Using metrics from complex networks to evaluate machine translation

Establishing metrics to assess machine translation (MT) systems automatically is now crucial owing to the widespread use of MT over the web. In this study we show that such evaluation can be done by modeling text as complex networks. Specifically, we extend our previous work by employing additional metrics of complex networks, whose results were used as input for machine learning methods and allowed MT texts of distinct qualities to be distinguished. Also shown is that the node-to-node mapping between source and target texts (English-Portuguese and Spanish-Portuguese pairs) can be improved by adding further hierarchical levels for the metrics out-degree, in-degree, hierarchical common degree, cluster coefficient, inter-ring degree, intra-ring degree and convergence ratio. The results presented here amount to a proof-of-principle that the possible capturing of a wider context with the hierarchical levels may be combined with machine learning methods to yield an approach for assessing the quality of MT systems. (C) 2010 Elsevier B.V. All rights reserved.

Insights on eukaryotic translation initiation factor 5A (eIF5A) in the brain and aging

Long-term memory, a persistent form of synaptic plasticity, requires translation of a subset of mRNA present in neuronal dendrites during a short and critical period through a mechanism not yet fully elucidated. Western blotting analysis revealed a high content of eukaryotic translation initiation factor 5A (eIF5A) in the brain of neonatal rats, a period of intense neurogenesis rate, differentiation and synaptic establishment, when compared to adult rats. Immunohistochemistry analysis revealed that eIF5A is present in the whole brain of adult rats showing a variable content among the cells from different areas (e.g. cortex, hippocampus and cerebellum). A high content of eIF5A in the soma and dendrites of Purkinje cells, key neurons in the control of motor long-term memory in the cerebellum, was observed. Detection of high eIF5A content was revealed in dendritic varicosities of Purkinje cells. Evidence is presented herein that a reduction of eIF5A content is associated to brain aging. (C) 2008 Elsevier B.V. All rights reserved.

T3 acutely increases GH mRNA translation rate and GH secretion in hypothyroid rats

Cytoskeleton controls the stability of transcripts, by mechanisms that involve mRNAs and eEF1A attachment to it. Besides, it plays a key role in protein synthesis and secretion, which seems to be impaired in somatotrophs of hypothyroid rats, whose cytoskeleton is disarranged. This study investigated the: eEF1A and GH mRNA binding to cytoskeleton plus GH mRNA translation rate and GH secretion, in sham-operated and thyroidectomized rats treated with T3 or saline, and killed 30 min thereafter. Thyroidectomy reduced: (a) pituitary F-actin content, and eEF1A plus GH mRNA binding to it; (b) GH mRNA recruitment to polysome; and (c) liver IGF-1 mRNA expression, indicating that GH mRNA stability and translation rate, as well as GH secretion were impaired. T3 acutely reversed all these changes, which points toward a nongenomic action of T3 on cytoskeleton rearrangement, which might contribute to the increase on GH mRNA translation rate and GH secretion. (C) 2009 Elsevier Ireland Ltd. All rights reserved.

Involvement of Eukaryotic Translation Initiation Factor 5A (eIF5A) in Skeletal Muscle Stem Cell Differentiation

The eukaryotic translation initiation factor 5A (eIF5A) contains a special amino acid residue named hypusine that is required for its activity, being produced by a post-translational modification using spermidine as substrate. Stem cells from rat skeletal muscles (satellite cells) were submitted to differentiation and an increase of eIF5A gene expression was observed. Higher content of eIF5A protein was found in satellite cells on differentiation in comparison to non-differentiated satellite cells and skeletal muscle. The treatment with NI-guanyl- 1,7-diaminoheptane (GC7), a hypusination inhibitor, reversibly abolished the differentiation process. In association with the differentiation blockage, an increase of glucose consumption and lactate production and a decrease of glucose and palmitic acid oxidation were observed. A reduction in cell proliferation and protein synthesis was also observed. L-Arginine, a spermidine precursor and partial suppressor of muscle dystrophic phenotype, partially abolished the GC7 inhibitory effect on satellite cell differentiation. These results reveal a new physiological role for eIF5A and contribute to elucidate the molecular mechanisms involved in muscle regeneration.

An investigation into feature construction to assist word sense disambiguation

Identifying the correct sense of a word in context is crucial for many tasks in natural language processing (machine translation is an example). State-of-the art methods for Word Sense Disambiguation (WSD) build models using hand-crafted features that usually capturing shallow linguistic information. Complex background knowledge, such as semantic relationships, are typically either not used, or used in specialised manner, due to the limitations of the feature-based modelling techniques used. On the other hand, empirical results from the use of Inductive Logic Programming (ILP) systems have repeatedly shown that they can use diverse sources of background knowledge when constructing models. In this paper, we investigate whether this ability of ILP systems could be used to improve the predictive accuracy of models for WSD. Specifically, we examine the use of a general-purpose ILP system as a method to construct a set of features using semantic, syntactic and lexical information. This feature-set is then used by a common modelling technique in the field (a support vector machine) to construct a classifier for predicting the sense of a word. In our investigation we examine one-shot and incremental approaches to feature-set construction applied to monolingual and bilingual WSD tasks. The monolingual tasks use 32 verbs and 85 verbs and nouns (in English) from the SENSEVAL-3 and SemEval-2007 benchmarks; while the bilingual WSD task consists of 7 highly ambiguous verbs in translating from English to Portuguese. The results are encouraging: the ILP-assisted models show substantial improvements over those that simply use shallow features. In addition, incremental feature-set construction appears to identify smaller and better sets of features. Taken together, the results suggest that the use of ILP with diverse sources of background knowledge provide a way for making substantial progress in the field of WSD.

Improvements on the Porter`s Stemming Algorithm for Portuguese

The amount of textual information digitally stored is growing every day. However, our capability of processing and analyzing that information is not growing at the same pace. To overcome this limitation, it is important to develop semiautomatic processes to extract relevant knowledge from textual information, such as the text mining process. One of the main and most expensive stages of the text mining process is the text pre-processing stage, where the unstructured text should be transformed to structured format such as an attribute-value table. The stemming process, i.e. linguistics normalization, is usually used to find the attributes of this table. However, the stemming process is strongly dependent on the language in which the original textual information is given. Furthermore, for most languages, the stemming algorithms proposed in the literature are computationally expensive. In this work, several improvements of the well know Porter stemming algorithm for the Portuguese language, which explore the characteristics of this language, are proposed. Experimental results show that the proposed algorithm executes in far less time without affecting the quality of the generated stems.

Structural modeling and mutational analysis of yeast eukaryotic translation initiation factor 5A reveal new critical residues and reinforce its involvement in protein synthesis

Eukaryotic translation initiation factor 5A (eIF5A) is a protein that is highly conserved and essential for cell viability. This factor is the only protein known to contain the unique and essential amino acid residue hypusine. This work focused on the structural and functional characterization of Saccharomyces cerevisiae eIF5A. The tertiary structure of yeast eIF5A was modeled based on the structure of its Leishmania mexicana homologue and this model was used to predict the structural localization of new site-directed and randomly generated mutations. Most of the 40 new mutants exhibited phenotypes that resulted from eIF-5A protein-folding defects. Our data provided evidence that the C-terminal alpha-helix present in yeast eIF5A is an essential structural element, whereas the eIF5A N-terminal 10 amino acid extension not present in archaeal eIF5A homologs, is not. Moreover, the mutants containing substitutions at or in the vicinity of the hypusine modification site displayed nonviable or temperature-sensitive phenotypes and were defective in hypusine modification. Interestingly, two of the temperature-sensitive strains produced stable mutant eIF5A proteins - eIF5A(K56A) and eIF5A(Q22H,L93F)- and showed defects in protein synthesis at the restrictive temperature. Our data revealed important structural features of eIF5A that are required for its vital role in cell viability and underscored an essential function of eIF5A in the translation step of gene expression.