We hope you have found our recent blog posts useful and that you’ve implemented the ideas to ensure your eCommerce site is ready for the Holiday season. Creating an optimized shopping experience for your customers will be a gift to your business accounts as well as your customers. This blog post is the last in the series that lays out best practices and tips as you prepare your eCommerce site for the biggest commercial period of the year.
You’ll be familiar with the default search language of Keywords; that condensed string of search terms, void of any connecting words. Many people who are used to using web search engines still choose to use keywords. CollegeHumor illustrates Keyword Search hilariously in their ‘If Google was a Guy’ videos.
But the tides are turning with the growing awareness of Natural Language Search. Being able to phrase questions as you would ask them if you were talking to someone, using everyday language, is increasingly expected of search engines.
It would be easy to assume that keyword search systems will deal as effectively with keyword searches as a Natural Language Processing (NLP) engine. But this is just not the case. Not only will a good NLP search system produce the right results first time for the growing numbers of users who are familiar with natural language searching, but it will also out-perform keyword search engines when it comes to seemingly simple keyword searches.
A good NLP search engine understands word inflection and is able to take versions of words and create a single concept. Changes such as tenses and pluralities are reduced down to a single word, normally a noun. For example, it would be equally natural for a customer to search for either ‘jackets’ or ‘jacket’. With some traditional keyword search systems, however, business users would have to manually create synonyms to say that jackets = jacket.
Irregular plurals such as ‘goose/geese’ or ‘lady/ladies’ can confuse a keyword search engine, whereas a good NLP system will understand all pluralities and grammar. EasyAsk’s search solution will even understand these language complexities across multiple languages.
Take the simple example of the keyword search, ‘red jacket’. A keyword search engine will match products that contain the words ‘red’ and ‘jacket’. It may even be configured to search in categories. But what if the category name is ‘jackets’ and the system doesn’t recognize ‘jacket’ as meaning the same thing? A good NLP system will match products containing the words ‘red’ and ‘jacket’, but it will also match if, for example, there is some text that describes the product as ‘redder than purple’. EasyAsk’s solution goes one step further. It will understand the keywords completely and realize that the words could be mapped to something else. It understands that red is a color attribute, and will bring back products that don’t necessarily contain the word ‘red’ but that are burgundy, scarlet, berry, crimson… It will also understand that ‘jacket’ is a category name, and that certain products may not include the category name, ‘jacket’, but instead a model name, such as a ‘red Bomber’.
The EasyAsk system on Nasco’s website responds perfectly to the simple keyword search ‘watercolor brush’:
EasyAsk’s system has recognized that ‘brushes’ in the product name matches ‘brush’ in the original search and that ‘watercolor’ describes the type of brushes. A simple keyword search system may have shown watercolor paints first in the product results and found no matches for ‘brush’.
Even if your customers are running relatively simple keyword searches, there is still lots more that can be done with the results. With EasyAsk, you have the ability to rank the products intelligently, using relevant scoring. Depending on where the match is found will affect the score for that product. For example, a product name match is more important than a match in the description as it could contain misleading text. You also have the ability to search the most important parts of the product first for matches, and only search the less important parts if nothing is found initially.
Yet another option is to configure EasyAsk to only search the highest weighted fields (e.g. product and category name), and if results are found, to stop there. If no matching products are found, however, EasyAsk will go back and try the lower weighted fields (e.g. product description and keywords). In this way, it is possible to only show the less-relevant results when there aren’t any highly relevant products to show. Results pages shouldn’t be filled with less-relevant information.
Reducing No Results
Great NLP engines will avoid ‘No Results’ for keyword searches by relaxing one or more of the terms. Where there may be no results for a search for ‘pink scissors’, one of the terms could be relaxed to show either pink products or scissors. EasyAsk even allows its users to configure the priority of which term is relaxed. In our example, EasyAsk would ignore the color and show scissors rather than pink products that aren’t scissors.
Search As You Type
Search As You Type (or SAYT) intelligently finds the most common searches and shows products before the user has even hit return. This function means that customers with a simple keyword search can type the first few letters and click the search or even go straight to the product that they had in mind.
SAYT in action at enacso.com
How does your business measure up?
NLP search systems understand everyday language with all of its complexities. Intelligent, intuitive NLP systems can take the simplest keyword searches and provide a vastly superior experience to that of a traditional keyword search engine. Does your search solution provide relevant products, however your customers decide to ask?
Let us show you the full range of features that the EasyAsk solution offers.