Humor prediction based on the statistical architecture aims to automatically judge whether a sentence/dialogue is laughing-provoking or not. Due to the complexity of the humor prediction, it is necessary to consider the contextual relevance on the text data when conducting semantic analysis, and the accuracy of Humor detection or comparison need to be further improved. This paper presents to the usage of Bi-directional Long-Short Term Memory (BiLSTM) to predict humor on textual data, and the predictions are reflected in the form of the comparison of humor between the two texts. In order to compare the two sentences in a mathematic way, the sentences are projected into vector space and the sentences are represented at the word-level granularity by using Global Vectors for Word Representation. To strengthen the influential factors of specific words in humor prediction, the similarity of humor vector (SoHV) is introduced as an additional feature, which is calculated from the Cosine similarity between the edit words and the original words in sentences. BiLSTM is compared with Convolutional Neural Network (CNN) on Humicroedit to illustrate its effectiveness in dealing with textual data. The experimental results illustrate that the proposed method significantly outperforms CNN by obtaining the accuracy of 51% and 34.2%, respectively. The effectiveness of BiLSTM is further verified on SemEval-2020 Task 7, which assesses Humor in Edited News Headlines. We are ranked 46th and 64th as to subtask-1 (classification task) and subtask-2 (regression task). Our analyses show that some particular words do have huge impact on the degree of humor in whole sentences and the BiLSTM model can have better cognition capacity by utilizing SoHV.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.