Recognizing and Handling Negations in Machine Learning

Recognizing and Handling Negations in Machine Learning

Kolthof, Daan (TU Delft Electrical Engineering, Mathematics and Computer Science; TU Delft Pattern Recognition and Bioinformatics)

Tax, David (mentor)
Reinders, Marcel (graduation committee)
Liem, Cynthia (graduation committee)

Delft University of Technology

2018-08-30

In several machine learning problems, a relatively small subproblem is present in which combinations of (negating) objects or structures result in a negation or otherwise other classification compared to when these (negating) objects are not present. To be more specific, a variant of the XOR problem is present in a small amount of objects in these classification problems. Examples of this could be negating words in textual sentiment classification or the presence of sarcasm when one wants to determine seriousness in speech. As negations are usually present in a small part of much larger datasets, it is important to recognize these relatively rare negation structures within objects' data. Correctly recognizion and handling negations could improve overall classification performance in machine learning problems that inhibit negations in some of their dataset objects. To lay the groundwork for solving these problems, the subproblem of recognizing negation words in sentiment classification is solved by employing a word embedding neural network to recognize text structure and to correctly classify these negations while at the same time this neural network is used to classify complete sentences in the problem of sentiment classification.

Machine Learning
Neural Networks
Natural Language Processing
sentiment analysis
sentiment classification
Word embedding
Negation Handling
Negation Recognition

http://resolver.tudelft.nl/uuid:caa1b4b3-59ca-4290-b95e-84190c54b787

Student theses

master thesis

© 2018 Daan Kolthof