[AI] Define Topic Score Update Logic

[sajz]

Topic Analysis and Score Updates Logic:

1. Identify Relevant Topics:

   • When a new message is posted, analyze its semantic content using SemanticVector.
   • Match the message’s semantic content with existing topics in the channel’s ASSOCIATED_WITH relationships by comparing keywords and the semantic_vector.

2. Amplify Scores for Matching Topics:

   • For each existing topic in the channel that appears in the message:
     • Apply the amplify_score function to calculate a small score_delta, increasing the topic’s overall_score based on relevance or semantic similarity.
     • Update message_count and refresh the last_updated timestamp.

3. Create New Topics:

   • For topics in the message that don’t already exist in the channel:
     • Create a new Topic node if it doesn’t exist, or establish an ASSOCIATED_WITH relationship to the channel.
     • Initialize fields like overall_score, message_count, and last_updated to begin tracking the topic’s relevance in this channel.

4. Diminish Scores for Unmentioned Topics:

   • For all existing topics that aren’t referenced in the message:
     • Use the diminish_score function to apply a gradual score_delta reduction, decreasing the overall_score based on the score_decay_rate.
     • This decay reflects the gradual fading relevance of less active topics in the channel.

Suggested Function Names:

1. Amplify Function (amplify_score):

   • amplify_score: This function calculates a small positive score_delta to incrementally increase the overall_score based on semantic similarity, with gradual progression towards a maximum limit.

2. Diminish Function (diminish_score):

   • diminish_score: This function calculates a small negative score_delta to gradually decrease the overall_score, applying a decay that starts slowly and accelerates toward zero for topics not mentioned over time.

-- Example

Topic	Initial Score	New Message Weight	Updated Score	Score Delta
topic_0	0.5	0.6	0.505	+0.005
topic_1	0.5	0.2	0.485	-0.015
topic_2	0.5	0.45	0.4975	-0.0025

The updated table includes the Score Delta for each topic, showing how much the overall_score changed based on the new message weights:

• Topic 0: Initial Score = 0.5, New Message Weight = 0.6, Updated Score = 0.505, Score Delta = +0.005
• Topic 1: Initial Score = 0.5, New Message Weight = 0.2, Updated Score = 0.485, Score Delta = -0.015
• Topic 2: Initial Score = 0.5, New Message Weight = 0.45, Updated Score = 0.4975, Score Delta = -0.0025

Adjusted score_delta Function for Topic overall_score Updates

Objective: Enhance the calculation of score_delta for updating the overall_score of topics in a channel, ensuring incremental changes are context-sensitive based on the initial score. This adjustment will use a non-linear function that:

• Limits rapid increments for very low (e.g., < 0.2) and very high (e.g., > 0.8) scores.
• Promotes normal growth between 0.2 and 0.8.
• Applies to both amplifying and diminishing the topic's overall_score.

Rationale:

• Stability of High-Scored Topics: Topics with an overall_score above 0.8 should not reach 1 too rapidly. This ensures dominant topics stabilize and do not skew the topic mapping unfairly.
• Controlled Growth for Low-Scored Topics: Topics with an overall_score below 0.2 should have limited growth to prevent sudden amplification due to noise or isolated messages.
• Consistent Normal Growth: The function should provide a standard growth pattern for scores between 0.2 and 0.8 to reflect typical topic relevance.

Function Design: Use a modified logistic function to calculate score_delta, allowing gradual increases that slow at the boundaries (near 0 and 1):

Topic	Initial Score	New Message Weight	Score Delta	Updated Score
topic_0	0.2	0.6	+0.000949	0.200949
topic_1	0.7	0.2	-0.022020	0.677980
topic_2	0.5	0.45	-0.001250	0.498750