Arun Jayapal

Arun is a dedicated spiritual researcher who transitioned into the IT industry, now thriving as a freelance NLP consultant. He holds a Ph.D. from the School of Computer Science and Statistics at Trinity College Dublin, focusing on Finding Diachronic Sense Changes by Unsupervised Methods under the guidance of Dr. Martin Emms (Sep 2013 - Apr 2018). Arun's comprehensive expertise extends to serving as a reviewer for prestigious conferences like CoLing and EACL.

Previously, Arun held prominent roles such as Assistant Director at EY's AI Practices Team (Bangalore) and Data Scientist at Onit India Pvt Ltd. Following his Ph.D., he ventured into a start-up, sayint, based in Hyderabad, where he excelled as a computational linguist specializing in conversational analytics. Arun's proficiency extends to his tenure at Ditech Networks, now Nuance Communications, where he played a pivotal role in de-normalizing ASR outcomes for voice messaging platforms.

Arun completed his master's degree in Speech and Language Processing at the University of Sheffield, UK, with a thesis titled "Entity Identification & Co-reference Resolution," under the mentorship of Dr. Rob Gaizauskas. His journey reflects a seamless fusion of academic rigor and practical expertise, contributing significantly to the fields of NLP and AI.

🔹 Sentiment Analysis: Analyze text data to determine the sentiment expressed (positive, negative, or neutral) towards a particular topic, product, or brand.

🔹 Text Classification: Classify text documents into predefined categories or topics, such as spam detection, news categorization, or customer support ticket routing.

🔹 Named Entity Recognition (NER): To identify and classify entities mentioned in text, such as people, organizations, locations, dates, and other relevant entities.

🔹 Text Summarization: To automatically summarize large volumes of text, extracting key information and reducing content while retaining its essence.

🔹 Language Translation: Provide services for machine translation, converting text from one language to another using NLP techniques and language models.

🔹 Topic Modeling: To extract underlying topics or themes from a collection of documents, enabling insights into the main subjects discussed.

🔹 Text Generation: Create models capable of generating human-like text based on a given prompt or context, useful for content generation, story writing, or dialogue creation.

🔹 Keyword Extraction: Develop systems to automatically extract relevant keywords or phrases from text, aiding in content analysis, search engine optimization (SEO), and information retrieval.

🔹 Text Mining and Analysis: For analyzing large volumes of text data to uncover patterns, trends, and insights, such as sentiment trends, topic trends, or user behavior analysis.

🔹 Document Similarity Analysis: To measure the similarity between text documents, enabling applications like plagiarism detection, document clustering, and recommendation systems.

🔹 Text Annotation and Labeling: Provide annotation and labeling services for training datasets used in supervised learning tasks, such as sentiment analysis, text classification, and named entity recognition.

🔹 Custom NLP Solutions: Offer bespoke NLP solutions tailored to specific business needs or industries, leveraging your expertise in natural language understanding and processing.

🔹 NLP Research and Development: Engage in research and development activities to advance the field of NLP, exploring new algorithms, techniques, and applications to push the boundaries of natural language understanding and processing.

Technology Stack

My proficiency extends across a wide range of technologies and tools, including:

Programming Languages: C++, Java, Python (2.7 & 3.5+), R, MATLAB, LaTeX
NLP/ML tools: LLM, NLTK, Thrax, scikit-learn, PyTorch, Stanford CoreNLP
Databases & Tools: ClickHouse, PostgreSQL, Celery (microservice), Flask & FastAPI (API development), Apache Airflow

Interests

Driven by a passion for innovation and knowledge advancement, my research interests include

Unsupervised learning using Bayesian modeling
Distributed semantics
Data mining and information retrieval
Machine translation and reinforcement learning

🔎 Are you searching for a seasoned Machine Learning / AI specialist capable of delivering tangible business outcomes? Look no further. I am dedicated to taking full ownership of projects, allowing you to focus on your core business while I lead the success of your AI initiatives.

Dynamic Generative model for Diachronic Sense Emergence Detection Proceedings of COLING 2016 by Martin Emms, Arun Jayapal held at Osaka, Japan in Dec. 2016 - A Gibbs sampling algorithm is developed for a diachronic model to estimate the parameters from raw time-stamped n-gram data with no sense annotation. We are able to demonstrate the inference of parameters which plausibly reflect the objective dynamics of sense use frequencies, in particular the emergence of a new sense.

An unsupervised EM method to infer time variation in sense probabilities Proceedings of ICON 2015: 12th International Conference on Natural Language Processing by Martin Emms, Arun Jayapal held at Trivandrum, India in Dec. 2015 - We developed an EM algorithm to estimate the parameters from raw time-stamped n-gram data with no sense annotation. We are able to demonstrate the inference of parameters which plausibly reflect the objective dynamics of sense use frequencies, in particular the emergence of a new sense.

Author Verification: Basic Stacked Generalization Applied To Predictions from a Set of Heterogeneous Learners by Erwan Moreau, Arun Jayapal, Gerard Lynch, Carl Vogel; PAN proceedings on Author verification, CLEF 2015 in Nov 2015 - Provides our system description for the author verification conducted by CLEF workshop.

Detecting change and emergence for multiword expressions by Martin Emms, Arun Jayapal; Proceedings of the 10th Workshop on Multiword Expressions (MWE 2014) held with EACL 2014 at Gotenburg, Sweden in Apr. 2014 - concerns unsupervised means to detect an emerging new 'MWE' usage of a given n-gram, such as the usage of smashed it to mean doing something excellently.

TCDSCSS: Dimensionality Reduction to Evaluate Texts of Varying Lengths-an IR Approach by Arun Jayapal, Martin Emms, John D. Kelleher; Proceedings of the 8th International Workshop on Semantic Evaluation (SemEval 2014), COLING held at Dublin, Ireland in Aug. 2014 - provides system description of the cross-level semantic similarity task for the SEMEVAL-2014 workshop, which uses dimensionality reduction technique.

Author Verification: Exploring a Large set of Parameters using a Genetic Algorithm by Erwan Moreau, Arun Jayapal, Carl Vogel; PAN proceedings on Author verification, CLEF 2015 in Sep. 2015 - concerns system description of Author verification task conducted by PAN for CLEF.

Vector space model and Overlap metric for Author Identification by Arun Jayapal, Binayak Goswami; PAN proceedings on Author verification, CLEF 2013 in Sep. 2013 - concerns system description of Author verification task conducted by PAN for CLEF. The system was developed using the very simple vector space model and worked reasonably well.

Similarity Overlap Metric and Greedy String Tiling for Plagiarism Detection by Arun Jayapal; PAN proceedings on Plagiarism detection, CLEF 2012 in Sep. 2012 - concerns system description of Plagiarism detection task conducted by PAN for CLEF. The system was judged first in the document retrieval task (one of two tasks involved in Plagiarism detection system).

USFD at KBP 2011: Entity linking, slot filling and temporal bounding by Amev Burman, Arun Jayapal, Sathish Kannan, Madhu Kavilikatta, Ayman Alhelbawy, Leon Derczynski, Robert Gaizauskas; KBP 2011 proceedings on Knowledge base Population for Text Analysis Conference in Jan. 2011 - concerns system description for the Question answering task and this was part of my Master's dissertation. My role was to come up the Named entitity recognition (NER) model and co-reference resolution model. For NER, I used stanford CRF classifier and for co-reference resolution, I had come-up with a rule-based classifier.