Proceedings of International Conferences of Experimental Linguistics, 2019
The paper assesses the consequences of gradience for approaches to variation based on the Princip... more The paper assesses the consequences of gradience for approaches to variation based on the Principles and Parameters model. In particular, the discussion focuses on recent crosslinguistic results obtained through magnitude estimation, a methodology particularly suited to the study of gradient acceptability/grammaticality. Results on superiority and relativised minimality effects in questions are discussed in the light of current theoretical assumptions regarding the locus of crosslinguistic variation.
... E03-1012: Zoltán Alexin; János Csirik; Tibor Gyimóthy; Károly Bibok; Csaba Hatvani; Gábor Pró... more ... E03-1012: Zoltán Alexin; János Csirik; Tibor Gyimóthy; Károly Bibok; Csaba Hatvani; Gábor Prószéky; László Tihanyi Annotated Hungarian ... E03-1024: Alexander Koller; Joachim Niehren; Stefan Thater Underspecification formalisms: Hole semantics as dominance constraints. ...
Meshing the heart and measurement surfaces can be time consuming, especially when dealing with co... more Meshing the heart and measurement surfaces can be time consuming, especially when dealing with complicated geometries or cardiac motion. To overcome this, a meshless method based on the method of fundamental solutions (MFS) has been adapted to non-invasive electrocardiographic imaging (ECGI). In the MFS, potentials are expressed as a summation over a discrete set of virtual point sources placed outside of the domain of interest (named 'pseudo-boundary'). It is well-known that optimal placement of the pseudoboundary can improve the efficacy of the MFS. Despite this, there have been no attempts to optimize their placement in the ECGI problem as far as we are aware. In the standard MFS, the sources are placed in two pseudo-boundaries constructed by inflating and deflating the heart and torso surfaces with respect to the geometric center of the heart. However, for some heart-torso geometries, this geometric center is a poor reference. We here show that adaptive placement of the pseudoboundaries (depending on the distance between the torso electrodes and the nearest heart locations) improves the conditioning of the inverse problem, making it less sensitive to the regularization process.
Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019
Recently, there has been an increasing interest in unsupervised parsers that optimize semanticall... more Recently, there has been an increasing interest in unsupervised parsers that optimize semantically oriented objectives, typically using reinforcement learning. Unfortunately, the learned trees often do not match actual syntax trees well. Shen et al. (2018) propose a structured attention mechanism for language modeling (PRPN), which induces better syntactic structures but relies on ad hoc heuristics. Also, their model lacks interpretability as it is not grounded in parsing actions. In our work, we propose an imitation learning approach to unsupervised parsing, where we transfer the syntactic knowledge induced by the PRPN to a Tree-LSTM model with discrete parsing actions. Its policy is then refined by Gumbel-Softmax training towards a semantically oriented objective. We evaluate our approach on the All Natural Language Inference dataset and show that it achieves a new state of the art in terms of parsing F-score, outperforming our base models, including the PRPN. 1
Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2016
We introduce a new task, visual sense disambiguation for verbs: given an image and a verb, assign... more We introduce a new task, visual sense disambiguation for verbs: given an image and a verb, assign the correct sense of the verb, i.e., the one that describes the action depicted in the image. Just as textual word sense disambiguation is useful for a wide range of NLP tasks, visual sense disambiguation can be useful for multimodal tasks such as image retrieval, image description, and text illustration. We introduce VerSe, a new dataset that augments existing multimodal datasets (COCO and TUHOI) with sense labels. We propose an unsupervised algorithm based on Lesk which performs visual sense disambiguation using textual, visual, or multimodal embeddings. We find that textual embeddings perform well when goldstandard textual annotations (object labels and image descriptions) are available, while multimodal embeddings perform well on unannotated images. We also verify our findings by using the textual and multimodal embeddings as features in a supervised setting and analyse the performance of visual sense disambiguation task. VerSe is made publicly available and can be downloaded at: https://github. com/spandanagella/verse.
Speakers and listeners in a dialogue establish mutual understanding by coordinating their linguis... more Speakers and listeners in a dialogue establish mutual understanding by coordinating their linguistic responses. When a visual scene is present, scan patterns on that scene are also coordinated. However, it is an open question which linguistic and scene factors affect coordination. In this paper, we investigate the coordination of scan patterns during the comprehension and generation of scene descriptions. We manipulate the animacy of the subject and the number of visual referents associated with it. By using Cross Recurrence Analysis, we demonstrate that coordination emerges only during linguistic processing, and that it is especially pronounced for inanimate unambiguous subjects. When the subject is referentially ambiguous (more than one visual object associated with it), scan pattern variability increases to the extent that the animacy effect is neutralized.
We present two magnitude estimation experiments on the role of resumptive pronouns in subject ext... more We present two magnitude estimation experiments on the role of resumptive pronouns in subject extraction in English and German. In the theoretical literature, resumptives have been claimed to remedy effects of island violations, but our results show that this is not the case. Furthermore, our data also confirm the existence of a that-trace effect for both languages, contrary to theoretical claims that this effect is not present in German. On a methodological level, our study demonstrates the importance of controlled experiments for obtaining theoretically crucial linguistic judgments, and the potential of the magnitude estimation paradigm for crosslinguistic comparisons.
The analysis of reading times can pro-vide insights into the processes that under-lie language co... more The analysis of reading times can pro-vide insights into the processes that under-lie language comprehension, with longer reading times indicating greater cognitive load. There is evidence that the language processor is highly predictive, such that prior context allows upcoming ...
@Book{HJCL:2008, editor = {Ron Artstein and Gemma Boleda and Frank Keller and Sabine Schulte im W... more @Book{HJCL:2008, editor = {Ron Artstein and Gemma Boleda and Frank Keller and Sabine Schulte im Walde}, title = {Coling 2008: Proceedings of the workshop on Human Judgements in Computational Linguistics}, month = {August}, year = {2008}, address = {Manchester, UK ...
Proceedings of International Conferences of Experimental Linguistics, 2019
The paper assesses the consequences of gradience for approaches to variation based on the Princip... more The paper assesses the consequences of gradience for approaches to variation based on the Principles and Parameters model. In particular, the discussion focuses on recent crosslinguistic results obtained through magnitude estimation, a methodology particularly suited to the study of gradient acceptability/grammaticality. Results on superiority and relativised minimality effects in questions are discussed in the light of current theoretical assumptions regarding the locus of crosslinguistic variation.
... E03-1012: Zoltán Alexin; János Csirik; Tibor Gyimóthy; Károly Bibok; Csaba Hatvani; Gábor Pró... more ... E03-1012: Zoltán Alexin; János Csirik; Tibor Gyimóthy; Károly Bibok; Csaba Hatvani; Gábor Prószéky; László Tihanyi Annotated Hungarian ... E03-1024: Alexander Koller; Joachim Niehren; Stefan Thater Underspecification formalisms: Hole semantics as dominance constraints. ...
Meshing the heart and measurement surfaces can be time consuming, especially when dealing with co... more Meshing the heart and measurement surfaces can be time consuming, especially when dealing with complicated geometries or cardiac motion. To overcome this, a meshless method based on the method of fundamental solutions (MFS) has been adapted to non-invasive electrocardiographic imaging (ECGI). In the MFS, potentials are expressed as a summation over a discrete set of virtual point sources placed outside of the domain of interest (named 'pseudo-boundary'). It is well-known that optimal placement of the pseudoboundary can improve the efficacy of the MFS. Despite this, there have been no attempts to optimize their placement in the ECGI problem as far as we are aware. In the standard MFS, the sources are placed in two pseudo-boundaries constructed by inflating and deflating the heart and torso surfaces with respect to the geometric center of the heart. However, for some heart-torso geometries, this geometric center is a poor reference. We here show that adaptive placement of the pseudoboundaries (depending on the distance between the torso electrodes and the nearest heart locations) improves the conditioning of the inverse problem, making it less sensitive to the regularization process.
Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 2019
Recently, there has been an increasing interest in unsupervised parsers that optimize semanticall... more Recently, there has been an increasing interest in unsupervised parsers that optimize semantically oriented objectives, typically using reinforcement learning. Unfortunately, the learned trees often do not match actual syntax trees well. Shen et al. (2018) propose a structured attention mechanism for language modeling (PRPN), which induces better syntactic structures but relies on ad hoc heuristics. Also, their model lacks interpretability as it is not grounded in parsing actions. In our work, we propose an imitation learning approach to unsupervised parsing, where we transfer the syntactic knowledge induced by the PRPN to a Tree-LSTM model with discrete parsing actions. Its policy is then refined by Gumbel-Softmax training towards a semantically oriented objective. We evaluate our approach on the All Natural Language Inference dataset and show that it achieves a new state of the art in terms of parsing F-score, outperforming our base models, including the PRPN. 1
Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, 2016
We introduce a new task, visual sense disambiguation for verbs: given an image and a verb, assign... more We introduce a new task, visual sense disambiguation for verbs: given an image and a verb, assign the correct sense of the verb, i.e., the one that describes the action depicted in the image. Just as textual word sense disambiguation is useful for a wide range of NLP tasks, visual sense disambiguation can be useful for multimodal tasks such as image retrieval, image description, and text illustration. We introduce VerSe, a new dataset that augments existing multimodal datasets (COCO and TUHOI) with sense labels. We propose an unsupervised algorithm based on Lesk which performs visual sense disambiguation using textual, visual, or multimodal embeddings. We find that textual embeddings perform well when goldstandard textual annotations (object labels and image descriptions) are available, while multimodal embeddings perform well on unannotated images. We also verify our findings by using the textual and multimodal embeddings as features in a supervised setting and analyse the performance of visual sense disambiguation task. VerSe is made publicly available and can be downloaded at: https://github. com/spandanagella/verse.
Speakers and listeners in a dialogue establish mutual understanding by coordinating their linguis... more Speakers and listeners in a dialogue establish mutual understanding by coordinating their linguistic responses. When a visual scene is present, scan patterns on that scene are also coordinated. However, it is an open question which linguistic and scene factors affect coordination. In this paper, we investigate the coordination of scan patterns during the comprehension and generation of scene descriptions. We manipulate the animacy of the subject and the number of visual referents associated with it. By using Cross Recurrence Analysis, we demonstrate that coordination emerges only during linguistic processing, and that it is especially pronounced for inanimate unambiguous subjects. When the subject is referentially ambiguous (more than one visual object associated with it), scan pattern variability increases to the extent that the animacy effect is neutralized.
We present two magnitude estimation experiments on the role of resumptive pronouns in subject ext... more We present two magnitude estimation experiments on the role of resumptive pronouns in subject extraction in English and German. In the theoretical literature, resumptives have been claimed to remedy effects of island violations, but our results show that this is not the case. Furthermore, our data also confirm the existence of a that-trace effect for both languages, contrary to theoretical claims that this effect is not present in German. On a methodological level, our study demonstrates the importance of controlled experiments for obtaining theoretically crucial linguistic judgments, and the potential of the magnitude estimation paradigm for crosslinguistic comparisons.
The analysis of reading times can pro-vide insights into the processes that under-lie language co... more The analysis of reading times can pro-vide insights into the processes that under-lie language comprehension, with longer reading times indicating greater cognitive load. There is evidence that the language processor is highly predictive, such that prior context allows upcoming ...
@Book{HJCL:2008, editor = {Ron Artstein and Gemma Boleda and Frank Keller and Sabine Schulte im W... more @Book{HJCL:2008, editor = {Ron Artstein and Gemma Boleda and Frank Keller and Sabine Schulte im Walde}, title = {Coling 2008: Proceedings of the workshop on Human Judgements in Computational Linguistics}, month = {August}, year = {2008}, address = {Manchester, UK ...
Uploads
Papers by Trọng Đỗ