anonymoush823d23d commited on
Commit
9cd83f6
·
1 Parent(s): 523096d

update dataset

Browse files
Files changed (3) hide show
  1. README.md +11 -0
  2. mapping_futures_exchanges.xlsx +0 -0
  3. prompt.txt +27 -0
README.md ADDED
@@ -0,0 +1,11 @@
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ ---
4
+
5
+ # FUTEXT: A Bilingual Dataset and Benchmark for Event-Driven Futures Price Forecasting
6
+
7
+ ## About This Repo
8
+
9
+ This repository includes: `FUTEXT-Base` and `FUTEXT-Struct`.
10
+
11
+ We also provide `mapping_futures_exchanges.xlsx` and Large Language Model (LLM) prompts used for generating structured text, which can be found in `prompt.txt` .
mapping_futures_exchanges.xlsx ADDED
Binary file (10.4 kB). View file
 
prompt.txt ADDED
@@ -0,0 +1,27 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Please read the provided title and news article above. Extract one or a few key events that have a major impact on global futures market prices, and provide any analyst opinions on these events. Also, extract the summary content from the news, if available. Each event description must be concise, within approximately 10 words, and specify the relevant futures commodities.
2
+ Output Format: JSON
3
+ {
4
+ "Summary": {
5
+ "EventSummary": "Objective summary content from the news article. If no summary is present, output an empty string ("").", "OpinionSummary": "Subjective opinion summary content from the news article. If no summary is present, output an empty string ("")."
6
+ },
7
+ "Events": {
8
+ "1": {
9
+ "EventOccurrence": ["Event Description", "Event Category", "Occurred/Not Occurred"],
10
+ "AnalystOpinion": "Description of the analyst's opinion. If no opinion is present, output an empty string ("").",
11
+ "Futures": ["Commodity1", "Commodity2"]
12
+ },
13
+ "2": { ... }
14
+ }
15
+ }
16
+ Event Category: Supply and Demand, Macroeconomics, Policy and Regulation, Weather and Natural Disasters, International Markets, Market Sentiment, Geopolitics, Technical Analysis
17
+ Futures Commodities: Canola, Cocoa, Coffee, Corn, Cotton, Feeder Cattle, Live Cattle, Lean Hogs, Palm Oil, Soybean, Soybean Meal, Soybean Oil, Sugar, Wheat, Aluminium, Copper, Gold, Lead, Nickel, Palladium, Platinum, Silver, Tin, Zinc, Brent, Gasoil, Heating Oil, RBOB Gasoline, Natural Gas, WTI.
18
+ Content Requirements:
19
+ 1.News Summary: Only output summary content that is explicitly stated in the news, such as sections starting with "In summary" or "Conclusion". If no such summary exists, output an empty string "".
20
+ 2.EventOccurrence: The event description must be an objective fact with a causal relationship to futures price changes. For example: "Mine shutdown" or "Increased purchasing intent from steel mills".
21
+ 3.Occurred/Not Occurred: Determine the timeliness of the event ("Occurred" or "Not Occurred") based on the news description.
22
+ 4.AnalystOpinion: Only describe the views and expected impacts from analysts. If no analyst opinion is mentioned, output an empty string "".
23
+ 5.Futures: List the futures commodities directly related to the event. If none are relevant, output an empty string "".
24
+ Special Attention:
25
+ 1.Distinguish Fact vs. Opinion: Ensure the EventOccurrence section describes objective facts, while the AnalystOpinion section only contains subjective views from analysts. Avoid mixing them.
26
+ 2.Do Not Fabricate Opinions: If the news does not contain views from analysts or institutions, the model must output an empty string "". All output must be strictly based on the provided news content.
27
+ Do Not Summarize Independently: Only extract pre-existing summary content from the news. Do not generate your own summary or make inferences.