Search

Author profiling is the task of inferring characteristics about individuals by analyzing content they share. Supervised machine learning still dominates automatic systems that perform this task, despite the popularity of prompting large language models to address natural language understanding tasks. One reason is that the classification instances consist of large amounts of posts, potentially a whole user profile, which may exceed the input length of Transformers. Even if a model can use a large context window, the entirety of posts makes the application of API-accessed black box systems costly and slow, next to issues which come with such "needle-in-the-haystack" tasks. To mitigate this limitation, we propose a new method for author profiling which aims at distinguishing relevant from irrelevant content first, followed by the actual user profiling only with relevant data. To circumvent the need for relevance-annotated data, we optimize this relevance filter via reinforcement learning with a reward function that utilizes the zero-shot capabilities of large language models. We evaluate our method for Big Five personality trait prediction on two Twitter corpora. On publicly available real-world data with a skewed label distribution, our method shows similar efficacy to using all posts in a user profile, but with a substantially shorter context. An evaluation on a version of these data balanced with artificial posts shows that the filtering to relevant posts leads to a significantly improved accuracy of the predictions., Comment: preprint, under review, supplementary material will be made available upon acceptance of the paper

An ideal proton exchange membrane should only permeate protons and be leak tight for fuels. Graphene is impermeable to water and poorly conducting to protons. Here, we chemically functionalized monolayer graphene to install sulfophenylated sp3 dislocations by diazotization. Selective to protons, transmembrane areal conductances are up to ~50 S/cm2, which is ~5000 fold higher than in pristine graphene. Mounted in a direct methanol fuel cell, sulfophenylated graphene resulted in power densities up to 1.6 W/mg or 123 mW/cm under standard cell operation (60 C), a value ~two-fold larger than micron-thick films of Nafion 117. The combination of sp3 dislocations and polar groups, therefore, allow the creation of hydrophilic ion paths through graphene and unveils a novel route to rationalize transmembrane hydron transport through 2D materials.

We report on the structural, electric and magnetic properties of $(\mathrm{Ni}_x\mathrm{Co}_{1-x})\mathrm{B}$ ferromagnetic nanotubes, displaying azimuthal magnetization. The tubes are fabricated using electroless plating in polycarbonate porous templates, with lengths several tens of micrometers, diameters from 100nm to 500nm and wall thicknesses from 10nm to 80nm. The resistivity is $\sim 1.5\times10^{-6}\mathrm{\Omega/m}$, and the anisotropic magnetoresistance~(AMR) of 0.2-0.3%, one order of magnitude larger~(resp. smaller) than in the bulk material, which we attribute to the resistance at grain boundaries. We determined the azimuthal anisotropy field from M(H) AMR loops of single tubes contacted electrically. Its magnitude is around 10mT, and tends to increase with the tube wall thickness, as well as the Co content. However, surprisingly it does not dependent much on the diameter nor on the curvature., Comment: 12 pages, 10 figures, includes methods and appendix

Lattice matched n-type AlInP(100) charge selective contacts are commonly grown on n-p GaInP(100) top absorbers in high-efficiency III-V multijunction solar or photoelectrochemical cells. The cell performance can be greatly limited by the electron selectivity and valance band offset at this heterointerface. Understanding of the atomic and electronic properties of the GaInP/AlInP heterointerface is crucial for the reduction of photocurrent losses in III-V multijunction devices. In our paper, we investigated chemical composition and electronic properties of n-GaInP/n-AlInP heterostructures by X-ray photoelectron spectroscopy (XPS). To mimic an in-situ interface experiment with in-situ stepwise deposition of the contact material, 1 nm - 50 nm thick n-AlInP(100) epitaxial layers were grown on n-GaInP(100) buffer layer on n-GaAs(100) substrates by metal organic vapor phase epitaxy. We observed (2x2)/c(4x2) low-energy electron diffraction patterns with characteristic diffuse streaks along the [01-1] direction due to P-P dimers on both AlInP(100) and GaInP(100) as-prepared surfaces. Atomic composition analysis confirmed P-rich termination on both surfaces. Angle-resolved XPS measurements revealed a surface core level shift of 0.9 eV in P 2p peaks and the absence of interface core level shifts. We assigned the surface chemical shift in the P2p spectrum to P-P bonds on a surface. We found an upward surface band bending on the (2x2)/c(4x2) surfaces most probably caused by localized mid-gap electronic states. Pinning of the Fermi level by localized electronic states remained in n-GaInP/n-AlInP heterostructures. A valence band offset of 0.2 eV was derived by XPS and band alignment diagram models for the n-n junctions were suggested.

Appraisal theories explain how the cognitive evaluation of an event leads to a particular emotion. In contrast to theories of basic emotions or affect (valence/arousal), this theory has not received a lot of attention in natural language processing. Yet, in psychology it has been proven powerful: Smith and Ellsworth (1985) showed that the appraisal dimensions attention, certainty, anticipated effort, pleasantness, responsibility/control and situational control discriminate between (at least) 15 emotion classes. We study different annotation strategies for these dimensions, based on the event-focused enISEAR corpus (Troiano et al., 2019). We analyze two manual annotation settings: (1) showing the text to annotate while masking the experienced emotion label; (2) revealing the emotion associated with the text. Setting 2 enables the annotators to develop a more realistic intuition of the described event, while Setting 1 is a more standard annotation procedure, purely relying on text. We evaluate these strategies in two ways: by measuring inter-annotator agreement and by fine-tuning RoBERTa to predict appraisal variables. Our results show that knowledge of the emotion increases annotators' reliability. Further, we evaluate a purely automatic rule-based labeling strategy (inferring appraisal from annotated emotion classes). Training on automatically assigned labels leads to a competitive performance of our classifier, even when tested on manual annotations. This is an indicator that it might be possible to automatically create appraisal corpora for every domain for which emotion corpora already exist., Comment: Published in WASSA 2021 at EACL 2021

Automatic emotion categorization has been predominantly formulated as text classification in which textual units are assigned to an emotion from a predefined inventory, for instance following the fundamental emotion classes proposed by Paul Ekman (fear, joy, anger, disgust, sadness, surprise) or Robert Plutchik (adding trust, anticipation). This approach ignores existing psychological theories to some degree, which provide explanations regarding the perception of events. For instance, the description that somebody discovers a snake is associated with fear, based on the appraisal as being an unpleasant and non-controllable situation. This emotion reconstruction is even possible without having access to explicit reports of a subjective feeling (for instance expressing this with the words "I am afraid."). Automatic classification approaches therefore need to learn properties of events as latent variables (for instance that the uncertainty and the mental or physical effort associated with the encounter of a snake leads to fear). With this paper, we propose to make such interpretations of events explicit, following theories of cognitive appraisal of events, and show their potential for emotion classification when being encoded in classification models. Our results show that high quality appraisal dimension assignments in event descriptions lead to an improvement in the classification of discrete emotion categories. We make our corpus of appraisal-annotated emotion-associated event descriptions publicly available., Comment: Accepted at COLING 2020

We prove that in each dimension $d$ there is a constant $w^\infty(d)\in \mathbb{N}$ such that for every $n\in \mathbb{N}$ all but finitely many $d$-polytopes with $n$ lattice points have width at most $w^\infty(d)$. We call $w^\infty(d)$ the finiteness threshold width and show that $d-2 \le w^\infty(d)\le O^*\left( d^{4/3}\right)$. Blanco and Santos determined the value $w^\infty(3)=1$. Here, we establish $w^\infty(4)=2$. This implies, in particular, that there are only finitely many empty $4$-simplices of width larger than two. The main tool in our proofs is the study of $d$-dimensional lifts of hollow $(d-1)$-polytopes., Comment: 21 pages, 10 figures; changes from previous version: several edits suggested by anonymous referees; this version has been accepted in Trans. of the Amer. Math. Soc

In 2012 Gubeladze (Adv.\ Math.\ 2012) introduced the notion of k-convex-normal polytopes to show that integral polytopes all of whose edges are longer than 4d(d+1) have the integer decomposition property. In the first part of this paper we show that for lattice polytopes there is no difference between k- and (k+1)-convex-normality (for k >= 3) and improve the bound to 2d(d+1). In the second part we extend the definition to pairs of polytopes and show that for rational polytopes P and Q, where the normal fan of P is a refinement of the normal fan of Q, if every edge e_P of P is at least d times as long as the corresponding edge e_Q of Q, then (P+Q) \cap \Z^d = (P\cap \Z^d) + (Q \cap \Z^d)., Comment: 10 pages, 8 figures