Hi Paul ,

Responding to your questions via Zoom: "Hey Morgan, my understanding is that Contegra is supposed to auto suggest a few options based on what the user is typing on SN and Full Text Search (similar to Dispute Documents tab). Am I right? I am testing on dev.islg and do not see any suggestions so far."

Yes, please do all the testing on dev.islg, and only in the Full Text Search. I suggest clearing your cache. This might help getting the search working. The team is still working on the Subject Navigator search. 

Contegra Auto-Suggest is suppose to suggest terms for the user based on what the user inputs into the search field (similar to Dispute Documents tab). However, unlike the Dispute Documents tab, Contegra's suggested results are based on machine learning tools that are suppose to generate suggest search terms based on their relevancy to content being searched. For example, if you search "good faith", I would expect suggested search terms like "abuse or process" and "estoppel" to appear as suggested terms, because these concepts would appears in the same context as "good faith". However, we'll see how effective these tools are, and whether they provide useful suggestions to users.

Please report your findings here, so the team can follow and comment on the results.

Thanks,

Morgan

Hi everyone:

As Morgan suggested above, I searched "good faith" under Full Text Search on dev.islg, but I do not get any Contegra Auto-Suggest. I cleared cache before I conducted this search using normal Chrome.
https://dev.investorstatelawguide.com/ResearchTools/FullTextSearch?toc=content&id=97&tab=r

image.png 84.9 KB • Download

However, if I do the same search on dev.islg on Chrome incognito, I get several suggestions. Nevertheless, it is not suggesting expected topics mentioned above like "abuse of process" or "estoppel." 
http://dev.investorstatelawguide.com/ResearchTools/FullTextSearch?toc=content&id=97&tab=r

image.png 126 KB • Download

Another example is searching "legitimate expectations." If Contegra "learned" the connection between legitimate expectations and fair and equitable treatment (they are closely related), I would expect one of the suggested terms/phrases to be "fair and equitable treatment" or "FET," but I do not see it in the example below. So far, it appears the Auto-Suggest is strictly based on terms typed in, not machine-learned connection.
http://dev.investorstatelawguide.com/ResearchTools/FullTextSearch?toc=content&id=97&tab=r

image.png 132 KB • Download

If the testing should be done differently, please let me know.

Thanks,

Paul

Hi Paul ,

Thanks for the testing above. I'm not sure why it wasn't working without incognito enabled. I just ran a couple of tests, and it's working fine for me. However, I noticed some other issues that I'll need to take a closer look at when I'm back in the office.

Ryan , let us know if you find any discrepancies. 

Thanks,

Morgan 

Hi Morgan and Paul  

My apologies for the delayed response. This is on my task list for this morning.

Thanks!

Ryan

Hi Morgan and Paul  

I'm seeing results when searching in Chrome normal mode, but noticed the following:

I'm also unsure as to how the relationship learning works as Paul pointed out in his last point.

Thanks!

Ryan

Hi Ryan ​ and Morgan ​,

Currently, we have set contegra auto suggestion only fresh full text search page. we have not integrated in result page.

If all will be Ok in fresh full text search page then will do same thing in result page.

Hi Ryan :

If the URL you indicated above is the one you are testing on (live site), I believe testing should be done on dev.islg.

I am still having trouble getting results on normal Chrome on both desktop and laptop, but maybe it is my browser setting. Incognito mode, however, is working for me.

Another trouble that I am having is some suggested topics as shown below. The last suggestion reads "fair links er2 expert report." I suspect it may be picking up some formatting issues that are not visible on the live site. Also, when I try to type "Fair and Equitable Treatment," nothing shows up. Since Fair and Equitable Treatment is a major topic, I would expect Contegra to suggest it. 

image.png 86.2 KB • Download

Re: relationship learning, it may be something that is not fully implemented yet, but Legitimate Expectations and Fair and Equitable Treatment are closely related topics that should be suggested (see the cross-reference in Subject Navigator below)
http://dev.investorstatelawguide.com/ResearchTools/SubjectNavigator?toc=content&id=50&kwList=27105,20913,&exList=,&search=&ci=20913&searchBranchLevel=#20913

image.png 142 KB • Download

Thanks,

Paul

Hi Paul  

Thanks for pointing that out - I've edited my post with the proper dev.islg URL. I was testing on dev as well, but had copied the wrong URL.

Ryan

Hi Paul ,

Another trouble that I am having is some suggested topics as shown below. The last suggestion reads "fair links er2 expert report." I suspect it may be picking up some formatting issues that are not visible on the live site.

it is visible on the live site also see below image.

live-site.png 63.5 KB • Download

Also, when I try to type "Fair and Equitable Treatment," nothing shows up. Since Fair and Equitable Treatment is a major topic, I would expect Contegra to suggest it. 

Contegra auto-suggest only displays a list of key phrases matching the letters entered in the search box, it will not display phrases are closely related topics. it is not like ML. In Contegra, Key phrases are selected using statistical algorithms and natural language processing technology that analyzes the content and identifies the most significant terms in each document and user will get result of suggestion display from those key phrases only.

second, I suggest user should type words rather than paste words in search box. I hope it will display result more meaning than copy/paste.


Thanks,
Jitesh

Hi Jitesh :

Ok, I see what you mean with "fair links er2 expert report." It does appear in one document on the live site, but since the auto suggest displays in all lower case letters, I thought it was an error ("fair links er2 expert report" appears on the PDF's abbreviations table in upper case letters as "FAIR LINKS ER 2    Expert Report on Damages of Fair Links of 23 May 2014").

Based on Morgan's first post, I thought Contegra would be "machine learned" beyond its actual capability, but clearly it is not. Yes, I have been typing terms and phrases, not copying/pasting.

Thanks,

Paul

Hi Jitesh ,

Could you comment on a few observations summarized the following video:

Contegra Full Text Search Observations (24-May-19) .mp4 17.4 MB • Download

Thanks,

Morgan

Hi Morgan ,

Do you know how the suggested searches are ordered? Is there a way to change their order so that they correspond to the relative number of hits in the dtSearch index? That way the suggested searches with most results are most likely to appear at the top of the list.

We will contact to contegra to change order, if possible, currently contegra not allowed to do change order.Second, suggested searches are not come from Dtsearch index. Contegra create new indexes from dtsearch index.It is from contegra Index.

 
Based on user type in search box, suggestion list will be display otherwise it will terminates, means may be there is no Boolean words in the Contegra Index. I think we have already remove "noise" words from Dtsearch,   Contegra index create based on dtsearch index.

User can type phases which he/she want to search if suggestion list not display than type phases and click on Search button, result will be display as per expectation. Auto suggest result are not in our hand, we are just display list from contegra based on search words, if user want result based on phases which you want like boolean operation than please type it and click on submit button. it will display result with boolean operator as well.

Please note, it will not display phrases are closely related topics. it is not like ML.
 
  

There was an instance where I entered " into the search field, and Contegra suggested """;"t, which is searched didn't produce any results. Why would something like this be suggested?

it is coming from contegra, we are sending search term and contegra will provide suggestion list based on search term (word).

Can we make it so that when a user selects a suggested search term, the system automatically runs the search query.

Yes, we have updated same in dev.islg.

Can we change the settings so that suggested searches are not presented to the user until they type 3 characters into the search field.

Yes, now suggested searches are not present until user type 3 characters into the search box.

Thanks,
Jitesh

Thanks Jitesh ,

Continuing the threads above:

Do you know how the suggested searches are ordered? Is there a way to change their order so that they correspond to the relative number of hits in the dtSearch index? That way the suggested searches with most results are most likely to appear at the top of the list.

We will contact to contegra to change order, if possible, currently contegra not allowed to do change order.Second, suggested searches are not come from Dtsearch index. Contegra create new indexes from dtsearch index.It is from contegra Index.

Ok. Sounds good. Hopefully there are some options here. At the very least, hopefully they can explain how suggested searches are currently ordered.

 
Based on user type in search box, suggestion list will be display otherwise it will terminates, means may be there is no Boolean words in the Contegra Index. I think we have already remove "noise" words from Dtsearch,   Contegra index create based on dtsearch index.

User can type phases which he/she want to search if suggestion list not display than type phases and click on Search button, result will be display as per expectation. Auto suggest result are not in our hand, we are just display list from contegra based on search words, if user want result based on phases which you want like boolean operation than please type it and click on submit button. it will display result with boolean operator as well.

Please note, it will not display phrases are closely related topics. it is not like ML.

Understood. I suppose this is just a limitation of the search. Please confirm that "noise" words are not excluded from the dtSearch and Contegra indexes.
   

There was an instance where I entered " into the search field, and Contegra suggested """;"t, which is searched didn't produce any results. Why would something like this be suggested?

it is coming from contegra, we are sending search term and contegra will provide suggestion list based on search term (word).

Ok. Given that the Contegra index is generated from the dtSearch index, it's odd that the suggest search doesn't show up in the results. Perhaps you could add this issue to your query when you contact Contegra.

Can we make it so that when a user selects a suggested search term, the system automatically runs the search query.

Yes, we have updated same in dev.islg.

Looks good.

Can we change the settings so that suggested searches are not presented to the user until they type 3 characters into the search field.

Yes, now suggested searches are not present until user type 3 characters into the search box.

Looks good.

Thanks,

Morgan

 

Hi Jitesh ,

Any updates on the items above. 

Thanks,

Morgan

Hi Morgan ,

Yesterday, contegra plan reset now We are focusing on Contegra working in the SQL database for the Subject Navigator.

Thanks,
Jitesh

Ok. Sounds good, Jitesh , but I understood that you were going to contact Contegra to get them to explain how suggested searches are ordered, and whether we can change that ordering. Could you please do so via email, or would you prefer me to do so?

Also, I asked you to confirm whether the "noise" terms are included in the dtSearch and Contegra indexes.

Thanks,

Morgan 

Hi Morgan ,

we have integrate contegra on Subject Navigator search also, please check it on dev site and let me know your feedback.

For "noise" terms, currently it is excluded in dtsearch so first we need to generate dtsearch index with "noise" terms, so can we do this excise on dev site? please let me know.

Thanks,
Jitesh

Hi Jitesh ,

Why are we excluding "noise" terms for the dtSearch index. I thought we fixed this problem a long-time ago? This is something Greg raise recently. Let's ensure all "noise" terms are included in the index (assuming this doesn't create any problem with daily index process).

I performed some testing on the Subject Navigator search, and the suggested searches do not correspond with the text of the branches in the Subject Navigator. Can you confirm that the Contegra index is getting populated by the SQL database index concerning the branch text? See video below.

Thanks,

Morgan

Problem with Contegra search in Subject Navigator (5-Jun-19).mp4 6.1 MB • Download

 

Hi Morgan ,

There was an issue with Contegra Indexing which we resolved on dev.islg. Now, The suggest search is corresponding with the text of branches in Subject Navigator tool. Please check and confirm.

Hi Morgan ,

At last, Greg requested to remove "noise" words from TLG dtsearch index.

image.png 320 KB • Download

I have also forward email chain for removing noise words conversation.

Thanks,
Jitesh

Hi Jitesh ,

In the email above, Greg requested that the noise words be deleted from the "noise words" file, which effectively includes the noise words in the dtSearch index. This ensures that dtSearch can differentiate between a search for "the cause" and "a cause" because "the" and "a" are included in the index. I believe the dtSearch index within ISLG has been setup this way. Could you confirm?

Could you also confirm that the same settings have been applied to the Contegra index (i.e., the noise words are deleted from the "noise words" file, and that terms like "the" and "a" are include in the index)?

Thanks,

Morgan 

Hi Harsh ,

Contegra seems to be working on the Subject Navigator, and it's actually working pretty well.

Paul and Ryan , take a look and let me know your feedback.
 
Morgan 

Hi Harsh and Jitesh ,

Further to the screenshot below, I noticed that Contegra is only picking up strings where the characters entered into the search field match the first word of the phrase. However, with the term "favourable", I would expect things like "most favourable" and "less favourable" to appear in the suggested searches.  Is there anything in the way Contegra is setup that is preventing it from suggesting searches where the relevant characters appear in the second or third word of the phrases?

image.png 152 KB • Download

Note that the example Contegra provided on its website: http://contegrasystems.com/products_/auto-suggest, suggests that it should be able to pick phrases where the searched term appears in the second or third word of the phrases:

image.png 260 KB • Download

Thanks,

Morgan

Hi Harsh and Jitesh :

I get the same result as Morgan. I do not get any hit when I search for "faith" (trying to see if "good faith" shows up) on dev.islg.

Thanks,

Paul

Hi Morgan and Paul ,

Yes, Your finding is right. We have pass Infix parameter with search term and its work as per your expectation.
Please check it and give your feedback.

Thanks,
Jitesh

Hi Jitesh : 

I tested "Fair" on Subject Navigator hoping to see "Fair and Equitable Treatment" as one of the suggestions (it is a large branch that you can see if you click on "F" once). However, as shown below, it is not one of the suggestions. The closest suggestion seems to be "breach of fair equitable treatment." I attach two screenshots to be comprehensive on what results I get from Contegra. Is Contegra sweeping through only the text of each entry as opposed to each branch title?

image.png 45.7 KB • Download

Please let me know.

Thanks,

Paul

Hi Paul and Jitesh ,

Yes, Paul 's, comment above is something I noticed as well, and I believe is related to the use of words that could be used as Boolean operators ("and", "not" and "or"). Any phrases that contain these words do not appear as suggested results, and the Contegra auto-suggest disables when you enter these words in the search field. See more in video below.

Problem with Boolean operators in Contegra auto-suggest (7-Jun-19).mp4 6.16 MB • Download

Jitesh , is it possible for you to make changes to ensure these Boolean operator words get included in suggested searches?

Thanks,

Morgan

Hi Morgan ,


Jitesh, is it possible for you to make changes to ensure these Boolean operator words get included in suggested searches?

Yes, but we need to include "noise" words in dtsearch than and than it will be available in contegra as dtsearch is a source of contegra index. so first we need to include "noise" words in dtsearch after will generate contegra index than it will be available in suggested searches.

currently dtsearch configuration removed "noise" word so indexes generated without "noise" words, Second if you want to include "noise" words in dtsearch indexes than it will be also include both site ISLG live and dev because there is dtsearch configuration use for both site; as dtsearch is common for all site in signal server. Please let me know can we include "noise" words in dtsearch?


Thanks,
Jitesh

Hi Jitesh ,

Are you saying that noise words are currently excluded from the dtSearch index? I'm confused by this, because we should have already been including noise words in the index. This is what Greg and I requested for TLG and ISLG during the email exchange above.

Also, the Full Text Search appears to be including noise words in the index. For example, performing searches for "a claim" and "to claim" produce different sets of results (see video below). If noise words were being excluded from the index, these searches would have overlapping results because the search would disregards the terms "a" and "to" as noise words. Could you please carefully explain how the current index is setup?

Problem with Boolean operators in Contegra auto-suggest (10-Jun-19).mp4 16.1 MB • Download

Thanks,

Morgan

Ok. Thanks Jitesh . 

Yes, assuming including all the noise words will not have a detrimental impact on search speeds, we want to remove all noise words similar to TLG.

Thanks,

Morgan

Hi Morgan ,

I have checked both TLG and ISLG dtsearch removed "noise" words.

Thanks,
Jitesh

Hi Jitesh ,

We're clearly not communicating clearly on this issue. We need to ensure that ISLG and TLG are including noise words in their indexes to ensure dtSearch is capable of differentiating between a search for "a claim" and "to claim".

In addition, we need to ensure the noise words and Boolean operators like "and", "not" and "or" are included in the Contegra Auto-Suggest search, and that including these terms in the search field does not interrupt of disable the auto-suggest functions.

Morgan

 

Hi Morgan ,

Sorry for inconvenience, we are looking into it. I will let you know solution as per your expectation because currently we are facing issue to generate Contegra Index same as we getting early because might be reach quota of plan.

Thanks,
Jitesh

Jitesh ​ , Did we restore the noise words for the dtSearch index on ISLG and TLG? Please ensure this done.

Thanks,

Morgan

Hi Jitesh and Harsh  

Are you available tomorrow morning at 10am EDT for a call? I'd just like to review the search issues with you and ensure we're all on the same page for the expected behaviour.

Thanks!

Ryan

Hi Ryan ​,

If possible then make a call at 9:00 AM EDT today. otherwise we are available at 10:00 AM EDT.

Hi Morgan ,

Yes , we have restored "noise" words for the dtsearch index on ISLG and TLG.

Thanks,
Jitesh

Hi Ryan ,

how we can connect? can you please share goto meeting link or other way to connect?

Hi Jitesh  

Here are some details for Zoom:

Topic: DEVIT
Time: Jun 13, 2019 9:00 AM America/Toronto

Join Zoom Meeting
https://zoom.us/j/848857067

One tap mobile
+16475580588,,848857067# Canada

Dial by your location
        +1 647 558 0588 Canada
        +1 669 900 6833 US (San Jose)
        +1 646 558 8656 US (New York)
Meeting ID: 848 857 067
Find your local number: https://zoom.us/u/aceDXZzO99

Hi Morgan ,

From yesterday, We are not able to generate contegra indexes due to quota of plan reach, it showing "unauthorized" error. we have already asked to Radomir about this error in past, he has told us it shows when quota of plan reach. 

So currently, dtsearch index included "noise" words and contegra index is old one due to currently not able to generate contegra Index because of quota of plan reach.


Thanks,
Jitesh

Hi Jitesh ,

I just ran a text on the Full Text Search, and the index is still excluding noise words. For example, I search for "a expropriation", and it produced results for "the expropriation", "to expropriation", "on expropriation", "direct expropriation", etc. This indicates that the search is disregarding "a" from the search and replacing it any random word. If we're including noise words in the dtSearch index, it should only produce results for "a expropriation".

https://www.investorstatelawguide.com/ResearchTools/DTSearchFTS?action=search&newchkvalue=&toc=content&id=97&q=a+expropriation&testfeild=test&c1t=Awards+on+Costs%2CDecisions+of+Annulment+Committee%2CDecisions+of+Reviewing+Court%2CDecisions+on+Arbitrator+Challenge%2CDecisions+on+Consolidation%2CDecisions+on+Interim+Measures+of+Protection%2CDecisions+on+Jurisdiction+or+Preliminary+Questions%2CDecisions+on+Place+of+Arbitration%2CDecisions+on++Rectification%2C+Supplementary+Reasons+or+Interpretation%2CDecisions+on+Stay+of+Execution%2CFinal+Awards+on+Jurisdiction%2C+Merits+or+Damages%2CNotices+of+Intent%2CNotices+of+%2F+Requests+for+Arbitration%2COther%2CPartial+Awards+or+Decisions+on+the+Merits%2CPleadings%2CProcedural+Orders%2CSeparate+Opinions%2C%2CICSID+Additional+Facility+%2CLondon+Court+of+International+Arbitration%2CStockholm+Chamber+of+Commerce%2CUNCITRAL%2C%2CFree+Trade+Agreements%2CInternational+Treaties+%26+Rules%2CNAFTA%2COther%2CRegional%2FSectoral+Agreements%2C&agreementID=1234&provisionSourceID=&languageId=test&addselection=&tempno=21&countryid=&nodeid=&selNodeId=&country_id=&treaty_id=&rules_id=&fuzziness_id=1&language_id=A&disputid=&provision_idsource=&TreatiesandOtherInstrumentsid=&RulesofArbitrationid=&DispId=c2%2Cc3%2Cc4%2Cc5%2Cc6%2Cc7%2Cc8%2Cc9%2Cc10%2Cc11%2Cc12%2Cc13%2Cc14%2Cc15%2Cc16%2Cc17%2Cc18%2Cc19%2Cc20%2Cc22%2Cc23%2Cc24%2Cc25%2Cc26%2Cc28%2Cc29%2Cc30%2Cc31%2Cc32%2Cc33&drpdwnId=%2CSelect+Agreement&listId=test&c1t_name=&traty_instrument_id=&checkboxesCheckedValuePass=cat%3D3%5Etype%3D1%2C2%2C3%2C4%2C5%2C6%2C7%2C8%2C9%2C10%2C11%2C12%2C13%2C14%2C15%2C16%2C17%2C18%2C19%2C20%2C21%2C22%2C23%5E%5Ecat%3D3%5Etype%3D17%5E%5Ecat%3D3%5Etype%3D8%5E%5Ecat%3D3%5Etype%3D13%5E%5Ecat%3D3%5Etype%3D12%5E%5Ecat%3D3%5Etype%3D23%5E%5Ecat%3D3%5Etype%3D3%5E%5Ecat%3D3%5Etype%3D5%5E%5Ecat%3D3%5Etype%3D4%5E%5Ecat%3D3%5Etype%3D22%5E%5Ecat%3D3%5Etype%3D11%5E%5Ecat%3D3%5Etype%3D15%5E%5Ecat%3D3%5Etype%3D7%5E%5Ecat%3D3%5Etype%3D1%5E%5Ecat%3D3%5Etype%3D2%5E%5Ecat%3D3%5Etype%3D18%5E%5Ecat%3D3%5Etype%3D6%5E%5Ecat%3D3%5Etype%3D16%5E%5Ecat%3D3%5Etype%3D21%5E%5Ecat%3D3%5Etype%3D10%5E%5Ecat%3D1%5E%5ERulesArbitHidden%3DICSID+%28ICS%29%7CICSID+Additional+Facility%7CLondon+Courts+of+International+Arbitration%7CStockholm+Chamber+of+Commerce%7CUNCITRAL%5E%5Ecat%3D2%5E%5EInvestTreatyHidden%3DBilateral+Investment%7CFree+Trade+Agreement%7CInternational+Treaties+%24%24+Rules%7CNAFTA%7COther%7CRegional%2FSectoral&InvestmentTreaty_CountryId=&InvestmentTreaty_TreatyId=&All_circDate1=01%2F01%2F1899&All_circDate2=6%2F13%2F2019&RulesArbitHidden=&All_filter_country_id=&All_agreement_id=&All_Arbitrator=&All_ArbitratorRules=&searchType=boolean&ftypo=1&fuzziness=1&language=A&circDate1=01%2F01%2F1899&circDate2=6%2F13%2F2019&c1=1&c1_param=cat%3D3%5Etype%3D1%2C2%2C3%2C4%2C5%2C6%2C7%2C8%2C9%2C10%2C11%2C12%2C13%2C14%2C15%2C16%2C17%2C18%2C19%2C20%2C21%2C22%2C23&c2=1&c2_param=cat%3D3%5Etype%3D17&c3=1&c3_param=cat%3D3%5Etype%3D8&c4=1&c4_param=cat%3D3%5Etype%3D13&c5=1&c5_param=cat%3D3%5Etype%3D12&c6=1&c6_param=cat%3D3%5Etype%3D23&c7=1&c7_param=cat%3D3%5Etype%3D3&c8=1&c8_param=cat%3D3%5Etype%3D5&c9=1&c9_param=cat%3D3%5Etype%3D4&c10=1&c10_param=cat%3D3%5Etype%3D22&c11=1&c11_param=cat%3D3%5Etype%3D11&c12=1&c12_param=cat%3D3%5Etype%3D15&c13=1&c13_param=cat%3D3%5Etype%3D7&c14=1&c14_param=cat%3D3%5Etype%3D1&c15=1&c15_param=cat%3D3%5Etype%3D2&c16=1&c16_param=cat%3D3%5Etype%3D18&c17=1&c17_param=cat%3D3%5Etype%3D6&c18=1&c18_param=cat%3D3%5Etype%3D16&c19=1&c19_param=cat%3D3%5Etype%3D21&c20=1&c20_param=cat%3D3%5Etype%3D10&c21=1&c21_param=cat%3D1%5E%5ERulesArbitHidden%3DICSID+%28ICS%29%7CICSID+Additional+Facility%7CLondon+Courts+of+International+Arbitration%7CStockholm+Chamber+of+Commerce%7CUNCITRAL&c22=1&c22_param=cat%3D1%5E%5ERulesArbitHidden%3DICSID+%28ICS%29&c23=1&c23_param=cat%3D1%5E%5ERulesArbitHidden%3DICSID+Additional+Facility&c24=1&c24_param=cat%3D1%5E%5ERulesArbitHidden%3DLondon+Courts+of+International+Arbitration&c25=1&c25_param=cat%3D1%5E%5ERulesArbitHidden%3DStockholm+Chamber+of+Commerce&c26=1&c26_param=cat%3D1%5E%5ERulesArbitHidden%3DUNCITRAL&c27=1&c27_param=cat%3D2%5E%5EInvestTreatyHidden%3DBilateral+Investment%7CFree+Trade+Agreement%7CInternational+Treaties+%24%24+Rules%7CNAFTA%7COther%7CRegional%2FSectoral&c28=1&c28_param=cat%3D2%5E%5EInvestTreatyHidden%3DBilateral+Investment&c29=1&c29_param=cat%3D2%5E%5EInvestTreatyHidden%3DFree+Trade+Agreement&c30=1&c30_param=cat%3D2%5E%5EInvestTreatyHidden%3DInternational+Treaties+%24%24+Rules&c31=1&c31_param=cat%3D2%5E%5EInvestTreatyHidden%3DNAFTA&c32=1&c32_param=cat%3D2%5E%5EInvestTreatyHidden%3DOther&c33=1&c33_param=cat%3D2%5E%5EInvestTreatyHidden%3DRegional%2FSectoral&agreementIDprov=&agreementSeq=1#b

image.png 154 KB • Download

 
In addition, the search is running extremely slowly. It took over a minute to load the results above. Please ensure this is resolved.

Thanks,

Morgan

Hi Jitesh ,

Understood, re the Contegra index. Did you have a discussion with Ryan about the disabling of the auto-suggest search when Boolean operators are entered into the search field? Can this be resolved without updating the Contegra index, or do we have to wait until the plan resets next month?

Morgan

Hi Morgan  

It sounds like this has been resolved, but the indexing isn't able to run until the plan resets which is why we're still seeing the same issue.

Ryan

Hi Jitesh ,

Lastly, I've become aware that dtSearch has the capability of ranking results by relevancy (currently it is setup only to rank results by frequency). Are you aware of the capability, and the possibility of introducing this as an option for users to rank results in the Full Text Search? 

Thanks,

Morgan

Hi Ryan ,

Does this mean entering "and", "or" and "not" in the search fields will still produce auto-suggest results, but no matches will be produced because the Contegra index currently excludes these terms?

Morgan

Hi Morgan   and Ryan ,

Morgan , As per your above comment that when you search for
"a expropriation", and it produced results for "the expropriation", "to expropriation", "on expropriation", "direct expropriation", etc.

Yesterday, We have included all noise words in DtSearch Indexing. Due to that reason, The DTsearch produced results like "the expropriation", "to expropriation" etc. 

Today we have again removed the noise words from DTsearch Indexing and now you can check the result in FTS tool.

One more point to that If we include all noise words in DTsearch then it is taking some time to populate the result.


Following, I have attached all noise words list. Please check and let us know which word you need to put as noise words.

noise.dat 541 Bytes • Download

Hi Morgan ,

Are you aware of the capability, and the possibility of introducing this as an option for users to rank results in the Full Text Search?

No Morgan, We didn't explore the dtsearch tool because it was already in functional condition so we have just created index so We don't have idea about the ranking functionality.



Thanks,
Jitesh

Hi Harsh ,

Understood. The FTS is working as expected now, but there's still a problem with the Subject Navigator search: https://www.investorstatelawguide.com/ResearchTools/SubjectNavigator?toc=content&id=50&tab=r&search=a+expropriation&searchType=boolean&stem=1&thes=&ftypo=1&fuzziness=1 

image.png 360 KB • Download

Also, if you haven't done so already, please ensure the same changes are made to TLG (BTW: we shouldn't have made any change to TLG without Greg's explicit authorization).

I think there is confusion on what is meant by "included". When I say I want the noise words include in the dtSearch, I mean that I want to ensure the noise words are included in the main index for the search so that the searched keyword always matches the words in the documents (even if a noise word is used). I believe this means deleting everything from the "noise words" file on dtSearch. Please ensure this is done across all the indexes on both ISLG and TLG.

Thanks,

Morgan 

Ok. Thanks Jitesh .

I'm having discussions with Rob Wiesenberg from Contegra about bringing them on-board to help with this process (and other ways to improve how we use search engines across the system). I'll let you and the rest of the team know as things develop.

Thanks,

Morgan

Hi Morgan ,

The Subject Navigator's noise word issue has resolved on www.islg. Due to some confusion, I think we will discuss this thing on Thursday's call.

Ok. Looks good Harsh . I'd like to avoid discussing this issue further. Going forward, please assume that this how we'll always want our searches setup.

I'll send an agenda out concerning the call on Thursday later today, but I would like to bring the discussion back to the issue concerning the use of Boolean operator terms "and", "not" and "or", and how those are interfering with the auto-suggest search.

Thanks,

Morgan

Hi Jitesh ,

Could you please give us some more information on how we're hitting the limits of our free Contegra/Watson?

Are we running the Contegra index daily at this point, or is it another limit we're hitting? If we're running this frequently do we need to run it as often? 

Thanks!

Ryan

Hi Ryan, 

We are generating contegra index based on dtsearch index and it is big size of dtsearch index so it was reach qouta of current plan. There is limited size conversion provide by current plan and we have big size if we are convert from small size than it will give many time to generate but dtsearch index is big one so it will reach single time. 

We are not hitting daily, as of now single/two time it will reach quota of current plan in month. You can also find free plan information from contegra IBM Watson website. 

I will provide more information from my desk. Thanks. 

Get Outlook for Android<https://aka.ms/ghei36>

Hi Jitesh ,

My understanding is that our quota is being limited by IBM Watson, rather than dtSearch. We are currently operating on the free plan (see screenshot below), which provides for 30,000 NLU items per month. Approximately how many NLU items is required to perform a Contegra index on each for the following:

Thanks,

Morgan

image.png 90.6 KB • Download

Hi Morgan ,

Yes, quota is being limited by IBM Watson, but dtsearch index is source for IBM Watson (contegra) , so it is related to NLU because if big size of dtsearch index than it will use more NLU to read characters of dtsearch indexes to create Contegra Index.

We will talk more about this in tomorrow call.

Thanks,
Jitesh

Hi Jitesh ,

Now that IBM Watson quota has reset, could we perform an index update for the auto-suggest to see if the fixes above implemented.

Thanks,

Morgan 

Hi Morgan ,

We have regenerate contegra indexes, You can check with your cases and give your feedback.

Thanks,
Jitesh

Hi Jitesh ,

As we discussed this morning, please ensure the indexes and fixes for the Subject Navigator are implemented on dev.islg as well, and then we'll follow-up with contegra about the issues we discussed concerning Boolean operators.

Thanks,

Morgan 

Hi Jitesh ,

Could you confirm whether the indexes and fixes where implemented on the Subject Navigator for dev.islg.

Thanks,

Morgan 

Hi Morgan ,

We have also regenerated contegra indexs for Subject Navigator on dev.islg.
Please check and provide your feedback.

Thanks,
Jitesh

Hi Jitesh

I've run a few tests and the noise words inclusion seems to be working after the index for me on dev.islg. Morgan  please confirm.

Thanks!

Ryan

Hi Ryan ,

Yes, the noise words seems to be included now. However, there appears to be a problem with the indexing that prevents the systems from including phrases like "Fair and Equitable Treatment". Further to the discussion during the team call last Thursday, Jitesh informed that that this is not within their control to fix, and we'll need to ask Contegra for a solution. I'll speak to Rob Weisenberg about this as part of their audit of the searches across the system. 

At the same, the auto-suggest search developed by the Machine Learning team is showing promise, so we may be able to use this for auto-suggest as an alternative.

We'll revisit this issue in the weeks ahead.

Thanks,

Morgan