Báo cáo khoa học: "Semi-Supervised Modeling for Prenominal Modifier Ordering" ppt

... Association for Computational Linguistics:shortpapers, pages 236–241, Portland, Oregon, June 19-24, 2011. c 2011 Association for Computational Linguistics Semi-Supervised Modeling for Prenominal Modifier ... a modifier s position, but it is clear that some modifier orderings sound more natu- ral than others, even if all are strictly speaking gram- matical. Determining methods fo...

Ngày tải lên: 30/03/2014, 21:20

... method for incor- porating word pronunciation information in a noisy channel model for spelling cor- rection. The proposed method builds an explicit error model for word pronuncia- tions. By modeling ... Pronunciation Modeling for Improved Spelling Correction Kristina Toutanova Computer Science Department Stanford University Stanford, CA 94305 USA Robert C. Moore Microsoft ... cor...

Ngày tải lên: 17/03/2014, 08:20

... Statistical Modeling for Unit Selection in Speech Synthesis Cyril Allauzen and Mehryar Mohri and Michael Riley ∗ AT&T ... and concatenation costs, essential for the design of the unit selection component. In contrast to these ap- proaches, we introduce a general statistical model- ing framework for unit selection inspired ... automata and transducers for the repre- sentation of...

Ngày tải lên: 23/03/2014, 19:20

... unseen for a given domain s, we are already performing an implicit form of smoothing (when computing the expected counts), since each docu- ment has a distribution over all topics, and therefore we ... corpus may be beneficial for translation. In this work, we consider the underlying latent topics of the documents (Blei et al., 2003). Topic modeling has received some use in SMT, for in...

Ngày tải lên: 19/02/2014, 19:20

... of NN for book. Then we maintain the top two hypotheses for span book interesting as shown below. The second most favorable label for interesting is still JJ, but in the context of VB for book. NN ... denote the top state for p as p.S = argmax s: M p (s)=∅ V (s.T ). Therefore, for each span p, we have a top hypothe- sis p.S.T , whose score is the highest among all the hypotheses...

Ngày tải lên: 20/02/2014, 12:20

... recall for the parsers mentioned in §3. The numbers are collated for reference only and are not directly comparable; see the text for details. 3 Parser Performance We first compare the performance ... to Table 1 for precision and recall for each model. Note that the CKK reordering requires not just category labels (e.g. NP) but also function labels (e.g. SB for subject); parser per...

Ngày tải lên: 07/03/2014, 22:20

... exactly the same structure of informa- tion states; but an implementor of a feature- based NLP system, for example, could secretly perform all of the work for those grammars in N, and no ... smallest set, I, such that, for every parametric type, p(Xt, ,Xn) E P, and for every tuple, (tt tn)EI n, p(tl, , tn)eI. • p(tl, ,tn) EI q(ut, ,urn) iff p EP q, and, for all l<i&...

Ngày tải lên: 08/03/2014, 05:21

... states for first modifier and rest of modifiers. • LEX+FCP: four distinct states, encoding: first modifier, previous modifier was a coor- dination, previous modifier was punctuation, and previous modifier ... algorithm can be formulated as a technique for inducing hidden structure from distributions computed by forward- backward recursions. Furthermore, we also present an inside-outside...

Ngày tải lên: 08/03/2014, 21:20

... 02-21 for training, 22 for development, and 23 for testing. Both DP and non-DP parsers use the same feature templates in Table 1. For Secs. 4.1-4.2, we use a baseline model trained with non-DP for ... bits of information from only the last (d + 1) trees on stack instead of the whole stack, because they provide all the relevant information for the features, and thus determine the c...

Ngày tải lên: 17/03/2014, 00:20

... en- try for “pushes” furthermore uses subj and obj for the subject and object argument, respectively. We combine here for simplicity the entries for “the” and “button” into “the button”. For generation, ... adjectives. We enforce this by keep- ing track of the current premodifier index of the RE in atoms of the form premod–index. Any newly generated RE node starts off with a premodifi...

Ngày tải lên: 17/03/2014, 00:20