Skip to content
The Chat Archive
claude-1· Anthropic Arena Human Preference 55K

SYSTEM : Cleanse job titles by: a. Identifying the core job role, reflecting the nature…

  1. User

    SYSTEM : Cleanse job titles by:

    a. Identifying the core job role, reflecting the nature and level of work. b. Standardizing job titles. c. Converting to industry-standard or recognized roles (e.g., engineer, analyst). #Guidelines

    1. Exclude non-essential job attributes (location, salary, benefits, company name, employment type).
    2. Remove advertisement terms (e.g., "apply now!") as they lack relevant job role information.
    3. Expand acronyms to their canonical forms (e.g., "sr" to "senior," "RN" to "registered nurse").
    4. Preserve seniority levels in job titles (e.g., "II" indicates seniority and should be retained).
    5. Retain department/division details if mentioned in the job title (e.g., "Engineer III Consultant-AI-ML Engineering").
    6. Maintain the word order in the clean job title as in the original.
    7. Assume the input is a job title; if not, try to summarize to a job title. For eg : "Earn by driving 250 trips" can be summarized as "Driver". Note while summarizing retain as much relvant information possible.
    8. If summarization is not possible, output: "not a job title" For example: Job Title: Registered Nurse - Operating Room | RN - OR | Chicago, Illinois | $3020.20 Weekly Clean Title: Registered Nurse - Operating Room

    Given a list of job titles provide the output in a json format: Maintain the order of output of clean titles as provided in the input of job titles. The output should be formatted as a JSON instance that conforms to the JSON schema below. Here is the output schema:

    {"title": "CleanTitles", "description": "Pydantic model to for a list of cleaned Job title response.", "type": "object", "properties": {"clean_job_titles": {"title": "Job Titles", "description": "List of clean titles in the order as provided in the input", "type": "array", "items": {"type": "string"}}}, "required": ["job_titles"]}
    

    JOB TITLES : ['Delivery Van Driver - At least 99.25/hour′,′Verizon−RetailSalesRepresentative−99.25/hour', 'Verizon - Retail Sales Representative - 43,000 - 56,000/Annually′,′SeniorOracleEPBCSTechnicalProductEngineer′,′Verizon−RetailSales/CustomerService−56,000/Annually', 'Senior Oracle EPBCS Technical Product Engineer', 'Verizon - Retail Sales/Customer Service - 43,000 - 56,000/Annually′,′CategorySourcingManager−OEMDevices′,′Supervisor−RevenueAssurance′,′EngrIIICslt−DataScience′,′Cslt−HRBP′,′SrClientExec−BusinessSales′,′Mgr−BusChnlMgmt′,′Saveupthisfall−Earnatleast56,000/Annually', 'Category Sourcing Manager - OEM Devices', 'Supervisor - Revenue Assurance', 'Engr III Cslt-Data Science', 'Cslt-HRBP', 'Sr Client Exec-Business Sales', 'Mgr-Bus Chnl Mgmt', 'Save up this fall - Earn at least 910 in your first 80 trips.', 'Drive with Uber - earn 3358driving50trips′,′3358 driving 50 trips', '13280 - Earn by driving 290 trips with Uber', 'Socio conductor - Genera hasta 17940conlaappdeUber′,′Driver−Earn17940 con la app de Uber', 'Driver - Earn 22110 with Uber', 'Part-time gig: Earn at least $1430 in your first 120 trips.' ]

    Generate output only as per the required JSON output. No other text except the json output should be generated.

    Assistant:

  2. Assistant

    {"clean_job_titles": ["Delivery Van Driver", "Retail Sales Representative", "Senior Oracle EPBCS Technical Product Engineer", "Retail Sales/Customer Service","Category Sourcing Manager", "Supervisor - Revenue Assurance","Engineer III Consultant-Data Science","Consultant-Human Resources Business Partner","Senior Client Executive-Business Sales","Manager-Business Channel Management","Driver","Driver","Driver","Driver","Driver","Driver"]}

Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.