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

... 117–120,Prague, June 2007.c2007 Association for Computational LinguisticsLearning to Rank Definitions to Generate Quizzes for Interactive Information PresentationRyuichiro Higashinaka and ... biographi-cal facts) to automatically generate “Whois this?” quizzes. The definitions are or-dered according to how difficult they makeit to name the person. Such ranking wouldenable users to interactively ... people’s biographical in-formation to users, and propose ranking definitions to automatically generate a “Who i s this?” quiz.Here, we use the term definitions of a person’ to mean a short series...

Báo cáo khoa học: "Learning From Collective Human Behavior to Introduce Diversity in Lexical Choice" potx

... shown that information overlap judgment is a difﬁcult task for human annotators. To avoid such a difﬁculty, weenforced our annotators to extract non-overlappingnuggets from a summary to make sure ... paper.The use of phrasal information as nuggets is an es-sential element to our experiments, since some head-line writers often try to use uncommon terms to re-fer to a factoid. For instance, two ... in order to capture a larger information content we need to reada greater number of summaries. But how manyheadlines should one read to capture a desired levelof information content? To answer...

Báo cáo khoa học: "Learning Semantic Links from a Corpus of Parallel Temporal and Causal Relations" doc

... that temporal information is cru-cial to causal relation identiﬁcation.1 IntroductionWorking out how events are tied together temporallyand causally is a crucial component for successfulnatural ... Inter-annotator agreement by task.linked causally in WordNet (Girju, 2003) were used to collect examples for annotation, with the result-ing corpora allowing machine learning models to achieve performance ... rules to identify con-joined event pairs. 1000 pairs were annotated bytwo annotators and adjudicated by a third. Table 1SADVPRBThenNPPRPtheyVPVP CC VPVBDtookNPDTtheNNartPP TO toNPNNPAcapulcoandbeganSVBDVP TO toVPVBtradeNPsome...

Báo cáo khoa học: "M AX S IM: A Maximum Similarity Metric for Machine Translation Evaluation" doc

... judges have to rank thesetranslations).During the workshop, Kappa values measured for inter- and intra-annotator agreement for rank andconstituent are substantially higher than those for adequacy ... need to be performed.2.3 ParaEvalThe ParaEval metric (Zhou et al., 2006) uses alarge collection of paraphrases, automatically ex-tracted from parallel corpora, to evaluate MT per-formance. To ... only n-gram information (Section 4.1) for ourmetric, while MAXSIMn+drepresents using both n-gram and dependency information. We also showthe breakdown of the correlation results into the Eu-roparl...

Báo cáo khoa học: "Comparison of Alignment Templates and Maximum Entropy Models for Natural Language Understanding" docx

... proven to be very powerful for statistical machine trans-lation tasks since they allow for many -to- manyalignments between source and target words (Ochet al., 1999). Alignment templates for NLU ... yields a one- to- one translation from source words to formal se-mantic concepts, i.e. the length of both sequencesmust be equal (I = J). Figure 5 depicts a one- to- one mapping applied to a sentence/concept ... redundant or non-informative, it turned out to perform well in practice. Experiments were car-ried out with different thresholds. It turned out that for the NLU task, a threshold of 2 for all featuresachieved...

Báo cáo khoa học: "Learning to Rank" doc

...

Báo cáo khoa học: "Learning to Rank Answers on Large Online QA Collections" pptx

... pa-per. To answer our ﬁrst research objective we willcompare the quality of the rankings provided by thiscomponent against the rankings generated by the IRmodel used for answer retrieval. To answer ... adapted andcombined for non-factoid answer ranking, and per-form a comparative feature analysis using a signif-icant amount of real-world data. For clarity, wegroup the features into four sets: ... collections to investigate features for answer ranking without theneed for costly human evaluation, (b) we can exploitlarge and noisy online QA collections to improve theaccuracy of answer ranking...

Tài liệu Báo cáo khoa học: An engineered disulﬁde bridge mimics the effect of calcium to protect neutral protease against local unfolding docx

... appliedspectroscopic methods, lead to an autoproteolyticallysusceptible (intermediate) state. Similarly, Panchalet al. [33] took advantage of autoproteolysis of theHIV protease to monitor local unfolding ... act as anindicator for the occurrence of the intermediate in theunfolding.Calcium ions are the main contributory factor to enzyme stabilityThe binding of calcium ions seems to be the mainsource ... coincided for both of the enzymes at all GdnHClconcentrations. The autoproteolysis rates were identi-cal to unfolding rates for G8C ⁄ N60C over the wholerange investigated (4–8 m GdnHCl) and for...

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

... weight vector w, takes in a foreign sen-tence f and returns a translated hypothesis h. Theargmax operates in vector space and our goal is to ﬁnd w leading to hypotheses on the Pareto Frontier.In ... =FindParetoFrontier({M(h)})8: foreach h ∈ {h}:if h ∈ {f}, set l=1, else l=0; Add (l, h) to T9: end for 10: w∗=OptimizationSubroutine(T , w)11: Add w∗ to W; Set w = w∗.12: end for 13: ... FindParetoFrontier({M(w)})ble results for PMO-PRO: for non-pareto hypothe-ses h /∈ {f}, we set label l =kMk(h)/K in-stead of l= 0, so the method not only learns to dis-criminate pareto vs....

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

... Romanized Chinese characters. Finally, we use the alignment information to generate transliteration feature for a Chinese token with respect to English words in the query. We extract person or location ... associate features to each Chinese token (characters or words) to reflect the likelihood of the token being part of the translation. We describe how to train a CRF model for identifying translations ... 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);...

Xem thêm