Application of Localized Similarity for Web Documents

Proceedings of the 2013 Conference on Empirical Methods in Natural Language Processing, pages 1399–1404, Seattle, Washington, USA, 18-21 October 2013. c⃝2013 Association for Computational Linguistics Application of Localized Similarity for Web Documents Peter Reberšek Zemanta Celovška cesta 32 Ljubljana, Slovenia peter.rebersek@zemanta.com Mateja Verliˇc Zemanta Celovška cesta 32 Ljubljana, Slovenia mateja.verlic@zemanta.com Abstract In this paper we present a novel approach to automatic creation of anchor texts for hyper- links in a document pointing to similar doc- uments. Methods used in this approach rank parts of a document based on the similarity to a presumably related document. Ranks are then used to automatically construct the best anchor text for a link inside original document to the compared document. A number of dif- ferent methods from information retrieval and natural language processing are adapted for this task. Automatically constructed anchor texts are manually evaluated in terms of relat- edness to linked documents and compared to baseline consisting of originally inserted an- chor texts. Additionally we use crowdsourc- ing for evaluation of original anchors and au- tomatically constructed anchors. Results show that our best adapted methods rival the preci- sion of the baseline method. 1 Introduction One of the features of hypertext documents are hy- perlinks that point to other resources – pictures, videos, tweets, or other hypertext documents. fairly familiar category of the latter is related arti-A cles; these usually appear at the end of a news article or a blog post with the title of the target document as anchor text. The target document is similar in con- tent to original document; it may tell the story from another point of view, it may be a more detailed ver- sion of a part of the events in the original document, etc. Another category are the in-text links; these ap- pear inside the main body of text and use some of the existing text as anchor. Ideally the anchor text is selected in such a way that it conveys some informa- tion about the target document; in reality sometimes just an adverb (e.g. here, there) is used, or even the destination URL may serve as anchor. Our goal is to develop a system that automatically constructs in-text links, i.e. for a query document finds a target document and an appropriate part of the text of the query document that serves as the an- chor text for the hyperlink. We want the target docu- ment to be similar in content to the query document and the anchor text to indicate that content. There are many potential uses for such a system, especially for simplifying and streamlining docu- ment creation. This includes authors of blogs that may use the system for adding related content from other sources without exhausting manual search for such material. It may also be used when writing a scientific paper, automatically adding citations to other relevant papers inside the main body. This ac- celerates the writing, again reducing the time spent searching for possible existing research in the field. A citation …

Content truncated to 500 words