International audience

We present a new question answering corpus in French designed to educational domain. To be useful in such domain, we have to propose more complex questions and to be able to justify the answers on validated material. We analyze some properties of this corpus. The last part of this paper will be devoted to present the first experiments we have carried out to demonstrate the value of this dataset for learning a Retrieval Augmented Genration framework. Different experiments are proposed, with an automatic evaluation. A human evaluation is proposed to confirm or infirm this automatic evaluation.

Transformer models have achieved significant improvements in multiple downstream tasks in recent years. One of the main contributions of Transformers is their ability to create new representations for out-of-vocabulary (OOV) words. In this paper, we have evaluated three categories of OOVs: (A) new domain-specific terms (e.g., “eucaryote’” in microbiology), (B) misspelled words containing typos, and (C) cross-domain homographs (e.g., “arm” has different meanings in a clinical trial and anatomy). We use three French domain-specific datasets on the legal, medical, and energetical domains to robustly analyze these categories. Our experiments have led to exciting findings that showed: (1) It is easier to improve the representation of new words (A and B) than it is for words that already exist in the vocabulary of the Transformer models (C), (2) To ameliorate the representation of OOVs, the most effective method relies on adding external morpho-syntactic context rather than improving the semantic understanding of the words directly (fine-tuning) and (3) We cannot foresee the impact of minor misspellings in words because similar misspellings have different impacts on their representation. We believe that tackling the challenges of processing OOVs regarding their specificities will significantly help the domain adaptation aspect of BERT.

Human-generated non-literal translations reflect the richness of human languages and are sometimes indispensable to ensure adequacy and fluency.Non-literal translations are difficult to produce even for human translators, especially for foreign language learners, and machine translations are still on the way to simulate human ones on this aspect.In order to foster the study on appropriate and creative non-literal translations, automatically detecting them in parallel corpora is an important step, which can benefit downstream NLP tasks or help to construct materials to teach translation.This article demonstrates that generic sentence representations produced by a pre-trained cross-lingual language model could be fine-tuned to solve this task.We show that there exists a moderate positive correlation between the prediction probability of being human translation and the non-literal translations' proportion in a sentence.The fine-tuning experiments show an accuracy of 80.16% when predicting the presence of non-literal translations in a sentence and an accuracy of 85.20% when distinguishing literal and non-literal translations at phrase level.We further conduct a linguistic error analysis and propose directions for future work.

Human translators often resort to different non-literal translation techniques besides the literal translation, such as idiom equivalence, generalization, particularization, semantic modulation, etc., especially when the source and target languages have different and distant origins. Translation techniques constitute an important subject in translation studies, which help researchers to understand and analyse translated texts. However, they receive less attention in developing Natural Language Processing (NLP) applications. To fill this gap, one of our long term objectives is to have a better semantic control of extracting paraphrases from bilingual parallel corpora. Based on this goal, we suggest this hypothesis: it is possible to automatically recognize different sub-sentential translation techniques. For this original task, since there is no dedicated data set for English-Chinese, we manually annotated a parallel corpus of eleven genres. Fifty sentence pairs for each genre have been annotated in order to consolidate our annotation guidelines. Based on this data set, we conducted an experiment to classify between literal and non-literal translations. The preliminary results confirm our hypothesis. The corpus and code are available. We hope that this annotated corpus will be useful for linguistic contrastive studies and for fine-grained evaluation of NLP tasks, such as automatic word alignment and machine translation.

This work uses parallel monolingual corpora for a detailed study of the task of sub-sentential paraphrase acquisition. We argue that the scarcity of this type of resource is compensated by the fact that it is the most suited type for studies on paraphrasing. We propose a large exploration of this task with experiments on two languages with five different acquisition techniques, selected for their complementarity, their combinations, as well as four monolingual corpus types of varying comparability. We report, under all conditions, a significant improvement over all techniques by validating candidate paraphrases using a maximum entropy classifier. An important result of our study is the identification of difficult-to-acquire paraphrase pairs, which are classified and quantified in a bilingual typology.

Question answering (QA) systems aim at finding answers to question posed in natural language using a collection of documents. When the collection is extracted from the Web, the structure and style of the texts are quite different from those of newspaper articles. We developed a QA system based on an answer validation process able to handle Web specificity. A large number of candidate answers are extracted from short passages in order to be validated according to question and passages characteristics. The validation module is based on a machine learning approach. It takes into account criteria characterizing both passage and answer relevance at surface, lexical, syntactic and semantic levels to deal with different types of texts. We present and compare results obtained for factual questions posed on a Web and on a newspaper collection. We show that our system outperforms a baseline by up to 48% in MRR.

Question answering (QA) aims at retrieving precise information from a large collection of documents. Different techniques can be used to find relevant information, and to compare these techniques, it is important to evaluate QA systems. The objective of an Answer Validation task is thus to judge the correctness of an answer returned by a QA system for a question, according to the text snippet given to support it. We participated in such a task in 2006. In this article, we present our strategy for deciding if the snippets justify the answers: a strategy based on our own QA system, comparing the answers it returned with the answer to judge. We discuss our results, then we point out the difficulties of this task.

Question answering (QA) aims at retrieving precise information from a large collection of documents, typically the Web. Different techniques can be used to find relevant information, and to compare these techniques, it is important to evaluate question answering systems. The objective of an Answer Validation task is to estimate the correctness of an answer returned by a QA system for a question, according to the text snippet given to support it. In this article, we present a lexical strategy for deciding if the snippets justify the answers, based on our own question answering system. We discuss our results, and show the possible extensions of our strategy.

We have developed a prototype of a question‐answering system which is intended to interact with an untrained user to give him information about the Yellow Pages of the French phone directory. We don't detail here the description of the system, because it is done in another paper in the same issue: “A question‐answering system for the French Yellow Pages”. In this paper we first present a typology of the different kinds of unforeseen situations the system may have to handle while interacting with an untrained user. Then, we detail different strategies we have developed to solve these problems, more specifically to correct misspelt words and to deal with the properly typed, but unknown, words. These mechanisms have been integrated into the parser, and try to maintain, as long as possible, its determinism.

This paper describes a dialogue‐based system which is intended as an intelligent natural language interface to the French Yellow Pages. We do not assume that the user knows how the Yellow Pages are organized, and we paraphrase his request, if necessary, so as to better search for the desired information. We do, however, assume that the reason the user is on line is to find an address and phone number for some supplier. There are three basic modules used in our system: parser, dialogue manager, and generator. The first two exist (and are constantly being extended); the generation module is still only a set of functional specifications which will be outlined later in this article.