We present CWRCzech, Click Web Ranking dataset for Czech, a 100M query-document Czech click dataset for relevance ranking with user behavior data collected from search engine logs of Seznam.cz. To the best of our knowledge, CWRCzech is the largest click dataset with raw text published so far. It provides document positions in the search results as well as information about user behavior: 27.6M clicked documents and 10.8M dwell times. In addition, we also publish a manually annotated Czech test for the relevance task, containing nearly 50k query-document pairs, each annotated by at least 2 annotators. Finally, we analyze how the user behavior data improve relevance ranking and show that models trained on data automatically harnessed at sufficient scale can surpass the performance of models trained on human annotated data. CWRCzech is published under an academic non-commercial license and is available to the research community at https://github.com/seznam/CWRCzech.

We propose a new end-to-end architecture for automatic speech recognition that expands the “listen, attend and spell” (LAS) paradigm. While the main word-predicting network is trained to predict words, the secondary, speller network, is optimized to predict word spellings from inner representations of the main network (e.g. word embeddings or context vectors from the attention module). We show that this joint training improves the word error rate of a word-based system and enables solving additional tasks, such as out-of-vocabulary word detection and recovery. The tests are conducted on LibriSpeech dataset consisting of 1000 h of read speech.

The way people describe where things are is one of the central questions of spatial information theory and has been the subject of considerable research. We investigate one particular type of location description, fictive motion (as in, The range runs along the coast). The use of this structure is known to highlight particular properties of the described entity, as well as to convey its configuration in physical space in an effective way. We annotated 496 fictive motion structures in seven corpora that represent different types of spatial discourse – news, travel blogs, texts describing outdoor pursuits and local history, as well as image and location descriptions. We analysed the results not only by examining the distribution of fictive motion structures across corpora, but also by exploring and comparing the semantic categories of verbs used in fictive motion. Our findings, first, add to our knowledge of location description strategies that go beyond prototypical locative phrases. They further reveal how the use of fictive motion varies across types of spatial discourse and reflects the nature of the described environment. Methodologically, we highlight the benefits of a cross-corpora analysis in the study of spatial language use across a variety of contexts.

The paper presents a new approach to extracting useful information from out-of-vocabulary (OOV) speech regions in ASR system output. The system makes use of a hybrid decoding network with both words and sub-word units. In the decoded lattices, candidates for OOV regions are identified as sub-graphs of sub-word units. To facilitate OOV word recovery, we search for recurring OOV s by clustering the detected candidate OOV s. The metrics for clustering is based on a comparison of the sub-graphs corresponding to the OOV candidates. The proposed method discovers repeating out-of-vocabulary words and finds their graphemic representation more robustly than more conventional techniques taking into account only one best sub-word string hypotheses.

Fictive motion (e.g. ‘The highway runs along the coast’) is a pervasive phenomenon in language that can imply both a static and a moving observer. In a corpus of alpine narratives, it is used in three types of spatial descriptions: conveying the actual motion of the observer, describing a vista and communicating encyclopaedic spatial knowledge. This study takes a knowledge-based approach to develop rules for automated extraction and classification of these types based on an annotated corpus of fictive motion instances. In particular, we identify the differences in the set of concepts involved into the production of the three types of descriptions, followed by their linguistic operationalization. Based on that, we build a set of rules that classify fictive motion with an overall precision of 0.87 and recall of 0.71. The article highlights the importance of examining spatially rich, naturally occurring corpora for the lines of work dealing with the automated interpretation of spatial information in texts, as well as, more broadly, investigation of spatial language involved into various types of spatial discourse.

Multilingual training of neural networks for ASR is widely studied these days. It has been shown that languages with little training data can benefit largely from multilingual resources. We have evaluated possible ways of adaptation of multilingual stacked bottle-neck hierarchy to target domain. This paper extends our latest work and focuses on the impact certain aspects have on the performance of an adapted neural network feature extractor. First, the performance of adapted multilingual networks preliminarily trained on different languages is studied. Next, the effect of different target units - phonemes vs. triphone states - used for multilingual NN training is evaluated. Then the impact of an increasing number of languages used for multilingual NN training is investigated. Here the condition of constant amount of data is added to separately control the influence of larger language variability and larger amount of data. The effect of adding languages from a different domain is also evaluated. Finally a study is performed where a language with the phonetic structure similar to the target's one is added to multilingual training data.

The paper addresses manual and semi-automatic approaches to building a multilingual phoneme set for automatic speech recognition. The first approach involves mapping and reduction of the phoneme set based on IPA and expert knowledge, the later one involves phoneme confusion matrix generated by a neural network. The comparison is done for 8 languages selected from GlobalPhone on three scenarios: 1) multilingual system with abundant data for all the languages, 2) multilingual systems excluding target language 3) multilingual systems with small amount of data for target languages. For 3), the multilingual system brought improvement for languages close enough to the others in the set.

In this paper we present novel language-independent bottleneck (BN) feature extraction framework. In our experiments we have used Multilingual Artificial Neural Network (ANN), where each language is modelled by separate output layer, while all the hidden layers jointly model the variability of all the source languages. The key idea is that the entire ANN is trained on all the languages simultaneously, thus the BN-features are not biased towards any of the languages. Exactly for this reason, the final BN-features are considered as language independent. In the experiments with GlobalPhone database, we show that Multilingual BN-features consistently outperform Monolingual BN-features. Also, cross-lingual generalization is evaluated, where we train on 5 source languages and test on 3 other languages. The results show that the ANN can produce very good BN-features even for unseen languages, in some cases even better than if we trained the ANN on the target language only.