Abstract—Users today need to express their informational
need in a way such that the search results can be further
analyzed to directly address the need, instead of merely
returning a list of lexical hits. In this paper we address the
problem of extracting sentiment metadata related to the user’s
topic or entity of interest along with the search results. We
propose a framework that provides the flexibility of specifying a
sentiment related informational need associated with a
particular topic of interest, along with the standard keyword
query. This information is then used to extract sentiment scores
for the expressed entity of interest based on the “hitlist” of most
relevant documents returned by the main search query. Our
experimental results based on product and movie review
datasets, demonstrate the advantages of embedding the
sentiment processing within the search engine framework.
Index Terms—Sentiment analysis, information retrieval,
query framework, text analytics, sentiment aware search.
The authors are with Oracle Text, USA (e-mail: shubhro.roy@oracle.com,
alexandra.czarlinska@oracle.com, asha.tarachandani@oracle.com).
[PDF]
Cite: Shubhro Jyoti Roy, Alexandra Czarlinska, and Asha Tarachandani, "Framework for Sentiment Aware Queries and Results in Search Using Oracle Text," International Journal of Knowledge Engineering vol. 1, no. 2, pp. 83-91, 2015.
