IBM Watson™ Ideas

Welcome to the IBM Watson™ Ideas Portal


We welcome and appreciate your feedback on IBM Watson™ Products to help make them even better than they are today!


If you are looking for troubleshooting help or wondering how to use our products and services, please check the IBM Watson™ documentation. Please do not use the Ideas Portal for reporting bugs - we ask that you report bugs or issues with the product by contacting IBM support.


Before you submit an idea, please perform a search first as a similar idea may have already been reported in the portal.


If a related idea is not yet listed, please create a new idea and include with it a description which includes expected behavior as well as why having this feature would improve the service and how it would address your use case.

WDS highlights should honour Advanced Search quotes

During Project Daisey development, we've come across the inconsistency between highlights returned by WDS and the query. 

Even if the WDS Query language query contains a compound term (which project daisey uses very frequently for financial terms that change meaning if used with NLQ which does not honor quotes) highlights highlight single terms. This makes it seem to the user that WDS does not honor the quotes duing WDS query search, even though it does.The real reason is  that  highlights do NOT honor the quotes, therefore the results look wrong.

E.g. If I search query="data capture"

I can get back highlights that emphasize either word alone:

 

    "text": [
                    "Cognitive tools are used to analyze existing documents, to accelerate <em>data</em> <em>capture</em> and organizing information on application. Existing business and IT value data is leveraged. Knowledge gaps are identified and a plan to close the gaps is prepared and executed • Assess business and IT value • In this step, applications are mapped to business capabilities or processes.",
                    "<em>capture</em> planning MEET IBM GBS Cloud Application Migration Services IBM Confidential 43 Asset work under way Cloud Innovate Thanks ",
                    "Advise on cloud I ensures IBM GBS Cloud Application Migration Services IBM Confidential - - - Microservices nt and Monitoring Secure cloud DevOps Migrate to cloud ions on the Cloud Integrated Cloud Platform Operations • Consistency of experience from IBM ensuring predictable outcomes • Assimilation of best practices & experiences ensuring superior output quality • Standard set of Tools facilitating <em>data</em>",
                    "<em>capture</em>, analysis and reporting • Efficiency that increases speed to value Ø Modernize for cloud Operations Guidelines Rationalize for cloud Cloud InnovateTM Methodology brings the IBM way to address Hybrid Cloud journey Cloud Innovate based on an end -to -end Method for cloud adoption meeting specific client demands Strategize / Mobilize Discovery & Analysis Design & Build ØP ops oa Applied To Secure",
                    "duplicate dete Data use pattern Minhash, LSH planning CTD, test scope partitioning detection Fine-grained Project simulation and risk assessment WBS and estimation * CRUD analysis qtr * * API candidate search Concept extraction, searchpatterns Test planning advisor Testing pattern advisor Mobile/web Coarse -grained practitioner support, Options text mining from catalog Transformation WBS and estimation <em>Data</em>"
                ],

 

The impact is that users think the query  MATCHED on either word from the compound word, i.e. that WDS (discovery query language) search does not honor quotes. . This is not the case, it does. However, when user sees highlights with single words highligted, it gives the impression that the system is not honoring quotes.

 

Project Daisey is a multi-year multi-million-collar Watson Delivery project in UK. This defect has big impact on the display of search results.

  • Guest
  • May 7 2019
  • Needs review
Why is it useful?
Who would benefit from this IDEA? Project Daisey in UK but also all other clients using WDS Query language for lond term searches
How should it work?
Idea Priority
Priority Justification
Customer Name
Submitting Organization
Submitter Tags
  • Attach files
  • SARA ELO DEAN commented
    31 May 06:01

    It turns out that there is a WDS defect related to the highlights. When a compound term is searched, it turns out that the one highlight is actually split into two,  reversed in order and they are created by an incorrect line feed between the two words of a compound search term: see example above for WDS query="data capture"  4 last highlights are actually two highlights split into 2 and reversed in order!

        "<em>capture</em> planning MEET IBM GBS Cloud Application Migration Services IBM Confidential 43 Asset work under way Cloud Innovate Thanks ",
                        "Advise on cloud I ensures IBM GBS Cloud Application Migration Services IBM Confidential - - - Microservices nt and Monitoring Secure cloud DevOps Migrate to cloud ions on the Cloud Integrated Cloud Platform Operations • Consistency of experience from IBM ensuring predictable outcomes • Assimilation of best practices & experiences ensuring superior output quality • Standard set of Tools facilitating <em>data</em>",

     

    and
                        "<em>capture</em>, analysis and reporting • Efficiency that increases speed to value Ø Modernize for cloud Operations Guidelines Rationalize for cloud Cloud InnovateTM Methodology brings the IBM way to address Hybrid Cloud journey Cloud Innovate based on an end -to -end Method for cloud adoption meeting specific client demands Strategize / Mobilize Discovery & Analysis Design & Build ØP ops oa Applied To Secure",
                        "duplicate dete Data use pattern Minhash, LSH planning CTD, test scope partitioning detection Fine-grained Project simulation and risk assessment WBS and estimation * CRUD analysis qtr * * API candidate search Concept extraction, searchpatterns Test planning advisor Testing pattern advisor Mobile/web Coarse -grained practitioner support, Options text mining from catalog Transformation WBS and estimation <em>Data</em>"

    I verified in the source document that the text indeed appears such that the above is true.

  • SARA ELO DEAN commented
    31 May 06:04

    There is  highlight issue however related to searches that are searching in a specific field. E.g.

    (metadata.dataroom_filename:"customer",metadata.dataroom_filename:("analysis"|"churn"))|("customer analysis"|"customer win"|"customer loss"|"customer churn"|"customer volume"|"top a customers")

    i.e. lot of compound terms, but single terms are searched against filename only.

    Here are matching snippets where you see that single words appear, in text snippets EVEN THOUGH they were search terms only against filename. Therefore user assumes that highlights are coming from the compound terms (e.g. "customer analysis") that are only partially matched. 

    Note that query does not have the term "customer" ever alone without being part of a compound term, except for filename search. However snippets show that customer, analysis match everywhere.

        "text": [
                        "Industry applications of techniques to discover the factors that were most predictive One media company, for example, used machine learning of <em>customer</em> <em>churn</em> and identified the 2 percent of <em>customers</em> causing almost 20 percent of overall <em>churn</em>.",
                        "In these applications, machine learning helps classify <em>customers</em> or observations into groups for predicting value, behavior, risk, or other metrics. It can be used to triage <em>customer</em> service calls; to segment <em>customers</em> based on risk, <em>churn</em>, and purchasing patterns; to identify fraud and anomalies in banking and cybersecurity; and to diagnose diseases from scans, biopsies, and other data.",
                        "s Our <em>analysis</em> filters business use cases by impact potential and by data richness.",
                        "SOURCE: McKinsey Global Institute <em>analysis</em> Seven of those 18 capabilities are well -suited to being implemented through the use of machine learning (Exhibit20). The first striking observation is that almost all activities require capabilities that correlate with what machine learning can do.",
                        "It can be difficult for decision makers and <em>customers</em> to commit to insights that are generated in a non- transparent way, especially where those insights are counterintuitive. Medical use cases could fall into this category."
                    ],