Week 3 - Taking a step back in the basics of NLP

1. What is Natural Language?

   • A set of agreed upon words

   • NLP is the merge with Ai and Linguistics

   • Natural Language Understanding (NLU) - understanding

   • Natural Language Generation (NLG) - generate a response

2. Tokenization 

   • word-tokenizer

   • Use import nltk to tokenize words

3. Stemming

4. Lemmatization

5. Named Entity Recognition (NER)

   • Extracts important entities from text

   • Use NER to extract top entities and query only the entities and not the whole dataset.

Week 4 - Text Classification with Machine Learning

1. Word Vectorization

• Turn the words to numbers and clean it

• Use Octoparse web scraper to gather Amazon data.

2. Recommender systems

• Use unstructured data

• Create actionable systems can be used to create better products

• What problem can we solve?

  • Songs for Spotify, movies on Netflix, videos on YT, related post on Twitter, similar dishes on UberEATS, and ads on Facebook.

3. Where do they get the Data?

• Explicit data

• Implicit data

• Google Tag Manager - tracks every click & contact you make with the website. 

• https://www.analyticsmania.com/post/google-tag-manager-use-cases/

4. Kinds of Recommenders

• Collaborative Filtering - Find similar users

• Content-Based Filtering

5. Word2Vec

• Bag of Words

• TF-IDF Vectorizer- NLP algorithm used to determine the term frequency of key words based on the number of times it appears in the corpus (group) of documents. 

• Every word will have its own TF & IDF scores (weight)

Notes:  

Look at LakeraAi to get information on hacking Ai to have it answer things it's not supposed to. 

Check out Dataiku and OpenAi Whisper.

Read chapter 3

Week 5 - Preprocess Text Data

1. Set up environment

2. Preprocess text

3. Get it in the format 'text'/'label'

4. Apply cleaning function

5. Stem and Lemma

6. TF/IDF, BoW and the Model

7. Create model with Vectorized data 

8. Evaluate NLP Model

9. Test model with new data 

10. Make predictions

*work on Modules 3,4 & 5

Week 6 - Data Processing

1. Basic cleaning

2. Lemma

3. Tokens (this is not Word2Vec)

4. Vectorize - turn words into numbers

   • Bag of Words

   • TF-IDF

5. ML Model

   • Binary classification

6. Create recommendation or Sentiment analysis with model

7. Word2Vec

   • Cluster in 2 dimensional space

   • Projector.tensorflow.org

   • Continuous Bag of Word (CBOW)

   • SkipGram

Additional Information:

1. Sequence to sequence learning

2. Transformer

3. Review NLTK Vader

4. Read Chapter 5 - Word Embedding

5. Use nlp-fundamentals to create Custom GPT web app on Streamlit

Week 7 - Review of YTD

1. Look for Text Classification datasets on Kaggle

   • Look for datasets that have columns with large amounts of text. 

   • Word2Vec

2. Word Embedding

   • ChatGPT only works with 'word embedding'.

   • W.E. is how you create a custom GPT

   • 
     

3. What are embeddings?

4. OpenAI’s text embeddings measure the relatedness of text strings. Embeddings are commonly used for:

   • Search (where results are ranked by relevance to a query string)

   • Clustering (where text strings are grouped by similarity)

   • Recommendations (where items with related text strings are recommended)

   • Anomaly detection (where outliers with little relatedness are identified)

   • Diversity measurement (where similarity distributions are analyzed)

   • Classification (where text strings are classified by their most similar label)

   • An embedding is a vector (list) of floating point numbers. The distance between two vectors measures their relatedness. Small distances suggest high relatedness and large distances suggest low relatedness.

5. Review Lamaindex

6. Review Langchain

7. AR-code.com

8. Examstopics.com

9. Review Mistrel Ai

10. Review Groc.com

    • uses Llama and Mixtral

11. AWS Access:

    • Sagemaker

    • Ask Dr. about the certification exams

12. Resume companies use CBOW to scan the resumes

    • Extract the skills from the job description 

    • Ask for the word density for each skill

    • I need my resume to reflect this key word density

8. Review Pinecone

HW:  Complete 'Word Embedding' in website

Week 8 - Azure for Personal ChatGPT

1. Azure

   • Figure out use cases

     1. Search internal documents with GPT

     2. Liquid resume

     3. Search real estate data

2. Semantic Search

3. Review Enterprise or Solutions Architecture 

Week 7 - Deep Learning

1. Ai

2. NLP

3. Colab - Learn to code a neural network

   • Can't have empty values or nulls 

   • Change everything to 1

4. Regression Model 

5. Classification Model

6. Make all data Numeric 

7. Scale the data

8. Turn non-numeric values into 0 or 1.  

   • 1-10 can be 1-5 = Bad, 6-10 = Good

   • Then Bad = 0 and Good = 1

9. You can over train the models

   • Overfitting

   • 
     

Week 8 - TensorFlow - Train Neural Networks 

1. Review the code 

2. complete 3 different types of models