{"title":"GETTING CLOSER TO THE THINGS WE ARE LOOKING FOR - A PROPOSAL FOR ASPECT-DRIVEN TEXT ANALYSIS","authors":"Kurt Englmeier, M. Schneider","doi":"10.33965/icwi2020_202012l013","DOIUrl":null,"url":null,"abstract":"Locating facts in texts is still a challenge for retrieval systems. Instead of measuring the relevance of the entire text for a search query, the focus of our approach of fact retrieval and text mining is on spotting only essential parts of text. We propose to enrich search queries by aspects supporting highly precise retrieval. An aspect denotes the unique interpretation of a set of words that appear in close proximity. We conceive an aspect representation as a hierarchical structure composed by its semantic elements that, in turn, constitute the aspect’s implicit meaning. In this paper, we present the definition and application of aspect blueprints that serve as standardized representations of aspects. Our prototypical system Contexter operates on these blueprints and locates facts in texts. In a semi-automatic way, the system tries to detect variants of blueprints that the user can confirm or reject. Here, we illustrate our approach and system in the realm of economic information.","PeriodicalId":254527,"journal":{"name":"Proceedings of the 19th International Conference on WWW/Internet","volume":"102 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2020-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Proceedings of the 19th International Conference on WWW/Internet","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.33965/icwi2020_202012l013","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Locating facts in texts is still a challenge for retrieval systems. Instead of measuring the relevance of the entire text for a search query, the focus of our approach of fact retrieval and text mining is on spotting only essential parts of text. We propose to enrich search queries by aspects supporting highly precise retrieval. An aspect denotes the unique interpretation of a set of words that appear in close proximity. We conceive an aspect representation as a hierarchical structure composed by its semantic elements that, in turn, constitute the aspect’s implicit meaning. In this paper, we present the definition and application of aspect blueprints that serve as standardized representations of aspects. Our prototypical system Contexter operates on these blueprints and locates facts in texts. In a semi-automatic way, the system tries to detect variants of blueprints that the user can confirm or reject. Here, we illustrate our approach and system in the realm of economic information.