"No one is harder on a talented person than the person themselves" - Linda Wilkinson ; "Trust your guts and don't follow the herd" ; "Validate direction not destination" ;
Showing posts with label NLP. Show all posts
Showing posts with label NLP. Show all posts

July 22, 2024

Experimentation, Always something to find a solution :)

Some questions and answers take days or weeks, and sometimes the approach moves from LLM to NLP, It's a blend of techniques to make things work.

  • How do we optimize RAG with internal documents, Original vs Summary vs Intents, What works best? 
  • How do we merge external data? How can we keep versions and relevance?
  • More than LLM work, The heavy lifting is for Data preprocessing/cleaning / Embedding on summary 
  • When to use LLM vs Multimodals?
  • What is the benchmark for our domain and how much do we meet it consistently?
  • The transition for LLM, LLM+KG, Creating the data mapping..

A lot of challenges but one at a time, Balancing Consistency, Accuracy, and Latency. If you want to solve real problems you can connect/explore potential learning experimentation opportunities / dedicate some learning hours. Please drop a note to career@proplens.ai

#learnings #NLP #Datascience #RAG #LLMs #perspectives #Datascience 

Keep Exploring!!!

March 04, 2024

Klara Chatbot - Devil in Details

  • It recites exact docs and passes me on to human support fast.
  • Good job on the team for making hallucination not possible - because it seems to spit out the same responses however I ask it, and refuses to go “out of bounds.”
  • As soon as I ask or instruct anything that is not a doc, I’m *boom* talking with a human agent.
  • Also, almost all questions I ask about payment terms or problems the chatbot tells me - in various ways - to talk to the merchant if I have a problem, not to Klarna
  • These assistants turn docs into chat text, that people read!
  • Klarna is a middleman. The customer buys from the merchant and Klarna sells defaulted payments to collections agencies!!
  • Klarna wants potential investors to believe they are buying into an “AI edge” company

Ref - Link

Keep Exploring!!!

March 03, 2024

Dense to Sparse - AI World

  • What we do in CNN - Convert Dense to Sparse with convolution and activations 
  • What we do in NLP - Text Preprocessing: Stemming / Lemmatization / Stop-word removal - Vectorization 
  • Topic Modelling - Words - Documents - Non-Negative Matrix Factorization 
  • ML Feature Engineering / Recommendations - PCA / SVD 

Everywhere we attempt to retain key features/vectors aligned to vision/text/features/topic modeling tasks. Converting Dense to Sparse is the way to get the signal from the noise :)

Keep Exploring!!!

Custom Chatbot vs OpenAI Chatbot

Build vs Develop on LLM

Custom Chatbot

Time- Data collection, labeling, classification, NER models

Build

  • Preprocessing
  • Lower case
  • Stemming, Lemmatization with POS Tags
  • Entities, NER

Inference

  • Intent
  • Topic Classification
  • Frame a response

Context - Limited to corpus

LLM Chatbot

Time - Prompts / Responses / Store / Retrieve

OpenAI

  • Prompt
  • Inference answers

Challenge

  • Air Canada hallucinations use cases
  • Context vs Hallucination


Keep Exploring!!!

October 11, 2023

Chatbot Lessons

Interesting Read

GenAI Customer Care Project Now Saving $4M/yr for Car Insurance Firm, Jerry

Key Lessons

  • Leverage Large Language Models (LLMs) to create a chatbot 
  • Open AI GPT-4 for complex queries
  • Open AI GPT-3.5 for initial sorting
  • Dataset - Messages are captured from chat and SMS through Twilio and stored on Jerry's servers
  • Routing Agent - Route based on requests - "Payments," "Policy," and "Opt-out" agents
  • Webhook / API - Handlebars to insert data from their database into the chatbot's responses

Key Learning's

  • Prompt engineering
  • Rapid iteration
  • Investment in testing
  • Version control

Drawback of LLM
  • Exposes your corporate data to the provider of your LLM 
  • LLMs have been shown to suffer from “hallucinations
Quick Lessons
  • FAQ from Cache
  • VectorDB for queries where answers can be located in docs
  • Intent recognition and call APIs based on OrderNumber#
  • Knowledge Graph if we have some meta data loaded
  • If all attempt fails LLM answer
  • Add guard rails wherever possible

Ref - Link

Dialogflow Notes Link



Keep Exploring!!!!

October 09, 2023

NLP Summit Key Talks Slides

NLP Summit Key Talks Slides

Patient Similarity through Representation Learning from Medical Records

1 - Problem Statement

2 - Data Sources and Challenges


3 - Multiple levels of Data Collection over extended periods

4 - Data Pipeline


5 - Representation

6 - Patient Similarity

7 - Classification

Operationalizing NLP Models

Model Development

Model Dev Cycle

Implementation Monitoring

MLOps Cycle

Suicide Classification for News Media using NLP

Suicide Stats

Topic Classification

Suicide Classification

Automated Extraction of Medical Risk Factors for Life Insurance Underwriting

Health Trends

Pipeline

Models

Functional Flow

How to Build a Question Answering Application with Haystack in 30 Minutes

5 - 1 Flow

5 -2 Pipelines

Keep Exploring!!!

August 23, 2023

Extract Reviews with Curl Command

 Google Chrome + Inspect Option 


Keep Exploring!!!

August 19, 2023

Aspect-based sentiment analysis - NLP

Aspect-based sentiment analysis is about identifying different aspects in a given topic. TextBlob will help us to calculate the polarity of the sentiments






  • context exclusion
  • intent extraction
  • co-reference resolution
  • sentiment analysis

Keep Exploring!!!

July 26, 2023

coreferences with GPT4

#Coreference resolution (CR) is a task in natural language processing (NLP) that identifies when two or more expressions in a text refer to the same entity. This can include pronouns like "he" or "she," definite nouns like "the car," and other expressions. 

#Coreference resolution helps in other NLP tasks such as machine translation, text summarization, and information extraction. It is required for understanding contextual information and improving comprehension of the text for machines.

Prompt Example

Extract coreferences from two sentences, and prepare the second sentence with coreference substituted. Here are two sentences. Return only second sentence with coreference

1. <First sentence - Steve jobs is great innovator. He founded apple>

2. <Second sentence - What are his key achievements>

Result

<Second sentence - What are Steve Jobs' key achievements>

Keep Exploring!!!


July 22, 2023

NLP use cases before and after ChatGPT

 


Many use cases will converge
Many startup ideas will converge
Who has the data is the King!!!



Keep Exploring!!!

February 26, 2023

NLP - NER - Entity Recognition

I have worked on NLP, and custom NER examples. This streamlit demo covers NER in domain context and multiple entities link






  • Name Entity Recognition - Extract Organizations, People, Locations, and many other entities from long, free-text financial documents.
  • Extract Financial Relationships - Automatically identify relationships between companies, products, and people – even when they are mentioned using aliases.
  • Classify Financial Text - Classify texts into 77 banking-related categories like credit reports, mortgages, money transfers and more.
  • Financial Sentiment Analysis - Identify positive, negative or neutral sentiments in financial news.
  • Financial De-identification - De-identify and mask sensitive personal information in documents and images.

Ref - Link

Keep Exploring!!!

February 23, 2023

Startup Analysis - hyperverge - KYV - Vision + OCR + NLP

Many times taking an idea, ideating it, and solving it end to end is key. KYC with Vision / Image / Data and NLP are very impressive.

Product Features

  • Real-time analysis of images and videos obtained from sources such as consumer photos, satellite images, surveillance cameras, industrial images, and documents. 
  • NLP solutions for automating and disrupting the Legal Document Analysis industry

Deep Learning Skills (Vision / NLP) - Our Perspectives

  • DL - Models built for Face detection, OCR, and Buildings / Signs Detection. Face, Object, Text, and Activity recognition.
  • NLP - One shot, Few shot, Self-Supervised approach, multi-task learning, and contrastive learning strategies
  • Plus a lot of custom embeddings/graph databases/custom models

Video KYC

  • Liveliness Check
  • Background Detection
  • Landmark validations
  • Key facial landmarks based on submitted docs
  • Similarity scores
  • Social media similar image scores 

Signature Check

  • Signature Font Size, Length, Height, features of it
  • Signature Font Style
  • Keypoints match, landmarks match, shape, texture

Face Verification

  • Landmarks
  • Landmark distances for iris, nose, cheek, chin
  • Mediapipe
  • Custom Segment and measure similarity
  • Classify face shapes/hairstyles

Keep Exploring!!!

February 22, 2023

Law and Order - Applied Vision, NLP, ML Use cases

Vision Use cases

  • Person Detection, Attribute Extraction
  • Real-time vehicle number plate validation
  • Attribute Extraction - Shirt, Dress Type
  • Age Estimation
  • Personal Re-identification
  • Vehicle Details Extraction
  • Vehicle color Extraction
  • Action Recognition - Detect Crowd, Fight
  • Anomaly Detection, Loitering - Notify suspicious movements

NLP Use cases

  • Using NLP to find similar cases from Digitized documents
  • Using OCR to digitize documents

ML Use cases

  • Cluster patterns of crimes
  • Cluster types of crime
  • Cluster patterns of offenders
  • Churn time for offenders
  • Average time after a conviction for repeat crimes

CCTV analysis is a reactive approach. Edge + AI + Data + NLP is the way forward. Be predictive, proactive, and prepared.

Analyze the below companies and build another product :)




Ref - Link, Link1

Mark43 Interesting Demo and Screenshots - Link




Crime Prediction using Machine Learning with a Novel Crime Dataset

  • News Link Collection using Manual Process
  • News Link Collection using Web Crawler
  • Filtering Crawled News Collection to Identify Crime News






Keep Exploring!!!

NLP, Recruitment, AI - Automated Matching :) Eightfold - Analysis

A ton of zero-shot / few shot / data labeling to customize and build models for

Some of the Key features / NLP overlaps

Custom Entity Recognition

  • Education
  • Place
  • Company 
  • Domain
  • Skills Extraction

Custom Embedding for

  • Similar Projects Search
  • Project Summarization
  • Creating custom embedding for each domain

NLP + Ranking

  • Ranking and retrieving based on location/salary/education
  • Skill Distribution / Contribution
  • Domain Extraction from Company names
Vision
  • OCR + Vision for Content Retrieval

Search / Retrieval

  • Vector database for Finetuning/indexing vector databases to search and retrieve the closest matches

From JD

  • Familiar with Language models, and transformers like BERT, GPT-3, T-5 etc.
  • Prior experience building and deploying machine learning models in production at scale
  • Familiarity with MLOps tools and pipelines (MLflow, Metaflow).
  • Integration with Workday, SuccessFactors, Taleo, PeopleSoft, iCIMS, SmartRecruiters,
  • REST APIs, microservices, data ingestion and processing systems, and distributed systems.