Báo cáo khoa học: "Minimum Risk Annealing for Training Log-Linear Models∗" doc

... 2006.c2006 Association for Computational LinguisticsMinimum Risk Annealing for Training Log-Linear Models∗David A. Smith and Jason EisnerDepartment of Computer ScienceCenter for Language and Speech ... machinery to training log-linear combinations of models for dependencyparsing and for machine translation (§6). Finally,we note the connections of minimum risk training to max-margin training and ... describe techniques for optimizing nonlinearfunctions such as precision or the BLEU metric. We presentexperiments training log-linear combinations of models for dependency parsing and for machine translation....

Tài liệu Báo cáo khoa học: "Minimum Cut Model for Spoken Lecture Segmentation" ppt

... three lectures is-used for estimating the optimal word block length for representing nodes, the threshold distances for discarding node edges, the number of uniformchunks for estimating tf-idf ... to specialized technical vocabularyand lack of in-domain spoken data for training. Finally, pedagogical considerations call for ﬂuenttransitions between different topics in a lecture,further ... scoring scheme used in theinformation-retrieval literature (Salton and Buck-ley, 1988). A transcript is split uniformly into Nchunks; each chunk serves as the equivalent ofdocuments in the tf-idf...

Tài liệu Báo cáo khoa học: "Web-Scale Features for Full-Scale Parsing" doc

... errorreduction. Results are for dependency parsing on the dev set for iters:5 ,training- k:1.tal errors break down by gold head. For example,the 12.1% total error reduction for attachments of anIN ... grammar6Their README speciﬁes training- k:5 iters:10 loss-type:nopunc decode-type:proj’, which we used for all ﬁnal ex-periments; we used the faster training- k:1 iters:5’ setting for most development ... k-best lists for the development set(WSJ section 22) and test set (WSJ section 23) usinga grammar trained on all the training data (WSJ sec-tions 2-21).8To get k-best lists for the training...

Tài liệu Báo cáo khoa học: "An Approximate Approach for Training Polynomial Kernel SVMs in Linear Time" doc

... Association for Computational LinguisticsAn Approximate Approach for Training Polynomial Kernel SVMs in Linear Time Yu-Chieh Wu Jie-Chi Yang Yue-Shi Lee Dept. of Computer Science and Information ... did not use feature conjunctions. However, the training and testing time costs for polynomial kernel SVM is far slow than the linear kernel. For example, it took one day to train the CoNLL-2000 ... adopt the mined patterns to re-represent the training/ testing examples. Subse-quently, we use the off-the-shelf linear kernel SVM algorithm to perform training and testing. Besides, to exponential-scaled...

Tài liệu Báo cáo khoa học: "A CONNECTIONIST PARSER FOR STRUCTURE UNIFICATION GRAMMAR" docx

... grouping of information, thus expressing the information in- terdependencies. The language which SUG pro- vides for specifying these descriptions allows par- tiality both in the information about ... thereby also forgetting the predications over the nodes. This forgetting operation abstracts away from the existence of the forgotten node in the phrase structure. Once a node is forgotten it ... Unification Grammar is a formaliza- tion of accumulating information about the phrase structure of a sentence until this structure is com- pletely described. This information is specified in...

Báo cáo khoa học: "SVD and Clustering for Unsupervised POS Tagging" docx

... map-free information-theoretic criterion—see Gao and Johnson (2008) for details. Although we find M-to-1 to be the most reliable criterion of the three, we include the other two criteria for completeness. ... Table 1 compares the per-formance of SVD2 to other leading models. Fol-lowing Gao and Johnson (2008), the number of induced tags is 17 for PTB17 evaluation and 50 for PTB45 evaluation. Thus, ... NVI scores (Reichart and Rappoport 2009) corres-ponding to the VI scores for SVD2 are 0.938 for PTB17 and 0.885 for PTB45. To examine the sensitivity of the algorithm to its four parameters,...

Báo cáo khoa học: "Stochastic Iterative Alignment for Machine Translation Evaluation" doc

... N for i = 1; i ≤ M; i = i +1 do for j = 1; j ≤ N; j = j +1 do for k = 1; k ≤ i; k = k +1 do for m = 1; m ≤ j; m = m +1 doscorei,j,k,m= max{scorei−1,j,k,m,scorei,j−1,k,m} ;end for end for scorei,j,i,j=maxn=1,M;p=1,N{scorei,j,i,j, ... |ref|; for i = 1; i ≤ M; i = i +1 do for j = 1; j ≤ N; j = j +1 do for k = 1; k ≤ i; k = k +1 do for m = 1; m ≤ j; m = m +1 doscorei,j,k,m= max{scorei−1,j,k,m, scorei,j−1,k,m};end for end for if ... 1993). Stochastic word matching is auniform replacement for both morphologicalprocessing and synonym matching. More im-portantly, it can be easily adapted for differ-ent kinds of languages, as...

Báo cáo khoa học: "A Debug Tool for Practical Grammar Development" doc

... example.1 IntroductionThere is an increasing need for syntactical parsers for practical usages, such as information extrac-tion. For example, Yakushiji et al. (2001) extractedargument structures from ... data for thedevelopers to clarify the defects of thegrammar statistically. We applied willexto a large-scale HPSG-style grammar asan example.1 IntroductionThere is an increasing need for ... corpora in XML format. Second, it recordsdata of grammar defects to allow developers to havea whole picture of parsing errors found in the targetcorpora to save debugging time and effort by priori-tizing...

Báo cáo khoa học: "Randomised Language Modelling for Statistical Machine Translation" doc

... hash of event {x, j} under hiBF[hi(x)] ← 1end for end for end for return BF3.1 Log-frequency Bloom ﬁlterThe efﬁciency of our scheme for storing n-gramstatistics within a BF relies on ... bound on qc(x) ∈ Strain for j = 1 to M AXQCOUNT do for i = 1 to k dohi(x) ← hash of event {x, j} under hiif BF[hi(x)] = 0 thenreturn j − 1end ifend for end for The probability of overestimating ... 3-grams,the actual error rate of the former is lower for mod-els with less memory. By testing for 2-grams priorto querying for the 3-grams, we can avoid perform-ing some queries that may otherwise have...

Báo cáo khoa học: "THE TEXT SYSTEM FOR NATURAL LANGUAGE GENERATION" doc

... used for TEXT i. 2. 3. 4. identification -requests for definitions attributive -requests for available information constituency -requests for definitions -requests for available information ... detailed attributive information is included. For entities that are very different, only generic class information is included. A combination of this information is included for entities falling ... constructed by a fairly simple process. For requests for definitions or available information, the area around the questioned object containing the information immediately associated with the...

Xem thêm