JobSearch

Brief: Job search assistant
Jump To: Parent Description
Code: GitHub

  • Job search has been greatly extended. It is now scraping jobs off the internet, and ordering them according to relevance to a candidate's profile, albeit inaccurately, so work is moving forward smoothly using MinorThird and the provided job labelled data set, which seems to work extremely well. I believe there is labelled resume data lying around here somewhere. We already have large resume and position copora.

    It already does almost everything else we need it to. Audience is getting to the point where it will not be difficult to write a simple statemachine to proxy all communications and review them. Used Sorcerer to locate a statemachine editor. Resume generation is working but is inaccurate until the task of extraction of position attributes using MinorThird is finished, since resumes are generated per position.

    The task of resume extraction will help in expanding resumes. FieldGoal/CLEAR are used to certify experience and familiarity with some given field in very quantitative terms.

    Here is some old stuff, an example negotiation session taken from Audience conversation training simulator:

    **** Imsorrywecouldntworksomethingouttoday63
    ****************************************
    <history>
    	Human-0	Good, and you?
    	Human-1	I'm doing well.
    	Human-0	Good.  I would like to discuss my pay now.
    	Human-1	Okay, let's move to a different room.
    	Human-0	What is your offer?
    	Human-1	Our offer is X (too low)
    	Human-0	At that rate I can't afford to work here.  Do you have a better offer?
    	Human-1	No
    	Human-0	Let me counter offer.  How about Z? (too high)
    	Human-1	We cannot do that.
    	Human-0	Okay, then what can you do?
    	Human-1	We offer X (same value as before)
    	Human-0	I'm sorry we couldn't work something out today.
    </history>
    Player: Human-1
    State: Imsorrywecouldntworksomethingouttoday63
    0) <Toggle>
    1) <Quit>
    2) <Delete>
    3) <Edit>
    4) <Add-Transposition>
    5) <Backup>
    5
    
    Other working features are we have a corpus of homepages and resumes and a statistical coauthoring system (functionality in coauthor) that helps the user write their homepage/resume using language models trained from the corpus. Resume generation is done using resumexml. There is a module for CLEAR which ensures the users familiarity with the subjects (based on the doctrinal hierarchy) parsed from the job description stored in the potential-job manager.