Before the annotation of the tdt corpus begins in earnest, we need to develop a preliminary list of topics from which we will select the core list of topics for further annotation.
After reviewing the supplied lists of seed articles, initial topics will be developed based on the determination of an identifiable seminal event that is found in some of the aricles. Once the decision is made that an article may be used as the seed for a potential topic, the seed document will be used as a query against the entire collection of stories; one can determine relatively well the possibility of the respective 'coverage' of the event, and subsequent topical framework surrounding the event within the available stories.
Please note: You do NOT need to create a topic from each individual seed story in a list - indeed, please don't.
Select those seeds for development that you feel have:
- identifiable seminal event - a 'reasonable' expectation that the event will be covered in the Mandarin, Arabic, and English language sources. There will be many stories that you may feel are too specific, and won't be covered - these should not be rejected out of hand, but please be cautious addressing stories that feel confident will only appear once. (most often human interest stories - boy loses dog in Montana, etc.) - not too broad in scope, and you MUST be able to point to a seminal event.
There are three primary display windows that you will be interacting with, two netscape windows, and the primary topic selection GUI, which has several components.
Three fields empty field - 'seen' - 'reject'
These are all fairly self-explanatory, save the 'reject' condition. You can use this option if you run across seeds that are very problematic, for instance, seeds that are advertisements, book reviews, or the like.
Show Query Results
shows a ranked list of documents that are the result of submitting the current article from the seedlist into a search engine for the entire corpus.
- use this check to see if there is a possibilty that this can become a reasonable topic to explore further, judging on an estimate of how many related articles you find.
If you are leaving and desire to continue working on the list when you return.
Mark as DONE
When you have completed reviewing the seeds on the list you are working on, and you wish to continue working on a new list - If you make an error and accidentally log out with this, just let one of the project managers know, and we'll reset it for you.
Add to Topic list
you will NOT be allowed to add a topic to the topic list until you enter information into each of these three fields, once these conditions have been satisfied, you can enter in a new topic.
title - self explanatory - minimum length eight characters
seminal event, - the specific event that you are identifying, indentifing the who, where, when, and what - minimum length 16 characters
topic description - a description in your own words describing the topic. Keep in mind that you will be working on a research page on this later, where you can expand on the information you would like to provide. - minimum length 32 characters
As was mentioned earlier, try to be fairly specific regarding the seminal event that you are choosing. Try to ensure that it is fairly limited in scope (for example, please do not use the US economic downturn as a topic - it is far too broad in scope for our purposes - a better option would be - mexico's second annual chess match. Make sure that you submit a query with the article - while you may be quite confident that it will be an exemplary topic, still submit a query and take a look at the returned files from the search engine just to confirm your intuition.
Topic Selection Quick Checklist
- Open netscape
- run start-tdt4-mktopics
- change encoding to simplified chinese (mandarin) - under 'view' toolbar
- review seeds
- potential topic identified?
- establish seminal event
- check topic list to ensure that no one has developed this topic
- use seed document as query
- examine ranked returns
- okay coverage of potential event/topic?
- have you developed better sense of the topic?
yes --> develop topic
- establish title
- establish who, what, where, when of seminal event
- create description in your own words about event, and topic, including additional information which you may decide is relevant.
BEFORE you update the topic list -
- refresh topic list window, re-check current list to re-establish that no one else has developed this topic.
- - if not, update the topic list with your new entry. You can return to edit this topic at any time when you are working on the same seedlist to udate your topic description.
after you have found 3 or 4 topics in seedlist start working on your topic research for the topics you have created.
The goal of topic research is to be able to provide additional information about the event and the topic. It is useful to have enough information in the document so that an individual who is unfamiliar with the subject matter have access to the reference material without having to do all of the related research themselves. magine you're taking an exam on a particular TDT topic, and you're allowed to have one piece of paper with all the relevant information about that topic on it - that piece of paper is what we'd like to create with the topic research task.
3043. Sri Lankan Gov't vs. Tamil RebelsTopic research from TDT2
3044. Kurd Separatist Abdullah Ocalan Arrested
3050. US Mid-Term Elections
3055. D'Alema's New Italian Government
3056. Chechnya Rebel Kidnapping and Beheading