Engineers and Contextual Search Experience

Engineers and Contextual Search Experience


Web search became part of our life. We don’t search anymore, we “google” everything . The visible simplicity of Google created a feeling that magic of search can transform and simplify any software product behavior. CAD, PLM and other enterprise software companies liked the idea as well. Search is certainly getting into mainstream. Open source search libraries such as Lucene and Solr created environment for easy implementation and distribution of  search products across multiple software solutions.

At the same time, not every search solution can lead to simplicity. So called “laundry list” of results can be very disappointing for customers and lead to many questions about results relevance. Data matters and data can be nasty. Especially when it comes to complex engineering design, and enterprise data management solutions. To index data located in enterprise software packages can be a tricky problem.

Even web is not a search paradise these days. Google is still web search king. Even so, the relevance of some Google results is questionable. The complexity of Web search multiplied by social networks, mobile, combined with commercial interests of web giants created complexity that can be compared to the complexity of enterprise software. In parallel, there is a clear trend is enterprise software to adopt successful ideas of social software and social collaboration.

Recent Mashable article Yahoo’s New Long Game: Contextual Search puts some lights on the innovation and possible ways to solve problem of relevance in web search results. This is my favorite passage:

When I look at things like contextual search, I get really excited,” Mayer said at the conference. Contextual search seeks to take in a variety of factors aside from a simple input to generate results that are tailored to a person’s time, place and patterns. For instance, a normal search for sushi might turn up a Wikipedia page or various websites about sushi. If one were to look up sushi from a phone through a contextualized mobile search, it could conceivably return nearby sushi restaurants with review, advertisements and coupons. The reason for Mayer to get excited is twofold: Nobody has yet mastered contextual search and it has the possibility of generating a ton of revenue.

Yahoo contextual search made me think about potential of such type of advanced search option for engineers. The specifics of engineering environment characterized by number of data dependencies, connected information and complexity to calculate the relevance search results. Engineering data can generate large volume of matches that hardly can be filtered based on simple filtering mechanisms. Think about document numbers, material names, design element names. Search for “shaft”, “tube” and “aluminum” can generate thousands of results that hardly can be distinguished, sorted and ordered.

This is a place where I think “contextual search” does fit in a perfect way. What can be used a context for search (query) mechanism? Actually, quite many elements of easy available data can be re-used – date, time, organization, project name, team, location, previously used assemblies, etc. Some of these elements can be captured from the environment (computer, browser, application) and some of them can be captured from directly from users via specific user interface (capturing semantic). Result – significant decrease in the number of search results and better relevance.

What is my conclusion? Search is not simple. Even Google simplicity is questionable when it comes to the reality of engineering and enterprise data. New algorithms and additional data analysis must be applied in order to improve the relevance of results. Contextual search is not completely new idea, but it can become the next big deal in improving of search and overall user experience. Just my thoughts…

Best, Oleg


Share This Post

  • You are right, Lucene and Solr are powerful technologies for search functionalities.
    It allows wild-card queries using * and ?, full text searches, tokenization, synonyms, default operators, boosts some terms, proximity for fields,… this allow to find what you are looking for even if you don’t type the exacts words. This is what we use in beCPG PLM.

  • beyondplm

    Philippe, thanks for your comment! Did you implement any “contextual” usage of Lucene/Solr?

  • Not really. We present contents that are in the site where the user is.
    When a user look for a component, we present the raw materials if he is in the composition and the packaging materials if he is in the packaging list.
    An example of contextual search could be:
    System displays different raw materials depending of the finished product user is making.
    ie: I do a light product or a free sugar drink so I put the sugar raw materials / raw materials with sugar at the end of the list

    Another interesting issue with solr is faceted navigation. Depending of your 1st search, the system shows a tag scope to adjust the search. An example is this french site that my colleague has done with Solr :
    If you search:
    Viking > click on other 138 > then you have a tag scope where Brest which is bigger
    After, you can do a search in the article.

  • beyondplm

    Philippe, thanks for sharing your story and link. Faceted navigation is important in many cases. However, in some cases it gets complicated because of data model, which is not allowing you to flatten all your data.