- Features
- Features
- Searchable Fields
- Libraries
- Facet Features
- Types of Facets
- Configure Facets
- Field Mapping
- Field Properties
- Configure Site
- Profile
- Team Management
- Sites
- Global Coverage
- Disaster Recovery
- Features
- Template Selection
- Keyword Suggestions
- In-Field Suggestions
- Top Search Queries
- Popular Products
- Trending Queries
- Promoted/Blacklisted Suggestions
- Real-time Preview
- Instantaneous Publishing
- Ranking Insights
- Clickstream Data
- Boost/Bury
- Sort
- Slot
- Pin
- Filter
- Landing Pages
- Redirects
- Set Banners
- Ranking Insights
- Create a Site Rule
- Managing Field Rules
- Managing Query Rules
- Campaign
- A top-down approach
- Faster and simpler
- Relevant Metrics for better analysis
- Interactive and easy-to-view preview
- Feature-filled Listing Page
- Promotions
- Banners
- Facets
- Redirects
- Add New Promotions
- Boost/Bury
- Sort
- Slot
- Pin
- Filter
- Landing Pages
- All Queries
- Query-based Banner
- Field-based Banner
- Creating Field-based Facets
- Create a Site Rule
- Overall Performance
- Query Report
- Zero Result Queries
- Product Reports
- In-field Suggestions
- Keyword Suggestions
- Top Search Queries
- Popular Products
Content Documentation
The relevancy of an eCommerce site is determined by the optimized results it produces for a search query. Even if a shopper misspells a query, writes unambiguously, or searches for a different product, they receive the closest query results.
Unbxd AI-powered algorithms are trained to handle shopper’s search query to generate results. However, we understand that each business has some unique instances where AI models may not be able to generate the best possible results. The Content section in Algorithms provides the flexibility to add site-specific dictionaries that can allow our algorithms to understand your business terminologies and provide relevant results. Some of the content terms that improve the relevancy of the search results are:
- Synonyms: Synonyms help your site understand search queries better, by matching keywords to similar words. For ex. ‘pants’ and ‘trousers’ are synonyms. So, whenever someone searches for either query, the result for both are returned.
- Phrases: Multi-word terms (phrases) that let your search understand the product-types in a query. For ex. “Shoe Polish” is a multi-word term that should deliver results for polish(product type) used on shoe and not shoes.
- Stemming: Allows users to override the default stemming in search. For ex. ‘”Bedding” if not overridden gets stemmed to “Bed” by default which are two different product types.
- Concepts: Concepts represent the terms in the search query which are considered important by the shoppers & the products in search results must definitely include these terms. For ex. ‘red sofa’ will segregate and deliver results for a sofa (product) in red color(attribute).
- Stop Words: Define terms that do not add value while ranking search results. For ex. ‘for’, ‘of’, ‘if’ etc.
- Spellcheck: Displaying results for the correct/intended query even if the shoppers misspell it. For ex. ‘rd drss’ will deliver results for ‘red dress’ automatically.
Let’s dig deeper into how to add, edit, delete each using the Unbxd console.
NOTE: You can create a query rule for a primary search keyword and additional keywords. To know more, refer to Create.
Synonyms
Synonyms help the shoppers to find their intended product by providing similar words for a query. Like ‘flip-flops’ is added as a synonym for ‘slippers’ or ‘pants’ for ‘trousers’. Adding them helps to avoid zero results for a query.
Let’s examine a few instances when synonyms are used.
- Terms with the same meaning: These are set of terms that have the same meaning i.e. synonyms as defined in the dictionary
women’s => ladies’
graph paper => grid paper - Geographically different: A lot of queries differ geographically in meaning and use.
Like, in India we may call it ‘Slippers’ but in the US it shifts to ‘Flip-Flops’ or ‘Pants’ to ‘trousers’. So, returning results for whatever a user types (slippers or flip-flops) regardless of their location should be of importance. - Related terms: These are a set of terms whose meanings are different, however, they are related hence we can show products for both the terms when a shopper is searching using a particular term
paper plates => plastic plates
abacus => calculator - Hyponyms: The synonyms describe a superset of the query :
desktop pc => computers
football gear => football cleats, football gloves, football helmet
* Hyponym is a word of more specific meaning than a general or superordinate term applicable to it. For example, a spoon is a hyponym for cutlery. - Spelling variants: These synonyms are used to handle different spelling variations for the same term (due to differences in dialects). For example,
Adapter => Adaptor
Fibre => Fiber
Color => Colour
Traveler => Traveller
Fone => Phone - Common Spelling mistakes: While Unbxd offers “did you mean” alternatives in the event of spelling errors, it is still possible to come up with synonyms for often misspelled searches in order to present customers with relevant results.
Purel => Purell (a brand) - Spacing variations: These are the synonyms that are used to handle different spacing variations for the same term.
Wifi => Wi-fi, Wi Fi
24 X 30 => 24 by 30, 24X30, 24 X30, 24X 30 - Abbreviations : There are instances where a shopper uses abbreviations to search for a particular type of product.
- Mk => Michael Kors (a brand of bags)
- NYC => New York City
Adding synonyms helps to avoid returning zero-results for similar searched terms.
Configure Synonyms
Now that we know various use-cases and types of synonyms, let us understand how to configure these synonyms using the Unbxd console.
To add synonyms (or a single synonym):
- Navigate to Content > Synonyms.
- Click the ‘hamburger’ icon > Bulk Upload Synonyms. You can upload a list of synonyms by either browsing on your computer or by using drag and drop. The supported upload format is .csv.
It is advisable for first-time users to upload a list of synonyms already known to your previous search system in order to handle all those instances in Unbxd search.
Once uploaded you can see the number of synonyms added.
- For further additions, click Add a Synonym to add a single synonym. Synonyms are categorized into One-Way synonyms and Two-way synonyms.
Type of Synonym |
Description |
One-Way Synonym |
When shoppers search for the keywords, it displays results for its synonyms as well. However, search queries using “one-way synonyms” will not include products that contain keywords. For instance: You’ve set ‘ivory’ as a one-way synonym for ‘white’. Now, typing in ‘White running shoes’, your search engine will search for ‘running shoes’ that are both white and ivory colored. But if you type in ‘Ivory running shoes’, your search engine will not search for running shoes in white. |
Two-Way Synonyms |
When shoppers search for the keywords, it displays results for the keyword and its synonyms and vice-versa is also possible. |
When a shopper searches for one of these words, all the other synonyms of the same are searched as well.
- You can search the keyword for which the synonyms are defined. Suppose, you search for the keyword red, you will get the one-way and two-way synonyms defined for the ‘color’: ‘red’.
- Once done, click ‘Save Changes’.
To delete a synonym:
- Navigate to Content> Synonyms.
- Search for the keyword and click the hamburger icon next to it.
- Click Delete where you receive a confirmation message for deletion.
- Click Yes if confirmed, or No otherwise.
What should synonyms not include:
- Empty record: entry where there is no value entered (whitespace)
- Symbols: entries where only symbols are typed as the value. The following characters are not accepted by our system : ‘ , ‘ ( comma) , ‘+’ (plus), ‘{‘, ‘}’ (curly braces), ‘*’ (Asterix),’&’ (Ampersand) , ‘\’ (backslash)
- Alphanumeric: entries with only alphanumeric values. It should have text characters.
- Stopwords: entries that contain just words that add no value like of, for, the, or any such stopwords.
- Redundant Synonyms: entries that already exist. Don’t add two similar inputs.
- Multi-term Synonyms: Synonyms having more than 2 keywords should be avoided .
Phrases
Phrases are multi-word terms (queries) that let your search understand the product-types using either a word of the phrase or the entire phrase. But if the search engine is not trained well, it can generate results based on any of the words from the searched query. Therefore, it is important to understand which word of the whole phrase should display results.
For instance, “Paint Brush” is a product-type that is made of two words ‘paint’ and ‘brush’. Searching for this query, shoppers are ideally looking for ‘brushes used for painting’. But if the search engine is not trained well due to lack of sufficient data, shoppers could see a mix of results for both ‘paint’ and ‘brushes’, if both exist in your catalog.
- Navigate to Content > Phrases.
- Click the ‘hamburger’ icon > Bulk Upload Phrases. You can upload a list of phrases by either browsing on your computer or by using drag and drop. The supported upload format is .csv.
Once uploaded you can see the number of phrases added. - For further additions, click Add a Phrase to add a single phrase. To display relevant results, specify the following for the phrases:
Terms |
Description |
Include the left-term |
Displays results matching the “full phrase” & “the left term” of it |
Include the right-term |
Displays results matching the “full phrase” & “the right term” of it |
Include the full phrase |
Displays results matching the exact full phrase |
For instance, if the phrase is ‘Shoe Polish’, we need results for ‘polish’ and not shoes. So, specify ‘Polish’ under ‘Include the right-term’. This way you train the engine to display results for polish and not shoes for the query ‘Shoe Polish’.
NOTE: If there are three or more word phrases, then the Results for only leftmost or rightmost terms can be included along with the full phrase. For Ex. if shoppers search for “Hand sanitizer dispenser”, they are looking for dispensers and not sanitizers. Include left -> Hand, Handsanitizerdispenser Include right -> Dispenser , Handsanitizerdispenser
What should phrases not include:
- Empty record: entry where there is no value entered (whitespace)
- Symbols: entries where only symbols are typed as the value. The following characters are not accepted by our system : ‘ , ‘ ( comma) , ‘+’ (plus), ‘{‘, ‘}’ (curly braces), ‘*’ (Asterix),’&’ (Ampersand) , ‘\’ (backslash)
- Alphanumeric: entries with only alphanumeric values. It should have text characters.
- Stopwords: entries that contain just words that add no value like of, for, the, or any such stopwords.
- Multiterm: entries that do not have more than one character. For phrases, add at least two words.
Stemming
Stemming truncates the query to its root form to sound semantically correct and deliver relevant results while having the same contextual meaning. Ex. It cuts down the query ‘painting brush’ to ‘paint brush’ form. You can add the stemmed versions of individual words to override the Unbxd stemming algorithm.
For instance, the search engine can stem “dressing” as “dress”. This would lead to incorrect results, this behaviour can be overridden by adding “dressing” and its acceptable form “dressing” to stemming dictionary.
To add/edit stemmed words:
- Navigate to Content > Stemming.
- Click the ‘hamburger’ icon > Bulk Upload Stemmed Words. You can upload a list of stemmed words by either browsing on your computer or by using drag and drop. The supported upload format is .csv.
- Once added, you would see the list of keywords and their stemmed forms. If you feel the stemmed word defined by Unbxd is not as per your requirement, you can edit it under ‘Stemmed Words’. Specify the entire word under both ‘Keywords’ and ‘Stemmed Words’ if the reduced form doesn’t mean the same. Like, Painting can be reduced to Paint but Leggings cannot reduce to Legs.
What should stemming not include:
- Empty record: entry where there is no value entered (whitespace)
- Stemword empty record: entries without a stem word for the root word.
- Symbols: entries where only symbols are typed as the value. The following characters are not accepted by our system : ‘ , ‘ ( comma) , ‘+’ (plus), ‘{‘, ‘}’ (curly braces), ‘*’ (Asterix),’&’ (Ampersand) , ‘\’ (backslash)
- Alphanumeric: entries with only alphanumeric values. It should have text characters.
- Stopwords: entries that contain just words that add no value like of, for, the, or any such stopwords.
- Single term: entries that do not have a stemmed term for the root word.
Concepts
The concepts are handy for improving the relevance for long-tail searches (high intent) where shoppers are aware of what they want to buy and type all the details of the product that they are looking for.
For example, a shopper searching for “Red polka dot half sleeve dresses” is clearly telling the search system that they are looking for the half sleeve (sleeve-type) “dresses” (product type) with polka dot ( pattern) in red color (attribute). The challenge with some of these high intent searches is that your eCommerce store may not have products matching the exact requirement of the shopper. For example, the store may carry full sleeve polka dot dresses in red color or half sleeve polka dot dresses in blue and black color, etc.
Defining concepts enables the search system to identify the important features from the search query (for example “dress” in this query). The search restricts the result set to the products matching these important features while treating other query terms as optional. This allows your store to show products that are similar to the shopper’s search but do not fulfill all the requirements.
What can be added as Concepts?
We recommend adding all the product types from the catalog as concepts in order to achieve higher precision for searches where customers are searching using product types.
NOTE: Although concepts help in improving the precision of search they must be used judiciously. Adding too-many concepts can reduce the recall and may lead to zero results.
Depending upon your use-case you can include other catalog terms as Concepts. For example, if your shoppers are highly brand conscious and search for products using brand names then you can add all the brands in the catalog as Concepts to ensure that only products from the brands that are being searched are included in the result set. Let’s take another example from an e-commerce store that sells computer games (for example, Xbox games, ps4 games, etc.). For these stores, when a customer is searching for “Xbox lego batman” they are only interested in games that are compatible with “Xbox”. Showing a PS4 game is of no use to the shopper. In such scenarios, you can define Xbox, PS4 platforms as concepts because shoppers are looking for high precision when searching using these terms.
NOTE: Our Named Entity Recognition (NER) algorithm offers an intelligent way of handling long tail searches.
To add concepts:
- Navigate to Content > Concepts.
- Click the ‘hamburger’ icon > Bulk Upload Concepts. You can upload a list of concept terms by either browsing on your computer or by using drag and drop. The supported upload format is .csv.
- Unbxd’s AI will identify the concepts from your catalog and you could edit them as needed.
What should concepts not include:
- Empty record: entry where there is no value entered (whitespace)
- Symbols: entries where only symbols are typed as the value. The following characters are not accepted by our system : ‘ , ‘ ( comma) , ‘+’ (plus), ‘{‘, ‘}’ (curly braces), ‘*’ (Asterix),’&’ (Ampersand) , ‘\’ (backslash)
- Alphanumeric: entries with only alphanumeric values. It should have text characters.
- Single term: entries that do not have a long-tail query (more than two words).
Stopwords
Stop words are a pool of words that do not help in the ranking of results and therefore should be excluded. Ex. at, to, and, for, etc.
Prepositions and conjunctions are instances of commonly used functional words and phrases. Removing them from the search query will enhance indexing. Ex. Shoes for women could simply search for ‘women shoes’ and remove the extra word ‘for’ while searching. This improves the search engine’s efficiency.
To add stop words:
- Navigate to Content > Stop Words.
- Click the ‘hamburger’ icon > Bulk Upload Stop Words. You can upload a list of concept terms by either browsing on your computer or by using drag and drop. The supported upload format is .csv.
- Keep adding stop words by clicking ‘Add Stop Words’.
Once done, save the changes.
What should stopwords not include:
- Empty record: entry where there is no value entered (whitespace)
- Alphanumeric: entries with only alphanumeric values. It should have text characters.
Spellcheck
Spellcheck checks the misspelled queries to avoid zero results. By enabling spellcheck, your site displays results even if they type the wrong spelling of their intended query.
Spellcheck works based on the following factors:
- Calculates Levenshtein distance which checks up to 2 characters to identify suggestions.
- Uses the rest of the query content to generate search results
- It can only show recommendations using the terms present in the catalog.
To enable/disable spellcheck:
- Navigate to Content > Spellcheck.
- Enable or Disable based on the queries.
- Once done, spellcheck is set.
- Did this answer your question?
On this page
- Common Applications
- How to implement
- Example
- How to apply boost