.png){kind=icon}
Return
Text-based emotion recognition using contextual phrase embedding model
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-023-14524-9.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
In this paper, the proposed approach categories the sentences in the dataset into the various topical documents using the TE-LSTM+SC model. As well as, the model generates semantic words related to topics that are fed into the word embedding like Skip-Gram and FrameNet to build the domain-specific lexicon. The topically related sentences in each document are contextually grouped using Skip-Phrase. Each sentence in contextual group is given to Semantic Role Labelling (SRL). SRL indentify the essential predicate-argument structures with the semantic labels like verb (V) tag or ARGM-NEG or ARGM-PRP or ARGM-CAU or structures with the semantic labels like verb (V) tag or ARGM-NEG or ARGM-PRP or ARGM-CAU or ARGM-MNR or ARGM-MOD. The selected predicate-argument structures are aggregated into a linear layer to form a semantic embedding. Simultaneously, the predicate-argument embedding is segmented to sub words by BERT. The sub-words are transformed to word level through a convolutional layer to acquire the contextual word representation. Finally, semantic embedding and word representation are integrated to efficiently find the emotion of the given sentence. The experimental result proved that the proposed approach outperforms all the state-of-art approaches.
Keywords:
Emotion recognition
Emotional cause
Skip phrase
FrameNet
Semantic role labelling
BERT
Journal
IF:
3
Papers:
2.0W
Citations:
3.2W
Organization
Cited Papers
.png){kind=icon}
ShareSave
ShareSaveShareSaveDepecheMood plus plus : A Bilingual Emotion Lexicon Built Through Simple Yet Powerful Techniques
OAAI
OAAIShareSaveShareSave
View more