@inproceedings{Baumann, author = {Baumann, Timo}, title = {Learning to Determine Who is the Better Speaker}, series = {Proceedings of the 9th International Conference on Speech Prosody, 13-16 June 2018, Poznań, Poland}, booktitle = {Proceedings of the 9th International Conference on Speech Prosody, 13-16 June 2018, Poznań, Poland}, publisher = {ISCA}, doi = {10.21437/SpeechProsody.2018-165}, pages = {819 -- 822}, abstract = {Speech can be more or less likable in various ways and comparing speakers by likability has important applications such as speaker selection or matching. Determining the likability of a speaker is a difficult task which can be simplified by breaking it down into pairwise preference decisions. Using a corpus of 5440 pairwise preference ratings collected previously through crowd-sourcing, we train classifiers to determine which of two speakers is "better". We find that modeling the speech feature sequences using LSTMs outperforms conventional methods that pre-aggregate feature averages by a large margin, indicating that the prosodic structure should be taken into account when determining speech quality. Our classifier reaches an accuracy of 97 \% for coarse-grained decisions, where differences between speech quality in both stimuli is relatively large.}, language = {en} } @inproceedings{Baumann, author = {Baumann, Timo}, title = {Highly Responsive Vocal Interaction through Incremental Processing}, series = {Vocal Interactivity in-and-between Humans, Animals and Robots (VIHAR), October 30-4, 2016}, volume = {6}, booktitle = {Vocal Interactivity in-and-between Humans, Animals and Robots (VIHAR), October 30-4, 2016}, number = {10}, editor = {Moore, Roger K. and Thill, Serge and Marxer, Ricard}, publisher = {Schloss Dagstuhl - Leibniz-Zentrum f{\"u}r Informatik}, address = {Dagstuhl, Germany}, doi = {10.4230/DagRep.6.10.154}, pages = {154 -- 194}, abstract = {This seminar was held in late 2016 and brought together, for the first time, researchers studying vocal interaction in a variety of different domains covering communications between all possible combinations of humans, animals, and robots. While each of these sub-domains has extensive histories of research progress, there is much potential for cross-fertilisation that currently remains underexplored. This seminar aimed at bridging this gap. In this report, we present the nascent research field of VIHAR and the major outputs from our seminar in the form of prioritised open research questions, abstracts from stimulus talks given by prominent researchers in their respective fields, and open problem statements by all participants.}, language = {en} } @inproceedings{MeyerSickendiekHusseinBaumann, author = {Meyer-Sickendiek, Burkhard and Hussein, Hussein and Baumann, Timo}, title = {Rhythmicalizer: Data Analysis for the Identification of Rhythmic Patterns in Readout Poetry}, series = {INFORMATIK 2017, Tagung vom 25.- 29. September 2017 in Chemnitz}, booktitle = {INFORMATIK 2017, Tagung vom 25.- 29. September 2017 in Chemnitz}, editor = {Eibl, Maximilian and Gaedke, Martin}, publisher = {Gesellschaft f{\"u}r Informatik}, address = {Bonn}, isbn = {978-3-88579-669-5}, doi = {10.18420/in2017_218}, pages = {2189 -- 2200}, abstract = {The most important development in modern and postmodern poetry is the replacement of traditional meter by new rhythmical patterns. Ever since Walt Whitman's Leaves of Grass (1855), modern (nineteenth-to twenty-first-century) poets have been searching for novel forms of prosody, accent, rhythm, and intonation. Along with the rejection of older metrical units such as the iamb or trochee, a structure of lyrical language was developed that renounced traditional forms like rhyme and meter. This development is subsumed under the term free verse prosody. Our project will test this theory by applying machine learning or deep learning techniques to a corpus of modern and postmodern poems as read aloud by the original authors. To this end, we examine "lyrikline", the most famous online portal for spoken poetry. First, about 17 different patterns being characteristic for the lyrikline-poems have been identified by the philological scholar of this project. This identification was based on a certain philological method including three different steps: a) grammetrical ranking; b) rhythmic phrasing; and c) mapping rubato and prosodic phrasing. In this paper we will show how to combine this philological and a digital analysis by using the prosody detection available in speech processing technology. In order to analyse the data, we want to use different tools for the following tasks: PoS-tagging, alignment, intonation, phrases and pauses, and tempo. We also analyzed the lyrikline-data by identifying the occurrence of the mentioned patterns. This analysis is a first step towards an automatic classification based on machine learning or deep learning techniques.}, language = {en} } @inproceedings{BaumannMeyerSickendiek, author = {Baumann, Timo and Meyer-Sickendiek, Burkhard}, title = {Large-scale Analysis of Spoken Free-verse Poetry}, series = {Proceedings of the Workshop on Language Technology Resources and Tools for Digital Humanities, Osaka, Japan, December 11-17 2016}, booktitle = {Proceedings of the Workshop on Language Technology Resources and Tools for Digital Humanities, Osaka, Japan, December 11-17 2016}, publisher = {The COLING 2016 Organizing Committee}, pages = {125 -- 130}, abstract = {Most modern and post-modern poems have developed a post-metrical idea of lyrical prosody that employs rhythmical features of everyday language and prose instead of a strict adherence to rhyme and metrical schemes. This development is subsumed under the term free verse prosody. We present our methodology for the large-scale analysis of modern and post-modern poetry in both their written form and as spoken aloud by the author. We employ language processing tools to align text and speech, to generate a null-model of how the poem would be spoken by a na{\"i}ve reader, and to extract contrastive prosodic features used by the poet. On these, we intend to build our model of free verse prosody, which will help to understand, differentiate and relate the different styles of free verse poetry. We plan to use our processing scheme on large amounts of data to iteratively build models of styles, to validate and guide manual style annotation, to identify further rhythmical categories, and ultimately to broaden our understanding of free verse poetry. In this paper, we report on a proof-of-concept of our methodology using smaller amounts of poems and a limited set of features. We find that our methodology helps to extract differentiating features in the authors' speech that can be explained by philological insight. Thus, our automatic method helps to guide the literary analysis and this in turn helps to improve our computational models.}, language = {en} } @inproceedings{HusseinMeyerSickendiekBaumann, author = {Hussein, Hussein and Meyer-Sickendiek, Burkhard and Baumann, Timo}, title = {Tonality in Language: The Generative Theory of Tonal Music as a Framework for Prosodic Analysis of Poetry}, series = {6th International Symposium on Tonal Aspects of Languages (TAL 2018), 18-20 June 2018, Berlin, Germany}, booktitle = {6th International Symposium on Tonal Aspects of Languages (TAL 2018), 18-20 June 2018, Berlin, Germany}, publisher = {ISCA}, doi = {10.21437/TAL.2018-36}, pages = {178 -- 182}, abstract = {This contribution focuses on structural similarities between tonality and cadences in music on the one hand, and rhythmical patterns in poetic languages respectively poetry on the other hand. We investigate two exemplary rhythmical patterns in modern and postmodern poetry to detect these tonality-like features in poetic language: The Parlando and the Variable Foot. German poems readout from the original poets are collected from the webpage of our partner lyrikline. We compared these rhythmical features with tonality rules, explained in two important theoretical volumes: The Generative Theory of Tonal Music and the Rhythmic Phrasing in English Verse. Using both volumes, we focused on a certain combination of four different features: The grouping structure, the metrical structure, the time-span-variation and the prolongation, in order to detect the two important rhythmical patterns which use tonality-like features in poetic language (Parlando and Variable Foot). Different features including pause and parser information are used in this classification process. The best classification result, calculated by the f-measure, for Parlando and Variable Foot is 0.69.}, language = {en} } @incollection{BaumannKenningtonHoughetal., author = {Baumann, Timo and Kennington, Casey and Hough, Julian and Schlangen, David}, title = {Recognising Conversational Speech: What an Incremental ASR Should Do for a Dialogue System and How to Get There}, series = {Dialogues with social robots: Enablements analyses and evaluation}, booktitle = {Dialogues with social robots: Enablements analyses and evaluation}, editor = {Jokinen, Kristiina and Wilcock, Graham}, publisher = {Springer}, address = {Singapore}, isbn = {978-981-10-2584-6}, doi = {10.1007/978-981-10-2585-3_35}, pages = {421 -- 432}, abstract = {Automatic speech recognition (ASR) is not only becoming increasingly accurate, but also increasingly adapted for producing timely, incremental output. However, overall accuracy and timeliness alone are insufficient when it comes to interactive dialogue systems which require stability in the output and responsivity to the utterance as it is unfolding. Furthermore, for a dialogue system to deal with phenomena such as disfluencies, to achieve deep understanding of user utterances these should be preserved or marked up for use by downstream components, such as language understanding, rather than be filtered out. Similarly, word timing can be informative for analyzing deictic expressions in a situated environment and should be available for analysis. Here we investigate the overall accuracy and incremental performance of three widely used systems and discuss their suitability for the aforementioned perspectives. From the differing performance along these measures we provide a picture of the requirements for incremental ASR in dialogue systems and describe freely available tools for using and evaluating incremental ASR.}, language = {en} } @inproceedings{KoehnBaumann, author = {K{\"o}hn, Arne and Baumann, Timo}, title = {Predictive Incremental Parsing Helps Language Modeling}, series = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, Osaka, Japan}, booktitle = {Proceedings of COLING 2016, the 26th International Conference on Computational Linguistics: Technical Papers, Osaka, Japan}, publisher = {The COLING 2016 Organizing Committee}, pages = {268 -- 277}, abstract = {Predictive incremental parsing produces syntactic representations of sentences as they are produced, e.g. by typing or speaking. In order to generate connected parses for such unfinished sentences, upcoming word types can be hypothesized and structurally integrated with already realized words. For example, the presence of a determiner as the last word of a sentence prefix may indicate that a noun will appear somewhere in the completion of that sentence, and the determiner can be attached to the predicted noun. We combine the forward-looking parser predictions with backward-looking N-gram histories and analyze in a set of experiments the impact on language models, i.e. stronger discriminative power but also higher data sparsity. Conditioning N-gram models, MaxEnt models or RNN-LMs on parser predictions yields perplexity reductions of about 6\%. Our method (a) retains online decoding capabilities and (b) incurs relatively little computational overhead which sets it apart from previous approaches that use syntax for language modeling. Our method is particularly attractive for modular systems that make use of a syntax parser anyway, e.g. as part of an understanding pipeline where predictive parsing improves language modeling at no additional cost.}, language = {en} } @inproceedings{Baumann, author = {Baumann, Timo}, title = {Large-Scale Speaker Ranking from Crowdsourced Pairwise Listener Ratings}, series = {Interspeech 2017 : 20-24 August 2017}, booktitle = {Interspeech 2017 : 20-24 August 2017}, publisher = {ISCA}, doi = {10.21437/Interspeech.2017-1697}, pages = {2262 -- 2266}, abstract = {Speech quality and likability is a multi-faceted phenomenon consisting of a combination of perceptory features that cannot easily be computed nor weighed automatically. Yet, it is often easy to decide which of two voices one likes better, even though it would be hard to describe why, or to name the underlying basic perceptory features. Although likability is inherently subjective and individual preferences differ frequently, generalizations are useful and there is often a broad intersubjective consensus about whether one speaker is more likable than another. However, breaking down likability rankings into pairwise comparisons leads to a quadratic explosion of rating pairs. We present a methodology and software to efficiently create a likability ranking for many speakers from crowdsourced pairwise likability ratings. We collected pairwise likability ratings for many (>220) speakers from many raters (>160) and turn these ratings into one likability ranking. We investigate the resulting speaker ranking stability under different conditions: limiting the number of ratings and the dependence on rater and speaker characteristics. We also analyze the ranking wrt. acoustic correlates to find out what factors influence likability. We publish our ranking and the underlying ratings in order to facilitate further research.}, language = {en} } @inproceedings{RohdeBaumann, author = {Rohde, Marcel and Baumann, Timo}, title = {Navigating the Spoken Wikipedia}, series = {Proceedings of the 7th Workshop on Speech and Language Processing for Assistive Technologies (SLPAT 2016)}, booktitle = {Proceedings of the 7th Workshop on Speech and Language Processing for Assistive Technologies (SLPAT 2016)}, doi = {10.21437/SLPAT.2016-2}, pages = {9 -- 13}, abstract = {The Spoken Wikipedia project unites volunteer readers of encyclopedic entries. Their recordings make encyclopedic knowledge accessible to persons who are unable to read (out of alexia, visual impairment, or because their sight is currently occupied, e. g. while driving). However, on Wikipedia, recordings are available as raw audio files that can only be consumed linearly, without the possibility for targeted navigation or search. We present a reading application which uses an alignment between the recording, text and article structure and which allows to navigate spoken articles, through a graphical or voice-based user interface (or a combination thereof). We present the results of a usability study in which we compare the two interaction modalities. We find that both types of interaction enable users to navigate articles and to find specific information much more quickly compared to a sequential presentation of the full article. In particular when the VUI is not restricted by speech recognition and understanding issues, this interface is on par with the graphical interface and thus a real option for browsing the Wikipedia without the need for vision or reading.}, language = {en} } @inproceedings{KoehnStegenBaumann, author = {K{\"o}hn, Arne and Stegen, Florian and Baumann, Timo}, title = {Mining the Spoken Wikipedia for Speech Data and Beyond}, series = {Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)}, booktitle = {Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC'16)}, publisher = {European Language Resources Association (ELRA)}, pages = {4644 -- 4647}, abstract = {We present a corpus of time-aligned spoken data of Wikipedia articles as well as the pipeline that allows to generate such corpora for many languages. There are initiatives to create and sustain spoken Wikipedia versions in many languages and hence the data is freely available, grows over time, and can be used for automatic corpus creation. Our pipeline automatically downloads and aligns this data. The resulting German corpus currently totals 293h of audio, of which we align 71h in full sentences and another 86h of sentences with some missing words. The English corpus consists of 287h, for which we align 27h in full sentence and 157h with some missing words. Results are publically available.}, language = {en} } @inproceedings{KirschSteinickeBaumann, author = {Kirsch, Kolja and Steinicke, Frank and Baumann, Timo}, title = {Entwicklung und Evaluierung einer Notenwender-App f{\"u}r Klaviernoten}, series = {Mensch und Computer 2016 - Tagungsband, 4. - 7. September 2016, Aachen}, booktitle = {Mensch und Computer 2016 - Tagungsband, 4. - 7. September 2016, Aachen}, publisher = {Gesellschaft f{\"u}r Informatik e.V.}, address = {Aachen}, doi = {10.18420/muc2016-mci-0159}, abstract = {Das st{\"a}ndige Umbl{\"a}ttern von Noten ist f{\"u}r Musiker ein wiederkehrendes Problem. Dieses wird h{\"a}ufig durch einen Assistenten des Musikers, dem sogenannten Notenwender, gel{\"o}st. Diese Unterst{\"u}tzung haben allerdings viele Musiker nur selten w{\"a}hrend des {\"U}bens. In diesem Artikel stellen wir eine Anwendung f{\"u}r mobile Ger{\"a}te vor, die auf verschiedene Arten das Umbl{\"a}ttern von Klavierpartituren unterst{\"u}tzt. In einer Studie mit professionellen Musikern und Klaviersch{\"u}lern wurden diese Arten gegeneinander abgewogen. Die Ergebnisse zeigen auf, dass computer-unterst{\"u}tztes Bl{\"a}ttern Vorteile gegen{\"u}ber herk{\"o}mmlichem Bl{\"a}ttern hat.}, language = {de} }