• Dowty's (1991) proto-role approach implements this observation–he defines properti... more • Dowty's (1991) proto-role approach implements this observation–he defines properties of prototypical agents and patients–arguments have these properties in varying degrees–the chances of being the subject (object) depends on the amount of its proto-agent (proto-patient) properties–the approach deliberately avoids selecting a prototypical agent and/or patient in every case
We present in this an updated version of the deliverable D3. 3A the actual state of work of SynAF... more We present in this an updated version of the deliverable D3. 3A the actual state of work of SynAF (Syntactic Annotation Framework), which will be soon submitted as a CD (end of September). SynAF has been reviewed in the line of joint discussions with the editors of LAF, LMF and MAF, discussions which started at the plenary ISO TC37 in Beijing (21-26 August 2006), and here we include some modifications suggested by participants to discussion on SynAF at a ISO meeting in Paris (Mai 2007) and at the plenary ISO TC37 ...
Proc. of the Workshop on Ideas and Strategies for Multilingual Grammar Development, ESSLLI 2003, 2003
In this paper we describe the development of a large grammar fragment of Modern Greek in a multil... more In this paper we describe the development of a large grammar fragment of Modern Greek in a multilingual context. The grammar is couched in the theoretical framework of HPSG (Head- Driven Phrase Structure Grammar; (Pollard and Sag, 1994)) and benefits from an organization of semantics based on MRS (Minimal Recursion Semantics; (Copestake et al., 1999), (Copestake et al., 2001)). MRS, a framework for computational semantics, in which the meaning of expressions is represented as a flat bag of Elementary Predications (EPs), combines naturally with ...
Proceedings of the 1st Workshop on Benchmarking: Past, Present and Future, 2021
Where have we been, and where are we going? It is easier to talk about the past than the future. ... more Where have we been, and where are we going? It is easier to talk about the past than the future. These days, benchmarks evolve more bottom up (such as papers with code). 1 There used to be more top-down leadership from government (and industry, in the case of systems, with benchmarks such as SPEC). 2 Going forward, there may be more top-down leadership from organizations like MLPerf 3 and/or influencers like David Ferrucci 4. Tasks such as reading comprehension become even more interesting as we move beyond English. Multilinguality introduces many challenges, and even more opportunities. 1.1.1 Bio Sam Bowman has been on the faculty at NYU since 2016, when he completed PhD with Chris Manning and Chris Potts at Stanford. At NYU, he is a member of the Center for Data Science, the Department of Linguistics, and Courant Institute's Department of Computer Science. His research focuses on data, evaluation techniques, and modeling techniques for sentence and paragraph
Research presented in the paper map PDT annotation 7 → RMRS structures manually annotated corpora... more Research presented in the paper map PDT annotation 7 → RMRS structures manually annotated corpora are valuable barrier: difference in formal descriptions → usage of a resource remains limited ⇒ precisely relate formalisms to overcome these limitations Benefits: flexibility, availability,
The limited availability of in-domain training data is a major issue in the training of applicati... more The limited availability of in-domain training data is a major issue in the training of application-specific neural machine translation models. Professional outsourcing of bilingual data collections is costly and often not feasible. In this paper we analyze the influence of using crowdsourcing as a scalable way to obtain translations of target in-domain data having in mind that the translations can be of a lower quality. We apply crowdsourcing with carefully designed quality controls to create parallel corpora for the educational domain by collecting translations of texts from MOOCs from English to eleven languages, which we then use to fine-tune neural machine translation models previously trained on general-domain data. The results from our research indicate that crowdsourced data collected with proper quality controls consistently yields performance gains over general-domain baseline systems, and systems fine-tuned with pre-existing in-domain corpora.
Massive open online courses have been growing rapidly in size and impact. TraMOOC aims at develop... more Massive open online courses have been growing rapidly in size and impact. TraMOOC aims at developing high-quality translation of all types of text genre included in MOOCs from English into eleven European and BRIC languages (DE, IT, PT, EL, DU, CS, BG, CR, PL, RU, ZH) that are hard to translate into and have weak MT support. Phrase-based and syntax-based SMT models will be developed for addressing language diversity and supporting the language-independent nature of the methodology. For a high quality, automatic translation approach and for adding value to existing infrastructure, extensive and advanced bootstrapping of new resources will be performed. An innovative multimodal automatic and human evaluation schema will further ensure translation quality. For human evaluation, an innovative, strict-access control, timeand cost-efficient crowdsourcing setup will be used. Translation experts, domain experts and end users will also be involved. Separate task mining applications will be e...
Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies Short Papers - HLT '08, 2008
This paper introduces a machine learning method based on bayesian networks which is applied to th... more This paper introduces a machine learning method based on bayesian networks which is applied to the mapping between deep semantic representations and lexical semantic resources. A probabilistic model comprising Minimal Recursion Semantics (MRS) structures and lexicalist oriented semantic features is acquired. Lexical semantic roles enriching the MRS structures are inferred, which are useful to improve the accuracy of deep semantic parsing. Verb classes inference was also investigated, which, together with lexical semantic information provided by VerbNet and PropBank resources, can be substantially beneficial to the parse disambiguation task.
Proceedings of the Second ACL Workshop on Effective Tools and Methodologies for Teaching Natural Language Processing and Computational Linguistics - TeachNLP '05, 2005
This paper describes the cooperation of four European Universities aiming at attracting more stud... more This paper describes the cooperation of four European Universities aiming at attracting more students to European master studies in Language and Communication Technologies. The cooperation has been formally approved within the framework of the new European program "Erasmus Mundus" as a Specific Support Action in 2004. The consortium also aims at creating a sound basis for a joint master program in the field of language technology and computer science.
Proceedings of the 11th International Conference on Parsing Technologies - IWPT '09, 2009
This paper presents a novel approach of incorporating fine-grained treebanking decisions made by ... more This paper presents a novel approach of incorporating fine-grained treebanking decisions made by human annotators as discriminative features for automatic parse disambiguation. To our best knowledge, this is the first work that exploits treebanking decisions for this task. The advantage of this approach is that use of human judgements is made. The paper presents comparative analyses of the performance of discriminative models built using treebanking decisions and state-of-the-art features. We also highlight how differently these features scale when these models are tested on out-of-domain data. We show that, features extracted using treebanking decisions are more efficient, informative and robust compared to traditional features.
Από την ποσοτική ανάλυση δεδοµένων αυθόρµητου λόγου προκύπτει µια προνοµιακή σχέση συγκεκριµένων ... more Από την ποσοτική ανάλυση δεδοµένων αυθόρµητου λόγου προκύπτει µια προνοµιακή σχέση συγκεκριµένων ρηµατικών εκφράσεων (λεκτικά και επιτονικά προσδιοριµένοι κλιτοί ρηµατικοί τύποι) µε συγκεκριµένες ενδολεκτικές δυνάµεις, καθώς και ότι η σχέση αυτή εξαρτάται, σε µεγαλύτερο ή µικρότερο βαθµό, από κοινωνιογλωσσικές µεταβλητές (φύλο, ηλικία, τάξη οµιλητή και ακροατή, µεταξύ τους σχέση, περιβάλλων χώρος, σειρά εµφάνισης λεκτικής πράξης στο λόγο). Οι ρηµατικές αυτές εκφράσεις πρέπει να ...
Page 1. Linguistically Interpreted Corpora 2004 (LINC-04) http://www.coli.uni-sb.de/linc04/progra... more Page 1. Linguistically Interpreted Corpora 2004 (LINC-04) http://www.coli.uni-sb.de/linc04/program. html 1 of 1 07/13/2004 11:12 AM 5th International Workshop on Linguistically Interpreted Corpora (LINC-04) Geneva, 29 August 2004 9:00 – 9:15 Welcome & Introduction Silvia Hansen-Schirra, Stephan Oepen, and Hans Uszkoreit 9:15 – 9:45 The HOLJ Corpus. Supporting Summarisation of Legal Texts Claire Grover, Ben Hachey, and Ian Hughson ...
Introduction-Motivation in natural language processing there has always been a need for compact s... more Introduction-Motivation in natural language processing there has always been a need for compact semantic representations which can express, among others, scope ambiguities, avoiding at the same time disjunctive specifications of many different possible readings; Minimal Recursion Semantics (MRS; Copestake, Flickinger, Sag, and Pollard (1999), Copestake (2002)) provides a framework which supports such semantic representations; Minimal Recoursion Semantics is not considered to be a semantic theory itself, but a meta ...
LFG: Nonlocal Dependencies early work within LFG on the characterization of longdistance dependen... more LFG: Nonlocal Dependencies early work within LFG on the characterization of longdistance dependencies was based on the then-standard assumption that the relationship between a\ ller" and a\ gap" was properly characterized in the c-structure; on analogy to", which relates a c-structure node to its corresponding f-structure, Kaplan and Bresnan 1982 introduced the double arrows to relate a local f-structure to the f-structure of a more distant, displaced constituent; this was done for 2 reasons: 1. rst, the c-structure was thought to provide the ...
Introduction-Motivation in natural language processing there has always been a need for compact s... more Introduction-Motivation in natural language processing there has always been a need for compact semantic representations which can express, among others, scope ambiguities, avoiding at the same time disjunctive specifications of many different possible readings; Minimal Recursion Semantics (MRS; Copestake, Flickinger, Sag, and Pollard (1999), Copestake (2002)) provides a framework which supports such semantic representations; Minimal Recoursion Semantics is not considered to be a semantic theory itself, but a meta ...
• Dowty's (1991) proto-role approach implements this observation–he defines properti... more • Dowty's (1991) proto-role approach implements this observation–he defines properties of prototypical agents and patients–arguments have these properties in varying degrees–the chances of being the subject (object) depends on the amount of its proto-agent (proto-patient) properties–the approach deliberately avoids selecting a prototypical agent and/or patient in every case
We present in this an updated version of the deliverable D3. 3A the actual state of work of SynAF... more We present in this an updated version of the deliverable D3. 3A the actual state of work of SynAF (Syntactic Annotation Framework), which will be soon submitted as a CD (end of September). SynAF has been reviewed in the line of joint discussions with the editors of LAF, LMF and MAF, discussions which started at the plenary ISO TC37 in Beijing (21-26 August 2006), and here we include some modifications suggested by participants to discussion on SynAF at a ISO meeting in Paris (Mai 2007) and at the plenary ISO TC37 ...
Proc. of the Workshop on Ideas and Strategies for Multilingual Grammar Development, ESSLLI 2003, 2003
In this paper we describe the development of a large grammar fragment of Modern Greek in a multil... more In this paper we describe the development of a large grammar fragment of Modern Greek in a multilingual context. The grammar is couched in the theoretical framework of HPSG (Head- Driven Phrase Structure Grammar; (Pollard and Sag, 1994)) and benefits from an organization of semantics based on MRS (Minimal Recursion Semantics; (Copestake et al., 1999), (Copestake et al., 2001)). MRS, a framework for computational semantics, in which the meaning of expressions is represented as a flat bag of Elementary Predications (EPs), combines naturally with ...
Proceedings of the 1st Workshop on Benchmarking: Past, Present and Future, 2021
Where have we been, and where are we going? It is easier to talk about the past than the future. ... more Where have we been, and where are we going? It is easier to talk about the past than the future. These days, benchmarks evolve more bottom up (such as papers with code). 1 There used to be more top-down leadership from government (and industry, in the case of systems, with benchmarks such as SPEC). 2 Going forward, there may be more top-down leadership from organizations like MLPerf 3 and/or influencers like David Ferrucci 4. Tasks such as reading comprehension become even more interesting as we move beyond English. Multilinguality introduces many challenges, and even more opportunities. 1.1.1 Bio Sam Bowman has been on the faculty at NYU since 2016, when he completed PhD with Chris Manning and Chris Potts at Stanford. At NYU, he is a member of the Center for Data Science, the Department of Linguistics, and Courant Institute's Department of Computer Science. His research focuses on data, evaluation techniques, and modeling techniques for sentence and paragraph
Research presented in the paper map PDT annotation 7 → RMRS structures manually annotated corpora... more Research presented in the paper map PDT annotation 7 → RMRS structures manually annotated corpora are valuable barrier: difference in formal descriptions → usage of a resource remains limited ⇒ precisely relate formalisms to overcome these limitations Benefits: flexibility, availability,
The limited availability of in-domain training data is a major issue in the training of applicati... more The limited availability of in-domain training data is a major issue in the training of application-specific neural machine translation models. Professional outsourcing of bilingual data collections is costly and often not feasible. In this paper we analyze the influence of using crowdsourcing as a scalable way to obtain translations of target in-domain data having in mind that the translations can be of a lower quality. We apply crowdsourcing with carefully designed quality controls to create parallel corpora for the educational domain by collecting translations of texts from MOOCs from English to eleven languages, which we then use to fine-tune neural machine translation models previously trained on general-domain data. The results from our research indicate that crowdsourced data collected with proper quality controls consistently yields performance gains over general-domain baseline systems, and systems fine-tuned with pre-existing in-domain corpora.
Massive open online courses have been growing rapidly in size and impact. TraMOOC aims at develop... more Massive open online courses have been growing rapidly in size and impact. TraMOOC aims at developing high-quality translation of all types of text genre included in MOOCs from English into eleven European and BRIC languages (DE, IT, PT, EL, DU, CS, BG, CR, PL, RU, ZH) that are hard to translate into and have weak MT support. Phrase-based and syntax-based SMT models will be developed for addressing language diversity and supporting the language-independent nature of the methodology. For a high quality, automatic translation approach and for adding value to existing infrastructure, extensive and advanced bootstrapping of new resources will be performed. An innovative multimodal automatic and human evaluation schema will further ensure translation quality. For human evaluation, an innovative, strict-access control, timeand cost-efficient crowdsourcing setup will be used. Translation experts, domain experts and end users will also be involved. Separate task mining applications will be e...
Proceedings of the 46th Annual Meeting of the Association for Computational Linguistics on Human Language Technologies Short Papers - HLT '08, 2008
This paper introduces a machine learning method based on bayesian networks which is applied to th... more This paper introduces a machine learning method based on bayesian networks which is applied to the mapping between deep semantic representations and lexical semantic resources. A probabilistic model comprising Minimal Recursion Semantics (MRS) structures and lexicalist oriented semantic features is acquired. Lexical semantic roles enriching the MRS structures are inferred, which are useful to improve the accuracy of deep semantic parsing. Verb classes inference was also investigated, which, together with lexical semantic information provided by VerbNet and PropBank resources, can be substantially beneficial to the parse disambiguation task.
Proceedings of the Second ACL Workshop on Effective Tools and Methodologies for Teaching Natural Language Processing and Computational Linguistics - TeachNLP '05, 2005
This paper describes the cooperation of four European Universities aiming at attracting more stud... more This paper describes the cooperation of four European Universities aiming at attracting more students to European master studies in Language and Communication Technologies. The cooperation has been formally approved within the framework of the new European program "Erasmus Mundus" as a Specific Support Action in 2004. The consortium also aims at creating a sound basis for a joint master program in the field of language technology and computer science.
Proceedings of the 11th International Conference on Parsing Technologies - IWPT '09, 2009
This paper presents a novel approach of incorporating fine-grained treebanking decisions made by ... more This paper presents a novel approach of incorporating fine-grained treebanking decisions made by human annotators as discriminative features for automatic parse disambiguation. To our best knowledge, this is the first work that exploits treebanking decisions for this task. The advantage of this approach is that use of human judgements is made. The paper presents comparative analyses of the performance of discriminative models built using treebanking decisions and state-of-the-art features. We also highlight how differently these features scale when these models are tested on out-of-domain data. We show that, features extracted using treebanking decisions are more efficient, informative and robust compared to traditional features.
Από την ποσοτική ανάλυση δεδοµένων αυθόρµητου λόγου προκύπτει µια προνοµιακή σχέση συγκεκριµένων ... more Από την ποσοτική ανάλυση δεδοµένων αυθόρµητου λόγου προκύπτει µια προνοµιακή σχέση συγκεκριµένων ρηµατικών εκφράσεων (λεκτικά και επιτονικά προσδιοριµένοι κλιτοί ρηµατικοί τύποι) µε συγκεκριµένες ενδολεκτικές δυνάµεις, καθώς και ότι η σχέση αυτή εξαρτάται, σε µεγαλύτερο ή µικρότερο βαθµό, από κοινωνιογλωσσικές µεταβλητές (φύλο, ηλικία, τάξη οµιλητή και ακροατή, µεταξύ τους σχέση, περιβάλλων χώρος, σειρά εµφάνισης λεκτικής πράξης στο λόγο). Οι ρηµατικές αυτές εκφράσεις πρέπει να ...
Page 1. Linguistically Interpreted Corpora 2004 (LINC-04) http://www.coli.uni-sb.de/linc04/progra... more Page 1. Linguistically Interpreted Corpora 2004 (LINC-04) http://www.coli.uni-sb.de/linc04/program. html 1 of 1 07/13/2004 11:12 AM 5th International Workshop on Linguistically Interpreted Corpora (LINC-04) Geneva, 29 August 2004 9:00 – 9:15 Welcome & Introduction Silvia Hansen-Schirra, Stephan Oepen, and Hans Uszkoreit 9:15 – 9:45 The HOLJ Corpus. Supporting Summarisation of Legal Texts Claire Grover, Ben Hachey, and Ian Hughson ...
Introduction-Motivation in natural language processing there has always been a need for compact s... more Introduction-Motivation in natural language processing there has always been a need for compact semantic representations which can express, among others, scope ambiguities, avoiding at the same time disjunctive specifications of many different possible readings; Minimal Recursion Semantics (MRS; Copestake, Flickinger, Sag, and Pollard (1999), Copestake (2002)) provides a framework which supports such semantic representations; Minimal Recoursion Semantics is not considered to be a semantic theory itself, but a meta ...
LFG: Nonlocal Dependencies early work within LFG on the characterization of longdistance dependen... more LFG: Nonlocal Dependencies early work within LFG on the characterization of longdistance dependencies was based on the then-standard assumption that the relationship between a\ ller" and a\ gap" was properly characterized in the c-structure; on analogy to", which relates a c-structure node to its corresponding f-structure, Kaplan and Bresnan 1982 introduced the double arrows to relate a local f-structure to the f-structure of a more distant, displaced constituent; this was done for 2 reasons: 1. rst, the c-structure was thought to provide the ...
Introduction-Motivation in natural language processing there has always been a need for compact s... more Introduction-Motivation in natural language processing there has always been a need for compact semantic representations which can express, among others, scope ambiguities, avoiding at the same time disjunctive specifications of many different possible readings; Minimal Recursion Semantics (MRS; Copestake, Flickinger, Sag, and Pollard (1999), Copestake (2002)) provides a framework which supports such semantic representations; Minimal Recoursion Semantics is not considered to be a semantic theory itself, but a meta ...
Uploads
Papers by Valia Kordoni