Báo cáo khoa học: "Domain Adaptation by Constraining Inter-Domain Variability of Latent Feature Representation" docx

... 2011. c 2011 Association for Computational Linguistics Domain Adaptation by Constraining Inter-Domain Variability of Latent Feature Representation Ivan Titov Saarland University Saarbruecken, Germany titov@mmci.uni-saarland.de Abstract We ... domain -adaptation setting is that some of the latent variables will cor- respond to clusters of features specific only to the so...

Ngày tải lên: 30/03/2014, 21:20

... extract feature vectors for all the words. We use context and orthographic features. In the sec- ond stage, using the dictionary probabilities of seen words, we identify pairs of words whose feature ... Linguistics Domain Adaptation for Machine Translation by Mining Unseen Words Hal Daum ´ e III University of Maryland Collge Park, USA hal@umiacs.umd.edu Jagadeesh Jagarlamudi Uni...

Ngày tải lên: 17/03/2014, 00:20

... To obtain a clearer picture of the adaptation process, we discard these 4 nouns, leaving a set of 21 nouns. 50 Dataset No. of MFS No. of No. of senses acc. training adaptation BC WSJ (%) examples ... (%) Percentage of adaptation examples added (%) a-c a r a-truePrior Figure 2: Adaptation process for all 21 nouns. of the BC training examples. At each adaptation iter- at...

Ngày tải lên: 08/03/2014, 02:21

... test data, often a small amount of speech data is human-transcribed. A LM is then built by interpolating the models estimated from large corpus of written language and the small corpus of transcribed ... domain, given a large corpus of out -of- domain training data and a small corpus of in-domain data. The main contribution of this ∗ Currently with Tallinn University of Techno...

Ngày tải lên: 23/03/2014, 16:20

... investigated in the presence of 100 m M KCl or DGP by acquiring a series of 1 H– 15 N HSQC spectra over a temperature span of 50 °C (from 30 to 80 °C). At 30 °C, the addition of DGP had little or no ... prior to analysis. The software package R 1 R 2_ DIFFUSION [31,40] was used to translate the centre of mass of the mean structure of the NMR ensemble [28] to the origin of co...

Ngày tải lên: 07/03/2014, 21:20

... formation of 3-hydroxy- adipate (4) proceeds by the loss of the acetyl side-chain carbon atoms of phenylacetate. A set of two relatively weak signals ( 3% signal intensity compared with the signals of ... intermediates of the pathway under study. Sequence similarity (40% identity and 11% of conserva- tive exchange by comparison with ChcB of Streptomyces collinus) of the...

Ngày tải lên: 23/03/2014, 18:20

... subtracting a property of enzyme A from a property of enzyme B or a property of enzyme B from a property of enzyme A if they are both mesophilic. Hence, the mean of the distribution is, by definition, ... using a set of 512 points located on a sphere around each atom. The sphere interaction radius of each atom is equal to the sum of the van der Waals radius of the atom type...

Ngày tải lên: 30/03/2014, 03:20

... comparable to that of Syn HO-1 in the presence of each of the reducing systems, although the ascorbate-supported activity of Syn HO-2 was stronger than that of Syn HO-1. The activity levels of both enzymes ... FEBS Absorption spectra of the complexes of heme and HOs resemble the spectrum of myoglobin with one exception, the HO of Drosophila melanogaster. In the HO of fru...

Ngày tải lên: 30/03/2014, 15:20

... impairment of mechanisms antagoniz- ing cell shrinkage may be a general feature of apoptosis. For example, renal tubular epithelial cell apoptosis was accompanied by a caspase-dependent cleavage of the ... tran- sient increase of intracellular Na + may be one of the first components of the mitogenic signal. Direct detec- tion by 23Na-NMR spectroscopy in quiescent 3T3 mouse fib...

Ngày tải lên: 18/02/2014, 16:20

... the seman- tic relatedness of words in Wikipedia. 1 Introduction The last years have seen a large amount of work in Natural Language Processing (NLP) using measures of semantic similarity and ... and relatedness. We believe that the extensive usage of such measures derives also from the availability of robust and freely avail- able software that allows to compute them (Pedersen et a...

Ngày tải lên: 20/02/2014, 12:20