Serial Experiments Lain | EPFL Graph Search

Serial Experiments Lain (stylized all lowercase) is a Japanese anime television series created and co-produced by Yasuyuki Ueda, written by Chiaki J. Konaka and directed by Ryūtarō Nakamura. Animated by Triangle Staff and featuring original character designs by Yoshitoshi Abe, the series was broadcast for 13 episodes on TV Tokyo and its affiliates from July to September 1998. The series follows Lain Iwakura, an adolescent girl in suburban Japan, and her relation to the Wired, a global communications network similar to the internet. Lain features surreal and avant-garde imagery and explores philosophical topics such as reality, identity, and communication. The series incorporates creative influences from computer history, cyberpunk, and conspiracy theories. Critics and fans have praised Lain for its originality, visuals, atmosphere, themes, and its dark depiction of a world fraught with paranoia, social alienation, and reliance on technology considered insightful of 21st century life.

CFD simulation of an aerosol mixing chamber

Andrea Giordano

In the last years, the correlation between air pollution and health issues related to respiratory, cardiovascular and digestive systems has become evident. Today, urban aerosols raise the interest of both scientific community and public opinion. METAS, the Swiss Federal Institute of Metrology, takes part in AeroTox, a European Union’s research project involving the development of a reference aerosol calibration infrastructure - a so-called mixing chamber. In this chamber, pure air and particles are injected on top and the resulting aerosol is sampled at the bottom. The quality of this aerosol is assessed according to its concentration homogeneity: the purpose of this master’s project is to improve it. In addition, two research questions were addressed. How much can the mixing chamber dimensions be reduced without affecting the concentration homogeneity? Dimensions are crucial because the mixing chamber must be transportable. Also, how much can the flow rates be reduced without affecting the concentration homogeneity? Computational Fluid Dynamics (CFD) simulations and experiments were employed. Numerical simulations were performed in COMSOL Multiphysics, implementing a particle tracing and a diluted species model. This allowed to investigate the structure of the flow and the involved mixing mechanisms: diffusion, convection and turbulent dispersion. However, only the diluted species model was successful. The simulated concentration at the outlet is perfectly homogeneous. Experiments were carried out using two particle size distributions: NaCl (size peak at 80 nm) and Polystyrene Latex (PSL, size peak at 900 nm). Empirical data validate simulations and show a concentration homogeneity within 5%. Furthermore, uncertainty on the measurements is of 4.24%: the simulated concentration homogeneity thus lies within the uncertainty of the experimental findings. Moreover, experiments show that salt particles reach a higher concentration homogeneity than PSL particles. Finally, in case of salt particles, experiments prove that the flow rates can be halved and even equalized and the length of the mixing chamber can be reduced to 50% without drastically affecting the concentration homogeneity.

2020

MMKG: Multi-modal Knowledge Graphs

Alberto Garcia Duran, Hongyu Li, Ye Liu

We present Mmkg, a collection of three knowledge graphs that contain both numerical features and (links to) images for all entities as well as entity alignments between pairs of KGs. Therefore, multi-relational link prediction and entity matching communities can benefit from this resource. We believe this data set has the potential to facilitate the development of novel multi-modal learning approaches for knowledge graphs. We validate the utility of Mmkg in the sameAs link prediction task with an extensive set of experiments. These experiments show that the task at hand benefits from learning of multiple feature types.

SPRINGER INTERNATIONAL PUBLISHING AG2019

Integrating Language Identification to improve Multilingual Speech Recognition

Holger Caesar

The process of determining the language of a speech utterance is called Language Identification (LID). This task can be very challenging as it has to take into account various language-specific aspects, such as phonetic, phonotactic, vocabulary and grammar-related cues. In multilingual speech recognition we try to find the most likely word sequence that corresponds to an utterance where the language is not known a priori. This is a considerably harder task compared to monolingual speech recognition and it is common to use LID to estimate the current language. In this project we present two general approaches for LID and describe how to integrate them into multilingual speech recognizers. The first approach uses hierarchical multilayer perceptrons to estimate language posterior probabilities given the acoustics in combination with hidden Markov models. The second approach evaluates the output of a multilingual speech recognizer to determine the spoken language. The research is applied to the MediaParl speech corpus that was recorded at the Parliament of the canton of Valais, where people switch from Swiss French to Swiss German or vice versa. Our experiments show that, on that particular data set, LID can be used to significantly improve the performance of multilingual speech recognizers. We will also point out that ASR dependent LID approaches yield the best performance due to higher-level cues and that our systems perform much worse on non-native data

Idiap2012

CFD simulation of an aerosol mixing chamber

Andrea Giordano

2020