Lifelong Learning for NLP Tasks

Student: Matkarimov Otabek

Supervisor: Ekaterina Artemova

Educational Programme: Data Science (Master)

Year of Graduation: 2020

This paper focuses on the lifelong learning problem of continuous learning based on streaming text data. We propose an episodic memory model for words that performs a replay of important words for each class to mitigate catastrophic forgetting on trained data. Text classification experiments demonstrate the advantages of reproducing important words over reproducing experience, allowing the classifier to continually extract knowledge from new data sets. The results of research on the effectiveness of keyword selection for each class are also presented. This work was made in Python with using the nltk, pymorphy2, sklearn libraries and the PyTorch machine learning library. Keywords: NLP, lifelong learning, classification tasks, Bag-of-words, TF-IDF, BERT, python.

Student Theses at HSE must be completed in accordance with the University Rules and regulations specified by each educational programme.

Summaries of all theses must be published and made freely available on the HSE website.

The full text of a thesis can be published in open access on the HSE website only if the authoring student (copyright holder) agrees, or, if the thesis was written by a team of students, if all the co-authors (copyright holders) agree. After a thesis is published on the HSE website, it obtains the status of an online publication.

Student theses are objects of copyright and their use is subject to limitations in accordance with the Russian Federation’s law on intellectual property.

In the event that a thesis is quoted or otherwise used, reference to the author’s name and the source of quotation is required.

Search all student theses