Difference between revisions of "The Real A Team"
(→Determining Author By Style Of Writing) |
(→Timing) |
||
| Line 64: | Line 64: | ||
| Blog Post|| 4,997 || 7 | | Blog Post|| 4,997 || 7 | ||
|} | |} | ||
| + | |||
| + | ====Timing Explanation==== | ||
| + | For the large flie, (250,234 characters) the parallel program had a significant speed up, however for the medium and small file, the serial program actually ran faster. I believe this is due to the overhead of setting up threads. | ||
Revision as of 18:29, 21 March 2016
GPU621/DPS921 | Participants | Groups and Projects | Resources | Glossary
Contents
Determining Author By Style Of Writing
A Team Members
- Adrian Sauvageot, All
- ...
Progress
Pre-Assignment
I decided to create a new program to test a theory I was told.
I was told by a professor that she believed that by taking a look at how a paper was written, she could tell if it was written by the same author. Further, she believed that a computer could tell if two pieces of text were written by the same author by looking at how it was written.
I decided to create a program that would analyze two pieces of text to try and determine if the same person wrote both pieces.
I decided to look at:
- average words/sentence
- average word length
- average sentences/paragraph
- average commas/sentence
- average colons/paragraph.
I then use this information to calculate how different two pieces are from each other. If they are within what I determined to be a 5% different writing style, I suggest the two pieces were written by the same person, otherwise I suggest they were written by two separate people.
To test this I ran the program on work by Shakespeare, One of my friends, and myself.
The program successfully was able to determine which author wrote each piece of text.
Assignment
The program I wrote relies on one single loop to run through a piece of text. It has no dependencies, so it can easily be parallelized using the methods discussed in this class.
Timing
To time the program I used various pieces of text. I used text from 3 authors, with varying lengths. I used 2 Shakespeare works (long - 46,956 words 250,234 characters), 2 assignments I completed for school (medium - 1,885 words 11,336 characters), and 2 blog posts that were written by the same author (short - 869 words 4,997 characters).
Serial Timing
| Author | Character Count | Time (milliseconds) |
|---|---|---|
| Shakespeare | 250,234 | 157 |
| Adrian Sauvageot | 11,336 | 7 |
| Blog Post | 4,997 | 3 |
OpenMP Timing
| Author | Character Count | Time (milliseconds) |
|---|---|---|
| Shakespeare | 250,234 | 72 |
| Adrian Sauvageot | 11,336 | 17 |
| Blog Post | 4,997 | 7 |
Timing Explanation
For the large flie, (250,234 characters) the parallel program had a significant speed up, however for the medium and small file, the serial program actually ran faster. I believe this is due to the overhead of setting up threads.
