Cristina Bosco

Developing Corpora for Sentiment Analysis: The Case of Irony and Senti-TUT

Senti-TUT-an ongoing Italian project that investigates sentiment and irony in online political discussions-illustrates how to develop corpora for mining and analyzing opinion and sentiment in social media.

Multiple-step Treebank Conversion: From Dependency to Penn Format

Whilst the degree to which a treebank subscribes to a specific linguistic theory limits the usefulness of the resource, the availability of more formats for the same resource plays a crucial role both in NLP and linguistics. Conversion tools and multi-format treebanks are useful for investigating portability of NLP systems and validity of annotation. Unfortunately, conversion is a quite complex task since it involves grammatical rules and linguistic knowledge to be incorporated into the converter program. n nThe paper focusses on a methodology for treebank conversion which consists in splitting the process in steps corresponding to the kinds of information that have to be converted, i.e. morphological, structural or relational syntactic. The advantage is the generation of a set of parallel treebanks featuring progressively differentiated formats. An application to the case of an Italian dependency-based treebank in a Penn like format is described.

The EVALITA Dependency Parsing Task: From 2007 to 2011

The aim of the EVALITA Parsing Task (EPT) is at defining and extending Italian state-of-the-art parsing by encouraging the application of existing models and approaches, comparing paradigms and annotation formats. Therefore, in all the editions, held respectively in 2007, 2009 and 2011, the Task has been organized around two tracks, namely Dependency Parsing and Constituency Parsing, exploiting the same data sets made available by the organizers in two different formats.

A Relation-Based Schema for Treebank Annotation

This paper presents a relation-based schema for treebank annotation, and its application in the development of a corpus of Italian sentences. The annotation schema keeps arguments and modifiers distinct and allows for an accurate representation of predicate-argument structure and subcategorization. The accuracy strongly depends on methods adopted for defining the relations which are tripartite feature structures that consist of a morpho-syntactic, a functional and a semantic component. We presents empirical evidence for these tripartite structures by illustrating phenomena faced in the development of an Italian treebank.

Debate on political reforms in Twitter: A hashtag-driven analysis of political polarization

Political debates about a reform may sparkle national controversies, by leading members of the community to polarize their opinions and sentiment about the topic addressed. With the rise of social media like Twitter users are encouraged to voice and share their strong and polarized views and in general people are exposed to broader viewpoints than they were before. The large amount of user-generated social data available is a great opportunity to investigate the communicative behaviors emerging in the context of such political debates and to shed some light on the way communities of users with different roles in the society and different political sentiment interact. In this paper we focussed on communications in Twitter around the reform of marriage in France in 2012 and 2013 - “Le Mariage Pour Tous” - which had been the subject of debate and controversy. We collected a corpus of tweets tagged by the hashtag #mariagepourtous, created to mark the messages about the reform. We applied different kinds of analysis on our dataset based on linguistic and non linguistic features of the observed data in order to investigate the communicative behavior in using subjective and evaluative language on a political topic. The analysis leaded also to reflect on the impact of different typologies of users involved in the virtual debate which included both political messages created by media organizations and by other individual users, from ordinary citizens to politicians or celebrities.

Looking Back to the EVALITA Constituency Parsing Task: 2007-2011

The aim of the EVALITA Parsing Task (EPT) is at defining and extending Italian state-of-the-art parsing by encouraging the application of existing models and approaches, comparing paradigms and annotation formats. Therefore, in all the editions, held respectively in 2007, 2009 and 2011, the Task has been organized around two tracks, namely Dependency Parsing and Constituency Parsing, exploiting the same data sets made available by the organizers in two different formats.

PartTUT: The Turin University Parallel Treebank

In this paper, we introduce an ongoing project for the development of a parallel treebank for Italian, English and French. The treebank is annotated in a dependency format, namely the one designed in the Turin University Treebank (TUT), hence the choice to call such new resource Par(allel)TUT. The project aims at creating a resource which can be useful in particular for translation research. Therefore, beyond constantly enriching the treebank with new and heterogeneous data, so as to build a dynamic and balanced multilingual treebank, the current stage of the project is devoted to the design of a tool for the alignment of data, which takes into account syntactic knowledge as annotated in this kind of resource. The paper focuses in particular on the study of translational divergences and their implications for the development of the alignment tool. The paper provides an overview of the treebank, with its current content and the peculiarities of the annotation format, the description of the classes of translational divergences which could be encountered in the treebank, together with a proposal for their alignment.

Harmonizing and Merging Italian Treebanks: Towards a Merged Italian Dependency Treebank and Beyond

In this paper we address the challenge of combining existing CoNLL-compliant dependency-annotated corpora with the final aim of constructing a bigger treebank for the Italian language. To this end, we defined a methodology for mapping different annotation schemes, based on: (i) The analysis of similarities and differences of considered source and target dependency annotation schemes; (ii) The analysis of the performance of state of the art dependency parsers trained on the source and target treebanks; (iii) The mapping of the source annotation scheme(s) onto a set of target (possibly underspecified) data categories. This methodology was applied in two different case studies. The first one was aimed at constructing a “Merged Italian Dependency Treebank” (MIDT) starting from existing Italian dependency treebanks, namely TUT and ISST–TANL. The second case study, still ongoing, consists in the conversion of the MIDT resource into the Stanford Dependencies de facto standard with the final aim of developing an “Italian Stanford Dependency Treebank” (ISDT).

Context and Multi-media Corpora

In the recent research work on multi-media corpora the notion of context appears to be crucial both on the theoretical level and the applied level. The aim of this paper is to analyze the three levels which are relevant to the treatment of contextual information in multi-media corpora (i.e. annotation, storage, retrieval), and the different solutions which are resorted to in recently implemented systems to meet the need for a multi-layered, multi-linked annotation, and a hierarchically organized retrieval of contextual data.

Morphological Future in Italian Children

Before we discuss the topic of acquisition, it will be useful to provide non-Italian readers with a basic description of the morphology of Future, its possible substitutions with other tenses (and forms), and its use in standard Italian,1 in order to illustrate some peculiarities of this tense in adults’ language. Adults’ language plays a role not only as the standard language to be acquired, but also in the actual interactions, where some children’s occurrences of Future are “primed” (cf. 2.4), according to textual constraints.