Sense representations for Portuguese: experiments with sense embeddings and deep neural language models

Sense representations have gone beyond word representations like Word2Vec, GloVe and FastText and achieved innovative performance on a wide range of natural language processing tasks. Although very useful in many applications, the traditional approaches for generating word embeddings have a strict drawback: they produce a single vector representation for a given word ignoring the fact that ambiguous words can assume different meanings. In this paper, we explore unsupervised sense representations which, different from traditional word embeddings, are able to induce different senses of a word by analyzing its contextual semantics in a text. The unsupervised sense representations investigated in this paper are: sense embeddings and deep neural language models. We present the first experiments carried out for generating sense embeddings for Portuguese. Our experiments show that the sense embedding model (Sense2vec) outperformed traditional word embeddings in syntactic and semantic analogies task, proving that the language resource generated here can improve the performance of NLP tasks in Portuguese. We also evaluated the performance of pre-trained deep neural language models (ELMo and BERT) in two transfer learning approaches: feature based and fine-tuning, in the semantic textual similarity task. Our experiments indicate that the fine tuned Multilingual and Portuguese BERT language models were able to achieve better accuracy than the ELMo model and baselines.

     

Alignment-based extraction of multiword expressions

Due to idiosyncrasies in their syntax, semantics or frequency, Multiword Expressions (MWEs) have received special attention from the NLP community, as the methods and techniques developed for the treatment of simplex words are not necessarily suitable for them. This is certainly the case for the automatic acquisition of MWEs from corpora. A lot of effort has been directed to the task of automatically identifying them, with considerable success. In this paper, we propose an approach for the identification of MWEs in a multilingual context, as a by-product of a word alignment process, that not only deals with the identification of possible MWE candidates, but also associates some multiword expressions with semantics. The results obtained indicate the feasibility and low costs in terms of tools and resources demanded by this approach, which could, for example, facilitate and speed up lexicographic work.     

Automatic induction of bilingual resources from aligned parallel corpora: application to shallow-transfer machine translation

The availability of machine-readable bilingual linguistic resources is crucial not only for rule-based machine translation but also for other applications such as cross-lingual information retrieval. However, the building of such resources (bilingual single-word and multi-word correspondences, translation rules) demands extensive manual work, and, as a consequence, bilingual resources are usually more difficult to find than “shallow” monolingual resources such as morphological dictionaries or part-of-speech taggers, especially when they involve a less-resourced language. This paper describes a methodology to build automatically both bilingual dictionaries and shallow-transfer rules by extracting knowledge from word-aligned parallel corpora processed with shallow monolingual resources (morphological analysers, and part-of-speech taggers). We present experiments for Brazilian Portuguese–Spanish and Brazilian Portuguese–English parallel texts. The results show that the proposed methodology can enable the rapid creation of valuable computational resources (bilingual dictionaries and shallow-transfer rules) for machine translation and other natural language processing tasks).     

Enzyme activity of horseradish peroxidase immobilized in chitosan matrices in alternated layers

The successful immobilization of enzymes such as horseradish peroxidase (HRP) in solid films is essential for applications in sensors and for fundamental studies aimed at identifying possible biotechnological devices. In this study we show that HRP can be immobilized in alternated layers with chitosan as the template material. The activity of HRP in HRP/chitosan films was preserved for several weeks, and could be detected optically upon monitoring the reaction with pyrogallol. The morphology of the film displayed stripes that disappeared after reaction with pyrogallol. Though the activity in the HRP/chitosan film was lower than in a homogeneous solution or in an LB film investigated earlier, the response was linear for a considerable period of time, which may be advantageous for sensing hydrogen peroxide.     

Controlled fabrication of gold nanoparticles biomediated by glucose oxidase immobilized on chitosan layer-by-layer films

The control of size and shape of metallic nanoparticles is a fundamental goal in nanochemistry, and crucial for applications exploiting nanoscale properties of materials. We present here an approach to the synthesis of gold nanoparticles mediated by glucose oxidase (GOD) immobilized on solid substrates using the Layer-by-Layer (LbL) technique. The LbL films contained four alternated layers of chitosan and poly(styrene sulfonate) (PSS), with GOD in the uppermost bilayer adsorbed on a fifth chitosan layer: (chitosan/PSS)₄/(chitosan/GOD). The films were inserted into a solution containing gold salt and glucose, at various pHs. Optimum conditions were achieved at pH 9, producing gold nanoparticles of ca. 30 nm according to transmission electron microscopy. A comparative study with the enzyme in solution demonstrated that the synthesis of gold nanoparticles is more efficient using immobilized GOD.     

Interaction of algal polysaccharide with lipid Langmuir monolayers

Algal polysaccharides may assist in the incorporation of metal ions from water and also have applications in health, interacting with biomembranes. In this paper, we have exploited the interactions at the molecular level of an extract of extracellular algal polysaccharide (EPS) with biomembrane models (lipid Langmuir monolayers). From surface pressure–area isotherms and Polarization Modulation Infrared Reflection–Absorption Spectroscopy (PM-IRRAS), we have inferred that EPS interacts with both kinds of lipids: a negatively charged and a positively charged, however with different mechanisms of action. EPS, being negatively charged, interacts with the polar heads of the positively charged lipid monolayer, while for a negatively charged lipid forming the Langmuir film, EPS penetrates into the alkyl chains of lipid in a mechanism driven mainly by hydrophobic interactions. The insights obtained in this paper are relevant because these help to understand the action of EPS in biomembranes and enable future applications of EPS–lipid hybrid Langmuir–Blodgett films for heavy metal ion sensors.     

Controlling the luminescence properties of poly(p-phenylene vinylene) entrapped in Langmuir and Langmuir–Blodgett films of stearic acid

Properties of hybrid films can be enhanced if their molecular architecture is controlled. In this paper, poly (p-phenylene vinylene) was mixed with stearic acid in order to form stable hybrid Langmuir monolayers. Surface properties of these films were investigated with measurements of surface pressure, and also with polarization modulation infrared reflection–absorption spectroscopy (PM-IRRAS). The films were transferred from the air–water interface to solid supports through the Langmuir–Blodgett technique, and the viability of the film as optical device was investigated with fluorescence spectroscopy. Comparing the fluorescent spectra for the polymer in solution, as a casting film, and as an LB film, the emission bands for LB films were narrower and appeared at lower wavelengths. The interactions between the film components and the design for the LB film may take advantage of the method to immobilize luminescent polymers in mixed ultrathin films adsorbed in solid matrices.