Search

In spite of the popularity of end-to-end diarization systems nowadays, modular systems comprised of voice activity detection (VAD), speaker embedding extraction plus clustering, and overlapped speech detection (OSD) plus handling still attain competitive performance in many conditions. However, one of the main drawbacks of modular systems is the need to run (and train) different modules independently. In this work, we propose an approach to jointly train a model to produce speaker embeddings, VAD and OSD simultaneously and reach competitive performance at a fraction of the inference time of a standard approach. Furthermore, the joint inference leads to a simplified overall pipeline which brings us one step closer to a unified clustering-based method that can be trained end-to-end towards a diarization-specific objective.

The interaction between light and matter in condensed matter excitations and electromagnetic resonators serves as a rich playground for fundamental research and lies at the core of photonic and quantum technologies. Herein, we present comprehensive experimental and theoretical studies of the photon-mediated hybridization of magnons and phonons in the terahertz (THz) range. We demonstrate the intriguing concept of composite states formed by distinct electric and magnetic quasiparticles strongly coupled to the same optical cavity modes. Specifically, we explore magnons excited in a slab of an antiferromagnetic crystal and phonons excited in a distinct specimen of an insulating material. The crystal slabs form an optical cavity with Fabry-P\'erot oscillations in the THz range. We demonstrate hybridized phonon-magnon polariton modes and their tunability by adjusting the distance between the slabs, showing that hybridization persists even at separations up to several millimeters. The experimental results are interpreted using both classical and quantum electrodynamical models. The quantum description allows us to quantify the degree of hybridization that is linked to a topological behavior of the electric field phasor, in agreement with the classical electrodynamics expectations. Importantly, the presented results refer to temperature conditions and cavities of millimeter size, paving the way for engineering realistic, frequency-tunable THz devices through the hybridization of electric (phononics) and magnetic (spintronics) elementary excitations of matter.

We generalize Miyanishi's theory of almost minimal models of log smooth surfaces with reduced boundary to the case of arbitrary log surfaces defined over an algebraically closed field. Given an MMP run of a log surface $(X,D)$ we define and construct its almost minimal model, whose underlying surface has singularities not worse than $X$ and which differs from a minimal model by a contraction of some curves supported in the boundary only. For boundaries of type $rD$, where $D$ is reduced and $r\in [0,1]\cap \mathbb{Q}$, we show that if $X$ is smooth or $r\in [0,\frac{1}{2}]$ then the construction respects $(1-r)$-divisorial log terminality and $(1-r)$-log canonicity. We show that the assumptions are optimal, too., Comment: 50 pages

The paper reports the results of an experiment aimed at testing to what extent ChatGPT 3.5 and 4 is able to answer questions regarding privacy policies designed in the new format that we propose. In a world of human-only interpreters, there was a trade-off between comprehensiveness and comprehensibility of privacy policies, leading to the actual policies not containing enough information for users to learn anything meaningful. Having shown that GPT performs relatively well with the new format, we provide experimental evidence supporting our policy suggestion, namely that the law should require fully comprehensive privacy policies, even if this means they become less concise.

This paper describes Adam Mickiewicz University's (AMU) solution for the 4th Shared Task on SlavNER. The task involves the identification, categorization, and lemmatization of named entities in Slavic languages. Our approach involved exploring the use of foundation models for these tasks. In particular, we used models based on the popular BERT and T5 model architectures. Additionally, we used external datasets to further improve the quality of our models. Our solution obtained promising results, achieving high metrics scores in both tasks. We describe our approach and the results of our experiments in detail, showing that the method is effective for NER and lemmatization in Slavic languages. Additionally, our models for lemmatization will be available at: https://huggingface.co/amu-cai., Comment: Slavic NLP 2023 @ EACL 2023

This paper presents Adam Mickiewicz University's (AMU) submissions to the constrained track of the WMT 2022 General MT Task. We participated in the Ukrainian $\leftrightarrow$ Czech translation directions. The systems are a weighted ensemble of four models based on the Transformer (big) architecture. The models use source factors to utilize the information about named entities present in the input. Each of the models in the ensemble was trained using only the data provided by the shared task organizers. A noisy back-translation technique was used to augment the training corpora. One of the models in the ensemble is a document-level model, trained on parallel and synthetic longer sequences. During the sentence-level decoding process, the ensemble generated the n-best list. The n-best list was merged with the n-best list generated by a single document-level model which translated multiple sentences at a time. Finally, existing quality estimation models and minimum Bayes risk decoding were used to rerank the n-best list so that the best hypothesis was chosen according to the COMET evaluation metric. According to the automatic evaluation results, our systems rank first in both translation directions., Comment: WMT 2022

The output structure of database-like tables, consisting of values structured in horizontal rows and vertical columns identifiable by name, can cover a wide range of NLP tasks. Following this constatation, we propose a framework for text-to-table neural models applicable to problems such as extraction of line items, joint entity and relation extraction, or knowledge base population. The permutation-based decoder of our proposal is a generalized sequential method that comprehends information from all cells in the table. The training maximizes the expected log-likelihood for a table's content across all random permutations of the factorization order. During the content inference, we exploit the model's ability to generate cells in any order by searching over possible orderings to maximize the model's confidence and avoid substantial error accumulation, which other sequential models are prone to. Experiments demonstrate a high practical value of the framework, which establishes state-of-the-art results on several challenging datasets, outperforming previous solutions by up to 15%.

As Deep Learning algorithms continue to evolve and become more sophisticated, they require massive datasets for model training and efficacy of models. Some of those data requirements can be met with the help of existing datasets within the organizations. Current Machine Learning practices can be leveraged to generate synthetic data from an existing dataset. Further, it is well established that diversity in generated synthetic data relies on (and is perhaps limited by) statistical properties of available dataset within a single organization or entity. The more diverse an existing dataset is, the more expressive and generic synthetic data can be. However, given the scarcity of underlying data, it is challenging to collate big data in one organization. The diverse, non-overlapping dataset across distinct organizations provides an opportunity for them to contribute their limited distinct data to a larger pool that can be leveraged to further synthesize. Unfortunately, this raises data privacy concerns that some institutions may not be comfortable with. This paper proposes a novel approach to generate synthetic data - FedSyn. FedSyn is a collaborative, privacy preserving approach to generate synthetic data among multiple participants in a federated and collaborative network. FedSyn creates a synthetic data generation model, which can generate synthetic data consisting of statistical distribution of almost all the participants in the network. FedSyn does not require access to the data of an individual participant, hence protecting the privacy of participant's data. The proposed technique in this paper leverages federated machine learning and generative adversarial network (GAN) as neural network architecture for synthetic data generation. The proposed method can be extended to many machine learning problem classes in finance, health, governance, technology and many more.

Surface roughness of particle monolayer is one of its fundamental characteristics, conventionally described in terms of power spectral density. We derive a general equation for this function and demonstrate its similarity to those applied in the theory of scattering. We validate our analytical results by comparison with numerical results for a system with parameters corresponding to typical experimental conditions. We also demonstrate a practical application of our approach for spectral analysis of the monolayer image. Our theoretical model provides a general framework for a cheap and easy parametrization of particle monolayers.

We address the challenging problem of Natural Language Comprehension beyond plain-text documents by introducing the TILT neural network architecture which simultaneously learns layout information, visual features, and textual semantics. Contrary to previous approaches, we rely on a decoder capable of unifying a variety of problems involving natural language. The layout is represented as an attention bias and complemented with contextualized visual information, while the core of our model is a pretrained encoder-decoder Transformer. Our novel approach achieves state-of-the-art results in extracting information from documents and answering questions which demand layout understanding (DocVQA, CORD, SROIE). At the same time, we simplify the process by employing an end-to-end model., Comment: Accepted at ICDAR 2021

This study is focused on applying genetic algorithms (GA) to model and band selection in hyperspectral image classification. We use a forensic-inspired data set of seven hyperspectral images with blood and five visually similar substances to test GA-optimised classifiers in two scenarios: when the training and test data come from the same image and when they come from different images, which is a more challenging task due to significant spectra differences. In our experiments we compare GA with a classic model optimisation through grid search. Our results show that GA-based model optimisation can reduce the number of bands and create an accurate classifier that outperforms the GS-based reference models, provided that during model optimisation it has access to examples similar to test data. We illustrate this with experiment highlighting the importance of a validation set.

We describe a method to construct completions of affine spaces into total spaces of $\mathbb{Q}$-factorial terminal Mori fiber spaces over the projective line. As an application we provide families of examples with non-rational, birationally rigid and non-stably rational general fibers., Comment: 24 pages

Research conducted in the framework of the J-PET project aims to develop a cost-effective total-body positron emission tomography scanner. As a first step on the way to construct a full-scale J-PET tomograph from long strips of plastic scintillators, a 24-strip prototype was built and tested. The prototype consists of detection modules arranged axially forming a cylindrical diagnostic chamber with the inner diameter of 360 mm and the axial field-of-view of 300 mm. Promising perspectives for a low-cost construction of a total-body PET scanner are opened due to an axial arrangement of strips of plastic scintillators, wchich have a small light attenuation, superior timing properties, and the possibility of cost-effective increase of the axial field-of-view. The presented prototype comprises dedicated solely digital front-end electronic circuits and a triggerless data acquisition system which required development of new calibration methods including time, thresholds and gain synchronization. The system and elaborated calibration methods including first results of the 24-module J-PET prototype are presented and discussed. The achieved coincidence resolving time equals to CRT = 490 $\pm$ 9 ps. This value can be translated to the position reconstruction accuracy $\sigma(\Delta l) =$ 18 mm which is fairly position-independent., Comment: Accepted for publication in IEEE Transactions on Instrumentation and Measurement

The J-PET tomograph is constructed from plastic scintillator strips arranged axially in concentric cylindrical layers. It enables investigations of positronium decays by measurement of the time, position, polarization and energy deposited by photons in the scintillators, in contrast to studies conducted so far with crystal and semiconductor based detection systems where the key selection of events is based on the measurement of the photons energies. In this article we show that the J-PET tomograph system is capable of exclusive measurements of the decays of ortho-positronium atoms. We present the first positronium production results, its lifetime distribution measurements and discuss estimation of the influence of various background sources. The tomograph s performance demonstrated here makes it suitable for precision studies of positronium decays including entanglement of the final state photons, positron annihilation lifetime spectroscopy plus molecular imaging diagnostics., Comment: 12 pages, 7 figures

We present high-statistic data on charged pion emission from Au+Au collisions at $\sqrt{s_{\rm{NN}}}$ = 2.4 GeV (corresponding to $E_{beam}$ = 1.23 A GeV) in four centrality classes in the range 0 - 40$\%$ of the most central collisions. The data are analyzed as a function of transverse momentum, transverse mass, rapidity, and polar angle. Pion multiplicity per participating nucleon decreases moderately with increasing centrality. The polar angular distributions are found to be non-isotropic even for the most central event class. Our results on pion multiplicity fit well into the general trend of the world data, but undershoot by $2.5 \sigma$ data from the FOPI experiment measured at slightly lower beam energy. We compare our data to state-of-the-art transport model calculations (PHSD, IQMD, PHQMD, GiBUU and SMASH) and find substantial differences between the measurement and the results of these calculations., Comment: 17 pages, 19 figures, 4 tables

In this article, a new method for the reconstruction of hit-position and hit-time of photons in long scintillator detectors is investigated. This research is motivated by the recent development of the positron emission tomography scanners based on plastic scintillators. The proposed method constitutes a new way of signal processing in Multi-Voltage-Technique. It is based on the determination of the degree of similarity between the registered signals and the synchronized model signals stored in a library. The library was established for a set of well defined hit-positions along the length of the scintillator. The Mahalanobis distance was used as a measure of similarity between the two compared signals. The method was validated on the experimental data measured using two-strips J-PET prototype with dimensions of 5x9x300 mm$^3$. The obtained Time-of-Flight (TOF) and spatial resolutions amount to 325~ps (FWHM) and 25~mm (FWHM), respectively. The TOF resolution was also compared to the results of an analogous study done using Linear Fitting method. The best TOF resolution was obtained with this method at four pre-defined threshold levels which was comparable to the resolution achieved from the Mahalanobis distance at two pre-defined threshold levels. Although the algorithm of Linear Fitting method is much simpler to apply than the Mahalanobis method, the application of the Mahalanobis distance requires a lower number of applied threshold levels and, hence, decreases the costs of electronics used in PET scanner.

We present an analysis of proton number fluctuations in $\sqrt{s_{NN}}$ = 2.4 GeV Au+Au collisions measured with the High-Acceptance DiElectron Spectrometer (HADES) at GSI. With the help of extensive detector simulations done with IQMD transport model events including nuclear clusters, various nuisance effects influencing the observed proton cumulants have been investigated. Acceptance and efficiency corrections have been applied as a function of fine grained rapidity and transverse momentum bins, as well as considering local track density dependencies. Next, the effects of volume changes within particular centrality selections have been considered and beyond-leading-order corrections have been applied to the data. The efficiency and volume corrected proton number moments and cumulants Kn of orders n = 1, . . . , 4 have been obtained as a function of centrality and phase-space bin, as well as the corresponding correlators C_n . We find that the observed correlators show a power-law scaling with the mean number of protons, i.e. $C_n \propto ^n$, indicative of mostly long-range multi-particle correlations in momentum space. We also present a comparison of our results with Au+Au collision data obtained at RHIC at similar centralities, but higher $\sqrt{s_{NN}}$., Comment: 35 pages, 33 figures

Time-Over-Threshold (TOT) technique is being used widely due to its implications in developing the multi channel readouts mainly when fast signal processing is required. Using TOT technique as a measure of energy loss instead of charge integration methods significantly reduces the signals readout cost by combining the time and energy information. Therefore, this approach can potentially be used in J-PET tomograph which is build from plastic scintillators characterized by fast light signals. The drawback in adopting this technique is lying in the non-linear correlation between input energy loss and TOT of the signal. The main motivation behind this work is to develop the relationship between TOT and energy loss and validate it with the J-PET tomograph. The experiment was performed using the $^{22}$Na beta emitter source placed in the center of the J-PET tomograph. One can obtain primary photons of two different energies: 511 keV photon from the annihilation of positron (direct annihilation or through the formation of para-Positronim atom or pick-off process of ortho-Positronium atoms), and 1275 keV prompt photon. This allows to study the correlation between TOT values and energy loss for energy range up to 1000 keV. As the photon interacts dominantly via Compton scattering inside the plastic scintillator, there is no direct information of primary photon energy. However, using the J-PET geometry one can measure the scattering angle of the interacting photon. Since, $^{22}$Na source emits photons of two different energies, it is required to know unambiguously the energy of incident photons and its corresponding scattering angle for the estimation of energy deposition. In this work, the relationship between Time Over Threshold and energy loss by interacting photons inside the plastic scintillators used in J-PET scanner is established for a energy deposited range 100-1000 keV

In living organisms the positron-electron annihilation (occurring during the PET imaging) proceeds in about 30% via creation of a metastable ortho-positronium atom. In the tissue, due to the pick-off and conversion processes, over 98% of ortho-positronia annihilate into two 511~keV photons. In this article we assess the feasibility for reconstruction of the mean ortho-positronium lifetime image based on annihilations into two photons. The main objectives of this work include: (i) estimation of the sensitivity of the total-body PET scanners for the ortho-positronium mean lifetime imaging using $2\gamma$ annihilations, and (ii) estimation of the spatial and time resolution of the ortho-positronium image as a function of the coincidence resolving time (CRT) of the scanner. Simulations are conducted assuming that radiopharmaceutical is labelled with $^{44}Sc$ isotope emitting one positron and one prompt gamma. The image is reconstructed on the basis of triple coincidence events. The ortho-positronium lifetime spectrum is determined for each voxel of the image. Calculations were performed for cases of total-body detectors build of (i) LYSO scintillators as used in the EXPLORER PET, and (ii) plastic scintillators as anticipated for the cost-effective total-body J-PET scanner. To assess the spatial and time resolution the three cases were considered assuming that CRT is equal to 140ps, 50ps and 10ps. The estimated total-body PET sensitivity for the registration and selection of image forming triple coincidences is larger by a factor of 12.2 (for LYSO PET) and by factor of 4.7 (for plastic PET) with respect to the sensitivity for the standard $2\gamma$ imaging by LYSO PET scanners with AFOV=20cm.

We propose a new shared task of semantic retrieval from legal texts, in which a so-called contract discovery is to be performed, where legal clauses are extracted from documents, given a few examples of similar clauses from other legal acts. The task differs substantially from conventional NLI and shared tasks on legal information extraction (e.g., one has to identify text span instead of a single document, page, or paragraph). The specification of the proposed task is followed by an evaluation of multiple solutions within the unified framework proposed for this branch of methods. It is shown that state-of-the-art pretrained encoders fail to provide satisfactory results on the task proposed. In contrast, Language Model-based solutions perform better, especially when unsupervised fine-tuning is applied. Besides the ablation studies, we addressed questions regarding detection accuracy for relevant text fragments depending on the number of examples available. In addition to the dataset and reference results, LMs specialized in the legal domain were made publicly available., Comment: Submitted to Findings of EMNLP

High-statistics $\pi^-\pi^-$ and $\pi^+\pi^+$ femtoscopy data are presented for Au+Au collisions at $\sqrt{s_\mathrm{NN}}=2.4$ GeV, measured with HADES at SIS18/GSI. The experimental correlation functions allow the determination of the space-time extent of the corresponding emission sources via a comparison to models. The emission source, parametrized as three-dimensional Gaussian distribution, is studied in dependence on pair transverse momentum, azimuthal emission angle with respect to the reaction plane, collision centrality and beam energy. For all centralities and transverse momenta, a geometrical distribution of ellipsoidal shape is found in the plane perpendicular to the beam direction with the larger extension perpendicular to the reaction plane. For large transverse momenta, the corresponding eccentricity approaches the initial eccentricity. The eccentricity is smallest for most central collisions, where the shape is almost circular. The magnitude of the tilt angle of the emission ellipsoid in the reaction plane decreases with increasing centrality and increasing transverse momentum. All source radii increase with centrality, largely exhibiting a linear rise with the number of participants, irrespective of transverse momentum. A substantial charge-sign difference of the source radii is found, appearing most pronounced at low transverse momentum. The extracted source parameters are consistent with the extrapolation of their energy dependence down from higher energies., Comment: 20 pages, 19 figures, 7 tables. arXiv admin note: text overlap with arXiv:1811.06213

We show that a complex planar curve homeomorphic to the projective line has at most four singular points. If it has exactly four then it has degree five and is unique up to a projective equivalence., Comment: 35 pages

We report a measurement of the ratios of branching fractions $\mathcal{R}(D) = {\cal B}(\bar{B} \to D \tau^- \bar{\nu}_{\tau})/{\cal B}(\bar{B} \to D \ell^- \bar{\nu}_{\ell})$ and $\mathcal{R}(D^{\ast}) = {\cal B}(\bar{B} \to D^* \tau^- \bar{\nu}_{\tau})/{\cal B}(\bar{B} \to D^* \ell^- \bar{\nu}_{\ell})$, where $\ell$ denotes an electron or a muon. The results are based on a data sample containing $772\times10^6$ $B\bar{B}$ events recorded at the $\Upsilon(4S)$ resonance with the Belle detector at the KEKB $e^+ e^-$ collider. The analysis utilizes a method where the tag-side $B$ meson is reconstructed in a semileptonic decay mode, and the signal-side $\tau$ is reconstructed in a purely leptonic decay. The measured values are $\mathcal{R}(D)= 0.307 \pm 0.037 \pm 0.016$ and $\mathcal{R}(D^{\ast})= 0.283 \pm 0.018 \pm 0.014$, where the first uncertainties are statistical and the second are systematic. These results are in agreement with the Standard Model predictions within $0.2$ and $1.1$ standard deviations, respectively, while their combination agrees with the Standard Model predictions within $1.2$ standard deviations., Comment: 12 pages, 6 figures

J-PET is the first positron-emission tomograph (PET) constructed from plastic scintillators. It was optimized for the detection of photons from electron-positron annihilation. Such photons, having an energy of 511 keV, interact with electrons in plastic scintillators predominantly via the Compton effect. Compton scattering is at most probable at an angle orthogonal to the electric field vector of the interacting photon. Thus registration of multiple photon scatterings with J-PET enables to determine the polarization of the annihilation photons. In this contribution we present estimates on the physical limitation in the accuracy of the polarization determination of $511$~keV photons with the J-PET detector., Comment: Submitted to Hyperfine Interactions

The quality of an As35S65 chalcogenide glass (ChG) grating fabricated by electron beam lithography (EBL) was characterized by optical scatterometry based on spectroscopic ellipsometry (SE) in the visible and near infrared spectral range and complementary techniques providing direct images, especially atomic force microscopy (AFM). The geometric dimensions and the shape of patterned grating lines were determined by fitting modeled values (calculated by the Fourier modal method) to SE experimental data. A simple power-dependent function with only one variable parameter was successfully used to describe the shape of the patterned lines. The result yielded by SE is shown to correspond to AFM measurement with high accuracy, provided that optical constants of ChG modified by EBL were used in the fitting procedure. The line edge roughness (LER) of the grating was also investigated by further fitting the SE data to find out that no LER is optically detectable in the spectral range used, which is essential for the functionality of optical tools fabricated by EBL., Comment: 10 pages, 9 figures

We present first data on sub-threshold production of K0 s mesons and {\Lambda} hyperons in Au+Au collisions at $\sqrt{s_{NN}}$ = 2.4 GeV. We observe an universal scaling of hadrons containing strangeness, independent of their corresponding production thresholds. Comparing the yields, their scaling, and the shapes of the rapidity and the pt spectra to state-of-the-art transport model (UrQMD, HSD, IQMD) predictions, we find that none of the latter can simultaneously describe all observables with reasonable \c{hi}2 values., Comment: 8 pages, 10 figures

We investigate identical pion HBT intensity interferometry for central Au+Au collisions at 1.23A GeV. High-statistics $\pi^-\pi^-$ and $\pi^+\pi^+$ data are measured with HADES at SIS18/GSI. The radius parameters, derived from the correlation function depending on relative momenta in the longitudinal-comoving system and parametrized as three-dimensional Gaussian distribution, are studied as function of transverse momentum. A substantial charge-sign difference of the source radii is found, particularly pronounced at low transverse momentum. The extracted Coulomb-corrected source parameters agree well with a smooth extrapolation of the center-of-mass energy dependence established at higher energies, extending the corresponding excitation functions down towards a very low energy. Our data would thus rather disfavour any strong energy dependence of the radius parameters in the low energy region., Comment: 7 pages, 5 figures, updated author and grant lists

Let $E\subseteq \mathbb{P}^2$ be a complex curve homeomorphic to the projective line. The Negativity Conjecture asserts that the Kodaira-Iitaka dimension of $K_X+\frac{1}{2}D$, where $(X,D)\to (\mathbb{P}^{2},E)$ is a minimal log resolution, is negative. We prove structure theorems for curves satisfying this conjecture and we finish their classification up to a projective equivalence by describing the ones whose complement admits no $\mathbb{C}^{**}$-fibration. As a consequence, we show that they satisfy the Strong Rigidity Conjecture of Flenner-Zaidenberg. The proofs are based on the almost minimal model program. The obtained list contains one new series of bicuspidal curves., Comment: 50 pages

J-PET is a detector optimized for registration of photons from the electron-positron annihilation via plastic scintillators where photons interact predominantly via Compton scattering. Registration of both primary and scattered photons enables to determinate the linear polarization of the primary photon on the event by event basis with a certain probability. Here we present quantitative results on the feasibility of such polarization measurements of photons from the decay of positronium with the J-PET and explore the physical limitations for the resolution of the polarization determination of 511 keV photons via Compton scattering. For scattering angles of about 82 deg (where the best contrast for polarization measurement is theoretically predicted) we find that the single event resolution for the determination of the polarization is about 40 deg (predominantly due to properties of the Compton effect). However, for samples larger than ten thousand events the J-PET is capable of determining relative average polarization of these photons with the precision of about few degrees. The obtained results open new perspectives for studies of various physics phenomena such as quantum entanglement and tests of discrete symmetries in decays of positronium and extend the energy range of polarization measurements by five orders of magnitude beyond the optical wavelength regime., Comment: 10 pages, 14 figures, submitted to EPJ C

In this paper the technological aspects of the FAZIA array will be explored. After a productive commissioning phase, FAZIA blocks started to measure and give very useful data to explore the physics of Fermi energy heavy-ion reactions. This was possible thanks to many technical measures and innovations developed in the commissioning phase and tuned during the first experimental campaigns. This paper gives a detailed description of the present status of the FAZIA setup from the electronic and mechanical point of view, trying also to trace a path for new improvements and refinements of the apparatus., Comment: 17 pages, 9 figures

The Jagiellonian Positron Emission Tomograph (J-PET) is the first PET device built from plastic scintillators. It is a multi-purpose detector designed for medical imaging and for studies of properties of positronium atoms in porous matter and in living organisms. In this article we report on the commissioning of the J-PET detector in view of studies of positronium decays. We present results of analysis of the positron lifetime measured in the porous polymer. The obtained results prove that J-PET is capable of performing simultaneous imaging of the density distribution of annihilation points as well as positron annihilation lifetime spectroscopy., Comment: 5 pages, 2 figures