Smart Calendar

From SWRC

Revision as of 05:09, 24 January 2017 by Kekeeo (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Introduction

Smart Calendar: From Email to Calendar

The Smart Calendar project aims to automatic schedule information extraction from user's emails, and update those information automatically to the user's electronic calendar. (2010 ~ )

Nowadays, emails are frequently used to exchange information between various people. Due to the emergence of smartphones, people can read and write emails everywhere. But processing information from email using smartphone is very tedious work for a user, considering the small display size and uncomfortable touchscreen input interface of a smartphone.

In this project, an approach to automatically extract schedule information from email is researched.





Details

Target information for Extraction

The below table shows target information for extraction, and their definitions:

Term Type Definition
Meeting Term representing the meeting itself.
Location A place. (Traditional NE meaning)
Organization An organization. (Traditional NE meaning)
Person A person. (Traditional NE meaning)
PhoneNumber A phone number.
Time Time and date.
Title Some kind of title.
Topic Some kind of topic.
Transport Transport (Ex. subway)
Webpage A webpage URL.
ETC Miscellaneous terms necessary for relation information extraction.


Relation Type Definition
startTime Arg1 (Meeting)'s starting time is Arg2(Time).
endTime Arg1 (Meeting)'s ending time is Arg2(Time).
hasAttendee Arg2 (Person) will attend to Arg1 (Meeting). Arg2 are decided to attend the meeting.
hasPresenter Arg2 (Person) will make a presentation at Arg1 (Meeting).
isOrganizedFor Arg2 (Person) is the target of Arg1 (Meeting).
hasReference If you want to know more about Arg1 (Meeting), please contact to Arg2(Homepage/PhoneNumber/Person/Organization).
hasSupervisor Arg1 (Meeting) is supervised by Arg2(Person/Organization).
hasTitle The title of Arg1 (Meeting) is Arg2(Title).
hasTopic The topic of Arg1 (Meeting) is Arg2(Topic).
isHeldAt The Meeting Arg1 (Meeting) will be held at Arg2(Location).
isRequiredToBring The Meeting Arg1 (Meeting) requires its attendee to bring Arg2(ETC).
locationLandmark To go to the meeting place of Arg1 (Meeting), you may use Arg2(Location) as landmark to go there.
movesBy To go to the meeting place of Arg1 (Meeting), it will be comfortable if you use Arg2(Transport) to get to the place.

Email corpus for Korean

Total 1,011 announcement emails are crawled from the web. Many of them seems to be a repetition of emails on the web bulletine board.

The annotation on those emails is carried out using COAT manual semantic annotation toolkit. The toolkit supports annotation/conjugation paradigm, thus making the resultant corpus reliable. The resultant annotated data are exported in XML format.

The corpus is downloadable at the following link:

Ver. 2011. 03. 02

Ver. 2011. 07. 19, manually modified by KyoungRyol Kim to fix some obvious errors

Ver. 2011. 09. 20. Manually attached per-meeting information for each time expressions. Plus, obvious errors (Ex. Term type "ETC" to time expression) are modified.

Distinguishing corpus documents based on their types

Case 1-1. StartTime/endTime expressions show up in itemized sentences( starttime:929 docs, endtime: 411 docs)

Case 1-2. StartTime/endTime epressions show up in natural language sentences ( starttime:187 docs, endTime: 29 docs)

Case 2-1. Email contains only one startTime expression ( 743docs)

Case 2-2. Email contains information about single startTime, and many different expressions refers to the same time. (53 docs)

Case 2-3. Email contains information about single startTime, but the information itself is spreaded across the whole email. (151 docs)

Case 2-4. Email contains information about multiple schedules. (64 docs)

Traversing Corpus Through Web

Traverse the Corpus.

Current status

Currently, a program for start/end time extraction of schedules for Korean is developed, and available as Chrome Extension and Android application.

Chrome Extension v.0.2.5 (Dec 28, 2010) : Manual Download Extension

Android Application integrated with K9-mail (Jan 2, 2011) : Download

Android Application v.0.1 integrated with K9-mail (May 16, 2011) : Smartphone Tablet (Galaxy-tab)

Web-based system integrated with NOCC mail client (May 17, 2011) : Link