Structurally informed methods for improved sentiment analysis
Universität Stuttgart / Institute for Natural Language Processing:
The University of Stuttgart is divided up into ten faculties and is an educational institution in high demand worldwide with 56 undergraduate programs and 20 postgraduate programs. The University meets the requirements of the working world with the internationalization of its broad range of courses on offer, numerous on-line offers in teaching and further education and its intensive support for start-up companies.
In a world of growing floods of information, humans need the assistance of mechanized information processing tools. One important source of information consists of written or spoken text. The Institute for Natural Language Processing (IMS) carries out basic and applied research and trains students to create tools for automated processing of spoken and written language.
Project: Structurally informed methods for improved sentiment analysis
Jonas Kuhn, Wiltrud Kessler
The task of sentiment analysis is to automatically identify the opinions people express in natural language about some topic (such as a product or a work of art). For instance, the sentence
"auto white balance is better than of the other canon cameras I have and they already do very-well"
from a product review expresses a positive opinion about a specific camera.
High-quality sentiment analysis of user generated content in social media is of great interest, but at the same time poses various challenges. As the example illustrates, nonstandard grammatical constructions and spellings preclude the direct use of standard tools for analyzing the sentence structure. Structural analysis can however be key to successful sentiment analysis, especially when more advanced syntactic and semantic phenomena are considered, which are beyond the scope of most current approaches. An example are comparative sentences in which the item under discussion is compared to a different item, a group of items ("the other canon cameras I have" in the example), or some expectation.
The goal of this project is to develop robust structurally informed models for sentiment analysis, addressing both the challenges from noisy, nonstandard language data and from structurally complex verbalization of opinions.
Wiltrud Kessler and Jonas Kuhn, 2013. Detection of Product Comparisons -- How Far Does an Out-of-the-box Semantic Role Labeling System Take You? In: Proceedings of EMNLP 2013: Conference on Empirical Methods in Natural Language Processing — October 18–21, 2013 — Seattle, USA
Kessler, W. and Kuhn, J. (2014a). A corpus of comparisons in product reviews. In Proceedings of the Ninth Language Resources and Evaluation Conference (LREC 2014), Reykjavik, Iceland.
Kessler, W. (2014). Improving the detection of comparison arguments in product reviews. In Proceedings of 44. Jahrestagung der Gesellschaft für Informatik e.V. (INFORMATIK 2014), Stuttgart, Germany.
Kessler, W. and Kuhn, J. (2014b). Detecting comparative sentiment expressions – a case study in annotation design decisions. In Proceedings of 12. Konferenz zur Verarbeitung Natürlicher Sprache (KONVENS 2014), Hildesheim, Germany.
Programs & Grants
100 years ago nobody would have imagined that it may make sense to talk to machines. Today, in the days of speech recognition and speech synthesis to be found in cars, computers, phones and many other devices this is already normal. But it doesn't stop there.