PicoSearch FAQ: Influencing the search

PicoSearch will retrieve the best documents for a query by considering the quantity, rarity, and density of the search terms in your documents (see ranking method below).

Now, a search engine is supposed to be unbiased and thorough in pointing a user to all the documents for certain words, but what if you want to steer the user to certain pages? This could be important, for example if you have pages for a special sales promotion that you want visitors to see first, or a central FAQ, or a top page for everyone who types a common word. In these and other situations you will want to adjust PicoSearch's document relevancy response, otherwise known as the ranking order.

You can influence PicoSearch's document fetching decisions by any of the following methods, which we discuss roughly in order of increasingly fine tuning. See which options makes sense for you. PicoSearch offers an unusually powerful set of features to help guarantee that any site can be optimized for absolutely best results. Please contact us if you need any guidance, because we really do enjoy helping our customers to harness the full potential of their very own customizable PicoSearch.

Exclude unwanted documents, or unwanted parts of documents. If your search is returning documents that you don't even want found, then those should be excluded from the search engine. But more, you may want to exclude sections of documents from searchability, either manually per document, or automatically by tag based rules that can eliminate, for example, navigation text that too many pages share and thus returns too many results. PicoSearch offers plenty of tools for all of these situations, just check out the FAQ for details.
Increase the weighting on titles and metas... or search them only! In the Index Modes section of your account manager, you can choose which parts of your documents to search, and also set relative importance weightings for them. Increasing the weighting can help a lot if you want to emphasize the searching of titles for example, and if you turn off the body text to search only titles then of course the search results will change the most. These titles and metas can be specific to PicoSearch, see FAQ on display options. Note that if you do choose to cut your search area down to something as small as titles only, an alternate ranking method may be better suited as well.

Search by default in ALL or EXACT PHRASE. The default search is for ANY words, which means the user can search for one or more words most easily. If you want multiple words to be found more often all together, then you might want to preset the search to start in FIND ALL words or FIND EXACT phrase mode instead of FIND ANY, and this will affect the ranking a lot. See the Initializing FAQ.

Try an alternate ranking method. In the Ranking Options & Promotions section of your account manager, you can try an alternate ranking formula. Sorting by Weighted Relevancy is the classic information retreival formula that works great on lots of varied length text, statistically finding the most focussed documents with the most terms, and the most important terms, first.
Occasionally for certain sites, it may be immediately obvious that Weighted Relevancy isn't the best formula. For example, if your documents are all very short with mostly overlapping vocabulary (this might happen if you're searching only page titles or metas) then the weighting statistics may not find the most difference between documents. Then ranking by Frequency Relevancy to favor the number of hits may be more intuitive and desirable. It's a quick switch to try, so it's an easy experiment.

Constrain the searches by Partitions or entry formats. Partitions allow you to search only subsections of your site, so you can steer a user to the right area for the topic of their searches before the searches even start. See the Partitions FAQ. Or you can set up your searches in database-like formats, where only certain key-word combinations can be chosen from. This is an advanced trick but very powerful; see this FAQ on Multiple Input queries.

Set deliberate Promote and Demote options. Once a document appears in the search results, you can force it immediately to the top or bottom of the results page by Promotion and Demotion. This technique is powerful and should be reserved for most important documents and directories; if you over-use it, you may lose track of which documents were to be prioritized over others. You can promote and demote by patterned entries in your account manager, or by tags that can specify trigger words within the HTML of specific files. For details, see this FAQ: How can I keep certain URLs on top in the Search Results (using promote and demote).

Try setting Synonym Groups, Spelling Variants, or Query Rules. Professional and Premium accounts get powerful features to augment the actual queries that the user types. See the "Synonyms, Spelling, Query Rules" section in the account manager. British/American spelling variants will bring in pages with the alternate spelling to what is being searched, such as theater in American for theatre in British. Setting your own user Synonym Groups will bring in pages with related words that you decide, such as other word forms of corporation to include corporate and incorporated. If you need to normalize or otherwise adjust the queries in rule-based ways, the Query Rules feature can accept completely customizable patterns in regular expressions. All of these possibilities can be used to change the ranking for all pages that have the affected words. Or if you want to ensure that a certain page comes up for a common word, you could make a user synonym to a less common word that only that page has.

Focus queries with Stopwords and Less Important words. In your account manager, you can control the influence of words in user queries in two major ways. Stopwords are unimportant (noise) words that are entirely removed from the search, whereas Less Important words are still searched and show in concordance. Now typically users don't type noisy words, and even if they do then PicoSearch automatically diminishes their statistical impact (see ranking method). But for websites with unusual vocabulary or user expectations, these features enable you to customize search results even further. See our FAQ for details.
Direct Searches can reroute specific queries. One absolute way to take control of a particular search is to send it directly to the only page you want a user to see, thus bypassing the normal list of search results. PicoSearch offers this feature as well, see the FAQ on Direct Searches.

Add search terms visibly to the body text, or less visibly to the titles or metas. For the body of the document just make sure the words you want are found abundantly, especially near the top, and PicoSearch will rate that document highly for them. Adding terms to the title and metas can also be effective but less obvious to the user. Refer to the Index Modes section of your account manager for which parts are being searched.

Invisibly add search terms to the body text. If you don't want to change the text that the reader will see, you can still lure a search to the document by adding hidden words that only PicoSearch will notice. These extra words can include synonyms and variant spellings, so it's a good trick to catch more searches. Note that Professional and Premium accounts get a User Synonyms entry box in the account manager, and that will apply to all pages and all queries. For particular files, the ADDSEARCHTEXT tag is covered in our FAQ on How to Add Text or Links to a Document.

Skip a term in some documents to make it stand out more in others. If a word is very common in all of your documents then it may become watered down, having less influence in ranking and bringing up unrelated documents. If you want to skip the term in all of your searches, see the Stopwords feature in your account manager. But if the word is still crucial to certain searches, you may find that a combination of peppering it more in one document plus masking it out in others is needed to force a search result. Skipping a term selectively will make the word more rare and significant for fewer documents overall, as you might imagine. See our FAQ on skipping text at How to Skip Text or Links in a Document.

Turn word stemming on/off. By default, PicoSearch finds all plural and singular forms of the nouns in your searches, automatically. This is no simple hack; it is full grammatical database lookup in 14 different languages. But if you want you can turn it off, in the "Synonyms, Spelling, Query Rules" section of your account manager. Note that equating other grammatical forms (such as industry and industrial) can be handled with the User Synonym Groups.

In special cases, randomize your search results! It may seem odd to randomize the results of a search engine, but this can serve to continually juggle the document rankings, like for competing customers on a portal. Randomizing is done before Promotions and Demotions so that additional ranking can be enforced. For details, see this FAQ: What if I want to alternate the top results?.