Architecture for text normalization using statistical machine translation techniques

This paper proposes an architecture, based on statistical machine translation, for developing the text normalization module of a text to speech conversion system. The main target is to generate a language independent text normalization module, based on data and flexible enough to deal with all situa-tions presented in this task. The proposed architecture is composed by three main modules: a tokenizer module for splitting the text input into a token graph (tokenization), a phrase-based translation module (token translation) and a post-processing module for removing some tokens. This paper presents initial exper-iments for numbers and abbreviations. The very good results obtained validate the proposed architecture.

Architecture for Text Sign Localization and Recognition

This paper describes a low complexity strategy for detecting and recognizing text signs automatically. Traditional approaches use large image algorithms for detecting the text sign, followed by the application of an Optical Character Recognition (OCR) algorithm in the previously identified areas. This paper proposes a new architecture that applies the OCR to a whole lightly treated image and then carries out the text detection process of the OCR output. The strategy presented in this paper significantly reduces the processing time required for text localization in an image, while guaranteeing a high recognition rate. This strategy will facilitate the incorporation of video processing-based applications into the automatic detection of text sign similar to that of a smartphone. These applications will increase the autonomy of visually impaired people in their daily life.

Multilingual number transcription for text-to-speech conversion

This paper describes the text normalization module of a text to speech fully-trainable conversion system and its application to number transcription. The main target is to generate a language independent text normalization module, based on data instead of on expert rules. This paper proposes a general architecture based on statistical machine translation techniques. This proposal is composed of three main modules: a tokenizer for splitting the text input into a token graph, a phrase-based translation module for token translation, and a post-processing module for removing some tokens. This architecture has been evaluated for number transcription in several languages: English, Spanish and Romanian. Number transcription is an important aspect in the text normalization problem.

Image processing methods for human brain connectivity analysis from in-vivo diffusion MRI

The structural connectivity of the brain is considered to encode species-wise and subject-wise patterns that will unlock large areas of understanding of the human brain. Currently, diffusion MRI of the living brain enables to map the microstructure of tissue, allowing to track the pathways of fiber bundles connecting the cortical regions across the brain. These bundles are summarized in a network representation called connectome that is analyzed using graph theory. The extraction of the connectome from diffusion MRI requires a large processing flow including image enhancement, reconstruction, segmentation, registration, diffusion tracking, etc. Although a concerted effort has been devoted to the definition of standard pipelines for the connectome extraction, it is still crucial to define quality assessment protocols of these workflows. The definition of quality control protocols is hindered by the complexity of the pipelines under test and the absolute lack of gold-standards for diffusion MRI data. Here we characterize the impact on structural connectivity workflows of the geometrical deformation typically shown by diffusion MRI data due to the inhomogeneity of magnetic susceptibility across the imaged object. We propose an evaluation framework to compare the existing methodologies to correct for these artifacts including whole-brain realistic phantoms. Additionally, we design and implement an image segmentation and registration method to avoid performing the correction task and to enable processing in the native space of diffusion data. We release PySDCev, an evaluation framework for the quality control of connectivity pipelines, specialized in the study of susceptibility-derived distortions. In this context, we propose Diffantom, a whole-brain phantom that provides a solution to the lack of gold-standard data. The three correction methodologies under comparison performed reasonably, and it is difficult to determine which method is more advisable. We demonstrate that susceptibility-derived correction is necessary to increase the sensitivity of connectivity pipelines, at the cost of specificity. Finally, with the registration and segmentation tool called regseg we demonstrate how the problem of susceptibility-derived distortion can be overcome allowing data to be used in their original coordinates. This is crucial to increase the sensitivity of the whole pipeline without any loss in specificity.

Rpp2, an essential protein subunit of nuclear RNase P, is required for processing of precursor tRNAs and 35S precursor rRNA in Saccharomyces cerevisiae

RPP2, an essential gene that encodes a 15.8-kDa protein subunit of nuclear RNase P, has been identified in the genome of Saccharomyces cerevisiae. Rpp2 was detected by sequence similarity with a human protein, Rpp20, which copurifies with human RNase P. Epitope-tagged Rpp2 can be found in association with both RNase P and RNase mitochondrial RNA processing in immunoprecipitates from crude extracts of cells. Depletion of Rpp2 protein in vivo causes accumulation of precursor tRNAs with unprocessed introns and 5′ and 3′ termini, and leads to defects in the processing of the 35S precursor rRNA. Rpp2-depleted cells are defective in processing of the 5.8S rRNA. Rpp2 immunoprecipitates cleave both yeast precursor tRNAs and precursor rRNAs accurately at the expected sites and contain the Rpp1 protein orthologue of the human scleroderma autoimmune antigen, Rpp30. These results demonstrate that Rpp2 is a protein subunit of nuclear RNase P that is functionally conserved in eukaryotes from yeast to humans.

A chloroplast processing enzyme functions as the general stromal processing peptidase

A highly specific stromal processing activity is thought to cleave a large diversity of precursors targeted to the chloroplast, removing an N-terminal transit peptide. The identity of this key component of the import machinery has not been unequivocally established. We have previously characterized a chloroplast processing enzyme (CPE) that cleaves the precursor of the light-harvesting chlorophyll a/b binding protein of photosystem II (LHCPII). Here we report the overexpression of active CPE in Escherichia coli. Examination of the recombinant enzyme in vitro revealed that it cleaves not only preLHCPII, but also the precursors for an array of proteins essential for different reactions and destined for different compartments of the organelle. CPE also processes its own precursor in trans. Neither the recombinant CPE nor the native CPE of chloroplasts process a preLHCPII mutant with an altered cleavage site demonstrating that both forms of the enzyme are sensitive to the same structural modification of the substrate. The transit peptide of the precursor of ferredoxin is released by a single cleavage event and found intact after processing by recombinant CPE and a chloroplast extract as well. These results provide the first direct demonstration that CPE is the general stromal processing peptidase that acts as an endopeptidase. Significantly, recombinant CPE cleaves in the absence of other chloroplast proteins, and this activity depends on metal cations, such as zinc.

Functional organization of spatial and nonspatial working memory processing within the human lateral frontal cortex

The present study used functional magnetic resonance imaging to demonstrate that performance of visual spatial and visual nonspatial working memory tasks involve the same regions of the lateral prefrontal cortex when all factors unrelated to the type of stimulus material are appropriately controlled. These results provide evidence that spatial and nonspatial working memory may not be mediated, respectively, by mid-dorsolateral and mid-ventrolateral regions of the frontal lobe, as widely assumed, and support the alternative notion that specific regions of the lateral prefrontal cortex make identical executive functional contributions to both spatial and nonspatial working memory.

Odor space and olfactory processing: Collective algorithms and neural implementation

Several basic olfactory tasks must be solved by highly olfactory animals, including background suppression, multiple object separation, mixture separation, and source identification. The large number N of classes of olfactory receptor cells—hundreds or thousands—permits the use of computational strategies and algorithms that would not be effective in a stimulus space of low dimension. A model of the patterns of olfactory receptor responses, based on the broad distribution of olfactory thresholds, is constructed. Representing one odor from the viewpoint of another then allows a common description of the most important basic problems and shows how to solve them when N is large. One possible biological implementation of these algorithms uses action potential timing and adaptation as the “hardware” features that are responsible for effective neural computation.

Cloning and functional characterization of a subunit of the transporter associated with antigen processing

The transporter associated with antigen processing (TAP) is essential for the transport of antigenic peptides across the membrane of the endoplasmic reticulum. In addition, TAP interacts with major histocompatibility complex class I heavy chain (HC)/β2-microglobulin (β2-m) dimers. We have cloned a cDNA encoding a TAP1/2-associated protein (TAP-A) corresponding in size and biochemical properties to tapasin, which was recently suggested to be involved in class I–TAP interaction (Sadasivan, B., Lehner, P. J., Ortmann, B., Spies, T. & Cresswell, P. (1996) Immunity 5, 103–114). The cDNA encodes a 448-residue-long ORF, including a signal peptide. The protein is predicted to be a type I membrane glycoprotein with a cytoplasmic tail containing a double-lysine motif (-KKKAE-COOH) known to maintain membrane proteins in the endoplasmic reticulum. Immunoprecipitation with anti-TAP1 or anti-TAP-A antisera demonstrated a consistent and stoichiometric association of TAP-A with TAP1/2. Class I HC and β2-m also were coprecipitated with these antisera, indicating the presence of a pentameric complex. In pulse–chase experiments, class I HC/β2-m rapidly dissociated from TAP1/2-TAP-A. We propose that TAP is a trimeric complex consisting of TAP1, TAP2, and TAP-A that interacts transiently with class I HC/β2-m. In peptide-binding assays using cross-linkable peptides and intact microsomes, TAP-A bound peptides only in the presence of ATP whereas binding of peptides to TAP1/2 was ATP-independent. This suggests a direct role of TAP-A in peptide loading onto class I HC/β2-m dimer.

Pattern of neuronal activity associated with conscious and unconscious processing of visual signals

Following striate cortex damage in monkeys and humans there can be residual function mediated by parallel visual pathways. In humans this can sometimes be associated with a “feeling” that something has happened, especially with rapid movement or abrupt onset. For less transient events, discriminative performance may still be well above chance even when the subject reports no conscious awareness of the stimulus. In a previous study we examined parameters that yield good residual visual performance in the “blind” hemifield of a subject with unilateral damage to the primary visual cortex. With appropriate parameters we demonstrated good discriminative performance, both with and without conscious awareness of a visual event. These observations raise the possibility of imaging the brain activity generated in the “aware” and the “unaware” modes, with matched levels of discrimination performance, and hence of revealing patterns of brain activation associated with visual awareness. The intact hemifield also allows a comparison with normal vision. Here we report the results of a functional magnetic resonance imaging study on the same subject carried out under aware and unaware stimulus conditions. The results point to a shift in the pattern of activity from neocortex in the aware mode, to subcortical structures in the unaware mode. In the aware mode prestriate and dorsolateral prefrontal cortices (area 46) are active. In the unaware mode the superior colliculus is active, together with medial and orbital prefrontal cortical sites.

Generation of CD8+ T cells specific for transporter associated with antigen processing deficient cells

Cells with impaired transporter associated with antigen processing (TAP) function express low levels of cell surface major histocompatibility complex (MHC) class I molecules, and are generally resistant to lysis by MHC class I restricted cytotoxic T lymphocytes (CTLs). Here we report the generation of MHC class I restricted CD8+ CTLs that surprisingly require target cell TAP deficiency for efficient recognition. C57BL/6 (B6) mice immunized with syngenic B7–1 (CD80) expressing TAP-deficient cells generated a potent CTL response against both TAP-deficient RMA-S tumor cells and TAP-deficient Con A blasts, whereas the corresponding TAP-expressing target cells were considerably less susceptible or resistant to lysis. The CTL epitopes recognized were expressed also by the human TAP-deficient cell line T2, transfected with appropriate MHC class I molecules. B6 mice immunized with B7–1-transfected TAP-deficient RMA-S cells were protected from outgrowth of a subsequent RMA-S tumor challenge. These findings are discussed in relation to the biochemical nature of MHC class I dependent CTL epitopes associated with impaired TAP function, as well as implications for immunotherapy and autoimmunity.

Sterols regulate processing of carbohydrate chains of wild-type SREBP cleavage-activating protein (SCAP), but not sterol-resistant mutants Y298C or D443N

SREBP cleavage activating protein (SCAP), a membrane-bound glycoprotein, regulates the proteolytic activation of sterol regulatory element binding proteins (SREBPs), which are membrane-bound transcription factors that control lipid synthesis in animal cells. SCAP-stimulated proteolysis releases active fragments of SREBPs from membranes of the endoplasmic reticulum and allows them to enter the nucleus where they activate transcription. Sterols such as 25-hydroxycholesterol inactivate SCAP, suppressing SREBP proteolysis and turning off cholesterol synthesis. We here report the isolation of Chinese hamster ovary cells with a point mutation in SCAP (Y298C) that renders the protein resistant to inhibition by 25-hydroxycholesterol. Like the previously described D443N mutation, the Y298C mutation occurs within the putative sterol-sensing domain, which is part of the polytopic membrane attachment region of SCAP. Cells that express SCAP(Y298C) continued to process SREBPs in the presence of 25-hydroxycholesterol and hence they resisted killing by this sterol. In wild-type Chinese hamster ovary cells the N-linked carbohydrate chains of SCAP were mostly in the endoglycosidase H-sensitive form when cells were grown in medium containing 25-hydroxycholesterol. In contrast, when cells were grown in sterol-depleted medium, these chains were converted to an endoglycosidase H-resistant form. 25-Hydroxycholesterol had virtually no effect in cells expressing SCAP(D443N) or SCAP(Y298C). The relation between this regulated carbohydrate processing to the SCAP-regulated proteolysis of SREBP remains to be explored.

Participation of the nuclear cap binding complex in pre-mRNA 3′ processing

Communication between the 5′ and 3′ ends is a common feature of several aspects of eukaryotic mRNA metabolism. In the nucleus, the pre-mRNA 5′ end is bound by the nuclear cap binding complex (CBC). This RNA–protein complex plays an active role in both splicing and RNA export. We provide evidence for participation of CBC in the processing of the 3′ end of the message. Depletion of CBC from HeLa cell nuclear extract strongly reduced the endonucleolytic cleavage step of the cleavage and polyadenylation process. Cleavage was restored by addition of recombinant CBC. CBC depletion was found to reduce the stability of poly(A) site cleavage complexes formed in nuclear extract. We also provide evidence that the communication between the 5′ and 3′ ends of the pre-mRNA during processing is mediated by the physical association of the CBC/cap complex with 3′ processing factors bound at the poly(A) site. These observations, along with previous data on the function of CBC in splicing, illustrate the key role played by CBC in pre-mRNA recognition and processing. The data provides further support for the hypothesis that pre-mRNAs and mRNAs may exist and be functional in the form of “closed-loops,” due to interactions between factors bound at their 5′ and 3′ ends.

Targeting HIV proteins to the major histocompatibility complex class I processing pathway with a novel gp120-anthrax toxin fusion protein

A challenge for subunit vaccines whose goal is to elicit CD8+ cytotoxic T lymphocytes (CTLs) is to deliver the antigen to the cytosol of the living cell, where it can be processed for presentation by major histocompatibility complex (MHC) class I molecules. Several bacterial toxins have evolved to efficiently deliver catalytic protein moieties to the cytosol of eukaryotic cells. Anthrax lethal toxin consists of two distinct proteins that combine to form the active toxin. Protective antigen (PA) binds to cells and is instrumental in delivering lethal factor (LF) to the cell cytosol. To test whether the lethal factor protein could be exploited for delivery of exogenous proteins to the MHC class I processing pathway, we constructed a genetic fusion between the amino-terminal 254 aa of LF and the gp120 portion of the HIV-1 envelope protein. Cells treated with this fusion protein (LF254-gp120) in the presence of PA effectively processed gp120 and presented an epitope recognized by HIV-1 gp120 V3-specific CTL. In contrast, when cells were treated with the LF254-gp120 fusion protein and a mutant PA protein defective for translocation, the cells were not able to present the epitope and were not lysed by the specific CTL. The entry into the cytosol and dependence on the classical cytosolic MHC class I pathway were confirmed by showing that antigen presentation by PA + LF254-gp120 was blocked by the proteasome inhibitor lactacystin. These data demonstrate the ability of the LF amino-terminal fragment to deliver antigens to the MHC class I pathway and provide the basis for the development of novel T cell vaccines.

In silico detection of control signals: mRNA 3′-end-processing sequences in diverse species

We have investigated mRNA 3′-end-processing signals in each of six eukaryotic species (yeast, rice, arabidopsis, fruitfly, mouse, and human) through the analysis of more than 20,000 3′-expressed sequence tags. The use and conservation of the canonical AAUAAA element vary widely among the six species and are especially weak in plants and yeast. Even in the animal species, the AAUAAA signal does not appear to be as universal as indicated by previous studies. The abundance of single-base variants of AAUAAA correlates with their measured processing efficiencies. As found previously, the plant polyadenylation signals are more similar to those of yeast than to those of animals, with both common content and arrangement of the signal elements. In all species examined, the complete polyadenylation signal appears to consist of an aggregate of multiple elements. In light of these and previous results, we present a broadened concept of 3′-end-processing signals in which no single exact sequence element is universally required for processing. Rather, the total efficiency is a function of all elements and, importantly, an inefficient word in one element can be compensated for by strong words in other elements. These complex patterns indicate that effective tools to identify 3′-end-processing signals will require more than consensus sequence identification.