Natural Language Processing | Build LLM Web App | RNN & LSTM

Learn sentence segmentation, the first step in the NLP pipeline, by splitting text into sentences using punctuation and regular expressions, via the R module in Python.

Tokenize a document by splitting on a regex that separates spaces and punctuation into tokens. The pattern matches commas and semicolons, and the method removes empty tokens to return tokens.

Explore regular expressions in Python for NLP automation and process automation, covering search, find all, compile, split, and replace techniques, with practical string indexing.

Apply stemming to normalize words to their base or root form using the porter stemmer. Note that some stems may lack meaning, as in intelligent, intelligently, and intelligence.

Explore lemmatization as an alternative to stemming, returning root words with meaning, with examples like intelligent and intelligence, and learn to use spaCy to lemmatize sentences.

Discover how stopwords, the most frequent but low-value words, are filtered from text using NLTK stopwords and the tokenize function, with dataset downloads and Punkt tokenizer prerequisites.

Explore parts of speech and POS tagging with spacy to analyze word roles in sentences, supporting language understanding and generation.

Explore dependency parsing, a syntax method that maps sentence structure as a directed graph of words with head and dependents, revealing subject, object, and modifier relations using spaCy.

Encode text into numerical representations using basic tokenization and subword methods like BPE and sentencepiece, then map words and subwords to dense embeddings and integer IDs for NLP models.

Explore lexical, syntactic, and referential ambiguities in natural language processing, learn how humans and NLP systems interpret ambiguity, and develop strategies to handle uncertainty for dependable LLM apps.

Explore lexical ambiguity in natural language processing, learn how word sense disambiguation using the Lesk algorithm and WordNet resolves multiple word meanings, with examples like bank.

Explore syntactic ambiguity when a sentence can be parsed in multiple ways, yielding different interpretations. Learn probabilistic context-free grammars, transformers with attention, and spaCy parsing to resolve such ambiguity.

Explore pragmatic ambiguity in dialogue systems, tracking context and intent with speech act recognition, user modeling, and coreference resolution to interpret questions like 'can you pass the salt' as requests.

Explore ellipsis handling in natural language processing by inferring missing verbs from context using Spacy and its pre-trained model, analyzing tokens and dependency relations to reveal implied sentence structure.

Examine structural ambiguity and metaphorical language in NLP, using dependency and semantic parsing to clarify meaning. Demonstrate spaCy and Displacy to visualize grammar and detect metaphor with embeddings like BERT.

Detect sarcasm in natural language processing by combining sentiment analysis with context, using neural networks and the transformers pipeline to analyze punctuation and word usage.

Explore a sentiment analysis case study with a word cloud to visualize customer feedback, showing positive and negative percentages and key terms driving insights.

Explore speech-to-text deployment in call centers, converting customer calls to text and analyzing greetings, sentiment, and insights with NLP libraries to automate quality checks and daily reports.

Learn how to summarize text with NLP by analyzing a predictive maintenance blog, removing stop words, scoring sentences by word frequency, and selecting top sentences before combining them.

Learn to classify messages as spam using machine learning, balance skewed data, and evaluate logistic regression against other models to achieve high accuracy.

Explore why recurrent neural networks matter by showing how loops create memory for sequence tasks like predicting the next word and time series forecasting through a persistent hidden state.

Learn how recurrent neural networks update the hidden state, using inputs, weights, and biases to compute outputs via softmax, trained by gradient descent on ground truth.

LSTM models improve next-word and time-series predictions by preserving long-term context with a memory cell and a four-gate architecture—input, forget, and output gates plus the cell state—addressing vanishing gradients.

Build a spam detection model using RNN and LSTM with TensorFlow and Keras, including tokenization, embedding, and preprocessing, achieving about 98% accuracy on the dataset.

Contrast LSTM and transformers, highlighting attention, multi-head self-attention, and positional encodings, then build a pdf Q&A app using a pre-trained Distilbert model and a QA pipeline.

learn to set up the infrastructure for a streamlined app using Anaconda on Windows, install Notepad for Python development, and run pip install streamlit from the Anaconda prompt.

Create your first web app with streamliner by displaying hello world, customizing font size and font, saving in a streamliner folder, and running locally to view in the browser.

Create a header and sub header for the app, implement success and error messages, and run from the command prompt to view backend results in a web app.

Learn to access a file stored in a folder, read its data from a comma separated values file, display the contents in a web app, and create graphs.

Learn to add a file upload button for processing, browse and read files from a folder, and use radio buttons and checkboxes with size limits and free vs paid options.

Build an NLP wordcloud app that extracts word frequencies from articles, removes stopwords, and visualizes the most frequent terms with larger fonts.

Deploy a locally run Streamlit word cloud app to a cloud platform such as Heroku by following a 12-step process, including git setup and pushing to a live site.

Learn to deploy a Streamlit app from a GitHub repository using the code, Procfile, requirements, and setup files, then create a Streamlit account and deploy to access the live app.

Build a chat bot like ChatGPT with Streamlit and an OpenAI API key, prioritizing data privacy and customizable prompts for topics, sections, and content.

Explore a curated list of related courses with a downloadable resource to access links easily, and view the complete course list in my Udonis profiling.