top of page
  • Writer's pictureAdmin

Catch tasks on mturk using HIT categories

Turk Guru extension/script now has a new mode that helps Amazon Mechanical Turk workers catch tasks using easily understandable categories. This acceptance mode accepts a HIT If the extension feels it is from one of the categories selected in the HIT types list. This mode gives mturk workers more control over their work without depending on keywords. This mode's accuracy isn't 100% as it uses Machine learning and some other issues(e.g., lack of data related to the content of a mturk HIT) that we face while we build our model. However, it will improve overtime as we improve our model and receive inputs from the mturk users.

Below are the steps to using it:-

  1. Click the Turk Guru console button on any mturk page.

  2. In the dialog that appears under the "HIT preferences" card block, pick the HIT acceptance mode with the message "Accept tasks that match the HIT types chosen below." using the select option.

  3. In the same dialog, scroll until you find a card block with the title "HIT Types" In that card, select the type of HITs you wish to work on. Selected chips have a checkmark next to them.

  4. Click save and then click the "Start turking" header button to catch HITs based on the categories selected.

The details on what a HIT from each HIT type requires a worker to do are as follows:-

  1. Annotation: Draw a box around objects.

  2. App Testing: Install an app or visit a website and test the features.

  3. Reading: Read an article, paragraph, essay, story, or scenario and then complete the tasks instructed by the requester.

  4. Audio Listening: Listen to an audio recording and do a miscellaneous task. An example of this type would be a HIT that requires the user to rate the quality of audio.

  5. Audio Transcribe: Type the words spoken in the audio.

  6. Audio Uploading: Record audio. An example of a HIT that matches this type would require the worker to speak words shown on the screen.

  7. Consumer Research: Answer a survey where the requester intends to know about consumers' views for a product. HITs with "consumer" in the title or description have a higher chance of matching this type. Users can see this as a sub-class of opinion type.

  8. Conversation: Chat with the bots. Some other mturk HITs related to evaluating a conversation may also match this type.

  9. Creativity: Do something that involves creativity, like drawing a logo design—significantly fewer HITs on mturk match this category.

  10. Data cleaning: Remove or improve bad quality data. An example of a HIT that matches this category would be a HIT requiring users to correct grammatical mistakes.

  11. Data entry: This type requires the worker to enter text in text fields as instructed by the requester. This type is for HITs that don't match any other category that requires the worker to type text.

  12. Data verification: This Type requires the worker to check or validate the work of another mturk worker.

  13. Decision making: Complete a task that requires the worker to make decisions. HITs that match this type on most occasions have a bonus, calculated by workers' decisions made in the job.

  14. Document file task: Read a pdf, excel, or some other file type and complete tasks as instructed by the requester.

  15. Experiment: Take part in an experiment. HITs with the keyword "experiment" in the title or description have a higher chance of matching this type.

  16. Game: Play a game as part of the HIT. This type doesn't include mturk HITs that require the user to answer a survey related to a game like Counter-Strike.

  17. Group Activity: Complete a task that also involves other workers at the same time. Workers generally need to connect as soon as HIT is accepted to be a part of the activity. Requester Yale Institute of Network & Science typically posts HITs in this category.

  18. Image Classification: Select a label from a list of labels after seeing an image. Requester MLDataLabeler post HITs in this category.

  19. Image Transcribe: Type the text from an image. Shopping receipt tasks don't match this category as there is a separate category for those types of HITs.

  20. Image Uploading: Upload an image as instructed by the requester.

  21. Non-English: Complete a task in a language other than English.

  22. Offline Task: Complete a task that requires going to a location near-by workers' location. An example of a HIT of this type would be when the worker needs to go to an offline store to gather information about a product.

  23. Opinion: Complete a survey that has the keyword "opinion" in the title or description. These HITs are generally re

  24. Qualification HIT: Complete a task that may make them eligible for jobs posted by the requester in the future.

  25. Receipt Transcription: Type details of products on a shopping receipt. Requesters like Panel, 411Richmond post HITs in this category.

  26. Review/Rate: Review, rate, or vote for an entity. This entity could be a product, person, text, etc.

  27. Rules-based QA: Follow the guidelines laid down by the requester and answer the questions strictly following them.

  28. Social media related: complete a task where Social media is involved. Jobs can be ranging from finding the sentiment of a tweet to answering a survey on social media usage. This type is matched before types Study, Experiment, Survey as this type gives more details about a task.

  29. Study: Complete a survey that has the keyword "study" in the title or description.

  30. Survey: Complete a questionnaire that doesn't match the types of Study, Experiment, Decision making, Opinion, Consumer Research.

  31. Text classification: Classify a sentence from the categories shown by the requester.

  32. Text Extraction: Extract a piece of text from a large text that is useful to the requester.

  33. Translation: Translate a video, audio, or text.

  34. Video or Screen Recording: Record a video either through a camera or record their Desktop or mobile screen. The mturk HITs that match this type may sometimes also require the worker to record the audio at the same time. An example of a requester that posts HITs in this type would be UserBob.

  35. Video Transcribe: Transcribe a video.

  36. Video watch: Watch a video and complete a task as instructed by the requester. This type doesn't include HITs that require the user to transcribe a video, as there is a separate type for those tasks.

  37. Web Scraping: Extract data after visiting a website.

  38. Writing based: Write an article, summary, or complete a typing task as instructed by the requester. Writing a review is not included in this type as there is a separate type for it.

433 views0 comments

Recent Posts

See All

Understanding HIT Acceptance modes

Turk Guru offers various modes that alter the filters used after the minimum reward filter, blocked requester's filter and the HIT group filter, filter out newly found HITs. Details of each of the mod

How Turk Guru works?

Turk Guru searches for new tasks(viz. HITs) every n second, where n is the number you decide beforehand. Then in each search, a series of filters filter out these newly found HITs, so you get to work


bottom of page