POC CHATBOT - February 2019 to October 2019
Assignment: Implementation of a neural network type conversational agent
Use case: Real estate
Tools used:
Keras
-Tensorflow
- Flask
- Spacy(NER, Part of Speech tagger)
- NLTK(Lemmatization, sentence tokenizer...)

Classification of some incidents (heating malfunction...)

NLP engineer · This is for a real estate company. Here, the aim is to redirect every single client complaint(via phone, via e-mail) to the right category(no electricity, heating malfunction, smoke detector breakdown...)

- A DataPreprocessing class has been written to clean up the data, implement topic modeling with tf-idf and only retain, for a given category, words that are similar to one another(word2Vec, cosine similarity matrix)

- A Model class has been written where a neural network has been built to implement the classification tasks. At testing, only relevant sentences(those that have context words in it) will be taken into account for further processing.

- A ClassificationCategory class has been written to instantiate the above classes, to save relevant variables(tf-idf dictionary, test data..) and to implement testing