Báo cáo khoa học: "Learning to Generate Naturalistic Utterances Using Reviews in Spoken Dialogue Systems" ppt

... Association for Computational LinguisticsLearning to Generate Naturalistic U tterances Using Reviews in Spok en Dialogue SystemsRyuichiro HigashinakaNTT Corporationrh@cslab.kecl.ntt.co.jpRashmi ... on which Minipar was trained, the outputof Minipar can be inaccurate, leading to failure in conversion. We check whether conversion is suc-cessful in the ﬁltering stage.2.4 FilteringThe goal ... attribute in subject position as a deﬁnitenoun phrase. However, the learned DSyntS canbe incorporated into SPaRKy using the semanticrepresentations to substitute learned DSyntSs intonodes in the...

Báo cáo khoa học: "Natural Language Generation as Planning Under Uncertainty for Spoken Dialogue Systems" pptx

... operat-ing in a stochastic environment, and optimiseit using Reinforcement Learning. The in- put to the module is a Communicative Goalsupplied by the Dialogue Manager. The CGconsists of a Dialogue ... evaluate a newmodel for NLG in Spoken Dialogue Systems asplanning under uncertainty. In Section 2 we arguefor applying RL to NLG problems and explain theoverall framework. In Section 3 we discuss ... planning task, wherethere are trade-offs and decisions to make, such aswhether to choose another NLG action (and whichone to choose) or to instead stop generating. Re-inforcement Learning...

Báo cáo khoa học: "User Simulations for context-sensitive speech recognition in Spoken Dialogue Systems" potx

... scoring ﬁrst in the InformationGain rank, validating our central hypothesis.The testing of the classiﬁer using each of theabove feature sets was performed on the remain-ing 25% of the Town-Info ... by the In- formation Gain measures that were calculated onthe training set using WEKA, as shown in table 3.8.1 Information GainQuite surprisingly, we note that the rank given bythe Information ... trained on 75% of the same Town-Info Corpusof 126 dialogues containing 60-best lists for 1554user utterances or a total of 93240 hypotheses. Theﬁrst layer was tested against a separate Town-InfoCorpus...

Tài liệu Báo cáo khoa học: "Learning to Compose Effective Strategies from a Library of Dialogue Components" doc

... easily be plugged in to test their usefulness in different dialogue contexts or for new domains.4 Training the Dialogue System4.1 Random strategy generation In its training mode, the dialogue system ... similar to ours, constisting of: generationof (limited) exploratory dialogue data, using a train-ing system; creating an MDP from these data andthe rewards assigned by the training users; off-linepolicy ... for automat-ically learning effective dialogue strategies,generated from a library of dialogue content, using reinforcement learning from user feed-back. This library includes greetings, so-cial...

Báo cáo khoa học: "Trainable Sentence Planning for Complex Information Presentation in Spoken Dialog Systems" pot

... combinestwo distinct clauses into a single sentence with acoordinating or subordinating conjunction (e.g.Above has decent decor BUT Carmine’s hasgood decor), while cue-word insertion insertsa cue ... domains. For example, the trainingmethod used in SPoT (Sentence Planner Train-able), as described in (Walker, Rambow and Ro-gati, 2002), was only shown to work in the traveldomain, for the information ... quality output in particular domains.A promising approach is trainable generation,which uses general-purpose linguistic knowledgeautomatically adapted to the application do-main. This paper...

Báo cáo khoa học: "Learning to Rank Deﬁnitions to Generate Quizzes for Interactive Information Presentation" doc

... learning.4.3 Training Ranking Models Using the reference ranking data, we trained a rank-ing model using the ranking SVM (Joachims, 2002)(with a linear kernel) that minimizes the pairwiseranking ... 0.13. Finally, taking the means ofranks given to each deﬁnition, we merged the indi-vidual rankings to create the reference rankings. Anexample of a reference ranking is the bottom one in Fig. ... authors.Ranking deﬁnitions is closely related to deﬁni-tional question answering and sentence ordering in multi-document summarization. In deﬁnitionalquestion answering, measures related to informationretrieval...

Tài liệu Báo cáo khoa học: "Learning to Translate with Multiple Objectives" doc

... information is added to the training set T (line 8), which is then optimizedby any conventional subroutine in line 10. We willfollow PRO in using a pairwise classiﬁer in line 10,which ﬁnds w∗that ... necessarily pareto-optimal. Algorithm 1shows the pseudocode. In line 3, we take a point h∗and check if it is dominating or dominated in the for-loop (lines 4-8). At least one pareto-optimal pointwill ... methods. The maindifference is that rather than trying to maximize asingle metric, we maximize the number of paretopoints, in order to expand the Pareto FrontierWe will explain PMO-PRO in terms...

Tài liệu Báo cáo khoa học: "Learning to Find Translations and Transliterations on the Web" doc

... while requiring minimal human intervention to prepare the training data. 3 Method To find translations for a given term on the Web, a promising approach is automatically learning to extract ... potentially be used to assist translators to find the most common translation for a given term, or to supplement a bilingual terminology bank (e.g., adding multilingual titles to existing Wikipedia); ... seeds to obtain the training and test data. Since we are using Wikipedia bilingual titles as the gold standard, we exclude any snippets from the wikipedia.org domain, so that we are not using...

Tài liệu Báo cáo khoa học: "Learning to Follow Navigational Directions" pdf

... set-ting it is situated in, motivating automated learn-ing approaches to acquiring meaning. Tradi-tional accounts of learning typically rely on lin-guistic annotation (Zettlemoyer and Collins, ... ourattention to just the utterances of the instructiongiver, ignoring the instruction follower. This is to reduce redundancy and noise in the data - the in- struction follower rarely introduces new informa-tion, ... tend to be interpreted as going to the north and southof landmarks, respectively. Interestingly, our sys-tem tends to move in the opposite cardinal direc-tion, i.e. the agent moves south in...

Tài liệu Báo cáo khoa học: "Learning to Recognize Tables in Free Text" pptx

... this paper due to space constraint. 443 Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line Line 1234567890123456789012345678901234567890123456789012345678901234567890 ... tables in Figure 1 to 3. Instead of resorting to an ad-hoc method to rec- ognize tables, we present a new approach in this pa- per that learns to recognize tables in plain text. Our learning ... kinds of characters and their rela- tive locations in a line and across lines to recognize tables. It is domain independent and does not rely on any domain-specific knowledge. We want to in- ...

Xem thêm