2020/12/01 by Shree Charran R, R, Shree Charran, Rahul Dubey +1
Computer Science · #Advanced Text Analysis Techniques #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Machine Learning (cs.LG) #Natural Language Processing Techniques #Text and Document Classification Technologies #Topic Modeling
paper · pdf · doi:10.48550/arxiv.2012.00633
openalex publication_date 2020/12/01 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Word Representations form the core component for almost all advanced Natural\nLanguage Processing (NLP) applications such as text mining, question-answering,\nand text summarization, etc. Over the last two decades, immense research is\nconducted to come up with one single model to solve all major NLP tasks. The\nmajor problem currently is that there are a plethora of choices for different\nNLP tasks. Thus for NLP practitioners, the task of choosing the right model to\nbe used itself becomes a challenge. Thus combining multiple pre-trained word\nembeddings and forming meta embeddings has become a viable approach to improve\ntackle NLP tasks. Meta embedding learning is a process of producing a single\nword embedding from a given set of pre-trained input word embeddings. In this\npaper, we propose to use Meta Embedding derived from few State-of-the-Art\n(SOTA) models to efficiently tackle mainstream NLP tasks like classification,\nsemantic relatedness, and text similarity. We have compared both ensemble and\ndynamic variants to identify an efficient approach. The results obtained show\nthat even the best State-of-the-Art models can be bettered. Thus showing us\nthat meta-embeddings can be used for several NLP tasks by harnessing the power\nof several individual representations.\n