Zoey
a Zodiac Oracle Experimental Yield
Created on May 14th 2024 2:54PM
Latest Version: 1.03 (May 23rd 6:38PM)
Click here to access her:
https://astrology-bot.streamlit.app/
Notes: This is the 3rd Version of our bot which is using OpenAI's Chatgpt-4o in tandem with Anthropic's Claude Haiku. Although the bot is technically following the rules, she still needs to be able to generate coherent text with minimal grammatical errors as well as playing with a larger set of words to avoid reader fatigue (sometimes she will repeat the same word). Users may play with this version but I doubt they will gain much insight from it at this stage.
a Zodiac Oracle Experimental Yield
Created on May 14th 2024 2:54PM
Latest Version: 1.03 (May 23rd 6:38PM)
Click here to access her:
https://astrology-bot.streamlit.app/
Notes: This is the 3rd Version of our bot which is using OpenAI's Chatgpt-4o in tandem with Anthropic's Claude Haiku. Although the bot is technically following the rules, she still needs to be able to generate coherent text with minimal grammatical errors as well as playing with a larger set of words to avoid reader fatigue (sometimes she will repeat the same word). Users may play with this version but I doubt they will gain much insight from it at this stage.
Timeline of Automated Storytelling Process
3rd Attempt - Suzie (Series of Unique Zodiac Interpretation Extractions)
Version 2.00 - TBA
Status: Privately Accessible
Notes: this time all data was transferred & neatly organized into a CSV file which included, glossary terms, sign rulership formula, & training data examples to help map out connections & relationships to various categories within the file. Ace parser was used for syntactic analysis along side Delph-in which is a deep linguistic processing tool specifically designed for head-driven phrase structure grammar initiatives. The LLM is unknown for now (potential candidates: Chatgpt-o1, DeepSeek r1)
Performance Rating: ??/15
Follows correct variable to word association, correct variable sequence order, basic entry field, filler words, LLM utilization, using training data examples, deep linguistic processing framework, grammatically correct output, coherent, user data entry field, sign rulership, ephemeris, accurately written representational theme, accurate visual depiction of theme, accurate descriptive summary of riddle to help with interpretation
Version 2.00 - TBA
Status: Privately Accessible
Notes: this time all data was transferred & neatly organized into a CSV file which included, glossary terms, sign rulership formula, & training data examples to help map out connections & relationships to various categories within the file. Ace parser was used for syntactic analysis along side Delph-in which is a deep linguistic processing tool specifically designed for head-driven phrase structure grammar initiatives. The LLM is unknown for now (potential candidates: Chatgpt-o1, DeepSeek r1)
Performance Rating: ??/15
Follows correct variable to word association, correct variable sequence order, basic entry field, filler words, LLM utilization, using training data examples, deep linguistic processing framework, grammatically correct output, coherent, user data entry field, sign rulership, ephemeris, accurately written representational theme, accurate visual depiction of theme, accurate descriptive summary of riddle to help with interpretation
2nd Attempt - Zoey (Zodiac Oracle Experimental Yield)
Version 1.01 - May 14th 2024 (2:54pm) bristow, va
Version 1.02 - May 15th (2:00am) bristow, va
Version 1.03 - May 23rd 2024 (6:38pm) bristow, va
Version 1.04 - TBA
Status: Publically Accessible
Notes: after being satisfied with my website design I decided to continue the project & purchased api keys from OpenAi & Anthropic, & attempted to apply logic directly into the prompts of both chatgpt-4o & claude haiku using a limited summarized version of the glossary files & python for the backend. While the bot was able to follow the directions after the 3rd update, the grammar output left a lot to be desired & remained clunky & difficult to decipher. A period of reflection was required to come up with a new solution.
Performance Rating: 5/15 (33.3%)
Follows correct variable to word association, correct variable sequence order, basic entry field, filler words, LLM utilization, using training data examples, deep linguistic processing framework, grammatically correct output, coherent, user data entry field, sign rulership, ephemeris, accurately written representational theme, accurate visual depiction of theme, accurate descriptive summary of riddle to help with interpretation
Version 1.01 - May 14th 2024 (2:54pm) bristow, va
Version 1.02 - May 15th (2:00am) bristow, va
Version 1.03 - May 23rd 2024 (6:38pm) bristow, va
Version 1.04 - TBA
Status: Publically Accessible
Notes: after being satisfied with my website design I decided to continue the project & purchased api keys from OpenAi & Anthropic, & attempted to apply logic directly into the prompts of both chatgpt-4o & claude haiku using a limited summarized version of the glossary files & python for the backend. While the bot was able to follow the directions after the 3rd update, the grammar output left a lot to be desired & remained clunky & difficult to decipher. A period of reflection was required to come up with a new solution.
Performance Rating: 5/15 (33.3%)
Follows correct variable to word association, correct variable sequence order, basic entry field, filler words, LLM utilization, using training data examples, deep linguistic processing framework, grammatically correct output, coherent, user data entry field, sign rulership, ephemeris, accurately written representational theme, accurate visual depiction of theme, accurate descriptive summary of riddle to help with interpretation
1st Attempt - Raw & Unfiltered Glossary
Version 1.00 - March 21st 2023 (4:36pm) bristow, va
Status: Privately Accessible
Notes: directly copying a random word from a text file that corresponded to each variable listed in the sequence using java. Although, It was completely unintelligble it served as the basic rough draft. I decided to postpone the project to build training data examples for my website which was created on March 26th 2023, bristow, va
Performance Rating: 4/15 (26.6%)
Follows correct variable to word association, correct variable sequence order, basic entry field, filler words, LLM utilization, using training data examples, deep linguistic processing framework, grammatically correct output, coherent, user data entry field, sign rulership, ephemeris, accurately written representational theme, accurate visual depiction of theme, accurate descriptive summary of riddle to help with interpretation.
Version 1.00 - March 21st 2023 (4:36pm) bristow, va
Status: Privately Accessible
Notes: directly copying a random word from a text file that corresponded to each variable listed in the sequence using java. Although, It was completely unintelligble it served as the basic rough draft. I decided to postpone the project to build training data examples for my website which was created on March 26th 2023, bristow, va
Performance Rating: 4/15 (26.6%)
Follows correct variable to word association, correct variable sequence order, basic entry field, filler words, LLM utilization, using training data examples, deep linguistic processing framework, grammatically correct output, coherent, user data entry field, sign rulership, ephemeris, accurately written representational theme, accurate visual depiction of theme, accurate descriptive summary of riddle to help with interpretation.
Copyright © 2024 by Ian Murphy
All rights reserved. This work, or parts thereof, may not be reproduced in any form without permission in writing from the author.
All rights reserved. This work, or parts thereof, may not be reproduced in any form without permission in writing from the author.