this paper, we investigate the feasibility of training monolingual Transformer-based language models for other languages, taking French as an example and evaluating our language models on part-of-speech tagging, dependency parsing, named entity recognition and natural language inference tasks. We show that the use of web crawled data is preferable to the use of Wikipedia data. More surprisingly, we show that a relatively small web crawled dataset (4GB) leads to results that are as good as those obtained using larger datasets (130+GB). Our best performing model CamemBERT reaches or improves the state of the art in all four downstream tasks.

Diachronic lexical information is not only important in the field of historical linguistics, but is also increasingly used in NLP, most recently for machine translation of low resource languages. Therefore, there is a need for fine-grained, large-coverage and accurate etymological lexical resources. In this paper, we propose a set of guidelines to generate such resources, for each step of the life-cycle of an etymological lexicon: creation, update, evaluation, dissemination, and exploitation. To illustrate the guidelines, we introduce EtymDB 2.0, an etymological database automatically generated from the Wiktionary, which contains 1.8 million lexemes, linked by more than 700,000 fine-grained etymological relations, across 2,536 living and dead languages. We also introduce use cases for which EtymDB 2.0 could represent a key resource, such as phylogenetic tree generation, low resource machine translation or medieval languages study.

In this paper we describe our work on the development and enrichment of OFrLex, a freely available, large-coverage morphological and syntactic Old French lexicon. We rely on several heterogeneous language resources to extract structured and exploitable information. The extraction follows a semi-automatic procedure with substantial manual steps to respond to difficulties encountered while aligning lexical entries from distinct language resources. OFrLex aims at improving natural language processing tasks on Old French such as part-of-speech tagging and dependency parsing. We provide quantitative information on OFrLex and discuss its reliability. We also describe and evaluate a semi-automatic, word-embedding-based lexical enrichment process aimed at increasing the accuracy of the resource. Results of this extension technique will be manually validated in the near future, a step that will take advantage of OFrLex’s viewing, searching and editing interface, which is already accessible online.

The French TreeBank developed at the University Paris 7 is the main source of morphosyntactic and syntactic annotations for French. However, it does not include explicit information related to named entities, which are among the most useful information for several natural language processing tasks and applications. Moreover, no large-scale French corpus with named entity annotations contain referential information, which complement the type and the span of each mention with an indication of the entity it refers to. We have manually annotated the French TreeBank with such information, after an automatic pre-annotation step. We sketch the underlying annotation guidelines and we provide a few figures about the resulting annotations.

Text simplification aims at making a text easier to read and understand by simplifying grammar and structure while keeping the underlying information identical. It is often considered an all-purpose generic task where the same simplification is suitable for all; however multiple audiences can benefit from simplified text in different ways. We adapt a discrete parametrization mechanism that provides explicit control on simplification systems based on Sequence-to-Sequence models. As a result, users can condition the simplifications returned by a model on attributes such as length, amount of paraphrasing, lexical complexity and syntactic complexity. We also show that carefully chosen values of these attributes allow out-of-the-box Sequence-to-Sequence models to outperform their standard counterparts on simplification benchmarks. Our model, which we call ACCESS (as shorthand for AudienCe-CEntric Sentence Simplification), establishes the state of the art at 41.87 SARI on the WikiLarge test set, a +1.42 improvement over the best previously reported score.

This paper investigates the impact of different types and size of training corpora on language models. By asking the fundamental question of quality versus quantity, we compare four French corpora by pre-training four different ELMos and evaluating them on dependency parsing, POS-tagging and Named Entities Recognition downstream tasks. We present and asses the relevance of a new balanced French corpus, CaBeRnet, that features a representative range of language usage, including a balanced variety of genres (oral transcriptions, newspapers, popular magazines, technical reports, fiction, academic texts), in oral and written styles. We hypothesize that a linguistically representative corpus will allow the language models to be more efficient, and therefore yield better evaluation scores on different evaluation sets and tasks. This paper offers three main contributions: (1) two newly built corpora: (a) CaBeRnet, a French Balanced Reference Corpus and (b) CBT-fr a domain-specific corpus having both oral and written style in youth literature, (2) five versions of ELMo pre-trained on differently built corpora, and (3) a whole array of computational results on downstream tasks that deepen our understanding of the effects of corpus balance and register in NLP evaluation.

Language model-based pre-trained representations have become ubiquitous in natural language processing. They have been shown to significantly improve the performance of neural models on a great variety of tasks. However, it remains unclear how useful those general models can be in handling non-canonical text. In this article, focusing on User Generated Content (UGC), we study the ability of BERT to perform lexical normalisation. Our contribution is simple: by framing lexical normalisation as a token prediction task, by enhancing its architecture and by carefully fine-tuning it, we show that BERT can be a competitive lexical normalisation model without the need of any UGC resources aside from 3,000 training sentences. To the best of our knowledge, it is the first work done in adapting and analysing the ability of this model to handle noisy UGC data.

Nous décrivons dans cet article notre travail de développement d’un lexique morphologique et syntaxique à grande échelle de l’ancien français pour le traitement automatique des langues. Nous nous sommes appuyés sur des ressources dictionnairiques et lexicales dans lesquelles l’extraction d’informations structurées et exploitables a nécessité des développements spécifiques. De plus, la mise en correspondance d’informations provenant de ces différentes sources a soulevé des difficultés. Nous donnons quelques indications quantitatives sur le lexique obtenu, et discutons de sa fiabilité dans sa version actuelle et des perspectives d’amélioration permises par l’existence d’une première version, notamment au travers de l’analyse automatique de données textuelles.

BERT is a recent language representation model that has surprisingly performed well in diverse language understanding benchmarks. This result indicates the possibility that BERT networks capture structural information about language. In this work, we provide novel support for this claim by performing a series of experiments to unpack the elements of English language structure learned by BERT. Our findings are fourfold. BERT’s phrasal representation captures the phrase-level information in the lower layers. The intermediate layers of BERT compose a rich hierarchy of linguistic information, starting with surface features at the bottom, syntactic features in the middle followed by semantic features at the top. BERT requires deeper layers while tracking subject-verb agreement to handle long-term dependency problem. Finally, the compositional scheme underlying BERT mimics classical, tree-like structures.

In this paper, we present the details of the neural dependency parser and the neural tagger submitted by our team ‘ParisNLP’ to the CoNLL 2018 Shared Task on parsing from raw text to Universal Dependencies. We augment the deep Biaffine (BiAF) parser (Dozat and Manning, 2016) with novel features to perform competitively: we utilize an indomain version of ELMo features (Peters et al., 2018) which provide context-dependent word representations; we utilize disambiguated, embedded, morphosyntactic features from lexicons (Sagot, 2018), which complements the existing feature set. Henceforth, we call our system ‘ELMoLex’. In addition to incorporating character embeddings, ELMoLex benefits from pre-trained word vectors, ELMo and morphosyntactic features (whenever available) to correctly handle rare or unknown words which are prevalent in languages with complex morphology. ELMoLex ranked 11th by Labeled Attachment Score metric (70.64%), Morphology-aware LAS metric (55.74%) and ranked 9th by Bilexical dependency metric (60.70%).

Les ressources lexicales électroniques ne contiennent quasiment jamais d’informations étymologiques. De telles informations, convenablement formalisées, permettraient pourtant de développer des outils automatiques au service de la linguistique historique et comparative, ainsi que d’améliorer significativement le traitement automatique de langues anciennes. Nous décrivons ici le processus que nous avons mis en œuvre pour extraire des données étymologiques à partir des notices étymologiques du wiktionary, rédigées en anglais. Nous avons ainsi produit une base multilingue de près d’un million de lexèmes et une base de plus d’un demi-million de relations étymologiques entre lexèmes.

We present the ParisNLP entry at the UD CoNLL 2017 parsing shared task. In addition to the UDpipe models provided, we built our own data-driven tokenization models, sentence segmenter and lexicon-based morphological analyzers. All of these were used with a range of different parsing models (neural or not, feature-rich or not, transition or graph-based, etc.) and the best combination for each language was selected. Unfortunately, a glitch in the shared task’s Matrix led our model selector to run generic, weakly lexicalized models, tailored for surprise languages, instead of our dataset-specific models. Because of this #ParsingTragedy, we officially ranked 27th, whereas our real models finally unofficially ranked 6th.

We focus on the identification of omission in statement pairs. We compare three annotation schemes, namely two different crowdsourcing schemes and manual expert annotation. We show that the simplest of the two crowdsourcing approaches yields a better annotation quality than the more complex one. We use a dedicated classifier to assess whether the annotators’ behavior can be explained by straightforward linguistic features. The classifier benefits from a modeling that uses lexical information beyond length and overlap measures. However, for our task, we argue that expert and not crowdsourcing-based annotation is the best compromise between annotation cost and quality.

In this paper, we present ongoing work for developing language resources and basic NLP tools for an undocumented variety of Romansh, in the context of a language documentation and language acquisition project. Our tools are meant to improve the speed and reliability of corpus annotations for noisy data involving large amounts of code-switching, occurrences of child-speech and orthographic noise. Being able to increase the efficiency of language resource development for language documentation and acquisition research also constitutes a step towards solving the data sparsity issues with which researchers have been struggling.

Neural part-of-speech tagging has achieved competitive results with the incorporation of character-based and pre-trained word embeddings. In this paper, we show that a state-of-the-art bi-LSTM tagger can benefit from using information from morphosyntactic lexicons as additional input. The tagger, trained on several dozen languages, shows a consistent, average improvement when using lexical information, even when also using character-based embeddings, thus showing the complementarity of the different sources of lexical information. The improvements are particularly important for the smaller datasets.

Nous présentons des travaux récents réalisés autour de MElt, système discriminant d’étiquetage en parties du discours. MElt met l’accent sur l’exploitation optimale d’informations lexicales externes pour améliorer les performances des étiqueteurs par rapport aux modèles entraînés seulement sur des corpus annotés. Nous avons entraîné MElt sur plus d’une quarantaine de jeux de données couvrant plus d’une trentaine de langues. Comparé au système état-de-l’art MarMoT, MElt obtient en moyenne des résultats légèrement moins bons en l’absence de lexique externe, mais meilleurs lorsque de telles ressources sont disponibles, produisant ainsi des étiqueteurs état-de-l’art pour plusieurs langues.

User-generated content presents many challenges for its automatic processing. While many of them do come from out-of-vocabulary effects, others spawn from different linguistic phenomena such as unusual syntax. In this work we present a French three-domain data set made up of question headlines from a cooking forum, game chat logs and associated forums from two popular online games (MINECRAFT & LEAGUE OF LEGENDS). We chose these domains because they encompass different degrees of lexical and syntactic compliance with canonical language. We conduct an automatic and manual evaluation of the difficulties of processing these domains for part-of-speech prediction, and introduce a pilot study to determine whether dependency analysis lends itself well to annotate these data. We also discuss the development cost of our data set.

We introduce DeLex, a freely-avaible, large-scale and linguistically grounded morphological lexicon for German developed within the Alexina framework. We extracted lexical information from the German wiktionary and developed a morphological inflection grammar for German, based on a linguistically sound model of inflectional morphology. Although the developement of DeLex involved some manual work, we show that is represents a good tradeoff between development cost, lexical coverage and resource accuracy.

In this paper, we describe and evaluate an unsupervised method for acquiring pairs of lexical entries belonging to the same morphological family, i.e., derivationally related words, starting from a purely inflectional lexicon. Our approach relies on transformation rules that relate lexical entries with the one another, and which are automatically extracted from the inflected lexicon based on surface form analogies and on part-of-speech information. It is generic enough to be applied to any language with a mainly concatenative derivational morphology. Results were obtained and evaluated on English, French, German and Spanish. Precision results are satisfying, and our French results favorably compare with another resource, although its construction relied on manually developed lexicographic information whereas our approach only requires an inflectional lexicon.

The Asfalda project aims to develop a French corpus with frame-based semantic annotations and automatic tools for shallow semantic analysis. We present the first part of the project: focusing on a set of notional domains, we delimited a subset of English frames, adapted them to French data when necessary, and developed the corresponding French lexicon. We believe that working domain by domain helped us to enforce the coherence of the resulting resource, and also has the advantage that, though the number of frames is limited (around a hundred), we obtain full coverage within a given domain.

This paper describes YaMTG (Yet another Multilingual Translation Graph), a new open-source heavily multilingual translation database (over 664 languages represented) built using several sources, namely various wiktionaries and the OPUS parallel corpora (Tiedemann, 2009). We detail the translation extraction process for 21 wiktionary language editions, and provide an evaluation of the translations contained in YaMTG.

In this paper, we describe our generic approach for transferring part-of-speech annotations from a resourced language towards an etymologically closely related non-resourced language, without using any bilingual (i.e., parallel) data. We first induce a translation lexicon from monolingual corpora, based on cognate detection followed by cross-lingual contextual similarity. Second, POS information is transferred from the resourced language along translation pairs to the non-resourced language and used for tagging the corpus. We evaluate our methods on three language families, consisting of five Romance languages, three Germanic languages and five Slavic languages. We obtain tagging accuracies of up to 91.6%.

The automatic development of semantic resources constitutes an important challenge in the NLP community. The methods used generally exploit existing large-scale resources, such as Princeton WordNet, often combined with information extracted from multilingual resources and parallel corpora. In this paper we show how Cross-Lingual Word Sense Disambiguation can be applied to wordnet development. We apply the proposed method to WOLF, a free wordnet for French still under construction, in order to fill synsets that did not contain any literal yet and increase its coverage.

We present some evaluation results for four French syntactic lexica, obtained through their conversion to the Alexina format used by the Lefff lexicon, and their integration within the large-coverage TAG-based FRMG parser. The evaluations are run on two test corpora, annotated with two distinct annotation formats, namely EASy/Passage chunks and relations and CoNLL dependencies. The information provided by the evaluation results provide valuable feedback about the four lexica. Moreover, when coupled with error mining techniques, they allow us to identify how these lexica might be improved.

In this article, we present a distributional analysis method for extracting nominalization relations from monolingual corpora. The acquisition method makes use of distributional and morphological information to select nominalization candidates. We explain how the learning is performed on a dependency annotated corpus and describe the nominalization results. Furthermore, we show how these results served to enrich an existing lexical resource, the WOLF (Wordnet Libre du Franc¸ais). We present the techniques that we developed in order to integrate the new information into WOLF, based on both its structure and content. Finally, we evaluate the validity of the automatically obtained information and the correctness of its integration into the semantic resource. The method proved to be useful for boosting the coverage of WOLF and presents the advantage of filling verbal synsets, which are particularly difficult to handle due to the high level of verbal polysemy.

Named entity recognition, which focuses on the identification of the span and type of named entity mentions in texts, has drawn the attention of the NLP community for a long time. However, many real-life applications need to know which real entity each mention refers to. For such a purpose, often refered to as entity resolution and linking, an inventory of entities is required in order to constitute a reference. In this paper, we describe how we extracted such a resource for French from freely available resources (the French Wikipedia and the GeoNames database). We describe the results of an instrinsic evaluation of the resulting entity database, named Aleda, as well as those of a task-based evaluation in the context of a named entity detection system. We also compare it with the NLGbAse database (Charton and Torres-Moreno, 2010), a resource with similar objectives.

Automatic approaches to creating and extending wordnets, which have become very popular in the past decade, inadvertently result in noisy synsets. This is why we propose an approach to detect synset outliers in order to eliminate the noise and improve accuracy of the developed wordnets, so that they become more useful lexico-semantic resources for natural language applications. The approach compares the words that appear in the synset and its surroundings with the contexts of the literals in question they are used in based on large monolingual corpora. By fine-tuning the outlier threshold we can influence how many outlier candidates will be eliminated. Although the proposed approach is language-independent we test it on Slovene and French that were created automatically from bilingual resources and contain plenty of disambiguation errors. Manual evaluation of the results shows that by applying a threshold similar to the estimated error rate in the respective wordnets, 67% of the proposed outlier candidates are indeed incorrect for French and a 64% for Slovene. This is a big improvement compared to the estimated overall error rates in the resources, which are 12% for French and 15% for Slovene.

In this paper, we propose a simple methodology for building or extending wordnets using easily extractible lexical knowledge from Wiktionary and Wikipedia. This method relies on a large multilingual translation/synonym graph in many languages as well as synset-aligned wordnets. It guesses frequent and polysemous literals that are difficult to find using other methods by looking at back-translations in the graph, showing that the use of a heavily multilingual lexicon can be a way to mitigate the lack of wide coverage bilingual lexicon for wordnet creation or extension. We evaluate our approach on French by applying it for extending WOLF, a freely available French wordnet.

Cet article est une prise de position concernant les plate-formes de type Amazon Mechanical Turk, dont l’utilisation est en plein essor depuis quelques années dans le traitement automatique des langues. Ces plateformes de travail en ligne permettent, selon le discours qui prévaut dans les articles du domaine, de faire développer toutes sortes de ressources linguistiques de qualité, pour un prix imbattable et en un temps très réduit, par des gens pour qui il s’agit d’un passe-temps. Nous allons ici démontrer que la situation est loin d’être aussi idéale, que ce soit sur le plan de la qualité, du prix, du statut des travailleurs ou de l’éthique. Nous rappellerons ensuite les solutions alternatives déjà existantes ou proposées. Notre but est ici double : informer les chercheurs, afin qu’ils fassent leur choix en toute connaissance de cause, et proposer des solutions pratiques et organisationnelles pour améliorer le développement de nouvelles ressources linguistiques en limitant les risques de dérives éthiques et légales, sans que cela se fasse au prix de leur coût ou de leur qualité.

Pour la plupart des langues utilisant l’alphabet latin, le découpage d’un texte selon les espaces et les symboles de ponctuation est une bonne approximation d’un découpage en unités lexicales. Bien que cette approximation cache de nombreuses difficultés, elles sont sans comparaison avec celles que l’on rencontre lorsque l’on veut traiter des langues qui, comme le chinois mandarin, n’utilisent pas l’espace. Un grand nombre de systèmes de segmentation ont été proposés parmi lesquels certains adoptent une approche non-supervisée motivée linguistiquement. Cependant les méthodes d’évaluation communément utilisées ne rendent pas compte de toutes les propriétés de tels systèmes. Dans cet article, nous montrons qu’un modèle simple qui repose sur une reformulation en termes d’entropie d’une hypothèse indépendante de la langue énoncée par Harris (1955), permet de segmenter un corpus et d’en extraire un lexique. Testé sur le corpus de l’Academia Sinica, notre système permet l’induction d’une segmentation et d’un lexique qui ont de bonnes propriétés intrinsèques et dont les caractéristiques sont similaires à celles du lexique sous-jacent au corpus segmenté manuellement. De plus, on constate une certaine corrélation entre les résultats du modèle de segmentation et les structures syntaxiques fournies par une sous-partie arborée corpus.

La détection et le typage des entités nommées sont des tâches pour lesquelles ont été développés à la fois des systèmes symboliques et probabilistes. Nous présentons les résultats d’une expérience visant à faire interagir le système à base de règles NP, développé sur des corpus provenant de l’AFP, intégrant la base d’entités Aleda et qui a une bonne précision, et le système LIANE, entraîné sur des transcriptions de l’oral provenant du corpus ESTER et qui a un bon rappel. Nous montrons qu’on peut adapter à un nouveau type de corpus, de manière non supervisée, un système probabiliste tel que LIANE grâce à des corpus volumineux annotés automatiquement par NP. Cette adaptation ne nécessite aucune annotation manuelle supplémentaire et illustre la complémentarité des méthodes numériques et symboliques pour la résolution de tâches linguistiques.

Après une brève analyse linguistique des adjectifs dénominaux en français, nous décrivons le processus automatique que nous avons mis en place à partir de lexiques et de corpus volumineux pour construire un lexique d’adjectifs dénominaux dérivés de manière régulière. Nous estimons à la fois la précision et la couverture du lexique dérivationnel obtenu. À terme, ce lexique librement disponible aura été validé manuellement et contiendra également les adjectifs dénominaux à base supplétive.

Nous présentons une nouvelle version de PerLex, lexique morphologique du persan, une version corrigée et partiellement réannotée du corpus étiqueté BijanKhan (BijanKhan, 2004) et MEltfa, un nouvel étiqueteur morphosyntaxique librement disponible pour le persan. Après avoir développé une première version de PerLex (Sagot & Walther, 2010), nous en proposons donc ici une version améliorée. Outre une validation manuelle partielle, PerLex 2 repose désormais sur un inventaire de catégories linguistiquement motivé. Nous avons également développé une nouvelle version du corpus BijanKhan : elle contient des corrections significatives de la tokenisation ainsi qu’un réétiquetage à l’aide des nouvelles catégories. Cette nouvelle version du corpus a enfin été utilisée pour l’entraînement de MEltfa, notre étiqueteur morphosyntaxique pour le persan librement disponible, s’appuyant à la fois sur ce nouvel inventaire de catégories, sur PerLex 2 et sur le système d’étiquetage MElt (Denis & Sagot, 2009).

Cet article présente MEltfr, un étiqueteur morpho-syntaxique automatique du français. Il repose sur un modèle probabiliste séquentiel qui bénéficie d’informations issues d’un lexique exogène, à savoir le Lefff. Evalué sur le FTB, MEltfr atteint un taux de précision de 97.75% (91.36% sur les mots inconnus) sur un jeu de 29 étiquettes. Ceci correspond à une diminution du taux d’erreur de 18% (36.1% sur les mots inconnus) par rapport au même modèle sans couplage avec le Lefff. Nous étudions plus en détail la contribution de cette ressource, au travers de deux séries d’expériences. Celles-ci font apparaître en particulier que la contribution des traits issus du Lefff est de permettre une meilleure couverture, ainsi qu’une modélisation plus fine du contexte droit des mots.

Nous présentons PerLex, un lexique morphologique du persan à large couverture et librement disponible, accompagné d’une chaîne de traitements de surface pour cette langue. Nous décrivons quelques caractéristiques de la morphologie du persan, et la façon dont nous l’avons représentée dans le formalisme lexical Alexina, sur lequel repose PerLex. Nous insistons sur la méthodologie que nous avons employée pour construire les entrées lexicales à partir de diverses sources, ainsi que sur les problèmes liés à la normalisation typographique. Le lexique obtenu a une couverture satisfaisante sur un corpus de référence, et devrait donc constituer un bon point de départ pour le développement d’un lexique syntaxique du persan.

Certaines ponctuations fortes sont « abusivement » utilisées à la place de ponctuations faibles, débouchant sur des phrases graphiques qui ne sont pas des phrases grammaticales. Cet article présente une étude sur corpus de ce phénomène et une ébauche d’outil pour repérer automatiquement les ponctuations fortes abusives.

Cet article aborde le phénomène de l’incomplétude des ressources lexicales, c’est-à-dire la problématique des inconnus, dans un contexte de traitement automatique. Nous proposons tout d’abord une définition opérationnelle de la notion d’inconnu. Nous décrivons ensuite une typologie des différentes classes d’inconnus, motivée par des considérations linguistiques et applicatives ainsi que par l’annotation des inconnus d’un petit corpus selon notre typologie. Cette typologie sera mise en oeuvre et validée par l’annotation d’un corpus important de l’Agence France-Presse dans le cadre du projet EDyLex.

Nous présentons NP, un système de reconnaissance d’entités nommées. Comprenant un module de résolution, il permet d’associer à chaque occurrence d’entité le référent qu’elle désigne parmi les entrées d’un référentiel dédié. NP apporte ainsi des informations pertinentes pour l’exploitation de l’extraction d’entités nommées en contexte applicatif. Ce système fait l’objet d’une évaluation grâce au développement d’un corpus annoté manuellement et adapté aux tâches de détection et de résolution.

Quotation extraction is an important information extraction task, especially when dealing with news wires. Quotations can be found in various configurations. In this paper, we focus on direct quotations introduced by a parenthetical clause, headed by a ""quotation verb"". Our study is based on a large French news wire corpus from the Agence France-Presse. We introduce and motivate an analysis at the discursive level of such quotations, which differs from the syntactic analyses generally proposed. We show how we enriched the Lefff syntactic lexicon so that it provides an account for quotation verbs heading a quotation parenthetical, especially those extracted from a news wire corpus. We also sketch how these lexical entries can be extended to the discursive level in order to model quotations introduced in a parenthetical clause in a complete way.

We introduce PerLex, a large-coverage and freely-available morphological lexicon for the Persian language. We describe the main features of the Persian morphology, and the way we have represented it within the Alexina formalism, on which PerLex is based. We focus on the methodology we used for constructing lexical entries from various sources, as well as the problems related to typographic normalisation. The resulting lexicon shows a satisfying coverage on a reference corpus and should therefore be a good starting point for developing a syntactic lexicon for the Persian language.

In this paper, we introduce the Lefff, a freely available, accurate and large-coverage morphological and syntactic lexicon for French, used in many NLP tools such as large-coverage parsers. We first describe Alexina, the lexical framework in which the Lefff is developed as well as the linguistic notions and formalisms it is based on. Next, we describe the various sources of lexical data we used for building the Lefff, in particular semi-automatic lexical development techniques and conversion and merging of existing resources. Finally, we illustrate the coverage and precision of the resource by comparing it with other resources and by assessing its impact in various NLP tools.

La couverture d’un analyseur syntaxique dépend avant tout de la grammaire et du lexique sur lequel il repose. Le développement d’un lexique complet et précis est une tâche ardue et de longue haleine, surtout lorsque le lexique atteint un certain niveau de qualité et de couverture. Dans cet article, nous présentons un processus capable de détecter automatiquement les entrées manquantes ou incomplètes d’un lexique, et de suggérer des corrections pour ces entrées. La détection se réalise au moyen de deux techniques reposant soit sur un modèle statistique, soit sur les informations fournies par un étiqueteur syntaxique. Les hypothèses de corrections pour les entrées lexicales détectées sont générées en étudiant les modifications qui permettent d’améliorer le taux d’analyse des phrases dans lesquelles ces entrées apparaissent. Le processus global met en oeuvre plusieurs techniques utilisant divers outils tels que des étiqueteurs et des analyseurs syntaxiques ou des classifieurs d’entropie. Son application au Lefff , un lexique morphologique et syntaxique à large couverture du français, nous a déjà permis de réaliser des améliorations notables.

Dans cet article, nous montrons comment nous avons converti les tables du Lexique-Grammaire en un format TAL, celui du lexique Lefff, permettant ainsi son intégration dans l’analyseur syntaxique FRMG. Nous présentons les fondements linguistiques de ce processus de conversion et le lexique obtenu. Nous validons le lexique obtenu en évaluant l’analyseur syntaxique FRMG sur le corpus de référence de la campagne EASy selon qu’il utilise les entrées verbales du Lefff ou celles des tables des verbes du Lexique-Grammaire ainsi converties.

Cet article décrit la construction d’un Wordnet Libre du Français (WOLF) à partir du Princeton WordNet et de diverses ressources multilingues. Les lexèmes polysémiques ont été traités au moyen d’une approche reposant sur l’alignement en mots d’un corpus parallèle en cinq langues. Le lexique multilingue extrait a été désambiguïsé sémantiquement à l’aide des wordnets des langues concernées. Par ailleurs, une approche bilingue a été suffisante pour construire de nouvelles entrées à partir des lexèmes monosémiques. Nous avons pour cela extrait des lexiques bilingues à partir deWikipédia et de thésaurus. Le wordnet obtenu a été évalué par rapport au wordnet français issu du projet EuroWordNet. Les résultats sont encourageants, et des applications sont d’ores et déjà envisagées.

Cet article compare le Lexique-Grammaire des verbes pleins et DICOVALENCE, deux ressources lexicales syntaxiques pour le français développées par des linguistes depuis de nombreuses années. Nous étudions en particulier les divergences et les empiètements des modèles lexicaux sous-jacents. Puis nous présentons le Lefff , lexique syntaxique à grande échelle pour le TAL, et son propre modèle lexical. Nous montrons que ce modèle est à même d’intégrer les informations lexicales présentes dans le Lexique-Grammaire et dans DICOVALENCE. Nous présentons les résultats des premiers travaux effectués en ce sens, avec pour objectif à terme la constitution d’un lexique syntaxique de référence pour le TAL.

Nous présentons une méthode de fouille d’erreurs pour détecter automatiquement des erreurs dans les ressources utilisées par les systèmes d’analyse syntaxique. Nous avons mis en oeuvre cette méthode sur le résultat de l’analyse de plusieurs millions de mots par deux systèmes d’analyse différents qui ont toutefois en commun le lexique syntaxique et la chaîne de traitement pré-syntaxique. Nous avons pu identifier ainsi des inexactitudes et des incomplétudes dans les ressources utilisées. En particulier, la comparaison des résultats obtenus sur les sorties des deux analyseurs sur un même corpus nous a permis d’isoler les problèmes issus des ressources partagées de ceux issus des grammaires.

Nous présentons dans cet article une approche générale pour la modélisation et l’analyse syntaxique des coordinations elliptiques. Nous montrons que les lexèmes élidés peuvent être remplacés, au cours de l’analyse, par des informations qui proviennent de l’autre membre de la coordination, utilisé comme guide au niveau des dérivations. De plus, nous montrons comment cette approche peut être effectivement mise en oeuvre par une légère extension des Grammaires d’Arbres Adjoints Lexicalisées (LTAG) à travers une opération dite de fusion. Nous décrivons les algorithmes de dérivation nécessaires pour l’analyse de constructions coordonnées pouvant comporter un nombre quelconque d’ellipses.

This paper reports a large-scale non-probabilistic parsing experiment with a deep LFG parser. We briefly introduce the parser we used, named SXLFG, and the resources that were used together with it. Then we report quantitative results about the parsing of a multi-million word journalistic corpus. We show that we can parse more than 6 million words in less than 12 hours, only 6.7% of all sentences reaching the 1s timeout. This shows that deep large-coverage non-probabilistic parsers can be efficient enough to parse very large corpora in a reasonable amount of time.

In this paper, we introduce a new lexical resource for French which is freely available as the second version of the Lefff (Lexique des formes fléchies du français - Lexicon of French inflected forms). It is a wide-coverage morphosyntactic and syntactic lexicon, whose architecture relies on properties inheritance, which makes it more compact and more easily maintainable and allows to describe lexical entries independantly from the formalisms it is used for. For these two reasons, we define it as a meta-lexicon. We describe its architecture, several automatic or semi-automatic approaches we use to acquire, correct and/or enrich such a lexicon, as well as the way it is used both with an LFG parser and with a TAG parser based on a meta-grammar, so as to build two large-coverage parsers for French. The web site of the Lefff is http://www.lefff.net/.

Cet article expose l’ensemble des outils que nous avons mis en oeuvre pour la campagne EASy d’évaluation d’analyse syntaxique. Nous commençons par un aperçu du lexique morphologique et syntaxique utilisé. Puis nous décrivons brièvement les propriétés de notre chaîne de traitement pré-syntaxique qui permet de gérer des corpus tout-venant. Nous présentons alors les deux systèmes d’analyse que nous avons utilisés, un analyseur TAG issu d’une méta-grammaire et un analyseur LFG. Nous comparons ces deux systèmes en indiquant leurs points communs, comme l’utilisation intensive du partage de calcul et des représentations compactes de l’information, mais également leurs différences, au niveau des formalismes, des grammaires et des analyseurs. Nous décrivons ensuite le processus de post-traitement, qui nous a permis d’extraire de nos analyses les informations demandées par la campagne EASy. Nous terminons par une évaluation quantitative de nos architectures.

Dans cet article, nous proposons un nouvel analyseur syntaxique, qui repose sur une variante du modèle Lexical-Functional Grammars (Grammaires Lexicales Fonctionnelles) ou LFG. Cet analyseur LFG accepte en entrée un treillis de mots et calcule ses structures fonctionnelles sur une forêt partagée. Nous présentons également les différentes techniques de rattrapage d’erreurs que nous avons mises en oeuvre. Puis nous évaluons cet analyseur sur une grammaire à large couverture du français dans le cadre d’une utilisation à grande échelle sur corpus variés. Nous montrons que cet analyseur est à la fois efficace et robuste.

Nous présentons dans cet article un nouveau formalisme linguistique qui repose sur les Grammaires à Concaténation d’Intervalles (RCG), appelé Méta-RCG. Nous exposons tout d’abord pourquoi la non-linéarité permet une représentation adéquate des phénomènes linguistiques, et en particulier de l’interaction entre les différents niveaux de description. Puis nous présentons les Méta-RCG et les concepts linguistiques supplémentaires qu’elles mettent en oeuvre, tout en restant convertibles en RCG classiques. Nous montrons que les analyses classiques (constituants, dépendances, topologie, sémantique prédicat-arguments) peuvent être obtenues par projection partielle d’une analyse Méta-RCG complète. Enfin, nous décrivons la grammaire du français que nous développons dans ce nouveau formalisme et l’analyseur efficace qui en découle. Nous illustrons alors la notion de projection partielle sur un exemple.

Le but de cet article est de montrer pourquoi les Grammaires à Concaténation d’Intervalles (Range Concatenation Grammars, ou RCG) sont un formalisme particulièrement bien adapté à la description du langage naturel. Nous expliquons d’abord que la puissance nécessaire pour décrire le langage naturel est celle de PTIME. Ensuite, parmi les formalismes grammaticaux ayant cette puissance d’expression, nous justifions le choix des RCG. Enfin, après un aperçu de leur définition et de leurs propriétés, nous montrons comment leur utilisation comme grammaires linguistiques permet de traiter des phénomènes syntagmatiques complexes, de réaliser simultanément l’analyse syntaxique et la vérification des diverses contraintes (morphosyntaxiques, sémantique lexicale), et de construire dynamiquement des grammaires linguistiques modulaires.