finished project

Author: cbartmann

.DS_Store

@@ -0,0 +1,3 @@
+User Name: [email protected]
+API Key: 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
+

Data/Key.txt

@@ -0,0 +1 @@
+{"system": "You are given the information of a news article and a user prompt. \nYou are part of the evaluation system of a recommender system for news articles based on user prompts.\nMore specific the recoomendation system is trying to match user prompts to news articles based on the topics that the user is looking for.\nYour task is to decide if the recommended article matches the user prompt in terms of the topics that the prompt is looking for.\nThe input information is given in german. \nThe different inputs of the news article are seperated by: ####\nMissing values are indicated by the string: nan\n\n{format_instructions}", "system_relaxed": "You are given the information of a news article and a user prompt. \nYou are part of the evaluation system of a recommender system for news articles based on user prompts.\nMore specific the recoomendation system is trying to match user prompts to news articles based on the topics that the user is looking for.\nYour task is to decide if the recommended article matches the user prompt in terms of the topics. In your decision process you are already satisfied if the topics the prompt is looking for are only remotly present in the news article.\nThe input information is given in german. \nThe different inputs of the news article are seperated by: ####\nMissing values are indicated by the string: nan\n\n{format_instructions}", "system_relaxed_2": "You are given the information of a news article and a user prompt. \nYou are part of the evaluation system of a recommender system that matches news articles based on user prompts based on topics.\nYour task is to decide if the recommended article matches the user prompt in terms of the topics that both handle. \nIn your decision process you are already not strict and are already satisfied if the topics the prompt is looking for are only slightly present in the news article.\nThe input information is given in german. \nThe different inputs of the news article are seperated by: ####\nMissing values are indicated by the string: nan\n\n{format_instructions}", "human": " The following is a news article and the corrpsoning prompt. All the information is given in german.\nThe title of the article:\n####\n{title}\n####\nThe text of the article:\n####\n{text}\n####\nThe general classification given by the dataset:\n####\n{general}\n####\nThe subtitle of the article:\n####\n{subtitle}\n####\nThe supertitle of the article:\n{dachzeile}\n###\nThe rubrik and resort of the article:\n{rubrik}, {resort}\n####\nFinally the propmt:\n{prompt}\n####\n", "schema": "Matches the recommended news article the given prompt in terms of topics of the news article that the user prompt is looking for  ?                             Answer True if yes,                            False if not or unknown.", "schema_relaxed": "Matches the recommended news article the given prompt in terms of topics of the news article that the user prompt is looking for in the slightest sense?                             Answer True if yes,                            False if not or unknown.", "schema_relaxed_2": "Matches the recommended news article the given prompt in terms of topics of the news article that the user prompt is looking for in the slightest sense?                             Answer True if yes,                            False if not or unknown."}

Data/cleaned_df.csv

@@ -0,0 +1 @@
+{"system_gen": " You are provided with the textual information and the title as well as subtitle of a new article.\nYour task is to create a Human prompt for the news article, specifically taking into acocount the topics that the article is talking about.\nThe propmt should be general format and not mention specific words or context of the article. Rather conctrate on the general topics described.\nThe prompt must be written in german. The input information is given in german. \nThe different inputs of the news article are seperated by: ####\nMissing values are indicated by the string: nan\n", "system_spe": " You are provided with the textual information and the title as well as subtitle of a new article.\nYour task is to create a Human prompt for the news article, specifically taking into acocount the topics that the article is talking about.\nThe propmt should be very specifict. Meaning is I use it in a query it will only return this article and no other. Conctrate on the specific topics described.\nThe prompt must be written in german. The input information is given in german. \nThe different inputs of the news article are seperated by: ####\nMissing values are indicated by the string: nan\n", "human_temp": " The following is a news article. All the information is given in german.\nThe title of the article:\n####\n{title}\n####\nThe text of the article:\n####\n{text}\n####\nThe general classification given by the dataset:\n####\n{general}\n####\nThe subtitle of the article:\n####\n{subtitle}\n####\nThe supertitle of the article:\n{dachzeile}\n###\nThe rubrik and resort of the article:\n{rubrik}, {resort}\n####\n"}

Data/topic_2-3_articles_2015.csv

@@ -0,0 +1 @@
+{"FALTER_20151111C499C672CF": [{"human": " The following is a news article. All the information is given in german.\nThe title of the article:\n####\nZampano Murdoch verliert Werbeerl\u00f6se\n####\nThe text of the article:\n####\n['Rupert Murdoch z\u00e4hlt zu den wichtigsten Medienmogulen der Welt, sein Imperium reicht vom konservativen Kampagnensender Fox bis hin zu den Qualit\u00e4tsbl\u00e4ttern Wall Street Journal und Times. Aber es l\u00e4uft nicht alles rund f\u00fcr den australischen Medienzampano.', 'Verglichen mit dem Vorjahreswert gingen die Erl\u00f6se seiner Mediengruppe News Corp im dritten Quartal um vier Prozent zur\u00fcck - auf zwei Milliarden Dollar, umgerechnet 1,8 Milliarden Euro. Seit Jahresbeginn ist die Aktie seiner News Corp um 18 Prozent gefallen. Grund f\u00fcr die Misere sind wegbr\u00f6ckelnde Werbeeinnahmen.']\n####\nThe general classification given by the dataset:\n####\nnan\n####\nThe subtitle of the article:\n####\nnan\n####\nThe supertitle of the article:\nMogulpackung Der Medienmarkt\n###\nThe rubrik and resort of the article:\nnan, Medien\n####\n", "ai_gen": "Ich such nach einen Artikel \u00fcber Nachrichtendienste", "ai_spec": "K\u00f6nntest du Artikel \u00fcber die Gesch\u00e4ftsentwicklung von der Mediengruppe News Corp anzeigen"}]}