Background & Task Overview

Our collection of news comes from Agence France Press, which has news bureaus in many Arabic-speaking nations.  The news coverage will be local, regional and international.  The news collection consists of 383,872 stories from May 1994 through December 2000 - 2,337 days worth of stories in all.

Your job will be to identify topics that are discussed in this news collection.  For this project, a topic is general -- not a specific event (like a particular plane crash or particular political event), but rather a broad category, like "energy resources" or "computer hacking".  Later, you'll be asked to review sets of stories that automatic processes have determined to be "on-topic" to determine whether they are indeed related to the topics you've identified.
 

TASK 1
Come up with a list of 50 potential topics that you believe are likely to be discussed in the news collection described above.  Each
topic should consist of a one-sentence description (in Arabic) of the information you are seeking.  Please word this description as a
question.  Some examples of topics you might come up with:

        -Have any computer hackers been charged with crimes in Jordan?
        -What measures are being taken to relieve air pollution in Cairo?

Use your knowledge of the collection's time period (1994-2000) and the regional focus (Jordan, Lebanon, Egypt, Syria, etc.) to help you
formulate potential topics.  Avoid creating topics which are punctual; that is, topics that relate to an event that occured at a specific place
and time.

You might find it helpful to consult resources like encyclopedias (which publish annual yearbooks covering major events and issues in a given
year) to help you think of likely topics.  Also, please consult the list I've given you containing the 79 topics that were developed for last
year's project (although these are Chinese news topics, some of them might be adaptable for this year's Arabic language focus.  In any case,
this document will give you a sense of what an appropriate topic looks like.)

TASK 2
Once each of you has developed a list of 50 potential topics, you will begin to search through the news collection to identify stories
that discuss your topics.  More details on the exact nature of topic searching will be discussed the next time we meet, but here is an
overview:

*Using a customized search engine, you'll submit a query consisting of a few keywords drawn from your topic description.
*The search engine will then display a list of 25 documents that are likely to be related to your topic.
*You will read each of these 25 documents, and determine which ones are indeed related to your topic.
*Depending on the number of documents related to your topic, you'll either
        -discard the topic (too few or too many documents)
OR
        -execute another search for this topic, and read the top 100 stories returned by the search engine

At this point, your work on this topic is complete and you'll move on to the next one.  Throughout this process, you'll keep careful notes about the search terms you've used, how you are defining the topic, and so on.  When we have identified & defined 25 topics that are suitable (neither too big nor too small), our work on the first phase is done.

TASK 3
You will be asked to review a set of stories for each of the 25 topics.  For each story, you'll decide whether it is related to the topic or not.  More details to come...
 

Timeline
 

Task	Description	Deadline
1	Compose preliminary topic lists	May 9, 2001
2	Create topic description pages for each topic	June 1, 2001
3	Identify 25 good topics from preliminary lists	June 1, 2001
4	Assess relevance of documents	September 28, 2001

How to Type in Arabic
We are using a version of the text editor emacs to create our Arabic language documents.  The editor is called armule (ar for Arabic).  To open a file using this text editor, type

armule filename

Meta-\    (Hold down meta key while typing backslash)

IMPORTANT: You must be on a blank line without any text entered before switching back and forth between Arabic-English!  Do not try to switch mid-sentence.

When typing in English, everything works as you might expect.  Lines wrap automatically, and to start a new line you hit Return.  When typing in Arabic, things are a little different.  Here are some pointers:

• Don't hit "RETURN" to start a new line!  Instead, use one of these two methods:
• Hit the space bar then type Cntrl-j (hold down control key while typing j)
• Hit Cntrl-e (hold down control key while typing e) to get to the absolute right margin of the line, and then hit <return>
• To undo what you've just done, type Cntrl-SHIFT-_ (hold down control key and shift key while typing dash)
• To delete the Arabic character preceding the cursor (to the left of the cursor), type Cntrl-d (hold down control key while typing d)

Creating Topic Description Pages

Each topic requires the creation of a topic description page, containing the following information:

Topic Number
Topic Title: Short phrase
Topic Author: Who created this topic
Description (Arabic & English): Brief, one-sentence description of the information you are seeking (in the form of a question).
Keywords (Arabic & English): Set of keywords you use to retrieve relevant documents.

To create the topic description pages, follow these steps:

1) Go to the Arabic directory by typing

cd trec-cl-ara/topics

Fatima : Topics 100-150 (ar-100 through ar-150)
Gordon: Topics 200-250 (ar-200 through ar-250)
Diana: Topics 300-350 (ar-300 through ar-350)
Nabih: Topics 400-450 (ar-400 through ar-450)

Fatima's preliminary topics
Gordon's preliminary topics
Diana's preliminary topics
Nabih's preliminary topics

armule ar-100

You'll use the arrow keys or the mouse to scroll from one field to another, but be sure to remember the tips about typing in Arabic when entering the data!

5) Once you have filled in each field, save your work by

pulling down File: Save Buffer

pulling down File: Exit Emacs