Using Wikipedia Titles to Create a Semantic Relation Searching System Supporting Paraphrase Matching

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

MOST106-2221-E019-072

Translated Name/計畫中文名

以維基百科條目建立支援同義改寫比對的語意關係查詢系統

Project Coordinator/計畫主持人

Chuan-Jie Lin

Funding Organization/主管機關

National Science and Technology Council

Department/Unit

Department of Computer Science and Engineering

Website

https://www.grb.gov.tw/search/planDetail?id=12487572

Year

2017

Start date/計畫起

01-08-2017

Expected Completion/計畫迄

31-07-2018

Bugetid/研究經費

539千元

ResearchField/研究領域

資訊科學--軟體

在許多自然語言處理技術以及應用系統中，語意資訊一直都是扮演最為關鍵的角色。目前常用的中文語意詞典有許多不足的地方，主要是語意詞典中專有名詞過少。以自動方式創建一個大型的專有名詞詞典，並能與現有語意詞典的語意關係建立連結，將對許多自然語言處理應用有莫大的幫助。本計畫擬以一年的時間，研究如何自動抽取維基百科中屬於專有名詞的條目、如何為各條目選取具上義關係的分類、如何為維基百科分類架構建立與WordNet 之間的關連、並同時解決同義改寫比對問題的方法。這項研究成果預計將建立一個大型的專有名詞語意資源，提供支援同義改寫比對的語意關係查詢系統，有助於其他自然語言應用系統的效能。建立流程保持自動化特性，可隨時由最新版維基百科內文重新建立出最新的專有名詞語意資源。預期完成的成果描述如下： (1) 連結維基百科類別之WordNet 上義詞群的找尋模組 (2) 選取維基百科條目具上義關係之維基類別的判斷模組 (3) 維基百科資訊框模板所對應之專名實體類別 (4) 找尋英文維基百科專有名詞條目的判斷模組 (5) 找尋中文維基百科專有名詞條目的判斷模組 (6) 大型的專有名詞語意資源 (7) 支援同義改寫比對的語意關係查詢系統 In many natural language processing applications, semantic information usually plays an important role. However, the scales of popular Chinese semantic resources are often not large enough, mostly because the lack of proper nouns. A study to automatically create a large scale of semantic dictionary containing tons of named entities and establish links to well-known semantic resources will greatly benefit NLP researches and applications. This one-year project plans to study how to automatically extract Wikipedia entries which are named entities, how to select categories which are hypernyms of the Wikipedia entries, how to establish links between Wikipedia categories and well-known semantic resources, and how to match paraphrases. This project will build a large-scale named-entity semantic resource, and a system providing semantic relation searching which also support paraphrase matching. The process will be fully automatic as much as possible so that the resource can be updated easily whenever there is a newer version of Wikipedia. The results achieved from this project are as follows. (1) A search module to build links from Wikipedia categories to WordNet synsets (2) A classifier to select Wikipedia categories which are hypernyms of Wikipedia entries (3) A mapping table from Wikipedia infobox templates to named entity types (4) Classifiers to find English Wikipedia entries which are named entities (5) Classifiers to find Chinese Wikipedia entries which are named entities (6) A large-scale named-entity semantic resource (7) A system providing semantic relation searching which also support paraphrase matching

DSpace CRIS

Using Wikipedia Titles to Create a Semantic Relation Searching System Supporting Paraphrase Matching

Details

Description