Upload 7 files

app.py

@@ -0,0 +1,146 @@
+import streamlit as st
+import os
+import frontmatter
+import json
+# import pyperclip
+
+def load_prompts():
+    prompts = []
+    prompt_dir = "prompt"
+
+    for filename in os.listdir(prompt_dir):
+        if filename.endswith(".md"):
+            with open(os.path.join(prompt_dir, filename), "r") as file:
+                prompt = frontmatter.load(file)
+                prompts.append({
+                    "id": prompt.get("id"),
+                    "title": prompt.get("title"),
+                    "author": prompt.get("author"),
+                    "date": prompt.get("date"),
+                    "language": prompt.get("language"),
+                    "task": prompt.get("task"),
+                    "version": prompt.get("version"),
+                    "addedby": prompt.get("addedby"),
+                    "keywords": prompt.get("keywords"),
+                    "content": prompt.content
+                })
+
+    return prompts
+
+def render_prompt_details(prompt):
+    st.markdown(f"# {prompt['title']}")
+    st.write("---")
+    st.write(f"**Author:** {prompt['author']}")
+    st.write(f"**Date:** {prompt['date']}")
+    st.write(f"**Language:** {prompt['language']}")
+    st.write(f"**Task:** {prompt['task']}")
+    st.write("---")
+    st.write(prompt["content"])
+
+def render_prompt_card(prompt):
+    card_style = """
+    <style>
+        .card {
+            background-color: #f8f9fa;
+            border: 1px solid #d9d9d9;
+            border-radius: 5px;
+            padding: 15px;
+            margin-bottom: 10px;
+            box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
+            display: flex;
+            justify-content: space-between;
+            align-items: center;
+        }
+        .card-content {
+            display: flex;
+            flex-direction: column;
+        }
+        .button-style {
+            background-color: #FFD700;
+            border: none;
+            color: white;
+            text-align: center;
+            text-decoration: none;
+            display: inline-block;
+            font-size: 16px;
+            margin: 4px 2px;
+            cursor: pointer;
+            border-radius: 4px;
+        }
+    </style>
+    """
+
+    st.markdown(card_style, unsafe_allow_html=True)
+
+    card_container = st.container()
+
+    with card_container:
+        with st.container():
+            col1, col2 = st.columns([4, 1])
+
+            with col1:
+                st.write(f"### **Card:** {prompt['title']}")
+                st.write(f"**Prompt Language:** {prompt['language']}")
+                st.write(f"**Annotation Task:** {prompt['task']}")
+                st.write(f"**Version:** {prompt['version']}")
+                st.write(f"**Keywords:** {prompt['keywords']}")
+
+            with col2:
+                if st.button("View details", key=f"view-details-{prompt['id']}"):
+                    st.session_state.selected_prompt_id = prompt["id"]
+                    st.experimental_rerun()
+
+            st.markdown("<style>.button-style {background-color: #FFD700;}</style>", unsafe_allow_html=True)
+
+    card_container.container().write("---")
+
+def main():
+    if "selected_prompt_id" not in st.session_state:
+        st.session_state.selected_prompt_id = None
+
+    st.title("🏷️ Annotation PromptCards Archive")
+    st.write("Welcome to the Prompt Archive! Click on the 'View details' button on each prompt card to see more information about the annotation prompt.")
+    st.write("---")
+
+    prompts = load_prompts()
+    language_list = list(set([prompt['language'] for prompt in prompts]))
+    task_list = list(set([prompt['task'] for prompt in prompts]))
+
+    st.sidebar.header("**Annotation PromptCards Archive**")
+
+    st.sidebar.write("A collection of prompts for annotation tasks in NLP. This is a work in progress. Please contribute your prompts via GitHub [[Upload]](https://github.com/chkla/PromptCards).")
+
+    # add a link to the GitHub repository
+    st.sidebar.write("---")
+    st.sidebar.write(f"**Total number of prompts:** {len(prompts)}")
+    st.sidebar.write("---")
+
+    st.sidebar.header("🧑🏽‍🚀 Explore:")
+    search = st.sidebar.text_input("Search by title")
+    language_filter = st.sidebar.selectbox("Filter by Language", ["All"] + language_list)
+    task_filter = st.sidebar.selectbox("Filter by Task", ["All"] + task_list)
+
+    if st.sidebar.button("Back to list"):
+        st.session_state.selected_prompt_id = None
+        st.experimental_rerun()
+
+    if st.session_state.selected_prompt_id is None:
+        filtered_prompts = [
+            prompt for prompt in prompts
+            if (search.lower() in prompt['title'].lower() or not search)
+            and (prompt['language'] == language_filter or language_filter == "All")
+            and (prompt['task'] == task_filter or task_filter == "All")
+        ]
+
+        for prompt in filtered_prompts:
+            render_prompt_card(prompt)
+    else:
+        prompt = next((p for p in prompts if p["id"] == st.session_state.selected_prompt_id), None)
+        if prompt:
+            render_prompt_details(prompt)
+
+    st.sidebar.write("---")
+    st.sidebar.write("Made with ❤️ and 🤖 by [Chkla](https://chkla.github.io/).")
+
+if __name__ == "__main__":
+    main()

prompt/gilardi-frame-problem-solution-2023.md

@@ -0,0 +1,72 @@
+---
+id: problem-solution-en-gilardi2023
+title: Problem/ Solution Frames
+author: Gilardi et al. 2023
+date: 11.4.2023
+language: en
+task: frames
+version: 1.0
+addedby: chkla
+keywords: frames, content moderation
+---
+
+## Prompt Description
+
+[Briefly describe the purpose of the prompt and the context in which it is intended to be used, especially in the context of artificial annotation with generative models.]
+
+## Prompt Text
+
+Content moderation can be seen from two different perspectives:<br>
+• Content moderation can be seen as a PROBLEM; for example, as a restriction of free speech<br>
+• Content moderation can be seen as a SOLUTION; for example, as a protection from harmful speech<br>
+<br>
+For each tweet in the sample, follow these instructions:<br>
+1. Carefully read the text of the tweet, paying close attention to details.<br>
+2. Classify the tweet as describing content moderation as a problem, as a solution, or neither.<br>
+<br>
+Tweets should be classified as describing content moderation as a PROBLEM if they emphasize negative effects of content moderation, such as restrictions to free speech, or the biases that can emerge from decisions regarding what users are allowed to post.<br>
+<br>
+Tweets should be classified as describing content moderation as a SOLUTION if they emphasize positive effects of content moderation, such as protecting users from various kinds of harmful content, including hate speech, misinformation, illegal adult content, or spam.<br>
+<br>
+Tweets should be classified as describing content moderation as NEUTRAL if they do not emphasize possible negative or positive effects of content moderation, for example if they simply report on the content moderation activity of social media platforms without linking them to potential advantages or disadvantages for users or stakeholders.
+
+## Language
+
+- Prompt Language: [Specify the language of the prompt, e.g., English]
+- Dataset Language: [Specify the language of the dataset to which the prompt is applied, e.g., English]
+
+## NLP Task
+
+- Task: [Specify the NLP task in more detail, e.g., sentiment analysis, named entity recognition, summarization]
+- Subtask: [If applicable, provide any subtask or variation related to the main NLP task, e.g., binary sentiment classification, multi-class sentiment classification]
+
+## Example Input and Output
+
+- Example 1
+  - Input: [Provide an example input for the prompt]
+  - Output: [Provide an example output for the given input]
+- Example 2
+  - Input: [Provide another example input for the prompt]
+  - Output: [Provide another example output for the given input]
+
+## Parameters and Constraints
+
+- Parameter 1: [Specify any parameters, such as temperature or token count]
+- Parameter 2: [Specify additional parameters or constraints if applicable]
+
+## Evaluation Metrics
+
+[List the evaluation metrics used to assess the quality of the generated artificial annotations, such as accuracy, F1 score, or BLEU score.]
+
+## Use Cases
+
+[List any specific use cases or applications for the prompt in artificial annotation, such as data annotation, semi-supervised learning, or active learning.]
+
+## Limitations and Potential Biases
+
+[Briefly discuss any limitations or potential biases associated with the prompt, as well as any steps taken to mitigate them, in the context of artificial annotation with generative models.]
+
+## Related Research and References
+
+[List any relevant research papers, articles, or resources that informed the creation of the prompt or are closely related to it, especially in the area of artificial annotation with generative models. Include proper citations where applicable.]
+

prompt/gilardi-moderation-relevance-2023.md

@@ -0,0 +1,65 @@
+---
+id: moderation-relevance-en-gilardi2023
+title: Content Moderation Relevance
+author: Gilardi et al. 2023
+date: 11.4.2023
+language: en
+task: relevance
+version: 1.0
+addedby: chkla
+keywords: content moderation, relevance
+---
+
+## Prompt Description
+
+[Briefly describe the purpose of the prompt and the context in which it is intended to be used, especially in the context of artificial annotation with generative models.]
+
+## Prompt Text
+For each tweet in the sample, follow these instructions:<br>
+1. Carefully read the text of the tweet, paying close attention to details.<br>
+2. Carefully read the text of the tweet, paying close attention to details.<br>
+<br>
+Tweets should be coded as RELEVANT when they directly relate to content moderation, as defined above. This includes tweets that discuss: social media platforms’ content moderation rules and practices, governments’ regulation of online content moderation, and/or mild forms of content moderation like flagging.<br>
+<br>
+Tweets should be coded as IRRELEVANT if they do not refer to content moderation, as defined above, or if they are themselves examples of moderated content. This would include, for example, a Tweet by Donald Trump that Twitter has labeled as “disputed”, a tweet claiming that something is false, or a tweet containing sensitive content. Such tweets might be subject to content moderation, but are not discussing content moderation. Therefore, they should be coded as irrelevant for our purposes.
+
+## Language
+
+- Prompt Language: [Specify the language of the prompt, e.g., English]
+- Dataset Language: [Specify the language of the dataset to which the prompt is applied, e.g., English]
+
+## NLP Task
+
+- Task: [Specify the NLP task in more detail, e.g., sentiment analysis, named entity recognition, summarization]
+- Subtask: [If applicable, provide any subtask or variation related to the main NLP task, e.g., binary sentiment classification, multi-class sentiment classification]
+
+## Example Input and Output
+
+- Example 1
+  - Input: [Provide an example input for the prompt]
+  - Output: [Provide an example output for the given input]
+- Example 2
+  - Input: [Provide another example input for the prompt]
+  - Output: [Provide another example output for the given input]
+
+## Parameters and Constraints
+
+- Parameter 1: [Specify any parameters, such as temperature or token count]
+- Parameter 2: [Specify additional parameters or constraints if applicable]
+
+## Evaluation Metrics
+
+[List the evaluation metrics used to assess the quality of the generated artificial annotations, such as accuracy, F1 score, or BLEU score.]
+
+## Use Cases
+
+[List any specific use cases or applications for the prompt in artificial annotation, such as data annotation, semi-supervised learning, or active learning.]
+
+## Limitations and Potential Biases
+
+[Briefly discuss any limitations or potential biases associated with the prompt, as well as any steps taken to mitigate them, in the context of artificial annotation with generative models.]
+
+## Related Research and References
+
+[List any relevant research papers, articles, or resources that informed the creation of the prompt or are closely related to it, especially in the area of artificial annotation with generative models. Include proper citations where applicable.]
+

prompt/gilardi-policy-frames-2023.md

@@ -0,0 +1,79 @@
+---
+id: policy-frames-en-gilardi2023
+title: Policy Frames Classification
+author: Gilardi et al. 2023
+date: 11.4.2023
+language: en
+task: frames
+version: 1.0
+addedby: chkla
+keywords: policy frames, content moderation
+---
+
+## Prompt Description
+
+[Briefly describe the purpose of the prompt and the context in which it is intended to be used, especially in the context of artificial annotation with generative models.]
+
+## Prompt Text
+
+Content moderation, as described above, can be linked to various other topics, such as health, crime, or equality.
+For each tweet in the sample, follow these instructions:<br>
+1. Carefully read the text of the tweet, paying close attention to details. 2. Classify the tweet into one of the topics defined below.<br>
+<br>
+The topics are defined as follows:<br>
+* ECONOMY: The costs, benefits, or monetary/financial implications of the issue (to an individual, family, community, or to the economy as a whole).<br>
+* Capacity and resources: The lack of or availability of physical, geographical, spatial, human, and financial resources, or the capacity of existing systems and resources to implement or carry out policy goals.<br>
+* MORALITY: Any perspective—or policy objective or action (including proposed action)that is compelled by religious doctrine or interpretation, duty, honor, righ- teousness or any other sense of ethics or social responsibility.<br>
+* FAIRNESS AND EQUALITY: Equality or inequality with which laws, punishment, rewards, and resources are applied or distributed among individuals or groups. Also the balance between the rights or interests of one individual or group compared to another individual or group.<br>
+* CONSTITUTIONALITY AND JURISPRUDENCE: The constraints imposed on or freedoms granted to individuals, government, and corporations via the Constitution, Bill of Rights and other amendments, or judicial interpretation. This deals specifically with the authority of government to regulate, and the authority of in- dividuals/corporations to act independently of government.<br>
+* POLICY PRESCRIPTION AND EVALUATION: Particular policies proposed for addressing an identified problem, and figuring out if certain policies will work, or if existing policies are effective.<br>
+* LAW AND ORDER, CRIME AND JUSTICE: Specific policies in practice and their enforcement, incentives, and implications. Includes stories about enforcement and interpretation of laws by individuals and law enforcement, breaking laws, loopholes, fines, sentencing and punishment. Increases or reductions in crime.<br>
+* SECURITY AND DEFENSE: Security, threats to security, and protection of one’s person, family, in-group, nation, etc. Generally an action or a call to action that can be taken to protect the welfare of a person, group, nation sometimes from a not yet manifested threat.<br>
+* HEALTH AND SAFETY: Health care access and effectiveness, illness, disease, sanitation, obesity, mental health effects, prevention of or perpetuation of gun violence, infrastructure and building safety.<br>
+* QUALITY OF LIFE: The effects of a policy on individuals’ wealth, mobility, access to resources, happiness, social structures, ease of day-to-day routines, quality of community life, etc.<br>
+* CULTURAL IDENTITY: The social norms, trends, values and customs constituting culture(s), as they relate to a specific policy issue.<br>
+* PUBLIC OPINION: References to general social attitudes, polling and demographic information, as well as implied or actual consequences of diverging from or “getting ahead of” public opinion or polls.<br>
+* POLITICAL: Any political considerations surrounding an issue. Issue actions or efforts or stances that are political, such as partisan filibusters, lobbyist involvement, bipartisan efforts, deal-making and vote trading, appealing to one’s base, mentions of political maneuvering. Explicit statements that a policy issue is good or bad for a particular political party.<br>
+* EXTERNAL REGULATION AND REPUTATION: The United States’ external relations with another nation; the external relations of one state with another; or relations between groups. This includes trade agreements and outcomes, comparisons of policy outcomes or desired policy outcomes.<br>
+* OTHER: Any topic that does not fit into the above categories.<br>
+
+## Language
+
+- Prompt Language: [Specify the language of the prompt, e.g., English]
+- Dataset Language: [Specify the language of the dataset to which the prompt is applied, e.g., English]
+
+## NLP Task
+
+- Task: [Specify the NLP task in more detail, e.g., sentiment analysis, named entity recognition, summarization]
+- Subtask: [If applicable, provide any subtask or variation related to the main NLP task, e.g., binary sentiment classification, multi-class sentiment classification]
+
+## Example Input and Output
+
+- Example 1
+  - Input: [Provide an example input for the prompt]
+  - Output: [Provide an example output for the given input]
+- Example 2
+  - Input: [Provide another example input for the prompt]
+  - Output: [Provide another example output for the given input]
+
+## Parameters and Constraints
+
+- Parameter 1: [Specify any parameters, such as temperature or token count]
+- Parameter 2: [Specify additional parameters or constraints if applicable]
+
+## Evaluation Metrics
+
+[List the evaluation metrics used to assess the quality of the generated artificial annotations, such as accuracy, F1 score, or BLEU score.]
+
+## Use Cases
+
+[List any specific use cases or applications for the prompt in artificial annotation, such as data annotation, semi-supervised learning, or active learning.]
+
+## Limitations and Potential Biases
+
+[Briefly discuss any limitations or potential biases associated with the prompt, as well as any steps taken to mitigate them, in the context of artificial annotation with generative models.]
+
+## Related Research and References
+
+[List any relevant research papers, articles, or resources that informed the creation of the prompt or are closely related to it, especially in the area of artificial annotation with generative models. Include proper citations where applicable.]
+

prompt/gilardi-stance-detection-2023.md

@@ -0,0 +1,64 @@
+---
+id: stance-detectionen-gilardi2023
+title: Stance Detection
+author: Gilardi et al. 2023
+date: 11.4.2023
+language: en
+task: stance
+version: 1.0
+addedby: chkla
+keywords: stance, content moderation
+---
+
+## Prompt Description
+
+[Briefly describe the purpose of the prompt and the context in which it is intended to be used, especially in the context of artificial annotation with generative models.]
+
+## Prompt Text
+
+In the context of content moderation, Section 230 is a law in the United States that protects websites and other online platforms from being held legally responsible for the content posted by their users. This means that if someone posts something illegal or harmful on a website, the website itself cannot be sued for allowing it to be posted. However, websites can still choose to moderate content and remove anything that violates their own policies.<br>
+<br>
+For each tweet in the sample, follow these instructions:<br>
+1. Carefully read the text of the tweet, paying close attention to details.<br>
+2. Classify the tweet as having a positive stance towards Section 230, a negative stance, or a neutral stance.<br>
+
+## Language
+
+- Prompt Language: [Specify the language of the prompt, e.g., English]
+- Dataset Language: [Specify the language of the dataset to which the prompt is applied, e.g., English]
+
+## NLP Task
+
+- Task: [Specify the NLP task in more detail, e.g., sentiment analysis, named entity recognition, summarization]
+- Subtask: [If applicable, provide any subtask or variation related to the main NLP task, e.g., binary sentiment classification, multi-class sentiment classification]
+
+## Example Input and Output
+
+- Example 1
+  - Input: [Provide an example input for the prompt]
+  - Output: [Provide an example output for the given input]
+- Example 2
+  - Input: [Provide another example input for the prompt]
+  - Output: [Provide another example output for the given input]
+
+## Parameters and Constraints
+
+- Parameter 1: [Specify any parameters, such as temperature or token count]
+- Parameter 2: [Specify additional parameters or constraints if applicable]
+
+## Evaluation Metrics
+
+[List the evaluation metrics used to assess the quality of the generated artificial annotations, such as accuracy, F1 score, or BLEU score.]
+
+## Use Cases
+
+[List any specific use cases or applications for the prompt in artificial annotation, such as data annotation, semi-supervised learning, or active learning.]
+
+## Limitations and Potential Biases
+
+[Briefly discuss any limitations or potential biases associated with the prompt, as well as any steps taken to mitigate them, in the context of artificial annotation with generative models.]
+
+## Related Research and References
+
+[List any relevant research papers, articles, or resources that informed the creation of the prompt or are closely related to it, especially in the area of artificial annotation with generative models. Include proper citations where applicable.]
+

prompt/gilardi-topic-detection-2023.md

@@ -0,0 +1,70 @@
+---
+id: topic-detection-gilardi2023
+title: Topic Detection
+author: Gilardi et al. 2023
+date: 11.4.2023
+language: en
+task: topic detection
+version: 1.0
+addedby: chkla
+keywords: content moderation, topic detection
+---
+
+## Prompt Description
+
+[Briefly describe the purpose of the prompt and the context in which it is intended to be used, especially in the context of artificial annotation with generative models.]
+
+## Prompt Text
+
+Tweets about content moderation may also discuss other related topics, such as:<br>
+1. Section 230, which is a law in the United States that protects websites and other online platforms from being held legally responsible for the content posted by their users (SECTION 230).<br>
+2. The decision by many social media platforms, such as Twitter and Facebook, to suspend Donald Trump’s account (TRUMP BAN).
+3. Requests directed to Twitter’s support account or help center (TWITTER SUPPORT).<br>
+4. Social media platforms’ policies and practices, such as community guidelines or terms of service (PLATFORM POLICIES).
+5. Complaints about platform’s policy and practices in deplatforming and content moderation or suggestions to suspend particular accounts, or complaints about accounts being suspended or reported (COMPLAINTS).<br>
+6. If a text is not about the SECTION 230, COMPLAINTS, TRUMP BAN, TWIT- TER SUPPORT, and PLATFORM POLICIES, then it should be classified in OTHER class (OTHER).<br>
+<br>
+For each tweet in the sample, follow these instructions:<br>
+1. Carefully read the text of the tweet, paying close attention to details.<br>
+2. Please classify the following text according to topic (defined by function of the text, author’s purpose and form of the text). You can choose from the following classes: SECTION 230, TRUMP BAN, COMPLAINTS, TWITTER SUPPORT, PLATFORM POLICIES, and OTHER<br>
+
+## Language
+
+- Prompt Language: [Specify the language of the prompt, e.g., English]
+- Dataset Language: [Specify the language of the dataset to which the prompt is applied, e.g., English]
+
+## NLP Task
+
+- Task: [Specify the NLP task in more detail, e.g., sentiment analysis, named entity recognition, summarization]
+- Subtask: [If applicable, provide any subtask or variation related to the main NLP task, e.g., binary sentiment classification, multi-class sentiment classification]
+
+## Example Input and Output
+
+- Example 1
+  - Input: [Provide an example input for the prompt]
+  - Output: [Provide an example output for the given input]
+- Example 2
+  - Input: [Provide another example input for the prompt]
+  - Output: [Provide another example output for the given input]
+
+## Parameters and Constraints
+
+- Parameter 1: [Specify any parameters, such as temperature or token count]
+- Parameter 2: [Specify additional parameters or constraints if applicable]
+
+## Evaluation Metrics
+
+[List the evaluation metrics used to assess the quality of the generated artificial annotations, such as accuracy, F1 score, or BLEU score.]
+
+## Use Cases
+
+[List any specific use cases or applications for the prompt in artificial annotation, such as data annotation, semi-supervised learning, or active learning.]
+
+## Limitations and Potential Biases
+
+[Briefly discuss any limitations or potential biases associated with the prompt, as well as any steps taken to mitigate them, in the context of artificial annotation with generative models.]
+
+## Related Research and References
+
+[List any relevant research papers, articles, or resources that informed the creation of the prompt or are closely related to it, especially in the area of artificial annotation with generative models. Include proper citations where applicable.]
+

requirements.txt

@@ -0,0 +1,3 @@
+streamlit
+frontmatter
+json