Entities' Sentiment Relevance

Abstract

Sentiment relevance detection problems occur when there is a sentiment expression in a text, and there is the question of whether or not the expression is related to a given entity or, more generally, to a given situation. The paper discusses variants of the problem, and shows that it is distinct from other somewhat similar problems occurring in the field of sentiment analysis and opinion mining. We experimentally demonstrate that using the information about relevancy significantly affects the final sentiment evaluation of the entities. We then compare a set of different algorithms for solving the relevance detection problem. The most accurate results are achieved by algorithms that use certain document-level information about the target entities. We show that this information can be accurately extracted using supervised classification methods.

Document-level Sentiment Inference with Social, Faction, and Discourse Context

“…Previous work has shown that notions related to salience (e.g., proximity to sentiment words) can help to detect sentiment targets (Ben-Ami et al, 2014). In our data, we found that an entity's occurrence pattern is highly indicative of being involved in sentiment, for example the most frequently mentioned entity is 3.4 times more likely to be polarized and an entity in the headline is two times more likely to be polarized.…”

“…(2) The sentiment label of the path e i ↑ nsubj ↓ ccomp ↓ nsubj ↓ e j , when it exists (e.g., McCully said any action against Henry is a matter entirely for TVNZ) (3) The sentiment label of path when the path does not contain any named entity (e.g., Nobel winner , Shirin Ebadi) (4) An indicator for the link nmod:against. Document Features Previous work has shown that notions related to salience (e.g., proximity to sentiment words) can help to detect sentiment targets (Ben-Ami et al, 2014). In our data, we found that an entity's occurrence pattern is highly indicative of being involved in sentiment, for example the most frequently mentioned entity is 3.4 times more likely to be polarized and an entity in the headline is two times more likely to be polarized.…”

Automatic Sentiment Analysis of Texts: The Case of Russian

“…However, the author can refer to an entity using the means of reference, for example, pronouns. In addition, if the entire text is devoted to the discussion of one entity, then it can be explicitly mentioned far from the sentiment location (Ben-Ami et al 2014).…”

Role-based Association of Verbs, Actions, and Sentiments with Entities in Political Discourse

“…Note that our method is especially relevant to political contexts although it may apply to other domains as well. For example, a solution applied to commercial discourse combines proximity and supervised classification to attain higher accuracy (Ben-ami, Feldman, & Rosenfeld, 2014). However, this solution explicitly ignores cases involving interactions between entities of the same type (e.g., two companies).…”