Merge pull request #97 from graviraja/applications/classification

xlm attn model added

README.md

@@ -305,6 +305,15 @@ To alleviate this problem XLM-RoBERTa model (which has been pre-trained on 100 l
 
 An F1 score of `0.59` was achieved. Methods to improve this will be explored later on.
 
+### Day 94: Sentimix with XLM-RoBERTa-LSTM-Attention
+
+The final output from XLM-RoBERTa model were used as input embeddings to bi-directional LSTM model. An attention layer, which takes the outputs from LSTM layer, produces a weighted representation of the input, which is then passed through a classifier for predicting the sentiment of the sentence.
+
+![codemix](assets/images/applications/sentiment/codemix_xlm_attn.png)
+
+An F1 score of `0.64` was achieved. 
+
+
 ### Day 21: Document Classification with Hierarchical Attention Network
 
 A Hierarchical Attention Network (HAN) considers the hierarchical structure of documents (document - sentences - words) and includes an attention mechanism that is able to find the most important words and sentences in a document while taking the context into consideration.

applications/classification/README.md

@@ -54,6 +54,13 @@ To alleviate this problem XLM-RoBERTa model (which has been pre-trained on 100 l
 
 An F1 score of `0.59` was achieved. Methods to improve this will be explored later on.
 
+
+### Sentimix with XLM-RoBERTa-LSTM-Attention.ipynb
+
+The final output from XLM-RoBERTa model were used as input embeddings to bi-directional LSTM model. An attention layer, which takes the outputs from LSTM layer, produces a weighted representation of the input, which is then passed through a classifier for predicting the sentiment of the sentence.
+
+An F1 score of `0.64` was achieved.
+
 # Document Classification
 
 ## Document Classification with Hierarchical Attention Network

applications/classification/sentiment_classification/README.md

@@ -142,3 +142,23 @@ Following metrics were achieved on test data.
 - `F1-Score`: 0.59
 
 Methods to improve this will be explored later on.
+
+
+### Sentimix with XLM-RoBERTa-LSTM-Attention.ipynb
+
+The final output from XLM-RoBERTa model were used as input embeddings to bi-directional LSTM model. An attention layer, which takes the outputs from LSTM layer, produces a weighted representation of the input, which is then passed through a classifier for predicting the sentiment of the sentence.
+
+![codemix](../../../assets/images/applications/sentiment/codemix_xlm_attn.png)
+
+
+Following metrics were achieved on test data.
+
+- `Precision`: 0.64
+
+- `Recall`: 0.64
+
+- `F1-Score`: 0.64
+
+#### References
+
+- [Google transliterated data](https://github.com/keshav22bansal/BAKSA_IITK)