Projects
Neural_Conditional_Random_Fields_for_Named_Entity_Recognition-2.pdfConditional Random Fields For NER
We use a BiLSTM to obtain feature representations and combine it with CRFs to obtain a globally normalised model with both transition and emission parameters. We train and alayse performance on the CoNLL-2003 shared task.
Sequence Tagging Using HMM.pdfSequence Tagging Using HMM Models
We use a trigram HMM model with emission and transition parameters along with Viterbi algorithm for decoding to perform sequence tagging on biological data.
Machine Translation Using IBM Models.pdfMachine Translation Using IBM Models
We translate from English to Spanish using traditional IBM Models. We also analyse the word alignments learnt by Model 2.
Feature Based Explanations for XAI
Part 1 of a 3 part tutorial on Explainable AI and its challenges and opportunities for data management research. Presented at ICDE 2022 and ACM SIGMOD 2022. I talk about Feature Based Explanation techniques including LIME, ANCHORS, MUSE and Shapley Values
