Identifying salient entities in web pages

Abstract

We propose a system that determines the salience of entities within web documents. Many recent advances in commercial search engines leverage the identification of entities in web pages. However, for many pages, only a small subset of entities are central to the document, which can lead to degraded relevance for entity triggered experiences. We address this problem by devising a system that scores each entity on a web page according to its centrality to the page content. We propose salience classification functions that incorporate various cues from document content, web search logs, and a large web graph. To cost-effectively train the models, we introduce a soft labeling methodology that generates a set of annotations based on user behaviors observed in web search logs. We evaluate several variations of our model via a large-scale empirical study conducted over a test set, which we release publicly to the research community. We demonstrate that our methods significantly outperform competitive baselines and the previous state of the art, while keeping the human annotation cost to a minimum.

Modeling Interestingness with Deep Neural Networks

“…The two other sources of semantic features are used as a point of comparison to the DSSM. One is a generative semantic model (Joint Transition Topic model, or JTT) (Gamon et al 2013). JTT is an LDA-style model (Blei et al 2003) that is trained jointly on source and target documents linked by browsing transitions.…”

“…In addition to the notion of relevance as described in Section 1, related to interestingness is also the notion of salience (also called aboutness) (Gamon et al 2013;2014;Parajpe 2009;Yih et al 2006). Salience is the centrality of a term to the content of a document.…”

“…In contrast to these approaches, we strive to predict what term a user is likely to be interested in when reading content, which may or may not be the same as the most popular content that is related to the current document. It has empirically been demonstrated in Gamon et al (2013) that popularity is in fact a rather poor predictor for interestingness. The task of contextual entity search, which is formulated as an information retrieval problem in this paper, is also related to research on entity resolution (Stefanidis et al 2013).…”

"Making the News": Identifying Noteworthy Events in News Articles

“…The authors suggested the use of centering constructs to keep track of the key entities, which change with discourse. Document-level importance of entities (which include events) was explored by Gamon et al (2013). The authors use the term salience to denote entity importance and graded entities into 3 categories -most salient, less salient, not salient.…”

From context to query

“…Information need expression: If an information need is detected, the query is constructed from the page's relevant entities. The task of extracting relevant entities is somewhat similar to extracting salient entities [1], but opposed to salience or aboutness, relevance incorporates the user's interests. The relevant terms are determined by a named entity recognizer, using an adapted ranking mechanism based on the relatedness of the candidate entities to interest topics in the user profile.…”