This paper presents our progress towards deploying a versatile communication platform in the task... more This paper presents our progress towards deploying a versatile communication platform in the task of highly multilingual live speech translation for conferences and remote meetings live subtitling. The platform has been designed with a focus on very low latency and high flexibility while allowing research prototypes of speech and text processing tools to be easily connected, regardless of where they physically run. We outline our architecture solution and also briefly compare it with the ELG platform. Technical details are provided on the most important components and we summarize the test deployment events we ran so far.
In this paper, we report on several software implementations that we have developed within Prague... more In this paper, we report on several software implementations that we have developed within Prague Arabic Dependency Treebank or some other projects concerned with Arabic Natural Language Processing. We try to guide the reader through some essential tasks and note the solutions that we have designed and used. We as well point to third-party computational systems that the research community might exploit in the future work in this field. Arabic, dependency grammar, treebank, language annotation and processing, application programming. 1.
Prague Arabic Dependency Treebank (PADT), recently published in its first version (Haji et al. 20... more Prague Arabic Dependency Treebank (PADT), recently published in its first version (Haji et al. 2004a) by the Linguistic Data Consortium, is both a collection of multi-level linguistic annotations over Modern Standard Arabic, and a suite of unique software implementations
In this paper, we report on several software implementations that we have developed within Prague... more In this paper, we report on several software implementations that we have developed within Prague Arabic Dependency Treebank or some other projects concerned with Arabic Natural Language Processing. We try to guide the reader through some essential tasks and note the solutions that we have designed and used. We as well point to third-party computational systems that the research community might exploit in the future work in this field.
This paper reports on the status of our ongoing project to extend, refine, and maintain the lexic... more This paper reports on the status of our ongoing project to extend, refine, and maintain the lexicon of the ElixirFM system. The latest release of ElixirFM has been enriched with valency frames for selected verbal lexemes. We briefly outline some other similar projects for Modern Standard Arabic and present crucial axioms of the approach we have adopted, which is Functional Generative Description. Our contribution focuses mainly on describing the annotation procedure, resources and tools, as well as on the representation of the valency phenomena in form of dependency trees in the lexicon. We also point at some intricate issues concerning reciprocal meaning associated with some verbs. Such a valency lexicon for Arabic can find broad use in various NLP tasks and applications, in addition to its reusability by the users of the language.
TrEd — the general-purpose tree editor being used within the two leading dependency treebanking p... more TrEd — the general-purpose tree editor being used within the two leading dependency treebanking projects in Czech and Arabic — has recently been extended with the MorphoTrees annotation context. On the case of Arabic, the novel concept of turning unorganized sets of complex morphological analyses into intuitive hierarchies is presented. Efficiency of disambiguation of MorphoTrees is discussed, and so is the independence of the idea on the language and the implementation.
ELITR (European Live Translator) project aims to create a speech translation system for simultane... more ELITR (European Live Translator) project aims to create a speech translation system for simultaneous subtitling of conferences and online meetings targetting up to 43 languages. The technology is tested by the Supreme Audit Office of the Czech Republic and by alfaview®, a German online conferencing system. Other project goals are to advance document-level and multilingual machine translation, automatic speech recognition, and automatic minuting.
The PADT project might be summarized as an open-ended activity of the Center for Computational Li... more The PADT project might be summarized as an open-ended activity of the Center for Computational Linguistics, the Institute of Formal and Applied Linguistics, and the Institute of Comparative Linguistics, Charles University in Prague, resting in multi-level annotation of Arabic language resources in the light of the theory of Functional Generative Description (Sgall et al., 1986; Hajicova and Sgall, 2003).
is is a summary of the author’s PhD dissertation defended on September 17, 2007 at the Faculty o... more is is a summary of the author’s PhD dissertation defended on September 17, 2007 at the Faculty of Mathematics and Physics, Charles University in Prague. e results comprised in the thesis were obtained within the author’s doctoral studies in Mathematical Linguistics during the years 2001–2007. e complete dissertation is available via http://sourceforge.net/projects/elixir-fm/.
Prague Arabic Dependency Treebank not only consists of multi-level linguistic annotations over th... more Prague Arabic Dependency Treebank not only consists of multi-level linguistic annotations over the language of Modern Standard Arabic, but even provides a variety of unique software implementations designed for general use in Natural Language Processing (NLP). This paper delivers an overview of the recent and most interesting results, findings and innovations within the project.
In this paper, we report on several software implementations that w e have developed within Pragu... more In this paper, we report on several software implementations that w e have developed within Prague Arabic Dependency Treebank or some other projects concerned wit h Arabic Natural Language Processing. We try to guide the reader through some essent ial tasks and note the solutions that we have designed and used. We as well point to third-party computational systems that the research community might exploit in the future work in this fiel d.
We describe our experience with providing automatic simultaneous spoken language translation for ... more We describe our experience with providing automatic simultaneous spoken language translation for an event with human interpreters. We provide a detailed overview of the systems we use, focusing on their interconnection and the issues it brings. We present our tools to monitor the pipeline and a web application to present the results of our SLT pipeline to the end users. Finally, we discuss various challenges we encountered, their possible solutions and we suggest improvements for future deployments.
The issue of information structure in language has been studied extensively both in the Prague Sc... more The issue of information structure in language has been studied extensively both in the Prague School of Linguistics and in the Functional Generative Description [FGD, 11, 5]. This theory of representation of linguistic meaning is the framework for a family of multi-level annotation projects, in particular the Prague Dependency Treebank for Czech [PDT, 2, 3] and the Prague Arabic Dependency Treebank [PADT, 4, 12]. Information structure — the question of ‘the given’ and ‘the new’ in an utterance and how it is expressed — is considered to contribute to the linguistic meaning, and its annotation in PDT is part of the third, the most detailed and abstract level of linguistic description. Next to determining which elements in a sentence are context-bound and which are non-bound (the elementary distinctive feature from which the topic–focus dichotomy is derived), attention is also paid to the resolution of anaphoric relations and to capturing the communicative dynamism of a proposition (a...
types than just strings. Inflected forms need not be merged with roots yet, and can retain the in... more types than just strings. Inflected forms need not be merged with roots yet, and can retain the internal structure: . . . ? paraVerbC Feminine Plural "u" FCuL −→ "u" >>| FCuL |<< "na" . . . ? merge "k t b" (Prefix "u" >| FCuL |< Suffix "na") −→ "uktubna" uktubna á . J » @ fem.pl. ‘write!’ The highlight of the Arabic morphology is that the ‘irregular’ inflection actually rests in strictly observing some additional rules, the nature of which is phonological. Therefore, surprisingly, ElixirFM does not even distinguish between verbal and nominal word formation when enforcing these rules. This reduces the number of paradigms to the prototypical 3 verbal and 5 nominal! Yet, the model is efficient. Nominal inflection is also driven by the information from the lexicon and by phonology. Note that the morphophonemic patterns and the Morphs a templates are actually extremely informative. We can use them as d...
This paper presents an automatic speech translation system aimed at live subtitling of conference... more This paper presents an automatic speech translation system aimed at live subtitling of conference presentations. We describe the overall architecture and key processing components. More importantly, we explain our strategy for building a complex system for end-users from numerous individual components, each of which has been tested only in laboratory conditions. The system is a working prototype that is routinely tested in recognizing English, Czech, and German speech and presenting it translated simultaneously into 42 target languages.
This research note reports on the work in progress which regards automatic transformation of phra... more This research note reports on the work in progress which regards automatic transformation of phrase-structure syntactic trees of Arabic into dependency-driven analytical ones. Guidelines for these descriptions have been developed at the Linguistic Data Consortium, University of Pennsylvania, and at the Faculty of Mathematics and Physics and the Faculty of Arts, Charles University in Prague, respectively. The transformation consists of (i) a recursive function translating the topology of a phrase tree into a corresponding dependency tree, and (ii) a procedure assigning analytical functions to the nodes of the dependency tree. Apart from an outline of the annotation schemes and a deeper insight into these procedures, model application of the transformation is given herein.
We present an implementation of a Haskell library for processing the Arabic language in the ArabT... more We present an implementation of a Haskell library for processing the Arabic language in the ArabTEX transliteration (Lagally, 2004), a non-trivial and multi-purpose notation for encoding Arabic orthographies and phonetic transcriptions in parallel. Our approach relies on the Pure Functional Parsing library developed in (Ljunglöf, 2002), which we accommodate to our problem and partly extend. In the general view, we describe two alternative algorithms for longest-match deterministic parsing and rewriting, present the monadic-style grammars formalizing the relation of the script and the sound in Arabic, and promote modular design in systems for modeling or processing natural languages.
This paper presents our progress towards deploying a versatile communication platform in the task... more This paper presents our progress towards deploying a versatile communication platform in the task of highly multilingual live speech translation for conferences and remote meetings live subtitling. The platform has been designed with a focus on very low latency and high flexibility while allowing research prototypes of speech and text processing tools to be easily connected, regardless of where they physically run. We outline our architecture solution and also briefly compare it with the ELG platform. Technical details are provided on the most important components and we summarize the test deployment events we ran so far.
In this paper, we report on several software implementations that we have developed within Prague... more In this paper, we report on several software implementations that we have developed within Prague Arabic Dependency Treebank or some other projects concerned with Arabic Natural Language Processing. We try to guide the reader through some essential tasks and note the solutions that we have designed and used. We as well point to third-party computational systems that the research community might exploit in the future work in this field. Arabic, dependency grammar, treebank, language annotation and processing, application programming. 1.
Prague Arabic Dependency Treebank (PADT), recently published in its first version (Haji et al. 20... more Prague Arabic Dependency Treebank (PADT), recently published in its first version (Haji et al. 2004a) by the Linguistic Data Consortium, is both a collection of multi-level linguistic annotations over Modern Standard Arabic, and a suite of unique software implementations
In this paper, we report on several software implementations that we have developed within Prague... more In this paper, we report on several software implementations that we have developed within Prague Arabic Dependency Treebank or some other projects concerned with Arabic Natural Language Processing. We try to guide the reader through some essential tasks and note the solutions that we have designed and used. We as well point to third-party computational systems that the research community might exploit in the future work in this field.
This paper reports on the status of our ongoing project to extend, refine, and maintain the lexic... more This paper reports on the status of our ongoing project to extend, refine, and maintain the lexicon of the ElixirFM system. The latest release of ElixirFM has been enriched with valency frames for selected verbal lexemes. We briefly outline some other similar projects for Modern Standard Arabic and present crucial axioms of the approach we have adopted, which is Functional Generative Description. Our contribution focuses mainly on describing the annotation procedure, resources and tools, as well as on the representation of the valency phenomena in form of dependency trees in the lexicon. We also point at some intricate issues concerning reciprocal meaning associated with some verbs. Such a valency lexicon for Arabic can find broad use in various NLP tasks and applications, in addition to its reusability by the users of the language.
TrEd — the general-purpose tree editor being used within the two leading dependency treebanking p... more TrEd — the general-purpose tree editor being used within the two leading dependency treebanking projects in Czech and Arabic — has recently been extended with the MorphoTrees annotation context. On the case of Arabic, the novel concept of turning unorganized sets of complex morphological analyses into intuitive hierarchies is presented. Efficiency of disambiguation of MorphoTrees is discussed, and so is the independence of the idea on the language and the implementation.
ELITR (European Live Translator) project aims to create a speech translation system for simultane... more ELITR (European Live Translator) project aims to create a speech translation system for simultaneous subtitling of conferences and online meetings targetting up to 43 languages. The technology is tested by the Supreme Audit Office of the Czech Republic and by alfaview®, a German online conferencing system. Other project goals are to advance document-level and multilingual machine translation, automatic speech recognition, and automatic minuting.
The PADT project might be summarized as an open-ended activity of the Center for Computational Li... more The PADT project might be summarized as an open-ended activity of the Center for Computational Linguistics, the Institute of Formal and Applied Linguistics, and the Institute of Comparative Linguistics, Charles University in Prague, resting in multi-level annotation of Arabic language resources in the light of the theory of Functional Generative Description (Sgall et al., 1986; Hajicova and Sgall, 2003).
is is a summary of the author’s PhD dissertation defended on September 17, 2007 at the Faculty o... more is is a summary of the author’s PhD dissertation defended on September 17, 2007 at the Faculty of Mathematics and Physics, Charles University in Prague. e results comprised in the thesis were obtained within the author’s doctoral studies in Mathematical Linguistics during the years 2001–2007. e complete dissertation is available via http://sourceforge.net/projects/elixir-fm/.
Prague Arabic Dependency Treebank not only consists of multi-level linguistic annotations over th... more Prague Arabic Dependency Treebank not only consists of multi-level linguistic annotations over the language of Modern Standard Arabic, but even provides a variety of unique software implementations designed for general use in Natural Language Processing (NLP). This paper delivers an overview of the recent and most interesting results, findings and innovations within the project.
In this paper, we report on several software implementations that w e have developed within Pragu... more In this paper, we report on several software implementations that w e have developed within Prague Arabic Dependency Treebank or some other projects concerned wit h Arabic Natural Language Processing. We try to guide the reader through some essent ial tasks and note the solutions that we have designed and used. We as well point to third-party computational systems that the research community might exploit in the future work in this fiel d.
We describe our experience with providing automatic simultaneous spoken language translation for ... more We describe our experience with providing automatic simultaneous spoken language translation for an event with human interpreters. We provide a detailed overview of the systems we use, focusing on their interconnection and the issues it brings. We present our tools to monitor the pipeline and a web application to present the results of our SLT pipeline to the end users. Finally, we discuss various challenges we encountered, their possible solutions and we suggest improvements for future deployments.
The issue of information structure in language has been studied extensively both in the Prague Sc... more The issue of information structure in language has been studied extensively both in the Prague School of Linguistics and in the Functional Generative Description [FGD, 11, 5]. This theory of representation of linguistic meaning is the framework for a family of multi-level annotation projects, in particular the Prague Dependency Treebank for Czech [PDT, 2, 3] and the Prague Arabic Dependency Treebank [PADT, 4, 12]. Information structure — the question of ‘the given’ and ‘the new’ in an utterance and how it is expressed — is considered to contribute to the linguistic meaning, and its annotation in PDT is part of the third, the most detailed and abstract level of linguistic description. Next to determining which elements in a sentence are context-bound and which are non-bound (the elementary distinctive feature from which the topic–focus dichotomy is derived), attention is also paid to the resolution of anaphoric relations and to capturing the communicative dynamism of a proposition (a...
types than just strings. Inflected forms need not be merged with roots yet, and can retain the in... more types than just strings. Inflected forms need not be merged with roots yet, and can retain the internal structure: . . . ? paraVerbC Feminine Plural "u" FCuL −→ "u" >>| FCuL |<< "na" . . . ? merge "k t b" (Prefix "u" >| FCuL |< Suffix "na") −→ "uktubna" uktubna á . J » @ fem.pl. ‘write!’ The highlight of the Arabic morphology is that the ‘irregular’ inflection actually rests in strictly observing some additional rules, the nature of which is phonological. Therefore, surprisingly, ElixirFM does not even distinguish between verbal and nominal word formation when enforcing these rules. This reduces the number of paradigms to the prototypical 3 verbal and 5 nominal! Yet, the model is efficient. Nominal inflection is also driven by the information from the lexicon and by phonology. Note that the morphophonemic patterns and the Morphs a templates are actually extremely informative. We can use them as d...
This paper presents an automatic speech translation system aimed at live subtitling of conference... more This paper presents an automatic speech translation system aimed at live subtitling of conference presentations. We describe the overall architecture and key processing components. More importantly, we explain our strategy for building a complex system for end-users from numerous individual components, each of which has been tested only in laboratory conditions. The system is a working prototype that is routinely tested in recognizing English, Czech, and German speech and presenting it translated simultaneously into 42 target languages.
This research note reports on the work in progress which regards automatic transformation of phra... more This research note reports on the work in progress which regards automatic transformation of phrase-structure syntactic trees of Arabic into dependency-driven analytical ones. Guidelines for these descriptions have been developed at the Linguistic Data Consortium, University of Pennsylvania, and at the Faculty of Mathematics and Physics and the Faculty of Arts, Charles University in Prague, respectively. The transformation consists of (i) a recursive function translating the topology of a phrase tree into a corresponding dependency tree, and (ii) a procedure assigning analytical functions to the nodes of the dependency tree. Apart from an outline of the annotation schemes and a deeper insight into these procedures, model application of the transformation is given herein.
We present an implementation of a Haskell library for processing the Arabic language in the ArabT... more We present an implementation of a Haskell library for processing the Arabic language in the ArabTEX transliteration (Lagally, 2004), a non-trivial and multi-purpose notation for encoding Arabic orthographies and phonetic transcriptions in parallel. Our approach relies on the Pure Functional Parsing library developed in (Ljunglöf, 2002), which we accommodate to our problem and partly extend. In the general view, we describe two alternative algorithms for longest-match deterministic parsing and rewriting, present the monadic-style grammars formalizing the relation of the script and the sound in Arabic, and promote modular design in systems for modeling or processing natural languages.
Uploads
Papers by Otakar Smrž