# 2024 04

[previous](2024-03.md)

### 4)
I am looking for a general strategy how to continue my masters-thesis.
The following two questions are guidelines:
1) [llm-general-physics-knowledge-extraction-question](../../topics/physics/knowledge-representation/llm-general-physics-knowledge-extraction-question.md)
2) [llm-resource-physics-knowledge-extraction-question](../../topics/physics/knowledge-representation/llm-resource-physics-knowledge-extraction-question.md)
I have finished a simple subtopic network generator:
* [master-simple-relations](../../topics/code/projects/master-simple-relations.md) 
### 5)
The results of the subtopic network should be [evaluated](../../topics/physics/knowledge-representation/physics-subtopic-network-evaluation.md).
One possibility of how to proceed would be to [create a list](../../topics/data/knowledge/most-important-terms-of-topic-question.md) of physics terms and then write an algorithm to classify them into a subtopic tree.

### 8)
Explored possibility to host large database files with [zenodo](https://zenodo.org/uploads/10936027) Other option is to create github repository for that purpose: [https://github.com/gratach/master-database-files](https://github.com/gratach/master-database-files)

### 11)
Created a [subtopic network search](../../topics/code/projects/master-subtopic-network-search.md) project. This project implements a [evaluation method](../../topics/data/knowledge/subtopic-network-search-evaluation-method.md) for subtopic networks. The next steps could be to apply this evaluation method to different types of large language models. Also the stability of a single search could be tested.

### 15)
Found a solution for the calculation of the statistical error of the percentage of found technical physics terms: [https://de.wikipedia.org/wiki/Binomialverteilung#M%C3%BCnzwurf](https://de.wikipedia.org/wiki/Binomialverteilung#M%C3%BCnzwurf) The error calculation of the average search path length is done like described in [this link](https://www3.physik.uni-stuttgart.de/studium/praktika/ep/pdf_dateien/Allgemeines/Fehlerrechnung.pdf)

### 16)
Implemented the [subtopic network search](../../topics/code/projects/master-subtopic-network-search.md) evaluation for the mistral model and the cosmosage model of [blablador](https://helmholtz-blablador.fz-juelich.de/). The cosmosage model is so bad at formatting the output correctly so a post processing with mistral was necessary.

### 17)
Regenerated subtopic network search data for the models gpt-3.5-turbo, gpt-4-turbo, mistral_7b_instruct_v_0_2 and cosmosage with a slightly changed query where I did not mention that I wanted to have the output formated as json because gpt-4-turbo was misinterpreting that information. Temperature change to 1 of model cosmosage leads to notable worse performance so I left the temperature at the default value.
[Compared models in table](../../topics/master-thesis/evaluation/subtopic-network-search-model-comparison.md).

### 19)
Created a [Project](../../topics/code/projects/master-adapted-subtopic-tree-generation.md) for subtopic tree generation which is adapted to a specific list of technical terms.

### 20)
Translated the [forum-concept-draft](../../topics/graph/debate/forum-concept-draft.md) into English

### 22)
[Evaluated](physics-term-adapted-subtopic-tree-evaluation.md) the results from the adapted subtopic tree generation

### 24)
Created [metric](subtopic-tree-average-number-of-viewed-terms-per-search-evaluation.md) for the measurement of the quality of a subtopic tree 

### 25)
Added the metric of the viewed bytes per search.

### 28)
Described [approach](../../topics/master-thesis/approaches/diverse-connections-between-existing-topics.md) of how to continue with the creation of diverse semantic connections in-between existing physics terms

[next](2024-05.md)

[Overview 2024](overview-2024.md)