Báo cáo khoa học: "Simple Supervised Document Geolocation with Geodesic Grids" pdf

... lo- cations for unseen documents/users. In this paper, we tackle document geolocation us- ing several simple supervised methods on the textual content of documents and a geodesic grid as a dis- crete ... describe several simple supervised methods for document geolocation using only the document s raw text as evi- dence. All of our methods predict locations in the context of ge...

Ngày tải lên: 23/03/2014, 16:20

... selected another 25 documents and added them to the previous 50 documents to get 75 documents. We made sure that every document participated in this experiment. The training documents for each ... selected 70 documents from the 348 documents as test data which roughly equaled the size of 1 fold in the baseline in Section 6.1. For the baseline in this section, all the rest of the doc...

Ngày tải lên: 30/03/2014, 21:20

... monolingual sentence, i  denotes the number of words that are aligned with i w . Since a word never collocates with itself, the alignment set is denoted as }&],1[|),{( ialiaiA ii  . ... 2010. c 2010 Association for Computational Linguistics Improving Statistical Machine Translation with Monolingual Collocation Zhanyi Liu 1 , Haifeng Wang 2 , Hua Wu 2 , Sheng Li 1 1...

Ngày tải lên: 20/02/2014, 04:20

... are presented in (Quirk, 2004) and (Ueffing and Ney, 2005). The former experimented with confidence esti- mation with several different learning algorithms; the latter uses word-level confidence measures ... viable using the ‘kernel trick’: finding a kernel function K in (1) with K(x i , x j ) = Φ(x i ) T Φ(x j ). We perform our ex- periments with the Radial Basis Function (RBF) kernel, a...

Ngày tải lên: 23/03/2014, 16:20

... (2009) stacks a POS tagger on an un- supervised clustering algorithm trained on large amounts of unlabeled data with mixed results. This work combines a new semi -supervised learning method to POS ... of supervised learning algorithms. Specifically we have experimented with support vector machines (SVMs), decision trees, bagging and random forests. Tri-training, explained below, is a...

Ngày tải lên: 07/03/2014, 22:20

... in dependency parsing applications. In general, semi -supervised learning can be mo- tivated by two concerns: first, given a fixed amount of supervised data, we might wish to leverage ad- ditional ... to work with the Brown algorithm due to its simplicity and prior suc- cess in other NLP applications (Miller et al., 2004; Liang, 2005). However, we expect that our approach can function wit...

Ngày tải lên: 17/03/2014, 02:20

... sms@tsinghua.edu.cn Abstract Searching documents that are similar to a query document is an important component in modern information retrieval. Some ex- isting hashing methods can be used for effi- cient document similarity ... via the kernel trick. However, KLSH is unsupervised, thus design- ing a data-specific kernel remains a big challenge. 2.3 Semi -Supervised Hashing Semi -Supervised...

Ngày tải lên: 30/03/2014, 21:20

... Understanding Conference (MUC) community has established within -document co- reference standards [MUC-7 1998]. Compared with within -document name disambiguation which can leverage highly reliable ... people are associated with the same name. In tracking person entities in a large document pool, it is important to determine whether multiple mentions of the same name across docu...

Ngày tải lên: 31/03/2014, 03:20

... positive documents are deemed positive or neutral, and vice versa for negative documents. As a way of overcoming these shortcomings, we propose to fuse a coarsely supervised model with a fully supervised model. Below, ... Preliminaries Let d be a document consisting of n sentences, s = (s i ) n i=1 , with a document sentence-sequence pair de- noted d = (d, s) . Let y d = (y d , y s...

Ngày tải lên: 20/02/2014, 05:20

... somewhat indicative of edge, etc. 5 Experimental Comparison with Unsupervised Learning In this section we compare GE training with meth- ods for unsupervised parsing. We use the WSJ10 corpus (as processed ... training of this model outperforms EM with 30 constraints and CE with 50 constraints. GE training of the full CRF outperforms EM with 10 constraints and CE with 20 constraint...

Ngày tải lên: 20/02/2014, 07:20