Fabio Tamburini

An Automatic System for Detecting Prosodic Prominence in American English Continuous Speech

A precise identification of prosodic phenomena and the construction of tools able to properly manage such phenomena are essential steps to disambiguate the meaning of certain utterances. In particular they are useful for a wide variety of tasks: automatic recognition of spontaneous speech, automatic enhancement of speech-generation systems, solving ambiguities in natural language interpretation, the construction of large annotated language resources, such as prosodically tagged speech corpora, and teaching languages to foreign students using Computer Aided Language Learning (CALL) systems. This paper presents a study on the automatic detection of prosodic prominence in continuous speech, with particular reference to American English, but with good prospects of application to other languages. Prosodic prominence involves two different prosodic features: pitch accent and stress accent. Pitch accent is acoustically connected with fundamental frequency (F0) movements and overall syllable energy, whereas stress exhibits a strong correlation with syllable nuclei duration and mid-to-high-frequency emphasis. This paper shows that a careful measurement of these acoustic parameters, as well as the identification of their connection to prosodic parameters, makes it possible to build an automatic system capable of identifying prominent syllables in utterances with performance comparable with the inter-human agreement reported in the literature. Two different prominence detectors were studied and developed: the first uses a training corpus to set up thresholds properly, while the second uses a pure unsupervised method. In both cases, it is worth stressing that only acoustic parameters derived directly from speech waveforms are exploited.

Prosodic prominence detection in speech

This paper presents work in progress on the automatic detection of prosodic prominence in continuous speech. Prosodic prominence involves two different phonetic features: pitch accents, connected with fundamental frequency (FO) movements and syllable overall energy, and stress, which exhibits a strong correlation with syllable nuclei duration and high-frequency emphasis. By measuring these acoustic parameters it is possible to build an automatic system capable of correctly identifying prominent syllables with an agreement with human-tagged data comparable with the inter-human agreement reported in the literature. These results were achieved without using any information apart from acoustic parameters.

A multimedia framework for second language teaching in self-access environments

Abstract This paper presents an account of a self-access language teaching scheme operated at the University of Bologna during the last three years. The main goal of the D.I.A.P.A.S.O.N. (Distributed, Interactive, And Personalised Audio-visual Study Over Network) project is to teach English to university students up to intermediate level, by building a self-access environment freely available to learners. Some laboratories equipped with multimedia workstations, video stations, satellite receivers and specifically targeted courseware were set up and tested in one complete year of work. Here I will analyse some topics concerned with self-access language learning through CALL (Computer Aided Language Learning) and the complete structure of the DIAPASON courseware. Some data collected on a sample of students during the year are presented and an interpretation of the data suggested. The main result emerging from the data is that students are very busy and tend to spend only the minimum required time to study foreign languages, organising this activity in few short sessions during the week.

A Parametric Architecture for Tags Clustering in Folksonomic Search Engines

Semantic search engines rely on the existence of a rich set of semantic connections between the concepts associated to documents and those used for the queries. With folksonomies, this is not always guaranteed. Creating clusters of folksonomic tags around terms of controlled ontological vocabularies is a potentially sophisticated approach, but algorithms abound for this clustering and no clear cut winner exists. In this paper we introduce FolksEngine, a parametric search engine for folksonomies allowing to specify any clustering algorithm as a three step process: the user’s query is expanded according to semantic rules associated to the terms of the query, the new query is then executed on the plain folksonomy search engine, and the results are ranked according to semantic rules associated to the folksonomic tags actually used for the documents.

Annotating large corpora for studying Italian derivational morphology

This paper presents ongoing research concerning the annotation of large corpora with morphological information. It aims at providing a general schema for inserting rich morphological information to enable complex corpus queries of word internal structure. Annotating real corpus data presents challenges that can hardly be managed with traditional linear analysis of word structure, but can effi ciently and correctly be handled with different, more complex, structures. For this reason, we propose Derivation Graphs as a new tool for representing the structure of complex words, and we discuss the theoretical consequences of this choice on the representation of affi xes, a crucial issue for all morphological models.

Of mice and terms: clustering algorithms on ambiguous terms in folksonomies

Developed using the principles of the Model-View-Controller architectural pattern, FolksEngine is a parametric search engine for folksonomies that allows us to test arbitrary search improvement algorithms by specifying them in three phases: expansion, where the original query is converted in multiple ones according to semantic rules associated to the query terms, search, executing the queries on a standard folksonomy search engine such as Delicious, and ranking, sorting the results according to rules. In this paper we extend our previous studies using FolksEngine and offer a new query expansion algorithms based on Natural Language Processing techniques, and a new view for the results based on Semantic Web technologies. We also describe some tests of the algorithms developed, in order to obtain a clear and effective evaluation of them.

Scheduling Data Flow Programs in Hard Real-Time Environments

Data Flow is a natural paradigm representing distributed programs able to express data and control dependencies between composing nodes. Our work concentrates on periodic Data-Flow programs having Hard Real-Time constraints. We propose a model and a stringbased representation for a relevant class of Data-Flow programs and present a two-phase algorithm able to compute a schedule for a given set of programs to be executed. The first phase computes a feasible solution (mapping and time assignments) using a Simulated Annealing technique, the second phase takes the mapping as input and computes the optimal solution, in terms of resource allocation, using a recursive descent Quasi-Newton method.

Speech analysis by Natural Language Processing techniques: a possible tool for very early detection of cognitive decline?

Background: The discovery of early, non-invasive biomarkers for the identification of “preclinical” or “pre-symptomatic” Alzheimers disease and other dementias is a key issue in the field, especially for research purposes, the design of preventive clinical trials, and drafting population-based health care policies. Complex behaviors are natural candidates for this. In particular, recent studies have suggested that speech alterations might be one of the earliest signs of cognitive decline, frequently noticeable years before other cognitive deficits become apparent. Traditional neuropsychological language tests provide ambiguous results in this context. In contrast, the analysis of spoken language productions by Natural Language Processing (NLP) techniques can pinpoint language modifications in potential patients. This interdisciplinary study aimed at using NLP to identify early linguistic signs of cognitive decline in a population of elderly individuals. Methods: We enrolled 96 participants (age range 50–75): 48 healthy controls (CG) and 48 cognitively impaired participants: 16 participants with single domain amnestic Mild Cognitive Impairment (aMCI), 16 with multiple domain MCI (mdMCI) and 16 with early Dementia (eD). Each subject underwent a brief neuropsychological screening composed by MMSE, MoCA, GPCog, CDT, and verbal fluency (phonemic and semantic). The spontaneous speech during three tasks (describing a complex picture, a typical working day and recalling a last remembered dream) was then recorded, transcribed and annotated at various linguistic levels. A multidimensional parameter computation was performed by a quantitative analysis of spoken texts, computing rhythmic, acoustic, lexical, morpho-syntactic, and syntactic features. Results: Neuropsychological tests showed significant differences between controls and mdMCI, and between controls and eD participants; GPCog, MoCA, PF, and SF also discriminated between controls and aMCI. In the linguistic experiments, a number of features regarding lexical, acoustic and syntactic aspects were significant in differentiating between mdMCI, eD, and CG (non-parametric statistical analysis). Some features, mainly in the acoustic domain also discriminated between CG and aMCI. Conclusions: Linguistic features of spontaneous speech transcribed and analyzed by NLP techniques show significant differences between controls and pathological states (not only eD but also MCI) and seems to be a promising approach for the identification of preclinical stages of dementia. Long duration follow-up studies are needed to confirm this assumption.

The Lemmatisation Task at the EVALITA 2011 Evaluation Campaign

This paper reports on the EVALITA 2011 Lemmatisation task, an initiative for the evaluation of automatic lemmatisation tools specifically developed for the Italian language. Despite lemmatisation is often considered a subproduct of a PoS-tagging procedure that does not cause any particular problem, there are a lot of specific cases, certainly in Italian and in some other highly inflected languages, in which, given the same lexical class, we face a lemma ambiguity. A relevant number of scholars and teams participated experimenting their systems on the data provided by the task organisers. The results are very interesting and the overall performances of the participating systems were very high, exceeding, on interesting cases, 99% of lemmatisation accuracy.

Morphology meets computational linguistics

The studies collected in this special issue of Lingue e Linguaggio aim to sketch a picture of the interfaces between morphology, computational linguistics, and corpus linguistics. Some of these studies are the outcome of the Galileo project Lessico e regole di formazione di parola: un approccio tipologico e computazionale, supported by the Universita Italo-Francese and the French Ministry of Foreign Affairs, and were presented at the workshop Morphology Meets Computational Linguistics, held at the University of Bologna on October 7-8, 2010. The papers published in this issue have been selected through a double-blind peer review process.