TASSY— A Text Annotation Survey System

TASSY is a web-based survey application built using Flask and Vue.js. The app allows combining classic survey functionality, such as posing single-choice, multiple-choice, and slider questions, with text annotation functionality, i.e., allowing participants to select and annotate words or phrases in a provided text.

Demo

http://tassy.blind-review.org

Note: In some browsers, disabling third-party cookies prevents the applications from runnig. Please enable cookies if you experience problems.

To run the application locally

1. Resolve all dependencies in requirements.txt.
2. Set up a local flask server.
3. Add this line app.run(0.0.0.0) at the end of the file flask_app.py.
4. Configure a local MySQL or SQLite server and make the appropriate changes in surveyapi.models.db to get the database up and running.

Application structure and hierarchies

• Models for basic survey questions, such as single-choice and multiple-choice questions, chips select, range sliders, and text highlights already exist in surveyapi.models
• To create new models, please perform the appropriate DB migrations as per the configuration of your flask and database servers.
• The following models (and corresponding tables) exist in the survey and can be used out-of-the-box:

1. SurveyRecord: Holds the parent object generated from a perticipant’s responses, such as DataConsent, Demographics, Ideologies and QualityControl.
2. SurveyGroups: Groups the sentences that are to be annotated. It also holds the max quota variable of the groups.
3. Survey: Container for various types of surveys, such as demographics, ideologies etc.
4. Question: Holds a survey question. It can be specified with the type of response the question should hold.
5. AnnotationSentences: Actual sentences that are to be annotated. Also holds other properties, such as publication, URLs, etc.
6. Annotations: Holds the words/phrases highlighted/annotated by the participants.

File seeder.py serves to seed and populate the platform database with questions etc.

Creating a simple demographic question with single choice select

Run the file seeder.py with these lines of code.

survey = Survey(name="demographic_questions")
questions = []

# 1
question_1 = Question(text='What is your gender?', type='radio', name='gender')
question_1.simple_choices = [
    SimpleChoice(text='Female'),
    SimpleChoice(text='Male'),
    SimpleChoice(text='Other/Prefer not to say')
]
questions.append(question_1)

survey.questions = questions
db.session.add(survey)
db.session.commit()

Creating a simple ideological question with a range slider input

Run the file seeder.py with these lines of code.

survey = Survey(name="ideology_questions")
ideology_questions = []
question_1 = Question(
    text='Do you consider yourself to be liberal, conservative, or somewhere in between?',
    type='range_slider',
    name='political_ideology'
)
question_1.range_slider_choices = [
    RangeSliderChoice(
        min_range=-10,
        max_range=10,
        label_left_side='Very liberal',
        label_right_side='Very conservative'
    )
]
ideology_questions.append(question_1)
survey.questions = ideology_questions
db.session.add(survey)
db.session.commit()

Seeding annotation sentences

In the file seeder.py, use the function seed_annotation_sentences().

    app = create_app()
    ctx = app.app_context()
    ctx.push()
    db.init_app(app)
    with open('sentences_grouped_reshuffled.csv') as f:
        reader = csv.DictReader(f)
        for row in reader:
            annotation_sentence = AnnotationSentences(
                id=uuid.uuid4().hex,
                text=row['sentence'],
                link=row['news_link'],
                type=row['type'],
                topic=row['topic'],
                outlet=row['outlet'],
                group_id=int(row['group_id'])
            )
            db.session.add(annotation_sentence)
            db.session.commit()
    ctx.pop()

seed_annotation_sentences()

The function expects a CSV file such as the following https://github.com/Media-Bias-Analysis-Group/SAQ_text_annotation_and_question_survey_tool/blob/master/sentences_grouped_reshuffled.csv

Seeding survey questions

In the file seeder.py, use the functions

1. seed_personal_questions()
2. seed_ideology_questions()
3. seed_annotation_questions()
4. create_groups()

Extracting the survey output

The output is extracted in CSV format. Use the command below to open the interactive python shell.

python -i db_helper.py

Get the db db, ctx = create_db_client()

Get all the survey worker records as CSV to_csv_survey_worker_records(db, ctx, after_date=optional)

Get all the survey annotations as CSV to_csv_all_annotations(db, ctx, after_date=optional)

---

Feel free to come forward with any questions or extend/improve the tool. Just contact us.