Towards more fine-grained sentiment and emotion analysis of text
Who is this presentation for?developers working with textual data, research staff
When consumers see a piece of text, what kinds of sentiment and emotions do they associate with it? While there is a long history of sentiment analysis, this talk describes a series of results on data-driven approaches to provide a more multi-faceted and fine-grained understanding of these associations.
The first part of the talk will focus on sentiment, i.e., methods to detect whether a text is perceived as more positive or negative. The key novelty that we propose involves using simple vector representations that capture more fine-grained sentiment information. For instance, the word “hot” is often positive when referring to music, but tends to be negative when referring to the temperature in a hotel room. Using simple techniques, we can create sentiment vector representations that capture such differences between different contexts. These can then readily be exploited by machine learning approaches such as deep neural networks. This has allowed us to better analyze reviews of products, hotels, and restaurants not only in English, but as well in numerous other languages.
Subsequently, the talk will focus on emotions and the perception of text. While there are several psychological theories of emotion, data-driven approaches can provide detailed ratings revealing to what extent a given word on average evokes a specific emotion. We can then connect these ratings with further data to automatically recommend appropriate fonts and color palettes to use when presenting specific pieces of information. For example, certain fonts and colors are perceived as more exciting, while others are more likely to convey trustworthiness. Hence, one can make informed choices that better accord with marketing-based desiderata.
Overall, these methods open up new opportunities for organizations to pay attention to what is being said about them in different markets, and to make smarter choices when presenting information to consumers.
Prerequisite knowledgeNo prior knowledge of these topics is required. The talk is designed such that one can take away the high level story without needing to understand the technical details. There will numerous visual examples.
What you'll learn
Gerard de Melo
Gerard de Melo is an Assistant Professor of computer science at Rutgers University, where he heads the Deep Data Lab, a team of researchers working on big data analytics, natural language processing, and web mining. Gerard’s research projects include UWN/MENTA, one of the largest multilingual knowledge bases, and Lexvo.org, an important hub in the Web of Data. Previously, he was a faculty member at Tsinghua University, one of China’s most prestigious universities, and a visiting scholar at ICSI/UC Berkeley. He serves as an editorial board member for Computational Intelligence, the Journal of Web Semantics, the Springer Language Resources and Evaluation journal, and the Language Science Press TMNLP book series. Gerard has published over 100 papers, with best paper or demo awards at WWW 2011, CIKM 2010, ICGL 2008, and the NAACL 2015 Workshop on Vector Space Modeling, as well as an ACL 2014 best paper honorable mention, a best student paper award nomination at ESWC 2015, and a thesis award for his work on graph algorithms for knowledge modeling. He holds a PhD in computer science from the Max Planck Institute for Informatics.
Leave a Comment or Question
Help us make this conference the best it can be for you. Have questions you'd like this speaker to address? Suggestions for issues that deserve extra attention? Feedback that you'd like to share with the speaker and other attendees?
Join the conversation here (requires login)
For conference registration information and customer service
For more information on community discounts and trade opportunities with O’Reilly conferences
For information on exhibiting or sponsoring a conference
View a complete list of Strata Data Conference contacts