Hello, everyone. Thank you for joining me today. My name is Mohit Mittal. And today I'm going to talk about NLP. NLP, Natural Language Processing, is a subfield of artificial intelligence that enables machines to understand, interpret, and generate human language. The need of NLP emerged due to the complexity of human language filled with ambiguity. context, idioms, and variations. Historically, NLP research dates back to 1950 when Alan Turning proposed the Turning Test to evaluate machine intelligence. Early NLP efforts involved rule based systems, but these had limitations due to the complexity of human language. Over the past two decades, the rapid growth in data, computing power, And machine learning algorithms have propelled NLP into real world applications. Today NLP is used to chat, is used in chatbots, voice assistants, search engines, translation services, sentiment analysis, shaping how humans interact with technology. This session will deep dive into the key techniques and advancements that have transformed NLP into what it is today. Growing impact of NLP. The NLP market has experienced remarkable growth. In 2018, the market was valued at around eight billions, and by 2022 it has surge to 26 billion. The projected CHER of 21.4% from 2023 to 20. 30 signifies that NLP is one of the fastest growing AI segments. To put this into perspective, the market size in 2018 was 8 billion, journal 8 billion, in 19 it grew up to 11. 6 billion. 2020. It was 16.2 billion, then 2021, it grew up to 20 point 21.3 billion and 2022, it's 26.4 billion. Healthcare has been a key driver in NLP adoption, contributing 15.2 of the 2% of the total market leveraging NLP for. medical records, analysis, predictive analysis, and clinical decision support. Cloud based NLP deployments now account for 60 percent of the industry, highlighting a shift towards more scalable and cost effective solutions. The increasing integration of NLP across finance, retail, and legal sectors further drives demand, making it a critical AI technology. Moving on to next slide where we are going to talk about advancements in large language models. What is LLM? Large language models or LLM are a subset of NLP models trained on massive data sets to understand and generate human like text. The evolution of LLMs can be tracked back to the earlier models where Word2VC 2013 which introduced vector based word representations. Later, transformer based architectures like BERT and GPT 3 revolutionized NLP by enabling deep conceptual understanding. GPT-4 launched in 2023. Further improved contextual reasoning, creativity, and logical consistency in generated text. LMS have been tested across various domains. If you talk about medical diagnosis, researching human level accuracy in medical text analysis is around 90 to 95%. Legal document review. where we are automating contract analysis with up to 92. percent of accuracy, conversational AI, enabling chatbots, virtual assistants to provide human like responses. With the rise of zero shot and few shot learning, LLMs can now perform new tasks without extensive retraining, making them highly adaptable. Moving on to next slide where we are going to talk about core NLP components is a it's a four four step ladder you can say where the top one is text tokenization. Tokenization is a process of breaking text into smaller units. which can be words, sub words, or characters. This step is essential for language models to understand and process it. There are different types of tokenization. Word based tokenization which splits text at word boundaries. Subword tokenizations like, BPE. Wordpiece breaks words into smaller meaningful units to handle unknown words. The another one is character based tokenization, which splits text into individual characters useful for language like Chinese. The next letter is part of speech tagging, POS tagging. POS tagging assigns grammatical label to words such as noun, verbs, adjectives. This helps NLP models understand sentence structure and meaning. For example, run can be a noun, a morning run. or a verb I run every day and POS tagging helps in differentiating them. The next is named entity recognition. We also call it NER. NER identifies real world entities such as names, locations, organization, and dates. Example, in Apple, INC is headquartered in California. NER identifies Apple Inc as a company and California as a location. That's how NER helps. The last one is pre processing. Pre processing involves cleaning and normalizing text to improve model accuracy. The common steps include removing stop words, lower casing text, and handling misspellings. Proper pre processing can enhance model performance by up to 30%. Now, what is tokenization in NLP? Tokenization is the fundamental step in NLP that converts unstructured text into structured data for processing. The choice of tokenization technique impacts model accuracy and efficiency. Traditional word based tokenization struggles with compound words and different language structures. Subword tokenization methods like Byte Pair Encoding, or BPE, have reduced vocabulary size and improved pre verb recognition. Effective tokenization enables models to handle languages with complex morphology and new words efficient. Moving on to next slide where we are going to talk about what is sentiment in NLP. Sentiment in NLP refers to the emotional tone expressed in a text. It can be classified into categories like positive, negative, neutral. Early sentiment analysis use predefined word. List that struggled with a sarcasm on Context. Machine learning approaches improved sentiment classification by learning and label data sets. Advanced model like BERT now understand contextual nuances and perform sentiment analysis by over 90% Advanced Sentiment Analysis. Multitask learning enhances accuracy by understanding overlapping emotions. Multilingual sentiment analysis ensures consistent sentiment detection across languages. Hybrid approaches combine rule based and deep learning models for better sarcastic Moving on to next slide, language generation breakthroughs. The introduction of transformer based models has revolutionized language generation. Attention mechanisms help models maintain context over long passages, improving coherence in generated text. Training techniques like teacher forcing have accelerated model learning by 40%, which nucleus sampling has reduced repetitive. Text generation by 60 percent real world application of NLP powered language generation include automatic text summarization, where new models now achieve near human performance. These advancements are paving the way of AI generated context in news, education, conversational AI applications. Now talking about implementation considerations. Deploying NLP at scale requires addressing several critical Considerations Computational requirements Optimized decoding algorithms can reduce inference latency by three times, making real time processing more efficient. Data quality High quality preprocessing can improve entity recognition accuracy by up to 12 to 15 percent. MLOps practices Automated model monitoring ensures model perform optimally and detect degration up to 92 percent accuracy. The ethical considerations, biased detection frameworks help ensure NLP systems are fair, unbiased. These factors are essential for organization looking to integrate NLP solutions effectively. Challenges and future directions for NLP. Despite advancements, NLP faces several key challenges. Factual accuracy. Like large language models sometimes generate incorrect information, but new fact checking mechanism have reduced errors by 45%. Domain specialization, custom NLP models for specific industries like healthcare, finance are achieving 85 percent accuracy in maintaining. Precise terminology, ethical AI development, privacy enhancing techniques such as federated learning are reducing data security risks by 60%. Addressing these challenges will shape the next NLP of NLP models. What's the future of NLP? NLP is rapidly evolving and will continue to transform industry in coming decade. The next phase will see improvements in multilingual capabilities, better context understanding and more advanced AI human collaboration. Responsible AI development, ethical data governance and scalable solutions will be the key factors in ensuring the positive impact of NLP. Thank you so much for, pulling your time out, giving me time to present my views on NLP. I am looking forward to your questions and future discussions. Thank you so much.