The authors present a time-delay neural network (TDNN) approach to phoneme recognition which is characterized by two important properties: (1) using a three-layer arrangement of simple computing units, a hierarchy can be constructed that allows for the formation of arbitrary nonlinear decision surfaces, which the TDNN learns automatically using error backpropagation; and (2) the time-delay arrangement enables the network to discover acoustic-phonetic features and the temporal relationships between them independently of position in time and therefore not blurred by temporal shifts in the input. As a recognition task, the speaker-dependent recognition of the phonemes B, D, and G in varying phonetic contexts was chosen. For comparison, several discrete hidden Markov models (HMM) were trained to perform the same task. Performance evaluation over 1946 testing tokens from three speakers showed that the TDNN achieves a recognition rate of 98.5% correct while the rate obtained by the best of the HMMs was only 93.7%. >

Phoneme recognition using time-delay neural networks

/pdf/phoneme-recognition-using-time-delay-neural-networks-3o10ydjad6.pdf

The description of a novel type of m-gram language model is given. The model offers, via a nonlinear recursive procedure, a computation and space efficient solution to the problem of estimating probabilities from sparse data. This solution compares favorably to other proposed methods. While the method has been developed for and successfully implemented in the IBM Real Time Speech Recognizers, its generality makes it applicable in other areas where the problem of estimating probabilities from sparse data arises.

Estimation of probabilities from sparse data for the language model component of a speech recognizer

Vimentin, a major constituent of the intermediate filament family of proteins, is ubiquitously expressed in normal mesenchymal cells and is known to maintain cellular integrity and provide resistance against stress. Vimentin is overexpressed in various epithelial cancers, including prostate cancer, gastrointestinal tumors, tumors of the central nervous system, breast cancer, malignant melanoma, and lung cancer. Vimentin’s overexpression in cancer correlates well with accelerated tumor growth, invasion, and poor prognosis; however, the role of vimentin in cancer progression remains obscure. In recent years, vimentin has been recognized as a marker for epithelial–mesenchymal transition (EMT). Although EMT is associated with several tumorigenic events, vimentin’s role in the underlying events mediating these processes remains unknown. By virtue of its overexpression in cancer and its association with tumor growth and metastasis, vimentin serves as an attractive potential target for cancer therapy; however, more research would be crucial to evaluate its specific role in cancer. Our recent discovery of a vimentin-binding mini-peptide has generated further impetus for vimentin-targeted tumor-specific therapy. Furthermore, research directed toward elucidating the role of vimentin in various signaling pathways would reveal new approaches for the development of therapeutic agents. This review summarizes the expression and functions of vimentin in various types of cancer and suggests some directions toward future cancer therapy utilizing vimentin as a potential molecular target.

/pdf/vimentin-in-cancer-and-its-potential-as-a-molecular-target-49tey909hk.pdf

Vimentin in cancer and its potential as a molecular target for cancer therapy.

A method for estimating the parameters of hidden Markov models of speech is described. Parameter values are chosen to maximize the mutual information between an acoustic observation sequence and the corresponding word sequence. Recognition results are presented comparing this method with maximum likelihood estimation.

Maximum mutual information estimation of hidden Markov model parameters for speech recognition

The Speech Recognition Group at IBM Research in Yorktown Heights has designed a real-time, isolated-utterance speech recognizer for natural language with a 5,000-word vocabulary based on the IBM Personal Computer (PC) AT model and two IBM Signal Processors realized in VLSI technology. The enrollment period for a new user is approximately 20 minutes. The basic vocabulary is chosen from the most common words in several collections of documents such as office memoranda and business letters. The system supports spelling and interactive personalization to augment this vocabulary. Signal processing, vector quantization, and acoustic matching algorithms are programmed on the IBM Signal Processors which fit into the PC AT chassis. The PC AT controls the Processors and implements the decoder stack search and the language model, as well as the application-specific interface. The modular architecture of the design is expandable to a 20,000-word vocabulary system by the addition of two more IBM Signal Processors housed in a PC Expansion Unit.

An IBM PC based large-vocabulary isolated-utterance speech recognizer

The role of differentiated trophoblast glycogen cells (GCs) in the ectoplacental cone (EPC) has not been elucidated yet. Recently, GC progenitors have been shown to be present from embryonic day 7.5 (E7.5), but glycogen is found in GC only from E10.5. Herein, we investigated the origin, localization and characterization of mouse GCs in EPC and their relationship with blood cells and trophoblast giant cells (TGCs) during placentation. Implantation sites (E5.5–E12.5) were processed for histological studies, histochemical detection (glycogen) and immunohistochemical staining (Ki67). Three-dimensional reconstruction of the EPC was obtained from suitably oriented embryos at E7.5. Our findings evidence that GCs are present and assembled in clusters from E6.5 to E12.5, and that they exhibit the classic vacuolated appearance and contain PAS-positive glycogen, which is amylase-sensitive and acetylation-resistant. In fact, only GCs were stained after acetylation, confirming unequivocally their presence in tissues. At E6.5, GCs showed numerous mitoses and vacuoles with scattered glycogen particles. At E7.5, GCs showed low numbers of mitoses and abundant vacuoles full of glycogen. During E7.5–E8.5, GCs were in close proximity to TGCs, and cells were intercalated by thin maternal blood spaces; placental GCs lost maternal blood contact during E9.5–E12.5. Our results indicate that GCs are originated and proliferate in the upper portion in the midregion of EPC at E6.5, and that at E7.5–E8.5 they show consistent glycogen deposits, which are likely metabolized to glucose. This compound may be directly transferred to circulating maternal blood, and used as a source of energy by GCs and TGCs during placentation.

Trophoblast glycogen cells differentiate early in the mouse ectoplacental cone: putative role during placentation

Trophoblastic giant cells reach their maximum size and exhibit a conspicuous synthetic and invasive activity during mouse placentation. The cytoskeleton, given the complex functions of the cells, shows a well-developed network of intermediate filament proteins. Immunohistochemistry combined with confocal and conventional immunofluorescence studies of intermediate filaments proteins cytokeratin and vimentin were performed in mice trophoblastic giant cells on days 9-11 of pregnancy. Specimens were fixed in phosphate-buffered formaldehyde and tissues were processed for routine paraffin embedding. Trophoblastic giant cells from antimesometrial, lateral or mesometrial uterine regions, through days 9-11 of pregnancy, expressed the same staining with both immunoperoxidase and immunofluorescent techniques. Cytokeratin filamentous structures were intensely immunoreactive and were detected throughout the cells cytoplasm; a few cells exhibited strongest fluorescence in the peripheral cytoplasm. Vimentin-positive staining was often distributed throughout the cells cytoplasm, most frequently and more intensely in the peripheral region; in some cells, it was present only in the peripheral regions. It is probable that expression of vimentin in midpregnancy trophoblastic giant cells may be associated with the rapid and conspicuous increase in size and synthetic activity of the cells and also with phagocytosis of degraded materials and invasion of decidual tissue.

Coexpression of cytokeratin and vimentin in mice trophoblastic giant cells.

This paper deals with two experiments with a large vocabulary isolated word recognizer. The first compares word error rates for 1) meaningful sentences belonging to actual documents and 2) random word lists from the same vocabulary. The error rate is considerably lower for random word lists. The second experiment investigates the performance of the recognition system on sentences containing words outside the vocabulary of the recognizer. Sentences from a 5000 word vocabulary task are recognized with a recognizer limited to a 2000 word subvocabulary. The error rate is only slightly higher than it would be if recognition of the full 5000 word vocabulary was allowed.

Some experiments with large-vocabulary isolated-word sentence recognition

The fine structure of trophoblast giant cells and their interaction with collagen at the antimesometrial region on the 9th day of pregnancy was examined in fed and acute fasted mice. Collagen fibrils and filamentous aggregates (disintegrating collagen fibrils) were observed in the extracellular space. Three types of intracellular vacuoles containing collagen fibrils were present: vacuole type A exhibited typical cross-banded collagen immersed in finely granular electron-translucent material; and vacuoles type B and C showed electron-opaque granular material containing, respectively, faint cross-banded collagen and narrow clear stripes often with faint periodicity. In fed animals vacuoles type B were absent and the others were less evident. Only fasted animals showed extracellular acid phosphatase activity on collagen fibrils, filamentous aggregates and confined regions of the extracellular space. Intracellular acid phosphatase activity was observed in vacuoles type B and in lysosomes. The results indicate that trophoblast giant cells are capable of breaking down extracellular collagen and also of internalizing collagen for intracellular degradation. It is likely that these events are part of the process of invasion of the uterine wall. However, in fasted mice, collagen breakdown is more pronounced, and it may therefore contribute to the provision of amino acids and other nutrients for the undernourished fetus.

Sima Godosevicius Katz

Papers

An IBM PC based large-vocabulary isolated-utterance speech recognizer

Trophoblast glycogen cells differentiate early in the mouse ectoplacental cone: putative role during placentation

Coexpression of cytokeratin and vimentin in mice trophoblastic giant cells.

Some experiments with large-vocabulary isolated-word sentence recognition

Extracellular and intracellular degradation of collagen by trophoblast giant cells in acute fasted mice examined by electron microscopy