Results 11 - 20 of 31,491

Table 1. Brief database schema and statistics of index terms per document

in unknown title
by unknown authors
"... In PAGE 6: ... Figure 3 shows the 1,296th example item which con- sists of @DOCUMENT(record delimiter), TITLE, AU- THOR, JOURNAL, VOLUME, NUMBER, PAGE START, PAGE END, PUBDATE(publish date), ABSTRACT and KEYWORDS. Table1 shows average term counts in every section in the text collection (a section is the equivalent to a eld or column in database world). Remind that to keep the lexicographical order of the terms in an inverted le, usually a vocabulary data structure is separated from the postings list, and in the KRISTAL-IRMS postings lists are further separated physically into two parts, postings of document identi ers with term frequencies and exact locations in the documents.... In PAGE 6: ... Remind that to keep the lexicographical order of the terms in an inverted le, usually a vocabulary data structure is separated from the postings list, and in the KRISTAL-IRMS postings lists are further separated physically into two parts, postings of document identi ers with term frequencies and exact locations in the documents. Now refer to Table1 where a document con- tains more than one hundred unique terms, which means... In PAGE 7: ...cess three disk data structures the vocabulary list, posting list and location list for each unique term from the input document. As shown in Table1 , KRISTAL-IRMS supports data types such as string (var char equivalent but with length limit of 2GB (Usually DB has 4K limit); KSTRING in KRISTAL nomenclature), character array (KCHAR[]), integer (KINT), and oat (KFLOAT), of which index type can be any one among INDEX AS IS (whole section value as an indexing unit), INDEX NUMERIC (indexing section value as a oating point), INDEX BY CHAR (char level in- dexing), INDEX BY TOKEN (token level indexing), several more indexing types for Korean and Chinese characters, and a few other types related to biological sequences such as DNA and proteins. In the auxiliary strategies, the non-contiguous nature of postings lists, being separated in the main index and auxil- iary index, may result in deteriorated postings access, com- pared with contiguous posting lists.... ..."

Table 1: Examples of indexing terms for different subword units.

in unknown title
by unknown authors 1998
Cited by 17

Table 1: Examples of nphone subword unit indexing terms.

in Towards Robust Methods For Spoken Document Retrieval
by Kenney Ng 1998
Cited by 14

Table 3: Examples of D2-phone subword unit indexing terms.

in Information Fusion For Spoken Document Retrieval
by Kenney Ng 2000
Cited by 5

Table 2: Summary of results on side collection of choosing different index terms.

in TREC-10 Experiments at University of Maryland CLIR and Video
by Kareem Darwish, David Doermann, Ryan Jones, Douglas Oard, Mika Rautiainen 2002
Cited by 1

Table 1: Manually extracted index terms and rele- vancy to exercise

in A Novel Approach to Semantic Indexing Based on Concept
by Bo-Yeong Kang Department

Table 1: Distribution of index terms extracted from documents

in Effectiveness of Complex Index Terms In Information Retrieval
by Tokunaga Takenobu, Ogibayasi Hironori, Tanaka Hozumi

Table 1: Distribution of index terms extracted from documents

in Effectiveness of Complex Index Terms in Information Retrieval
by Tokunaga Takenobu, Ogibayasi Hironori, Tanaka Hozumi

Table 5. Maximum number of index terms per article

in
by unknown authors

Table 5. The results of combining the HMM and RM models, using both word and syllable indexing terms; the RM model was constructed with the IR system using syllable indexing terms.

in Extractive Chinese Spoken Document Summarization Using Probabilistic Ranking Models
by Yi-ting Chen, Suhan Yu, Hsin-min Wang, Berlin Chen
"... In PAGE 11: ... Compared with the results in Table 3, the summarization model implemented with syllable indexing terms is considerably better than the one implemented with word indexing terms, especially at lower summarization ratios. Finally, the results derived by combining the HMM and RM models, as well as by using both word and syllable indexing terms, are shown in Table5 . Compared with the results in Table 4, the fusion of these two kinds of indexing information clearly yields additional performance gains.... ..."
Results 11 - 20 of 31,491