https://github.com/arshitarora/online-psycologist-sentiment-analysis-

Online Psychologist is an AI sentiment analysis tool that understands your emotion in the form of input sentences and helps you analyze your emotions by categorizing them into anger, fear, joy, sadness or neutral.
https://github.com/arshitarora/online-psycologist-sentiment-analysis-

Last synced: 24 days ago
JSON representation

Online Psychologist is an AI sentiment analysis tool that understands your emotion in the form of input sentences and helps you analyze your emotions by categorizing them into anger, fear, joy, sadness or neutral.

• Host: GitHub
• URL: https://github.com/arshitarora/online-psycologist-sentiment-analysis-
• Owner: ArshitArora
• Created: 2022-12-01T06:08:49.000Z (about 2 years ago)
• Default Branch: main
• Last Pushed: 2022-12-01T06:16:34.000Z (about 2 years ago)
• Last Synced: 2024-09-18T10:49:17.199Z (4 months ago)
• Language: Jupyter Notebook
• Size: 28.9 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Online-Psycologist-Sentiment-Analysis-

Online Psychologist is an AI sentiment analysis tool that understands your emotion in the form of input sentences and helps you analyze your emotions by categorizing them into anger,  fear, joy, sadness or neutral.

# Emotion Classification in Short Messages

Multi-class sentiment analysis problem to classify texts into five emotion categories: joy, sadness, anger, fear, neutral. A fun weekend project to go through different text classification techniques. This includes dataset preparation, traditional machine learning with scikit-learn, LSTM neural networks and transfer learning using BERT (tensorflow keras).

# Datasets

**Summary Table**

|     Dataset    | Year |  Content  |     Size     | Emotion categories | Balanced |

| :--------------: | :--: | :-------: | ------------ | ------------------ | :-------: |

|dailydialog| 2017 | dialogues |102k sentences|neutral, joy, surprise, sadness, anger, disgust, fear| No |

|emotion-stimulus|2015|dialogues|2.5k sentences|sadness, joy, anger, fear, surprise, disgust| No |

|isear|1990|emotional situations|7.5k sentences|joy, fear, anger, sadness, disgust, shame, guilt| Yes |

links: [dailydialog](http://yanran.li/dailydialog.html), [emotion-stimulus](http://www.site.uottawa.ca/~diana/resources/emotion_stimulus_data), [isear](http://www.affective-sciences.org/index.php/download_file/view/395/296/)

## Combined dataset

Dataset was combined from dailydialog, isear, and emotion-stimulus to create a balanced dataset with 5 labels: joy, sad, anger, fear, and neutral. The texts mainly consist of short messages and dialog utterances.

# Experiments

### Traditional Machine Learning:

* Data preprocessing: noise and punctuation removal, tokenization, stemming

* Text Representation: TF-IDF

* Classifiers: Naive Bayes, Random Forrest, Logistic Regrassion, SVM

| Approach            | F1-Score |

| :------------------ | :------: |

| Naive Bayes         | 0.6702   |

| Random Forrest      | 0.6372   |

| Logistic Regression | 0.6935   | 

| SVM                 | 0.7271   | 

### Neural Networks

* Data preprocessing: noise and punctuation removal, tokenization

* Word Embeddings: pretrained 300 dimensional word2vec ([link](https://fasttext.cc/docs/en/english-vectors.html))

* Deep Network: LSTM, biLSTM, CNN 

| Approach            | F1-Score |

| :------------------ | :------: |

| LSTM + w2v_wiki     | 0.7395   |

| biLSTM + w2v_wiki   | 0.7414   |

| CNN + w2v_wiki      | 0.7580   |

### Transfer learning with BERT

Finetuning BERT for text classification

| Approach            | F1-Score |

| :------------------ | :------: |

| finetuned BERT      | 0.8320   |