Search

Cross-lingual word embeddings are vector representations of words in different languages where words with similar meaning are represented by similar vectors, regardless of the language. Recent developments which construct these embeddings by aligning monolingual spaces have shown that accurate alignments can be obtained with little or no supervision. However, the focus has been on a particular controlled scenario for evaluation, and there is no strong evidence on how current state-of-the-art systems would fare with noisy text or for language pairs with major linguistic differences. In this paper we present an extensive evaluation over multiple cross-lingual embedding models, analyzing their strengths and limitations with respect to different variables such as target language, training corpora and amount of supervision. Our conclusions put in doubt the view that high-quality cross-lingual embeddings can always be learned without much supervision., Comment: 11 pages, 2 figures, 7 tables. Camera-ready submitted to LREC 2020

I introduce a stability notion, dynamic stability, for two-sided dynamic matching markets where (i) matching opportunities arrive over time, (ii) matching is one-to-one, and (iii) matching is irreversible. The definition addresses two conceptual issues. First, since not all agents are available to match at the same time, one must establish which agents are allowed to form blocking pairs. Second, dynamic matching markets exhibit a form of externality that is not present in static markets: an agent's payoff from remaining unmatched cannot be defined independently of what other contemporaneous agents' outcomes are. Dynamically stable matchings always exist. Dynamic stability is a necessary condition to ensure timely participation in the economy by ensuring that agents do not strategically delay the time at which they are available to match.

Cross-lingual embeddings represent the meaning of words from different languages in the same vector space. Recent work has shown that it is possible to construct such representations by aligning independently learned monolingual embedding spaces, and that accurate alignments can be obtained even without external bilingual data. In this paper we explore a research direction that has been surprisingly neglected in the literature: leveraging noisy user-generated text to learn cross-lingual embeddings particularly tailored towards social media applications. While the noisiness and informal nature of the social media genre poses additional challenges to cross-lingual embedding methods, we find that it also provides key opportunities due to the abundance of code-switching and the existence of a shared vocabulary of emoji and named entities. Our contribution consists of a very simple post-processing step that exploits these phenomena to significantly improve the performance of state-of-the-art alignment methods., Comment: Accepted to ICWSM 2020. 11 pages, 1 appendix. Pre-trained embeddings available at https://github.com/pedrada88/crossembeddings-twitter

We study the scattering of transient, high-frequency, narrow-band quasi-Rayleigh elastic waves by through-thickness holes in aluminum plates, in the framework of ultrasonic nondestructive testing (NDT) based on full-field optical detection. Sequences of the instantaneous two-dimensional (2-D) out-of-plane displacement scattering maps are measured with a self-developed PTVH system. The corresponding simulated sequences are obtained by means of an FC(Gram) elastodynamic solver introduced recently, which implements a full three-dimensional (3D) vector formulation of the direct linear-elasticity scattering problem. A detailed quantitative comparison between these experimental and numerical sequences, which is presented here for the first time, shows very good agreement both in the amplitude and the phase of the acoustic field in the forward, lateral and backscattering areas. It is thus suggested that the combination of the PTVH system and the FC(Gram) elastodynamic solver provides an effective ultrasonic inspection tool for plate-like structures, with a significant potential for ultrasonic NDT applications., Comment: 46 pages, 16 figures, corresponding author Jos\'e Carlos L\'opez-V\'azquez, jclopez@uvigo.es. Changes: 1st, 4th, 5th paragraphs (intro), 3rd, 4th paragraphs (sec. 4); [59-60] cited only in appendixes; old ref. [52] removed; misprints corrected in the uncertainty of c_L (subsec. 3.1), citation to fig. 10 (sec. 4), size of images (caption fig.15); reference to Lam\'e constants removed in subsec. 3.4

A buyer wishes to purchase a durable good from a seller who in each period chooses a mechanism under limited commitment. The buyer's valuation is binary and fully persistent. We show that posted prices implement all equilibrium outcomes of an infinite-horizon, mechanism selection game. Despite being able to choose mechanisms, the seller can do no better and no worse than if he chose prices in each period, so that he is subject to Coase's conjecture. Our analysis marries insights from information and mechanism design with those from the literature on durable goods. We do so by relying on the revelation principle in Doval and Skreta (2020).

We first present detailed measurements of the rounding behavior around the superconducting transition temperature, $T_c$, of the in-plane electrical conductivity, magnetoconductivity and magnetization, including the low and moderate magnetic field regimes, in a high-quality single crystal and a thin film of the prototypical optimally-doped YBa$_2$Cu$_3$O$_{7-\delta}$ (OPT Y-123), in which the inhomogeneity effects are minimized. Then, it is presented a comparison of these experimental data with the phenomenological Ginzburg-Landau (GL) approach that takes into account the unavoidable contribution of the fluctuating pairs, the only theoretical scenario allowing at present to analyze these roundings at the quantitative level. These analyses demonstrate that the measured rounding effects around $T_c$ may be explained quantitative and consistently in terms of the GL scenario, even up to the rounding onset temperatures if the quantum localization, associated with the shrinkage of the superconducting wavefunction, is taken into account. The implications of our results on the pseudogap physics of optimally-doped cuprates are also discussed., Comment: Version of the preprint of the paper published in Superconductor Science & Technology 32, (2019) 045009, with an extended introduction and more references

Word segmentation is the task of inserting or deleting word boundary characters in order to separate character sequences that correspond to words in some language. In this article we propose an approach based on a beam search algorithm and a language model working at the byte/character level, the latter component implemented either as an n-gram model or a recurrent neural network. The resulting system analyzes the text input with no word boundaries one token at a time, which can be a character or a byte, and uses the information gathered by the language model to determine if a boundary must be placed in the current position or not. Our aim is to use this system in a preprocessing step for a microtext normalization system. This means that it needs to effectively cope with the data sparsity present on this kind of texts. We also strove to surpass the performance of two readily available word segmentation systems: The well-known and accessible Word Breaker by Microsoft, and the Python module WordSegment by Grant Jenks. The results show that we have met our objectives, and we hope to continue to improve both the precision and the efficiency of our system in the future., Comment: 11 pages, 4 figures, 5 tables, accepted in Journal of the Association for Information Science and Technology

We provide tools to analyze information design problems subject to constraints. We do so by extending the insight in Le Treust and Tomala (2019) to the case of multiple inequality and equality constraints. Namely, that an information design problem subject to constraints can be represented as an unconstrained information design problem with a additional states, one for each constraint. Thus, without loss of generality, optimal solutions induce as many posteriors as the number of states and constraints. We provide results that refine this upper bound. Furthermore, we provide conditions under which there is no duality gap in constrained information design, thus validating a Lagrangian approach. We illustrate our results with applications to mechanism design with limited commitment (Doval and Skreta, 2022a) and persuasion of a privately informed receiver (Kolotilin et al., 2017).

We develop a tool akin to the revelation principle for dynamic mechanism-selection games in which the designer can only commit to short-term mechanisms. We identify a canonical class of mechanisms rich enough to replicate the outcomes of any equilibrium in a mechanism-selection game between an uninformed designer and a privately informed agent. A cornerstone of our methodology is the idea that a mechanism should encode not only the rules that determine the allocation, but also the information the designer obtains from the interaction with the agent. Therefore, how much the designer learns, which is the key tension in design with limited commitment, becomes an explicit part of the design. Our result simplifies the search for the designer-optimal outcome by reducing the agent's behavior to a series of participation, truthtelling, and Bayes' plausibility constraints the mechanisms must satisfy., Comment: Added an omitted assumption in Section 4 (see footnote 21 and the proof of Proposition 4.1)

Cross-lingual word embeddings are becoming increasingly important in multilingual NLP. Recently, it has been shown that these embeddings can be effectively learned by aligning two disjoint monolingual vector spaces through linear transformations, using no more than a small bilingual dictionary as supervision. In this work, we propose to apply an additional transformation after the initial alignment step, which moves cross-lingual synonyms towards a middle point between them. By applying this transformation our aim is to obtain a better cross-lingual integration of the vector spaces. In addition, and perhaps surprisingly, the monolingual spaces also improve by this transformation. This is in contrast to the original alignment, which is typically learned such that the structure of the monolingual spaces is preserved. Our experiments confirm that the resulting cross-lingual embeddings outperform state-of-the-art models in both monolingual and cross-lingual evaluation tasks., Comment: 11 pages, 4 tables, 1 figure. EMNLP 2018 camera-ready