3784500541

240 Jarosław Gramacki. Artur Gramacki

Tabela 10. Przykładowy dokument, który należy podsumować: Wprowadzenie do systemu Oracle Text. 26 zdań w języku angielskim. Tekst pokazano z wyraźnym zaznaczeniem wchodzących w jego skład zdań

^1	Oracle Text, Oracle‘s integrated full-text retrieval technolog} , is part of the Oradei lg Standard and Enterprise Edi-tions.
2	Oracle Text uses standard SQL to index, search, and analyze text and documents stored in the Oracle database, in files. and on the Web.
3	Oracle Text can perform linguistic analysis on documents: search text using a variety of stratcgies including keyword searching. contextual ąueries. Boolean operations, pattem inatching. inixed thernatic queries, HTML/XML section searching. etc.
4	Oracle Text excels at mixed queries. those tliat involve stnictured relational attributes as well as text.
5	Oracle Text can render search results in various fonnats including unfomiatted text. HTML with term higlrlighting. and original document format.
6	Oracle Text supports multiple languages and uses advanced relevance-ranking technolog} to improve search quality.
7	Over the last decade, organizations liave imested heavily in systems that enable rapid access to stnictured data stored in database systems.
8	However, this data represents a fraction of all corporate information.
9	A far larger volume exists as text - in documents. web pages. manuals. rcports. email. faxes. and presentations.
10	Tliese valuable sources of business infonnation are oftcn inaccessiblc and not managcd in a cost-effective manner.
11	Users accessing organization infonnation - w hether they are employees visiting an intranet portal or btiyers browsing a catalog - need sophisticated support from text search infrastructure to find what they want.
12	Text is undemtilized in many organizations.
13	Text assets are no longer static. physical entities.
14	Current technology allows companies to create globally interconnected systems that storę text information drawn from many sources.
15	Important text assets may be hidden because it's difficult to find them.
16	Poor search quality is expensive.
17	nlocking the value of an organization*s textual infonnation has been a long tenn challenge.
18	Historically. text has been seen to require a different set of technologies for retrieval and management than other business data.
19	This misperception has burdened organizations with multiple storage and retrieval systems. and also multiple devel-oprncnl emironments.
20	This has stood in the way of effectively integrating all of the corporation’s information assets.
21	As a legacy of this misperception. many companies today buy different products for soh ing their text searching needs and their structured data (database) searching needs.
22	Not only is this approach costly over the lifecycle of purchasing, integrating. operating and maintaining different products. but it also results in poor performance and a high latency in development of applications.
23	Further. puryeyors of specialty servers can seldoin deliver the high reliability. throughput and multi-platform scal-ability of an enterprise database.
24	Wliat if it were possible to extend the power and advantages of relational database systems to all corporate information. including text and other unstructured data ?
25	Aftcr all. text data is real data tliat warrants the infrastmcture of a real database and proven tools for application development.
26	In this w lute paper. we look at such an approach in the fonn of Oracle Text.

4.1. Struktura TSM

Zawartość każdego podsumowywanego dokumentu można przedstawić w postaci macierzy TDM opisywanej we wcześniejszych rozdziałach. Różnica jest tylko taka, że macierz, zamiast dokumentów, w kolumnach zawierać będzie zdania⁸. Macierz taką nazywa się macierzą term-zdanie (ang. Term Sentence Matrix, TSM). Jest ona punktem wyjścia do dalszej analizy. Indeksuje

Oczywiście większe dokumenty można podsumowywać poprzez wybór większych ich fragmentów. przykład akapitów.