Proposal Daniel Michlits h Research Seminar System Analyses
Agenda Related works Problem statement Movie Review Mining and Summarization Experiments Summary of the base article Research questions
base article title: „Movie Review Mining and Summarization“ authors: Li Zhuang, Feng Jing and Xiao- Yan Zhu published by: CIKM’06 November 5-11,2006, Arlington,Virginia
Summary Related works: Most of the existing work is focused on produkt reviews Subjective classification distinguishes between sentences that contain objective information and sentences that describe opinions Sentiment classification Defines the semantic orientation of words
Summary Problem statement Definine classes for features Divide the classes into two groups Elements: OA (overall), ST (screenplay), CH (character design), VP (vision effects), MS (music and sound),… People: PPR (producer), PDF (director), PSC (screenwriter), PAC (actor and actress),…
Summary Movie Review Mining and Summarization Generating a Keyword list including the most important feature & opinion words relating the keywords to the element classes nouns, such as movie names or people names, can also be feature words Mining explicit feature-opinion pairs Using the keyword list and dependency relation templates Mining implicit feature-opinion pairs Only in short sentences at the beginning or end of the review
Summary Movie Review Mining and Summarization Generating the final summary Collecting sentences that express opinions on a specific feature class Identifying semantic orientation of the relevant opinion in each sentence is identified Creating an organized sentence list
Summary
Research Questions 1.How much are movie reviews influencing the decision of consumers? Do they also influence the perception during a movie? 2. Which keywords have the biggest impact on the consumer? 3.Which are the most important elements of movie reviews regarding the consumer purchase decision process?
Summary Movie Review Mining and Summarization Generating the final summary Collecting sentences that express opinions on a specific feature class Identifying semantic orientation of the relevant opinion in each sentence is identified Creating an organized sentence list