timelog.md

Timelog

• Task structure extraction from text
• Lewis Grimmer
• 2350871G
• Jeff Dalton

Week 1

24 September 2021 - 6 hours

• 4 hours Read the project guidance notes
• 1 hour Created Github repository for project, made use of moodle example
• 1 hour Created dissertation from template, edited

Week 2

30 September 2021 - 6 hours

• 6 hours Background reading on extraction from text, task structure extraction

Week 3

7 October 2021 - 6 hours

• 0.5 hour Meeting with supervisor
• 3.5 hours Background reading on papers given by supervisor
• 2.5 hours Initial analysis and system design

Week 4

14 October 2021 - 6 hours

• 0.5 hour Meeting with supervisor
• 1 hour Meeting with alexa challenge team lead
• 4.5 hours Research entity linking

Week 5

21 October 2021 - 7 hours

• 7 hours Create GENRE initial model

Week 6

28 October 2021 - 6 hours

• 6 hours Evaluate GENRE as baseline model

Week 7

4 November 2021 - 6 hours

• 0.5 hour Meeting with supervisor
• 1 hour Meeting with alexa challenge team lead
• 4.5 hours Create GPT-J initial model

Week 8

11 November 2021 - 8 hours

• 0.5 hour Meeting with supervisor
• 2.5 hours Create user stories and requirements
• 5 hours Run GPT-J model on instruction examples

Week 9

18 November 2021 - 6 hours

• 0.5 hour Meeting with supervisor
• 2.5 hours Research NER
• 3 hours Design system using NER pipeline

Week 10

25 November 2021 - 6 hours

• 3 hours Data design for NER system
• 3 hours Implementation research and testing for NER system

Week 11

2 December 2021 - 7 hours

• 7 hours Create and analyse TYN dataset

December Project period

6-10 December 2021 - 10 hours

• 10 hours Create labelled data using simple string matching

13-17 December 2021 - 10 hours

• 10 hours Create labelled data using Jaccard matching

Week 3

26 January 2022 - 1 hour

• 1 hour Meeting with supervisor

Week 4

3 February 2022 - 5 hours

• 1 hour Create human labelled data using prodigy
• 4 hours Train CNN model on simple labelled data

Week 5

10 February 2022 - 6 hours

• 1 hour Create human labelled data using prodigy
• 5 hours Train CNN model on Jaccard labelled data

Week 6

17 February 2022 - 8 hours

• 2 hours Create human labelled data using prodigy
• 2 hours Train CNN model on prodigy data
• 4 hours Train transformer model on prodigy data

Week 7

24 February 2022 - 5.5 hours

• 0.5 hour Meeting with supervisor
• 5 hours Work on model access from scripts

Week 8

3 March 2022 - 6 hours

• 3 hours Read example projects
• 2 hours Writing dissertation introduction section
• 1 hours Begin background section

Week 9

10 March 2022 - 6.5 hours

• 0.5 hour Meeting with supervisor
• 2 hours Continue background section
• 3 hours Writing dissertation analysis section
• 1 hour Begin design section

Week 10

15 March 2022 - 6 hours

• 3 hours Continue design section
• 3 hours Writing implementation section

17 March 2022 - 6 hours

• 6 hours Continue implementation section

Week 11

21 March 2022 - 6 hours

• 6 hours Begin evaluation section

23 March 2022 - 6 hours

• 6 hours Editing from to supervisor comments

24 March 2022 - 6 hours

• 1 hours Meeting with supervisor
• 3 hours Editing from supervisor comments
• 2 hours Continuing evaluation section