Ariadne Maria Brito Rizzoni Carvalho

Politeness and Bias in Dialogue Summarization: Two Exploratory Studies

In this chapter, two empirical pilot studies on the role of politeness in dialogue summarization are described. In these studies, a collection of four dialogues was used. Each dialogue was automatically generated by the NECA system and the politeness of the dialogue participants was systematically manipulated. Subjects were divided into groups who had to summarize the dialogues from a particular dialogue participant’s point of view or the point of view of an impartial observer. In the first study, there were no other constraints. In the second study, the summarizers were restricted to summaries whose length did not exceed 10% of the number of words in the dialogue that was being summarized.

A relevance feedback approach for the author name disambiguation problem

This paper presents a new name disambiguation method that exploits user feedback on ambiguous references across iterations. An unsupervised step is used to define pure training samples, and a hybrid supervised step is employed to learn a classification model for assigning references to authors. Our classification scheme combines the Optimum-Path Forest (OPF) classifier with complex reference similarity functions generated by a Genetic Programming framework. Experiments demonstrate that the proposed method yields better results than state-of-the-art disambiguation methods on two traditional datasets.

Lappin and leass’ algorithm for pronoun resolution in portuguese

This paper presents a variant of Lappin and Leass’ Algorithm for pronoun resolution in Portuguese texts; the algorithm resolves third person pronominal anaphora, as well as reflexive and reciprocal pronouns. It relies on salience measures, derived from the syntactic structure of the sentence, and on a simple discourse representation model. The algorithm, as well as its evaluation with legal and literary corpora, are presented.

Semi-automatic Generation of Extended Finite State Machines from Natural Language Standard Documents

Many requirement documents are written in natural language and, therefore, may contain problems such as inconsistencies and ambiguities. To minimize these problems, there is a trend in Software Engineering to use models to represent systems. These models are obtained from textual requirements. However, manual modelling is a complex task and, in order to do it semi-automatically, one has to deal with problems such as the kind of model to be generated, the automation degree to be achieved, and the quality of the document that must be processed. We propose a methodology to semi-automatically generate Extended Finite State Machines (EFSMs) from natural language standard documents. We used Natural Language Processing (NLP) techniques and tools to extract information from the document, and implemented a prototype which generates EFSMs. The generated EFSMs were validated with a model checking tool, and manually evaluated by comparing them with the manually generated models.

Spelling error patterns in brazilian portuguese

Fifty years after Damerau set up his statistics for the distribution of errors in typed texts, his findings are still used in a range of different languages. Because these statistics were derived from texts in English, the question of whether they actually apply to other languages has been raised. We address this issue through the analysis of a set of typed texts in Brazilian Portuguese, deriving statistics tailored to this language. Results show that diacritical marks play a major role, as indicated by the frequency of mistakes involving them, thereby rendering Dameraus original findings mostly unfit for spelling correction systems, although still holding them useful, should one set aside such marks. Furthermore, a comparison between these results and those published for Spanish show no statistically significant differences between both languages—an indication that the distribution of spelling errors depends on the adopted character set rather than the language itself.

A comparative analysis of centering-based algorithms for pronoun resolution in Portuguese

Pronominal anaphora resolution consists in finding a referent for a given pronoun. Although being essential for many natural language processing systems, such as automatic translators, text generators and summarizers, there is a myriad of issues regarding this task, particularly when there ismore than one possible referent for a given pronoun. Over the years, several approaches have been proposed to deal with this challenge, usually taking only syntactic information into account. On the other hand, methods based on Centering Theory rely on concepts such as coherence, for instance, to do the job. In this work, we describe our implementation and evaluation of existing centering-based algorithms for pronominal resolution in Portuguese. As a result, we indicate both the strong and weak points of each of the tested algorithms, thereby helping other researchers to make a more informed decision about which method to use.

Syntactic Analysis for Ellipsis Handling in Coordinated Clauses

This work is intended as an investigation into elliptical phenomena in natural language. We argue that some types of ellipsis can be resolved at the syntactic level since they are subject to syntactic constraints. We have dealt with four of the major types of ellipsis found in Portuguese, namely: Null VP, Gapping, Stripping and Sluicing. We have used Island Constraints in order to decide on the grammaticality of the sentence. Finally, we have developed and implemented a syntactically-based algorithm that recovers the elidedconstituents and reconstructs the elliptical clause, when applicable. The linguistic data in this work is drawn primarily from Portuguese, but we believe that the results can also be applied to other languages, such as English.

Large-Scale Micro-Blog Authorship Attribution: Beyond Simple Feature Engineering

With the ever-growing use of social media, authorship attribution plays an important role in avoiding cybercrime, and helping the analysis of online trails left behind by cyber pranks, stalkers, bullies, identity thieves and alike. In this paper, we propose a method for authorship attribution in micro-blogs with efficiency one hundred to a thousand times faster than state-of-the-art counterparts. The method relies on a powerful and scalable feature representation approach taking advantage of user patterns in micro-blog messages, and also on a custom-tailored pattern classifier adapted to deal with big data and high-dimensional data. Finally, we discuss search-space reduction when analyzing hundreds of online suspects and millions of online micro messages, which makes this approach invaluable for digital forensics and law enforcement.

A multi-dimensional annotation scheme for behaviour in dialogues

In this paper we introduce a multi-dimensional annotation scheme for emotional and behavioural assessment in dialogue summaries. To test the soundness both of the annotation scheme and corresponding guidelines, reliability studies with nine independent annotators were carried out. As an illustration of the utility of our scheme, we have applied it to an already published study and verified whether the same conclusions hold. We hope that, in using our scheme, researchers will be able to save a lot of time and effort that, otherwise, would be spent in planning, developing and testing a scheme of their own.

Semi-Automatic Checklist Quality Assessment of Natural Language Requirements for Space Applications

Problems with the specification of software requirements documents are a common cause of software defects. In the space applications domain, such defects are very costly, especially when detected after deployment in the field. It is imperative to ensure that software requirements are well written to avoid the introduction of these defects. The quality of software requirements is frequently assessed via checklists, based on standards for space application software, and on problems found in previous projects. Given the importance of quality assessment, and the fact that it is manually performed by domain experts, we propose to develop a semi-automatic, natural language processing tool, to diminish the reviewers effort in the assessment, and to reduce errors in this process.