Improved Emoji Identification for Sentence using LSTM in RNN

Bulla Premamayudu


We proposed an improved sentence classification using emoji. Word embedding are used to assign emoji for sentences. The benchmark existing models speak to the word as a position vector in word embedding space. These researchers convert each word in the sentence into their word vectors and afterward take a mean of word vectors. They were used pre-trained word embeddings for representing words in a sentence. Thus, existing proposals cannot catch the multiple aspects and context of the sentence. Furthermore, due to taking the normal of word vectors of given sentence, existing strategies perform less precision at sentence order for accurate emoji. In this paper, we proposed an improved sentence classification which is used LSTM units in RNN model and train the existing embeddings during model creation based on training set. This proposal improved word embedding vector when the model can be updated with new training examples. It likewise has generally high segregating power, on account of sentence is represented in various feature space by representing to each feature dimension with potential highlights. Experimental results on various data sets show that an accuracy of about100% on training set and about 98% on test set. Overall, the proposed model can make moderately incredible various interpretable highlights and word vector embeddings and gives the extension to actualize new strategies to improve the sentence characterization. Further, it very well may be utilized to improve the exhibition of sentence representation, document classification and representation and text characterization.


Text mining; Sentiment Analysis; Word Embeddings; Recurrent Neutral Networks



  • There are currently no refbacks.

Creative Commons License
This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

View IJICT Stats