Path forward with Machine Learning project

Hi Morgan  

For the User Behaviour tracking, I think this is a good idea would the intention to log this anonymously, and/or is this possible?

Thanks!

Ryan

Great. Thanks Ryan . Yes, ideally we would log the data anonymously or users are randomly assigned IDs that can't be traced back to their name and organization. Maintaining user anonymity regarding their search queries is an important component our license agreement with subscribers.

Devaang , let me know when you're available to have a call on this issue (perhaps Ryan and some others from Industrial could join as well). Note that I'd like to keep this separate from the team call on Thursdays.

Thanks,

Morgan 

Hi Devaang and Jitesh ,

Could we please get your comments on the approach for the ML project above. 

Thanks,

Morgan 

Hi Morgan,

Tracking and recording user behavior is an important aspect for the ML approach. This combined with other input elements (SME Tagging, User Interest Areas) will help in returning pin-point results to user searches. Yet, the need for user activity secrecy is paramount to maintain privacy and legal compliance. This scenario puts us in a situation where we are able to grab search strings but not able to track this back to the same user. To address this situation, we would need to construct a complex 'software' workaround (not an ML application) that would first capture and store user behavior as encrypted information which cannot be read (decoded) easily. Yet, we are still trying to determine an appropriate technical approach to address this.

On other thoughts, such a stringent 'No user tracking' condition impedes the working of the ML model and therefore accuracy. In general, all global applications actually "Ask for the User Consent" by explicitly explaining their "Privacy Policy". This covers - what type of data will be captured, how it will be stored, how it will be secured, how it will be used to provide more efficient services (in our case, this is "Searches"), etc. Such a "Consent" can be a part of the "User Profile" page where it becomes mandatory for the user to make a choice when they setup their profile information, areas of interest, etc.

We can talk more about this next Thursday (Dec 5th) maybe, 30 minutes prior to the weekly call. Do let me know what works for you and I shall do my best to accommodate.

Best
Devaang

Hi Devaang ,

Thank you for the note above. Why don't we setup a call next week to discuss, and then we can update the team during the call on Thursday. I have a feeling this will be a longer discussion, and I don't want it to hijack the other priority items that need discussion on Thursday. Are you and Dhrumil available for a call at one of the following times (all Vancouver time)?

Monday at 7:30am Vancouver time
Monday at 8:30pm Vancouver time
Tuesday at 7:00am Vancouver time
Wednesday at 7:30am Vancouver time

Let me know, and I'll send you details for the call.

Also, please note that I understand that 'no user tracking' condition impedes our ability to provide inputs for the ML, but unfortunately that is a strict constraint. Our users are performing research on highly confidential, sensitive matters and it's extremely important that we abide by the terms of our subscription agreements, which require complete anonymity to any search data inputted by the user.

Thanks,

Morgan

Perfect. Thanks Devaang . I'll send you and Dhrumil a calendar invite shortly.

Have a good weekend.

Morgan 

Hello all,

Devaang and Dhrumil , my apologies for cutting off the call this morning. My daughter hadn't seen me in a few days, because of some extra hours at the office this week, and she was a pretty upset when I told her I her I was on a call. In any event, I think we're in alignment on the approach for further developing and implementing the ML model.

Devaang is going to document a more detailed plan in the days ahead; however, the broad approach will be similar to what was discussed above. The front-end search for ISLG will be built using dtSearch as originally planned. However, we will also develop a platform for using search strings inputted by users as inputs for developing and refining the ML model. In addition, Devaang is proposing that we build an interface on the admin site that allows subject matter experts to rank and sort search results as further inputs for the ML model. 

Two important points to emphasize:

Please share any questions, comments or concerns. Otherwise, we'll discuss more when Devaang has completed documenting a detailed plan.

Thanks,

Morgan

Hi Devaang ,

Further to comment above, please let me know when you have the document prepared outlining the plan for integrating the ML input tools into the new ISLG application. We will be starting to shift attention to front-end searches in the New Year, and we'll want to ensure the plan is integrated into that workflow.

Thanks,

Morgan 

Hi Morgan ,

We will submit this by tomorrow eod. As You are aware about that its a complex process and  We have to finalize the steps that we have to take and decide the architecture that we have to follow so our team was planning for that and also they were study about the things that we have to implement to follow our decided approach so by tomorrow our team will be able to provide you the detailed documentation on this.

Let me know in case of any concern.

OK. Thanks for update Dhrumil . Look forward to hearing from you tomorrow.

Morgan

Hi Morgan ,

I have submitted the document to the Devaang   and he will review it and send it. I have called him but He is not available for two days so once he is back on Monday he will review it and send it to you.


Let me know in case of any concerns. 

OK. Sounds good. Thanks for the update,  Dhrumil . 

Morgan 

Hello Dhrumil and Devaang ,

I still haven't heard from anyone about the document above. Could you please provide me with an update. Let's move this process forward.

Thanks,

Morgan

Ok. Great. Thanks Devaang .

Morgan

Hi Devaang ,

Any update on the document above?

Thanks,

Morgan 

Ok,  Devaang . We'll discuss more tomorrow.

Morgan