Web and Text Analytics Quiz

Web and Text Analytics 2024-25, Week 1

• The course covers web analytics, text analytics, and the semantic web.
• Web analytics involves measuring, collecting, analyzing, and reporting web data to understand and optimize web usage.
• Text analytics, also known as text mining, is the process of drawing meaning from written communication. A key component is natural language processing (NLP).
• Semantic web is a concept coined by Tim Berners-Lee for a web of data that machines can process, meaning much of the data is machine-readable.

Web Analytics

• Analytics platforms track website activity, including the number of visitors, time spent on the site, pages visited, and how users arrived at the site.

Digital Analytics

• In 2012, the Web Analytics Association changed its name to the Digital Analytics Association.
• Companies that previously provided web analytics tools now provide digital analytics tools.

Marketing Analytics

• Google Analytics is a central service within the Google Marketing Platform.
• It reports website traffic.

Google Analytics

• Google Analytics uses cookies to track website visitors.
• It provides insights into user behavior, including page views, time on site, and other metrics.

Text Analytics

• Text mining (or text analysis) is a process for converting unstructured text into meaningful information.
• It uses AI and NLP.

Natural Language Processing (NLP)

• NLP allows computers to "read" and understand text, mimicking human comprehension of language.

• Useful in both understanding existing text and generating novel text.

The World Wide Web

• Tim Berners-Lee proposed the concept of the World Wide Web in 1989.

Linked Data Web

• The linked data web, also known as the semantic web, enables computers to better understand data.

Large Language Models (LLMs) development

• Large Language Models (LLMs) are notable for their general-purpose language understanding and generation capabilities.
• They achieve these abilities through large datasets.
• LLMs comprise artificial neural networks (primarily transformers).

Google's BERT

• Google's open-source BERT framework for natural language processing.
• BERT's training used a massive dataset of 3.3 billion words.
• Training accelerated using 64 TPU processors.

LLM Training Costs

• Training LLMs requires significant computational resources and costly hardware.

Fine-tuning LLMs

• Fine-tuning can retrain a foundation model on new data.
• This is useful for adapting models to specific tasks, like medical applications.

Prompt Engineering

• Prompt engineering modifies inputs to improve a model's outputs.
• It requires less computing power than the data used in fine-tuning.

Retrieval Augmentation Generation (RAG)

• Fine-tuning LLMs might be unnecessary or impossible in applications with frequently changing data.
• In these cases, in-context learning or retrieval augmentation are viable alternatives.

Applications of LLMs

• There is a need for applications that utilize LLM models.

The Evolution of NLP

• NLP has evolved from rules-based methodologies in the 1950s to deep learning approaches in the 2020s.

Common NLP Tasks

• Text/document classification, sentiment analysis, information retrieval, parts-of-speech tagging, machine translation, conversational agents, knowledge graphs, text summarization, topic modelling, text generation, spell checking, grammar correction, and speech-to-text are common NLP tasks.

Text Classification

• Text classification involves assigning categories to unstructured text data.

• Use cases, for example include predicting disease outcomes from clinical notes.

Sentiment Analysis

• Sentiment analysis involves the analysis of emotions and opinions within text, commonly applied to reviews, posts, and support tickets.

Sentiment Analysis in Tweets

• Examples of sentiment analysis within Twitter posts.

Brand Reputation Management

• Monitoring public sentiment about a brand enables businesses to manage their reputation effectively.

Text Extraction

• Text extraction is used to pull out critical information from documents, including keywords, entity names, and more details.

Named Entity Recognition

• An NLP tool that allows identification and extraction of entities, such as companies, persons, and more.

Text Summarization

• Summarizing large texts using NLP.

Machine Translation

• Machine translation, like Google Translate, translates text to different languages. The results are more sophisticated than simple word replacements.

Rephrasing in NLP

• A study investigated how well chatbots can rephrase physician questions from public forums.

Application of NLP in Practical Arenas

• Discusses practical applications of NLP to text, including steps like noise removal, normalization, and vectorization.